.\" t '\" vim:set syntax=groff: .\"Copyright (C) 2009-2015 Kaz Kylheku . .\"All rights reserved. .\" .\"BSD License: .\" .\"Redistribution and use in source and binary forms, with or without .\"modification, are permitted provided that the following conditions .\"are met: .\" .\" 1. Redistributions of source code must retain the above copyright .\" notice, this list of conditions and the following disclaimer. .\" 2. Redistributions in binary form must reproduce the above copyright .\" notice, this list of conditions and the following disclaimer in .\" the documentation and/or other materials provided with the .\" distribution. .\" 3. The name of the author may not be used to endorse or promote .\" products derived from this software without specific prior .\" written permission. .\" .\"THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR .\"IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED .\"WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. .\" .\" Useful groff definitions. .\" .\" Some constants that depend on troff/nroff mode: .ie n \{\ .ds vspc 1 .\} .el \{\ .ds vspc 0.5 .\} .\" Mount numeric fonts when not running under man2html .if !\n(M2 \{\ . fp 4 CR . fp 5 CI .\} .\" Base font .nr fsav 1 .\" start of code block: switch to monospace font .de cblk . ft 4 .. .\" end of code block: restore font .de cble . ft 1 .. .\" typeset arguments in monospace .\" .code x y z -> \f[CR]x y z\f[] .de code \f[4]\\$*\f[] .. .\" like .code typesets meta-syntax .\" which is done in angle brackets in nroff or oblique .\" courier in PDF/HTML. .de meta . ie n \{\ <\\$*> . \} . el \{\ \f[5]\\$*\f[] . \} .. .\" like .meta but tack on second argument with no space. .de metn . ie n \{\ <\\$1>\\$2 . \} . el \{\ \f[5]\\$1\f[]\\$2 . \} .. .\" like .code but wraps in quotes .\" .str x y z -> \f[CR]"x y z"\f[]. .de str \f[4]"\\$*"\f[] .. .\" wrap first argument in quotes, tack no second one with no space .\" .strn x y -> \f[CR]"x"\f[]y. .de strn \f[4]"\\$1"\f[]\\$2 .. .\" like .IP but use monospace too .de coIP . IP "\\f[4]\\$*\\f[]" .. .\" Directive heading .de dir . SS* The \f[4]\\$1\f[] directive .. .\" Multiple directive heading .de dirs . ds s " . while (\\n[.$]>2) \{\ . as s \f[4]\\$1\f[], . shift . \} . if (\\n[.$]>1) \{\ . as s \f[4]\\$1\f[] . shift . \} . if (\\n[.$]>0) \{\ . as s and \f[4]\\$1\f[] . \} . SS* The \\*s directives .. .\" heading with code in position 1 .de c1SS . ds s \\f[4]\\$1\\f[] . shift . as s " \\$* . SS* \\*s .. .\" utility macro for gathering material into "s" string. .\" a pair of arguments "@ arg" becomes arg set in code .\" a pair of arguments "@, arg" becomes "arg," where .\" arg is set in code, followed by comma not in code. .de gets . ds s " . while (\\n[.$]>0) \{\ . ie "\\$1"@" \{\ . shift . as s \f[4]\\$1\f[] . shift . \} . el \{\ . ie "\\$1"@," \{\ . shift . as s \f[4]\\$1\f[], . shift . \} . el \{\ . as s \\$1 . shift . \} . \} . \} .. .\" a macro for gathering material into "s" .\" a pair of arguments "< arg" is typeset like .\" .meta arg. "<< arg arg" is like .metn arg arg. .ie n \{\ . de getm . ds s " . while (\\n[.$]>0) \{\ . ie "\\$1"<" \{\ . shift . as s <\\$1> . shift . \} . el \{\ . ie "\\$1"<<" \{\ . shift . as s <\\$1>\\$2 . shift . shift . \} . el \{\ . ie "\\$1">>" \{\ . shift . as s \\$1<\\$2> . shift . shift . \} . el \{\ . ie "\\$1"<>" \{\ . shift . as s \\$1<\\$2>\\$3 . shift . shift . shift . \} . el \{\ . as s \\$1 . shift . \} . \} . \} . \} . \} . . .\} .el \{\ . de getm . ds s " . while (\\n[.$]>0) \{\ . ie "\\$1"<" \{\ . shift . as s \\f5\\$1\\f4 . shift . \} . el \{\ . ie "\\$1"<<" \{\ . shift . as s \\f5\\$1\\f4\\$2 . shift . shift . \} . el \{\ . ie "\\$1">>" \{\ . shift . as s \\$1\\f5\\$2\\f4 . shift . shift . \} . el \{\ . ie "\\$1"<>" \{\ . shift . as s \\$1\\f5\\$2\f4\\$3 . shift . shift . shift . \} . el \{\ . as s \\$1 . shift . \} . \} . \} . \} . \} . . .\} .\" typeset all arguments by the last one in monospace, followed .\" by the last one in previous font with no space .\" .codn x y ... z -> \f[CR]x y ...\f[]z .de codn . ds s " . while (\\n[.$]>2) \{\ . as s \\$1 . shift . \} \f[4]\\*s\\$1\f[]\\$2 .. .\" .cod1 a b c -> abc where a is typeset as code .de cod1 \\$1\f[4]\\$2\f[]\$3 .. .\" .cod2 a b -> ab where b is typeset as code .de cod2 \\$1\f[4]\\$2\f[] .. .\" .cod3 a b c -> abc where a and c are typeset as code .de cod3 \f[4]\\$1\f[]\\$2\f[4]\\$3\f[] .. .\" Syntax section markup .de synb . TP* Syntax: . cblk .. .de syne . cble .. .\" Used for meta-variables in syntax blocks .de mets . nr fsav \\n[.f] . getm \\$* . \"workaround for man2html: . as s \\f\\n[fsav] \\*s . ft \\n[fsav] .. .\" Used for meta-variables in inline blocks .de meti . nr fsav \\n[.f] . getm \\$* . \"workaround for man2html: . as s \\f\\n[fsav] \\*s . ft \\n[fsav] .. .\" Used for meta-variables in .coIP .de meIP . nr fsav \\n[.f] . getm \\$* . \"workaround for man2html: . as s \\f\\n[fsav] .coIP \\*s . ft \\n[fsav] .. .\" Description section .de desc . TP* Description: .. .\" Section counters: heading, section, paragraph. .nr shco 0 1 .nr ssco 0 1 .nr spco 0 1 .\" wrapper for .SH .de SH* . SH \\n+[shco] \\$* . rs . nr ssco 0 . nr spco 0 . sp \*[vspc] . ns .. .\" wrapper for .SS .de SS* . SS \\n[shco].\\n+[ssco] \\$* . rs . nr spco 0 . sp \*[vspc] . ns .. .\" wrapper for .TP .de TP* . ds s \\$1 . shift . TP \\$* \\*s . sp \*[vspc] . ns .. .\" numbered paragraph .de NP* . ie \n(M2 \{\ . M2SS 2 H4 "\\n[shco].\\n[ssco].\\n+[spco] \\$*" . \} . el \{\ . TP* "\f[B]\\n[shco].\\n[ssco].\\n+[spco] \\$*\f[]" . \} . PP .. .\" process arguments using .gets so that some material .\" is typeset as code. Then pass to .SS* section macro. .de coSS . gets \\$* . SS* \\*s .. .\" like coSS but targeting NP* .de coNP . gets \\$* . NP* \\*s .. .\" like coSS but use monospace IP .de ccIP . gets \\$* . IP "\\*s" .. .\" TXR name .ds TX \f[B]TXR\f[] .ds TL \f[B]TXR Lisp\f[] .\" Start of man page: .TH TXR 1 2015-06-21 "Utility Commands" "TXR Data Processing Language" "Kaz Kylheku" .SH* NAME \*(TX \- text processing language (version 109) .SH* SYNOPSIS .cblk .meti txr >> [ options ] < query-file < data-files .. .cble .sp .SH* DESCRIPTION \*(TX is a language oriented toward processing text from files or streams, using multiple programming paradigms. A \*(TX script is called a query, and it specifies a pattern which matches (a prefix of) an entire file, or multiple files. Patterns can consists of large chunks of multi-line free-form text, which is matched literally against material in the input sources. Free variables occurring in the pattern (denoted by the .code @ symbol) are bound to the pieces of text occurring in the corresponding positions. If the overall match is successful, then \*(TX can do one of two things: it can report the list of variables which were bound, in the form of a set of variable assignments which can be evaluated by the .B eval command of the POSIX shell language, or generate a custom report according to special directives in the query. Patterns can be arbitrarily complex, and can be broken down into named pattern functions, which may be mutually recursive. \*(TX patterns can work horizontally (characters within a line) or vertically (spanning multiple lines). Multiple lines can be treated as a single line. In addition to embedded variables which implicitly match text, the \*(TX query language supports a number of directives, for matching text using regular expressions, for continuing a match in another file, for searching through a file for the place where an entire sub-query matches, for collecting lists, and for combining sub-queries using logical conjunction, disjunction and negation, and numerous others. Furthermore, embedded within \*(TX is a powerful Lisp dialect. \*(TL supports functional and imperative programming, and provides data types such as symbols, strings, vectors, hash tables with weak reference support, lazy lists, and arbitrary-precision (bignum integers). .SH* ARGUMENTS AND OPTIONS Options which don't take an argument may be combined together. The .code -v and .code -q options are mutually exclusive. Of these two, the one which occurs in the rightmost position in the argument list dominates. The .code -c and .code -f options are also mutually exclusive; if both are specified, it is a fatal error. .meIP >> -D var=value Bind the variable .meta var to the value .meta value prior to processing the query. The name is in scope over the entire query, so that all occurrence of the variable are substituted and match the equivalent text. If the value contains commas, these are interpreted as separators, which give rise to a list value. For instance .code -Da,b,c creates a list of the strings .strn "a" , .str "b" and .strn "c" . (See Collect Directive bellow). List variables provide a multiple match. That is to say, if a list variable occurs in a query, a successful match occurs if any of its values matches the text. If more than one value matches the text, the first one is taken. .meIP >> -D var Binds the variable .meta var to an empty string value prior to processing the query. .coIP -q Quiet operation during matching. Certain error messages are not reported on the standard error device (but the if the situations occur, they still fail the query). This option does not suppress error generation during the parsing of the query, only during its execution. .coIP -d .coIP --debugger Invoke the interactive \*(TX debugger. See the DEBUGGER section. .coIP -n .coIP --noninteractive This option affects behavior related to \*(TX's .code *std-input* stream. Normally, if this stream is connected to a terminal device, it is automatically marked as having the real-time property when \*(TX starts up (see the functions .code stream-set-prop and .codn real-time-stream-p ). The .code -n option suppresses this behavior; the .code *std-input* stream remains ordinary. The \*(TX pattern language reads standard input via a lazy list, created by applying the .code lazy-stream-cons function to the .code *std-input* stream. If that stream is marked real-time, then the lazy list which is returned by that function has behaviors that are better suited for scanning interactive input. A more detailed explanation is given under the description of this function. .coIP -v Verbose operation. Detailed logging is enabled. .coIP -b This is a deprecated option, which is silently ignored. In \*(TX versions prior to 90, the printing of variable bindings (see .code -B option) was implicit behavior which was automatically suppressed in certain situations. The -b option suppressed it unconditionally. .coIP -B If the query is successful, print the variable bindings as a sequence of assignments in shell syntax that can be .IR eval -ed by a POSIX shell. II the query fails, print the word "false". Evaluation of this word by the shell has the effect of producing an unsuccessful termination status from the shell's .I eval command. .coIP "-l or --lisp-bindings" This option implies .codn -B . Print the variable bindings in Lisp syntax instead of shell syntax. .meIP -a < num This option implies .codn -B . The decimal integer argument .meta num specifies the maximum number of array dimensions to use for list-valued variable bindings. The default is 1. Additional dimensions are expressed using numeric suffixes in the generated variable names. For instance, consider the three-dimensional list arising out of a triply nested collect: .cblk ((("a" "b") ("c" "d")) (("e" "f") ("g" "h"))). .cble Suppose this is bound to a variable V. With .codn -a 1 , this will be reported as: .cblk V_0_0[0]="a" V_0_1[0]="b" V_1_0[0]="c" V_1_1[0]="d" V_0_0[1]="e" V_0_1[1]="f" V_1_0[1]="g" V_1_1[1]="h" .cble With .codn -a 2 , it comes out as: .cblk V_0[0][0]="a" V_1[0][0]="b" V_0[0][1]="c" V_1[0][1]="d" V_0[1][0]="e" V_1[1][0]="f" V_0[1][1]="g" V_1[1][1]="h" .cble The leftmost bracketed index is the most major index. That is to say, the dimension order is: .codn "NAME_m_m+1_..._n[1][2]...[m-1]" . .meIP -c < query Specifies the query in the form of a command line argument. If this option is used, the query-file argument is omitted. The first non-option argument, if there is one, now specifies the first input source rather than a query. Unlike queries read from a file, (non-empty) queries specified as arguments using -c do not have to properly end in a newline. Internally, \*(TX adds the missing newline before parsing the query. Thus .code -c .str @a is a valid query which matches a line. Example: Shell script which uses \*(TX to read two lines .str 1 and .str 2 from standard input, binding them to variables .code a and .codn b . Standard input is specified as .code - and the data comes from shell "here document" redirection: .RS .IP code: .cblk \ #!/bin/sh txr -B -c "@a @b" - <> --compat= number Requests \*(TX to behave in a manner that is compatible with the specified version of \*(TX. This makes a difference in situations when a release of \*(TX breaks backward compatibility. If some version N+1 deliberately introduces a change which is backward incompatible, then .code -C N can be used to request the old behavior. The requested value of N can be too low, in which case \*(TX will complain and exit with an unsuccessful termination status. This indicates that \*(TX refuses to be compatible with such an old version. Users requiring the behavior of that version will have to install an older version of \*(TX which supports that behavior, or even that exact version. If the option is specified more than once, the behavior is not specified. For more information, see the COMPATIBILITY section. .meIP >> --gc-delta= number The .meta number argument to this option must be a decimal integer. It represents a megabyte value, the "GC delta": one megabyte is 1048576 bytes. The "GC delta" controls an aspect of the garbage collector behavior. See the .code gc-set-delta function for a description. .coIP --help Prints usage summary on standard output, and terminates successfully. .coIP --license Prints the software license. This depends on the software being installed such that the LICENSE file is in the data directory. Use of \*(TX implies agreement with the liability disclaimer in the license. .coIP --version Prints program version standard output, and terminates successfully. .coIP --args The .code --args option provides a way to encode multiple arguments as a single argument, which is useful on some systems which have limitations in their implementation of the "hash bang" mechanism. For details about its special syntax, See Hash Bang Support below. .coIP --lisp This option influences the treatment of query files which do not have a suffix indicating their type: they are treated as \*(TL source. Moreover, if .code --lisp is specified, and an unsuffixed file does not exist, then \*(TX will add the .str .tl suffix and try the file again. In the same situation, if .code --lisp is not present, \*(TX will first try adding the .str .txr suffix. If that fails, then .str .tl suffix will be tried. Note that .code --lisp influences how the argument of the .code -f option is treated, but only if it precedes that option. It has no effect on the .code -c option. The argument of .code -c is always \*(TX pattern language code. Lisp code can be evaluated using the .codn -e , .code -p , or .code -P options. .coIP -- Signifies the end of the option list. .coIP - This argument is not interpreted as an option, but treated as a filename argument. After the first such argument, no more options are recognized. Even if another argument looks like an option, it is treated as a name. This special argument .code - means "read from standard input" instead of a file. The query file, or any of the data files, may be specified using this option. If two or more files are specified as .codn - , the behavior is system-dependent. It may be possible to indicate EOF from the interactive terminal, and then specify more input which is interpreted as the second file, and so forth. .PP After the options, the remaining arguments are files. The first file argument specifies the query, and is mandatory if the .code -f option has not been specified. A file argument consisting of a single .code - means to read the standard input instead of opening a file. A file argument which begins with an exclamation symbol means that the rest of the argument is a shell command which is to be run as a coprocess, and its output read like a file. .PP \*(TX begins by reading the query. The entire query is scanned, internalized and then begins executing, if it is free of syntax errors. The reading of data, on the other hand, is lazy. A file isn't opened until the query demands material from that file, and then the contents are read on demand, not all at once. The suffix of the query file is significant. If the query file name has no suffix, or if it has a .str .txr suffix, then it is assumed to be in the \*(TX query language. If it has the .str .tl suffix, then it is assumed to be \*(TL. The .code --lisp option changes the treatment of unsuffixed query file names, causing them to be interpreted as \*(TL . If an unsuffixed query file name is specified, and cannot be opened, then \*(TX will add the .str .txr suffix and try again. If that fails, it will be tried with the .str .tl suffix, and treated as \*(TL . If the .code --lisp option has been specified, then \*(TX tries only the .str .tl suffix. A \*(TL file is processed as if by the .code load macro: forms from the file are read and evaluated. If the forms do not terminate the \*(TX process or throw an exception, and there are no syntax errors, then \*(TX terminates successfully after evaluating the last form. If syntax errors are encountered in a form, then \*(TX terminates unsuccessfully. \*(TL is documented in the section TXR LISP. If no files arguments are specified on the command line, it is up to the query to open a file, pipe or standard input via the .code @(next) directive prior to attempting to make a match. If a query attempts to match text, but has run out of files to process, the match fails. .SH* STATUS AND ERROR REPORTING \*(TX sends errors and verbose logs to the standard error device. The following paragraphs apply when \*(TX is run without enabling verbose mode with .codn -v , or the printing of variable bindings with .code -B or .codn -a . If the command line arguments are incorrect, or the query has a malformed syntax, \*(TX issues an error diagnostic and terminates with a failed status. If the query fails due to a mismatch, \*(TX terminates with a failed status. No diagnostics are issued. If the query is well-formed, and matches, then \*(TX issues no diagnostics, and terminates with a successful status. In verbose mode (option .codn -v ), \*(TX issues diagnostics on the standard error device even in situations which are not erroneous. In bindings-printing mode (options .code -B or .codn -a) , \*(TX prints the word .code false if the query fails, and exits with a failed termination status. If the query succeeds, the variable bindings, if any, are output on standard output. .SH* BASIC QUERY SYNTAX AND SEMANTICS .SS* Comments A query may contain comments which are delimited by the sequence .code @; and extend to the end of the line. No whitespace can occur between the .code @ and .codn ; . A comment which begins on a line swallows that entire line, as well as the newline which terminates it. In essence, the entire comment line disappears. If the comment follows some material in a line, then it does not consume the newline. Thus, the following two queries are equivalent: .IP 1. .cblk \ @a@; comment: match whole line against variable @a @; this comment disappears entirely @b .cble .IP 2. .cblk \ @a @b .cble .PP The comment after the .code @a does not consume the newline, but the comment which follows does. Without this intuitive behavior, line comment would give rise to empty lines that must match empty lines in the data, leading to spurious mismatches. Instead of the .code ; character, the .code # character can be used. This is an obsolescent feature. .SS* Hash Bang Support \*(TX has several features which support use of the "hash bang" convention for creating apparently stand-alone executable programs. If the first line of a query begins with the characters .codn #! , that entire line is deleted from the query. This allows for \*(TX queries to be turned into standalone executable programs in the POSIX environment. Shell example: create a simple executable program called .str "twoline.txr" and run it. This assumes \*(TX is installed in .codn /usr/bin . .cblk $ cat > hello.txr #!/usr/bin/txr @(bind a "Hey") @(output) Hello, world! @(end) $ chmod a+x hello.txr $ ./hello.txr Hello, world! .cble When this plain hash bang line is used, \*(TX receives the name of the script as an argument. Therefore, it is not possible to pass additional options to \*(TX. For instance, if the above script is invoked like this .cblk $ ./hello.txr -B .cble the -B option isn't processed by \*(TX, but treated as an additional argument, just as if .cblk .meti txr < scriptname -B .cble had been executed directly. This behavior is useful if the script author wants not to expose the \*(TX options to the user of the script. However, the hash bang line can use the .code -f option: .cblk #!/usr/bin/txr -f .cble Now, the name of the script is passed as an argument to the .code -f option, and \*(TX will look for more options after that, so that the resulting program appears to accept \*(TX options. Now we can run .cblk $ ./hello.txr -B Hello, world! a="Hey" .cble The .code -B option is honored. On some operating systems, it is not possible to pass more than one argument through the hash bang mechanism. That is to say, this will not work. .cblk #!/usr/bin/txr -B -f .cble To support systems like this, \*(TX supports the special argument .codn --args . With .codn --args , it is possible to encode multiple arguments into one argument. The .code --args option must be followed by a separator character, chosen by the programmer. The characters after that are split into multiple arguments on the separator character. The .code --args option is then removed from the argument list and replaced with these arguments, which are processed in its place. Example: .cblk #!/usr/bin/txr --args:-B:-f .cble The above has the same behavior as .cblk #!/usr/bin/txr -B -f .cble on a system which supports multiple arguments in hash bang. The separator character is the colon, and so the remainder of that argument, .codn -B:-f , is split into the two arguments .codn "-B -f" . .SS* Whitespace Outside of directives, whitespace is significant in \*(TX queries, and represents a pattern match for whitespace in the input. An extent of text consisting of an undivided mixture of tabs and spaces is a whitespace token. Whitespace tokens match a precisely identical piece of whitespace in the input, with one exception: a whitespace token consisting of precisely one space has a special meaning. It is equivalent to the regular expression .codn "@/[ ]+/" : match an extent of one or more spaces (but not tabs!). Multiple consecutive spaces do not have this meaning. Thus, the query line .str "a b" (one space between .code a and .codn b ) matches .str "a b" with any number of spaces between the two letters. For matching a single space, the syntax .code "@\e " can be used (backslash-escaped space). It is more often necessary to match multiple spaces than to exactly match one space, so this rule simplifies many queries and adds inconvenience to only few. In output clauses, string and character literals and quasiliterals, a space token denotes a space. .SS* Text Query material which is not escaped by the special character .code @ is literal text, which matches input character for character. Text which occurs at the beginning of a line matches the beginning of a line. Text which starts in the middle of a line, other than following a variable, must match exactly at the current position, where the previous match left off. Moreover, if the text is the last element in the line, its match is anchored to the end of the line. An empty query line matches an empty line in the input. Note that an empty input stream does not contain any lines, and therefore is not matched by an empty line. An empty line in the input is represented by a newline character which is either the first character of the file, or follows a previous newline-terminated line. Input streams which end without terminating their last line with a newline are tolerated, and are treated as if they had the terminator. Text which follows a variable has special semantics, discussed in the section Variables below. A query may not leave a line of input partially matched. If any portion of a line of input is matched, it must be entirely matched, otherwise a matching failure results. However, a query may leave unmatched lines. Matching only four lines of a ten line file is not a matching failure. The .code eof directive can be used to explicitly match the end of a file. In the following example, the query matches the text, even though the text has an extra line. .IP code: .cblk \ Four score and seven years ago our .cble .IP data: .cblk \ Four score and seven years ago our forefathers .cble .PP In the following example, the query .B fails to match the text, because the text has extra material on one line that is not matched: .IP code: .cblk \ I can carry nearly eighty gigs in my head .cble .IP data: .cblk \ I can carry nearly eighty gigs of data in my head .cble .PP Needless to say, if the text has insufficient material relative to the query, that is a failure also. To match arbitrary material from the current position to the end of a line, the "match any sequence of characters, including empty" regular expression .code @/.*/ can be used. Example: .IP code: .cblk \ I can carry nearly eighty gigs@/.*/ .cble .IP data: .cblk \ I can carry nearly eighty gigs of data .cble .PP In this example, the query matches, since the regular expression matches the string "of data". (See Regular Expressions section below). Another way to do this is: .IP code: .cblk \ I can carry nearly eighty gigs@(skip) .cble .SS* Special Characters in Text Control characters may be embedded directly in a query (with the exception of newline characters). An alternative to embedding is to use escape syntax. The following escapes are supported: .meIP >> @\e newline A backslash immediately followed by a newline introduces a physical line break without breaking up the logical line. Material following this sequence continues to be interpreted as a continuation of the previous line, so that indentation can be introduced to show the continuation without appearing in the data. .meIP >> @\e space A backslash followed by a space encodes a space. This is useful in line continuations when it is necessary for some or all of the leading spaces to be preserved. For instance the two line sequence .cblk abcd@\e @\e efg .cble is equivalent to the line .cblk abcd efg .cble The two spaces before the .code @\e in the second line are consumed. The spaces after are preserved. .coIP @\ea Alert character (ASCII 7, BEL). .coIP @\eb Backspace (ASCII 8, BS). .coIP @\et Horizontal tab (ASCII 9, HT). .coIP @\en Line feed (ASCII 10, LF). Serves as abstract newline on POSIX systems. .coIP @\ev Vertical tab (ASCII 11, VT). .coIP @\ef Form feed (ASCII 12, FF). This character clears the screen on many kinds of terminals, or ejects a page of text from a line printer. .coIP @\er Carriage return (ASCII 13, CR). .coIP @\ee Escape (ASCII 27, ESC) .meIP @\ex < hex-digits A .code @\ex immediately followed by a sequence of hex digits is interpreted as a hexadecimal numeric character code. For instance .code @\ex41 is the ASCII character A. If a semicolon character immediately follows the hex digits, it is consumed, and characters which follow are not considered part of the hex escape even if they are hex digits. .meIP @\e < octal-digits A .code @\e immediately followed by a sequence of octal digits (0 through 7) is interpreted as an octal character code. For instance .code @\e010 is character 8, same as .codn @\eb . If a semicolon character immediately follows the octal digits, it is consumed, and subsequent characters are not treated as part of the octal escape, even if they are octal digits. .PP Note that if a newline is embedded into a query line with .code @\en, this does not split the line into two; it's embedded into the line and thus cannot match anything. However, .code @\en may be useful in the .code @(cat) directive and in .codn @(output) . .SS* Character Handling and International Characters \*(TX represents text internally using wide characters, which are used to represent Unicode code points. The query language, as well as all data sources, are assumed to be in the UTF-8 encoding. In the query language, extended characters can be used directly in comments, literal text, string literals, quasiliterals and regular expressions. Extended characters can also be expressed indirectly using hexadecimal or octal escapes. On some platforms, wide characters may be restricted to 16 bits, so that \*(TX can only work with characters in the BMP (Basic Multilingual Plane) subset of Unicode. \*(TX does not use the localization features of the system library; its handling of extended characters is not affected by environment variables like .code LANG and .codn L_CTYPE . The program reads and writes only the UTF-8 encoding. If \*(TX encounters an invalid bytes in the UTF-8 input, what happens depends on the context in which this occurs. In a query, comments are read without regard for encoding, so invalid encoding bytes in comments are not detected. A comment is simply a sequence of bytes terminated by a newline. In lexical elements which represent text, such as string literals, invalid or unexpected encoding bytes are treated as syntax errors. The scanner issues an error message, then discards a byte and resumes scanning. Certain sequences pass through the scanner without triggering an error, namely some UTF-8 overlong sequences. These are caught when when the lexeme is subject to UTF-8 decoding, and treated in the same manner as other UTF-8 data, described in the following paragraph. Invalid bytes in data are treated as follows. When an invalid byte is encountered in the middle of a multibyte character, or if the input ends in the middle of a multibyte character, or if a character is extracted which is encoded as an overlong form, the UTF-8 decoder returns to the starting byte of the ill-formed multibyte character, and extracts just that byte, mapping it to the Unicode character range U+DC00 through U+DCFF. The decoding resumes afresh at the following byte, expecting that byte to be the start of a UTF-8 code. Furthermore, because \*(TX internally uses a null-terminated character representation of strings which easily interoperates with C language interfaces, when a null character is read from a stream, \*(TX converts it to the code U+DC00. On output, this code converts back to a null byte, as explained in the previous paragraph. By means of this representational trick, \*(TX can handle textual data containing null bytes. .SS* Regular Expression Directives In place of a piece of text (see section Text above), a regular expression directive may be used, which has the following syntax: .cblk @/RE/ .cble where the RE part enclosed in slashes represents regular expression syntax (described in the section Regular Expressions below). Long regular expressions can be broken into multiple lines using a backslash-newline sequence. Whitespace before the sequence or after the sequence is not significant, so the following two are equivalent: .cblk @/reg \e ular/ @/regular/ .cble There may not be whitespace between the backslash and newline. Whereas literal text simply represents itself, regular expression denotes a (potentially infinite) set of texts. The regular expression directive matches the longest piece of text (possibly empty) which belongs to the set denoted by the regular expression. The match is anchored to the current position; thus if the directive is the first element of a line, the match is anchored to the start of a line. If the regular expression directive is the last element of a line, it is anchored to the end of the line also: the regular expression must match the text from the current position to the end of the line. Even if the regular expression matches the empty string, the match will fail if the input is empty, or has run out of data. For instance suppose the third line of the query is the regular expression .codn @/.*/ , but the input is a file which has only two lines. This will fail: the data has no line for the regular expression to match. A line containing no characters is not the same thing as the absence of a line, even though both abstractions imply an absence of characters. Like text which follows a variable, a regular expression directive which follows a variable has special semantics, discussed in the section Variables below. .SS* Variables Much of the query syntax consists of arbitrary text, which matches file data character for character. Embedded within the query may be variables and directives which are introduced by a .code @ character. Two consecutive .code @@ characters encode a literal .codn @ . A variable matching or substitution directive is written in one of several ways: .cblk .mets >> @ sident .mets <> @{ bident } .mets >> @* sident .mets <> @*{ bident } .mets >> @{ bident <> / regex /} .mets >> @{ bident >> ( fun >> [ arg ... ])} .mets >> @{ bident << number } .cble The forms with an .code * indicate a long match, see Longest Match below. The last two three forms with the embedded regexp .cblk .meti <> / regex / .cble or .meta number or function have special semantics; see Positive Match below. The identifier .code t cannot be used as a name; it is a reserved symbol which denotes the value true. An attempt to use the variable .code @t will result in an exception. The symbol .code nil can be used where a variable name is required syntactically, but it has special semantics, described in a section below. A .meta sident is a "simple identifier" form which is not delimited by braces. A .meta sident consists of any combination of one or more letters, numbers, and underscores. It may not look like a number, so that for instance .code 123 is not a valid .metn sident , but .code 12A is valid. Case is sensitive, so that .code FOO is different from .codn foo , which is different from .codn Foo . The braces around an identifier can be used when material which follows would otherwise be interpreted as being part of the identifier. When a name is enclosed in braces it is a .metn bident . The following additional characters may be used as part of .meta bident which are not allowed in a .metn sident : .cblk ! $ % & * + - < = > ? \e _ ~ .cble The rule still holds that a name cannot look like a number so .code +123 is not a valid .meta bident but these are valid: .codn a->b , .codn *xyz* , .codn foo-bar . The syntax .code @FOO_bar introduces the name .codn FOO_bar , whereas .code @{FOO}_bar means the variable named .str FOO followed by the text .strn _bar . There may be whitespace between the .code @ and the name, or opening brace. Whitespace is also allowed in the interior of the braces. It is not significant. If a variable has no prior binding, then it specifies a match. The match is determined from some current position in the data: the character which immediately follows all that has been matched previously. If a variable occurs at the start of a line, it matches some text at the start of the line. If it occurs at the end of a line, it matches everything from the current position to the end of the line. .SS* Negative Match If a variable is one of the plain forms .cblk .mets >> @ sident .mets <> @{ bident } .mets >> @* sident .mets <> @*{ bident } .cble then this is a "negative match". The extent of the matched text (the text bound to the variable) is determined by looking at what follows the variable, and ranges from the current position to some position where the following material finds a match. This is why this is called a "negative match": the spanned text which ends up bound to the variable is that in which the match for the trailing material did not occur. A variable may be followed by a piece of text, a regular expression directive, a function call, a directive, another variable, or nothing (i.e. occurs at the end of a line). These cases are discussed in detail below. .NP* Variable Followed by Nothing If the variable is followed by nothing, the negative match extends from the current position in the data, to the end of the line. Example: .IP code: .cblk \ a b c @FOO .cble .IP data: .cblk \ a b c defghijk .cble .IP result: .cblk \ FOO="defghijk" .cble .NP* Variable Followed by Text For the purposes of determining the negative match, text is defined as a sequence of literal text and regular expressions, not divided by a directive. So for instance in this example: .cblk @a:@/foo/bcd e@(maybe)f@(end) .cble .PP the variable @a is considered to be followed by .strn ":@/foo/bcd e" . If a variable is followed by text, then the extent of the negative match is determined by searching for the first occurrence of that text within the line, starting at the current position. The variable matches everything between the current position and the matching position (not including the matching position). Any whitespace which follows the variable (and is not enclosed inside braces that surround the variable name) is part of the text. For example: .IP code: .cblk \ a b @FOO e f .cble .IP data: .cblk \ a b c d e f .cble .IP result: .cblk \ FOO="c d" .cble .PP In the above example, the pattern text .str "a b " matches the data .strn "a b " . So when the .code @FOO variable is processed, the data being matched is the remaining .strn "c d e f" . The text which follows .code @FOO is .strn " e f" . This is found within the data .str "c d e f" at position 3 (counting from 0). So positions 0-2 .cblk ("c d") .cble constitute the matching text which is bound to FOO. .NP* Variable Followed by a Function Call or Directive If the variable is followed by a function call, or a directive, the extent is determined by scanning the text for the first position where a match occurs for the entire remainder of the line. (For a description of functions, see FUNCTIONS.) For example: .cblk @foo@(bind a "abc")xyz .cble Here, .code foo will match the text from the current position to where .str "xyz" occurs, even though there is a .code @(bind) directive. Furthermore, if more material is added after the xyz, it is part of the search. Note the difference between the following two: .cblk @foo@/abc/@(func) @foo@(func)@/abc/ .cble In the first example, the variable foo matches the text from the current position until the match for the regular expression abc. .code @(func) is not considered when processing .codn @foo . In the second example, the variable foo matches the text from the current position until the position which matches the function call, followed by a match for the regular expression. The entire sequence .code @(func)@/abc/ is considered. .NP* Consecutive Variables If an unbound variable specified a fixed-width match or a regular expression, then the issue of consecutive variables does not arise. Such a variable consumes text regardless of any context which follows it. However, what if an unbound variable with no modifier is followed by another variable? The behavior depends on the nature of the other variable. If the other variable also has no modifier, this is a semantic error which will cause the query to fail. A diagnostic message will be issued, unless operating in quiet mode via .codn -q . The reason is that there is no way to bind two consecutive variables to an extent of text; this is an ambiguous situation, since there is no matching criterion for dividing the text between two variables. (In theory, a repetition of the same variable, like .codn @FOO@FOO , could find a solution by dividing the match extent in half, which would work only in the case when it contains an even number of characters. This behavior seems to have dubious value). An unbound variable may be followed by one which is bound. The bound variable is replaced by the text which it denotes, and the logic proceeds accordingly. It is possible for a variable to be bound to a regular expression. If .code x is an unbound variable and .code y is bound to a regular expression .codn RE , then .code @x@y means .codn @x@/RE/ . A variable .code v can be bound to a regular expression using, for example, .codn @(bind v #/RE/) . The .code @* syntax for longest match is available. Example: .IP code: .cblk \ @FOO:@BAR@FOO .cble .IP data: .cblk \ xyz:defxyz .cble .IP result: .cblk \ FOO=xyz, BAR=def .cble .PP Here, .code FOO is matched with .strn "xyz" , based on the delimiting around the colon. The colon in the pattern then matches the colon in the data, so that .code BAR is considered for matching against .strn "defxyz" . .code BAR is followed by .codn FOO , which is already bound to .strn "xyz" . Thus .str "xyz" is located in the .str "defxyz" data following .strn "def" , and so BAR is bound to .strn "def" . If an unbound variable is followed by a variable which is bound to a list, or nested list, then each character string in the list is tried in turn to produce a match. The first match is taken. An unbound variable may be followed by another unbound variable which specifies a regular expression or function call match. This is a special case called a "double variable match". What happens is that the text is searched using the regular expression or function. If the search fails, than neither variable is bound: it is a matching failure. If the search succeeds, than the first variable is bound to the text which is skipped by the search. The second variable is bound to the text matched by the regular expression or function. Examples: .IP code: .cblk \ @foo@{bar /abc/} .cble .IP data: .cblk \ xyz@#abc .cble .IP result: .cblk \ foo="xyz@#", BAR="abc" .cble .PP .NP* Consecutive Variables Via Directive Two variables can be de facto consecutive in a manner shown in the following example: .cblk @var1@(all)@var2@(end) .cble This is treated just like the variable followed by directive. No semantic error is identified, even if both variables are unbound. Here, .code @var2 matches everything at the current position, and so .code @var1 ends up bound to the empty string. Example 1: .code b matches at position 0 and .code a binds the empty string: .IP code: .cblk \ @a@(all)@b@(end) .cble .IP data: .cblk \ abc .cble .IP result: .cblk \ a="" b="abc" .cble .PP Example 2: .code *a specifies longest match (see Longest Match below), and so it takes everything: .IP code: .cblk \ @*a@(all)@b@(end) .cble .IP data: .cblk \ abc .cble .IP result: .cblk \ a="abc" b="" .cble .PP .NP* Longest Match The closest-match behavior for the negative match can be overridden to longest match behavior. A special syntax is provided for this: an asterisk between the .code @ and the variable, e.g: .IP code: .cblk \ a @*{FOO}cd .cble .IP data: .cblk \ a b cdcdcdcd .cble .IP result: .cblk \ FOO="b cdcdcd" .cble .PP .IP code: .cblk \ a @{FOO}cd .cble .IP data: .cblk \ a b cdcdcd .cble .IP result: .cblk \ FOO="b " b="" .cble .PP In the former example, the match extends to the rightmost occurrence of .strn "cd" , and so .code FOO receives .strn "b cdcdcd" . In the latter example, the .code * syntax isn't used, and so a leftmost match takes place. The extent covers only the .strn "b " , stopping at the first .str "cd" occurrence. .SS* Positive Match There are syntactic variants of variable syntax which have an embedded expression enclosed with the variable in braces: .cblk .mets >> @{ bident <> / regex /} .mets >> @{ bident >> ( fun >> [args ...])} .mets >> @{ bident << number } .cble These specify a variable binding that is driven by a positive match derived from a regular expression, function or character count, rather than from trailing material (which is regarded as a "negative" match, since the variable is bound to material which is .B skipped in order to match the trailing material). In the .cblk .meti <> / regex / .cble form, the match extends over all characters from the current position which match the regular expression .metn regex . (see Regular Expressions section below). In the .cblk .meti >> ( fun >> [ args ...]) .cble form, the match extends over characters which are matched by the call to the function, if the call succeeds. Thus .code @{x (y z w)} is just like .codn "@(y z w)" , except that the region of text skipped over by .code @(y z w) is also bound to the variable .codn x . See FUNCTIONS below. In the .meta number form, the match processes a field of text which consists of the specified number of characters, which must be non-negative number. If the data line doesn't have that many characters starting at the current position, the match fails. A match for zero characters produces an empty string. The text which is actually bound to the variable is all text within the specified field, but excluding leading and trailing whitespace. If the field contains only spaces, then an empty string is extracted. This syntax is processed without consideration of what other syntax follows. A positive match may be directly followed by an unbound variable. .coSS Special Symbols @ nil and @ t Just like in the Common Lisp language, the names .code nil and .code t are special. .code nil symbol stands for the empty list object, an object which marks the end of a list, and boolean false. It is synonymous with the syntax .code () which may be used interchangeably with .code nil in most constructs. In \*(TL, .code nil and .code t cannot be used as variables. When evaluated, they evaluate to themselves. In the \*(TX pattern language, .code nil can be used in the variable binding syntax, but does not create a binding; it has a special meaning. It allows the variable matching syntax to be used to skip material, in ways similar to the .codn skip directive. The .code nil symbol is also used as a .code block name, both in the \*(TX pattern language and in \*(TL. A block named .code nil is considered to be anonymous. .SS* Keyword Symbols Names whose names begin with the .code : character are keyword symbols. These also may not be used as variables either and stand for themselves. Keywords are useful for labeling information and situations. .SS* Regular Expressions Regular expressions are a language for specifying sets of character strings. Through the use of pattern matching elements, regular expression is able to denote an infinite set of texts. \*(TX contains an original implementation of regular expressions, which supports the following syntax: .coIP . The period is a "wildcard" that matches any character. .coIP [] Character class: matches a single character, from the set specified by special syntax written between the square brackets. This supports basic regexp character class syntax. POSIX notation like .code [:digit:] is not supported. The regex tokens .codn \es , .code \ed and .code \ew are permitted in character classes, but not their complementing counterparts. These tokens simply contribute their characters to the class. The class .code [a-zA-Z] means match an uppercase or lowercase letter; the class .code [0-9a-f] means match a digit or a lowercase letter; the class .code [^0-9] means match a non-digit, and so forth. There are no locale-specific behaviors in \*(TX regular expressions; .code [A-Z] denotes an ASCII/Unicode range of characters. The class .code [\ed.] means match a digit or the period character. A .code ] or .code - can be used within a character class, but must be escaped with a backslash. A .code ^ in the first position denotes a complemented class, unless it is escaped by backslash. In any other position, it denotes itself. Two backslashes code for one backslash. So for instance .code [\e[\e-] means match a .code [ or .code - character, .code [^^] means match any character other than .codn ^ , and .code [\e^\e\e] means match either a .code ^ or a backslash. Regex operators such as .codn * , .code + and .code & appearing in a character class represent ordinary characters. The characters .codn - , .code ] and .code ^ occurring outside of a character class are ordinary. Unescaped .code / characters can appear within a character class. The empty character class .code [] matches no character at all, and its complement .code [^] matches any character, and is treated as a synonym for the .code . (period) wildcard operator. .ccIP @, \es @ \ew and @ \ed These regex tokens each match a single character. The .code \es regex token matches a wide variety of ASCII whitespace characters and Unicode spaces. The .code \ew token matches alphabetic word characters; it is equivalent to the character class .codn [A-Za-z_] . The .code \ed token matches a digit, and is equivalent to .codn [0-9] . .ccIP @, \eS @ \eW and @ \eD These regex tokens are the complemented counterparts of .codn \es , .code \ew and .codn \ed . The .code \eS token matches all those characters which .code \es does not match, .code \eW matches all characters that .code \ew does not match and .code \eD matches nondigits. .coIP empty An empty expression is a regular expression. It represents the set of strings consisting of the empty string; i.e. it matches just the empty string. The empty regex can appear alone as a full regular expression (for instance the \*(TX syntax .code @// with nothing between the slashes) and can also be passed as a subexpression to operators, though this may require the use of parentheses to make the empty regex explicit. For example, the expression .code a| means: match either .codn a , or nothing. The forms .code * and .code (*) are syntax errors; though not useful, the correct way to match the empty expression zero or more times is the syntax .codn ()* . .coIP nomatch The nomatch regular expression represents the empty set: it matches no strings at all, not even the empty string. There is no dedicated syntax to directly express nomatch in the regex language. However, the empty character class .code [] is equivalent to nomatch, and may be considered to be a notation for it. Other representations of nomatch are possible: for instance, the regex .code ~.* which is the complement of the regex that denotes the set of all possible strings, and thus denotes the empty set. A nomatch has uses; for instance, it can be used to temporarily "comment out" regular expressions. The regex .code ([]abc|xyz) is equivalent to .codn (xyz) , since the .code []abc branch cannot match anything. Using .code [] to "block" a subexpression allows you to leave it in place, then enable it later by removing the "block". .coIP (R) If .code R is a regular expression, then so is .code (R). The contents of parentheses denote one regular expression unit, so that for instance in .codn (RE)* , the .code * operator applies to the entire parenthesized group. The syntax .code () is valid and equivalent to the empty regular expression. .coIP R? Optionally match the preceding regular expression .codn R . .coIP R* Match the expression .code R zero or more times. This operator is sometimes called the "Kleene star", or "Kleene closure". The Kleene closure favors the longest match. Roughly speaking, if there are two or more ways in which .code R1*R2 can match, than that match occurs in which .code R1* matches the longest possible text. .coIP R+ Match the preceding expression .code R one or more times. Like .codn R* , this favors the longest possible match: .code R+ is equivalent to .codn RR* . .coIP R1%R2 Match .code R1 zero or more times, then match .codn R2 . If this match can occur in more than one way, then it occurs such that .code R1 is matched the fewest number of times, which is opposite from the behavior of .codn R1*R2 . Repetitions of .code R1 terminate at the earliest point in the text where a non-empty match for .code R2 occurs. Because it favors shorter matches, .code % is termed a non-greedy operator. If .code R2 is the empty expression, or equivalent to it, then .code R1%R2 reduces to . codn R1* . So for instance .code (R%) is equivalent to .codn (R*) , since the missing right operand is interpreted as the empty regex. Note that whereas the expression .code (R1*R2) is equivalent to .codn (R1*)R2 , the expression .code (R1%R2) is .B not equivalent to .codn (R1%)R2 . .coIP ~R Match the opposite of the following expression .codn R ; that is, match exactly those texts that .code R does not match. This operator is called complement, or logical not. .coIP R1R2 Two consecutive regular expressions denote catenation: the left expression must match, and then the right. .coIP R1|R2 match either the expression .code R1 or .codn R2 . This operator is known by a number of names: union, logical or, disjunction, branch, or alternative. .coIP R1&R2 Match both the expression .code R1 and .code R2 simultaneously; i.e. the matching text must be one of the texts which are in the intersection of the set of texts matched by .code R1 and the set matched by .codn R2 . This operator is called intersection, logical and, or conjunction. .PP Any character which is not a regular expression operator, a backslash escape, or the slash delimiter, denotes one-position match of that character itself. Any of the special characters, including the delimiting .codn / , and the backslash, can be escaped with a backslash to suppress its meaning and denote the character itself. Furthermore, all of the same escapes as are described in the section Special Characters in Text above are supported - the difference is that in regular expressions, the .code @ character is not required, so for example a tab is coded as .code \et rather than .codn @\et . Octal and hex character escapes can be optionally terminated by a semicolon, which is useful if the following characters are octal or hex digits not intended to be part of the escape. Only the above escapes are supported. Unlike in some other regular expression implementations, if a backlash appears before a character which isn't a regex special character or one of the supported escape sequences, it is an error. This wasn't true of historic versions of \*(TX. See the Compatibility section. .IP "Precedence table, highest to lowest:" .TS tab(!); l l l. Operators!Class!Associativity \f[4](R) []\f[]!primary! \f[4]R? R+ R* R%...\f[]!postfix!left-to-right \f[4]R1R2\f[]!catenation!left-to-right \f[4]~R ...%R\f[]\f[]\f[]!unary!right-to-left \f[4]R1&R2\f[]!intersection!left-to-right \f[4]R1|R2\f[]!union!left-to-right .TE .PP The .code % operator is like a postfix operator with respect to its left operand, but like a unary operator with respect to its right operand. Thus .code a~b%c~d is .cblk a(~(b%(c(~d)))) .cble , demonstrating right-to-left associativity, where all of .code b% may be regarded as a unary operator being applied to .codn c~d . Similarly, .code a?*+%b means .codn (((a?)*)+)%b , where the trailing .code %b behaves like a postfix operator. In \*(TX, regular expression matches do not span multiple lines. The regex language has no feature for multi-line matching. However, the .code @(freeform) directive allows the remaining portion of the input to be treated as one string in which line terminators appear as explicit characters. Regular expressions may freely match through this sequence. It's possible for a regular expression to match an empty string. For instance, if the next input character is .codn z , facing a the regular expression .codn /a?/ , there is a zero-character match: the regular expression's state machine can reach an acceptance state without consuming any characters. Examples: .IP code: .cblk \ @A@/a?/@/.*/ .cble .IP data: .cblk \ zzzzz .cble .IP result: .cblk \ A="" .cble .PP .IP code: .cblk \ @{A /a?/}@B .cble .IP data: .cblk \ zzzzz .cble .IP result: .cblk \ A="", B="zzzz" .cble .PP .IP code: .cblk \ @*A@/a?/ .cble .IP data: .cblk \ zzzzz .cble .IP result: .cblk \ A="zzzzz" .cble .PP In the first example, variable .code @A is followed by a regular expression which can match an empty string. The expression faces the letter .code "z" at position 0 in the data line. A zero-character match occurs there, therefore the variable .code A takes on the empty string. The .code @/.*/ regular expression then consumes the line. Similarly, in the second example, the .code /a?/ regular expression faces a .codn "z" , and thus yields an empty string which is bound to .codn A . Variable .code @B consumes the entire line. The third example requests the longest match for the variable binding. Thus, a search takes place for the rightmost position where the regular expression matches. The regular expression matches anywhere, including the empty string after the last character, which is the rightmost place. Thus variable .code A fetches the entire line. For additional information about the advanced regular expression operators, NOTES ON EXOTIC REGULAR EXPRESSIONS below. .SS* Directives The general syntax of a directive is: .cblk .mets >> @ expr .cble where .meta expr stands for a parenthesized list of subexpressions. A subexpression is a symbol, number, string literal, character literal, quasiliteral, regular expression, or a parenthesized expression. So, examples of syntactically valid directives are: .cblk @(banana) @(a b c (d e f)) @( a (b (c d) (e ) )) @("apple" #\eb #\espace 3) @(a #/[a-z]*/ b) @(_ `@file.txt`) .cble A symbol has a slight more permissive lexical syntax than the .meta bident in the syntax .cblk .meti <> @{ bident } .cble introduced earlier. The .code / (slash) character may be part of an identifier, or even constitute an entire identifier. In fact a symbol inside a directive is a .metn lident . This is discussed in the Symbol Tokens section under TXR LISP. A symbol must not be a number; tokens that look like numbers are treated as numbers and not symbols. .SS* Character Literals Character literals are introduced by the .code #\e syntax, which is either followed by a character name, the letter .code x followed by hex digits, the letter .code o followed by octal digits, or a single character. Valid character names are: .cblk nul linefeed return alarm newline esc backspace vtab space tab page pnul .cble For instance .code #\eesc denotes the escape character. This convention for character literals is similar to that of the Scheme language. Note that .code #\elinefeed and .code #\enewline are the same character. The .code #\epnul character is specific to \*(TX and denotes the .code U+DC00 code in Unicode; the name stands for "pseudo-null", which is related to its special function. For more information about this, see the section "Character Handling and International Characters". .SS* String Literals String literals are delimited by double quotes. A double quote within a string literal is encoded using .cblk \e" .cble and a backslash is encoded as .codn \e\e . Backslash escapes like .code \en and .code \et are recognized, as are hexadecimal escapes like .code \exFF or .code \exxabc and octal escapes like .codn \e123 . Ambiguity between an escape and subsequent text can be resolved by using trailing semicolon delimiter: .str "\exabc;d" is a string consisting of the character .code "U+0ABC" followed by .strn "d" . The semicolon delimiter disappears. To write a literal semicolon immediately after a hex or octal escape, write two semicolons, the first of which will be interpreted as a delimiter. Thus, .str "\ex21;;" represents .strn "!;" . If the line ends in the middle of a literal, it is an error, unless the last character is a backslash. This backslash is a special escape which does not denote a character; rather, it indicates that the string literal continues on the next line. The backslash is deleted, along with whitespace which immediately precedes it, as well as leading whitespace in the following line. The escape sequence .str "\e " (backslash space) can be used to encode a significant space. Example: .cblk "foo \e bar" "foo \e \e bar" "foo\e \e bar" .cble The first string literal is the string .strn "foobar" . The second two are .strn "foo bar" . .SS* Word List Literals A word list literal (WLL) provides a convenient way to write a list of strings when such a list can be given as whitespace-delimited words. There are two flavors of the WLL: the regular WLL which begins with .cblk #" .cble (hash, double-quote) and the splicing list literal which begins with .cblk #*" .cble (hash, star, double-quote). Both types are terminated by a double quote, which may be escaped as .cblk \e" .cble in order to include it as a character. All the escaping conventions used in string literals can be used in word literals. Unlike in string literals, whitespace (tabs and spaces) is not significant in word literals: it separates words. Whitespace may be escaped with a backslash in order to include it as a literal character. Just like in string literals, an unescaped newline character is not allowed. A newline preceded by a backslash is permitted. Such an escaped backslash, together with any leading and trailing unescaped whitespace, is removed and replaced with a single space. Example: .cblk #"abc def ghi" --> notates ("abc" "def" "ghi") #"abc def \e ghi" --> notates ("abc" "def" "ghi") #"abc\e def ghi" --> notates ("abc def" "ghi") #"abc\e def\e \e \e ghi" --> notates ("abc def " " ghi") .cble A splicing word literal differs from a word literal in that it does not produce a list of string literals, but rather it produces a sequence of string literals that is merged into the surrounding syntax. Thus, the following two notations are equivalent: .cblk (1 2 3 #*"abc def" 4 5 #"abc def") (1 2 3 "abc" "def" 4 5 ("abc" "def")) .cble The regular WLL produced a single list object, but the splicing WLL expanded into multiple string literal objects. .SS* String Quasiliterals Quasiliterals are similar to string literals, except that they may contain variable references denoted by the usual .code @ syntax. The quasiliteral represents a string formed by substituting the values of those variables into the literal template. If .code a is bound to .str "apple" and .code b to .strn "banana" , the quasiliteral .code `one @a and two @{b}s` represents the string .strn "one apple and two bananas" . A backquote escaped by a backslash represents itself. Unlike in directive syntax, two consecutive .code @ characters do not code for a literal .codn @ , but cause a syntax error. The reason for this is that compounding of the .code @ syntax is meaningful. Instead, there is a .code \e@ escape for encoding a literal .code @ character. Quasiliterals support the full output variable syntax. Expressions within variables substitutions follow the evaluation rules of \*(TL when the quasiliteral occurs in \*(TL, and the rules of the \*(TX pattern language when the quasiliteral occurs in the pattern language. Quasiliterals can be split into multiple lines in the same way as ordinary string literals. .SS* Quasiword List Literals The quasiword list literals (QLL-s) are to quasiliterals what WLL-s are to ordinary literals. (See the above section Word List Literals.) A QLL combines the convenience of the WLL with the power of quasistrings. Just as in the case of WLL-s, there are two flavors of the QLL: the regular QLL which begins with .code #` \ (hash, backquote) and the splicing QLL which begins with .code #*` \ (hash, star, backquote). Both types are terminated by a backquote, which may be escaped as .code \e` \ in order to include it as a character. All the escaping conventions used in quasiliterals can be used in QLL. Unlike in quasiliterals, whitespace (tabs and spaces) is not significant in QLL: it separates words. Whitespace may be escaped with a backslash in order to include it as a literal character. A newline is not permitted unless escaped. An escaped newline works exactly the same way as it does in word list literals (WLL-s). Note that the delimiting into words is done before the variable substitution. If the variable a contains spaces, then .code #`@a` nevertheless expands into a list of one item: the string derived from .codn a . Examples: .cblk #`abc @a ghi` --> notates (`abc` `@a` `ghi`) #`abc @d@e@f \e ghi` --> notates (`abc` `@d@e@f` `ghi`) #`@a\e @b @c` --> notates (`@a @b` `@c`) .cble A splicing QLL differs from an ordinary QLL in that it does not produce a list of quasiliterals, but rather it produces a sequence of quasiliterals that is merged into the surrounding syntax. .SS* Numbers \*(TX supports integers and floating-point numbers. An integer constant is made up of digits .code 0 through .codn 9 , optionally preceded by a .code + or .code - sign. Examples: .cblk 123 -34 +0 -0 +234483527304983792384729384723234 .cble An integer constant can also be specified in hexadecimal using the prefix .code #x followed by an optional sign, followed by hexadecimal digits: .code 0 through .code 9 and the upper or lower case letters .code A through .codn F : .cblk #xFF ;; 255 #x-ABC ;; -2748 .cble Similarly, octal numbers are supported with the prefix .code #o followed by octal digits: .cblk #o777 ;; 511 .cble and binary numbers can be written with a .code #b prefix: .cblk #b1110 ;; 14 .cble A floating-point constant is marked by the inclusion of a decimal point, the exponential "e notation", or both. It is an optional sign, followed by a mantissa consisting of digits, a decimal point, more digits, and then an optional exponential notation consisting of the letter .code "e" or .codn "E" , an optional .code + or .code - sign, and then digits indicating the exponent value. In the mantissa, the digits are not optional. At least one digit must either precede the decimal point or follow. That is to say, a decimal point by itself is not a floating-point constant. Examples: .cblk .123 123. 1E-3 20E40 .9E1 9.E19 -.5 +3E+3 .cble Examples which are not floating-point constant tokens: .cblk . ;; consing dot 123E ;; the symbol 123E 1.0E- ;; syntax error: invalid floating point constant 1.0E ;; syntax error: invalid floating point constant 1.E ;; 1; consing dot; symbol E .e ;; consing dot followed by symbol e .cble In \*(TX there is a special "dotdot" token consisting of two consecutive periods. An integer constant followed immediately by dotdot is recognized as such; it is not treated as a floating constant followed by a dot. That is to say, .code 123.. does not mean .code 123. . (floating point .code 123.0 value followed by dot token). It means .code 123 .. (integer .code 123 followed by .code .. token). Dialect note: unlike in Common Lisp, .code 123. is not an integer, but the floating-point number .codn 123.0 . .SS* Comments Comments of the form .code @; were already covered. Inside directives, comments are introduced just by a .code ; character. Example: .cblk @(foo ; this is a comment bar ; this is another comment ) .cble This is equivalent to .codn "@(foo bar)" . .SS* Directives-driven Syntax Some directives not only denote an expression, but are also involved in surrounding syntax. For instance, the directive .cblk @(collect) .cble not only denotes an expression, but it also introduces a syntactic phrase which requires a matching .code @(end) directive. In other words, .code @(collect) is not only an expression, but serves as a kind of token in a higher level phrase structure grammar. Usually if this type of "syntactic directive" occurs alone in a line, not preceded or followed by other material, it is involved in a "vertical" (or line oriented) syntax. If such a directive is embedded in a line (has preceding or trailing material) then it is in a horizontal syntactic and semantic context (character-oriented). There is an exception: the definition of a horizontal function looks like this: .cblk @(define name (arg))body material@(end) .cble Yet, this is considered one vertical item, which means that it does not match a line of data. (This is necessary because all horizontal syntax matches something within a line of data, which is undesirable for definitions.) Many directives exhibit both horizontal and vertical syntax, with different but closely related semantics. A few are vertical only, and some are horizontal only. A summary of the available directives follows: .coIP @(eof) Explicitly match the end of file. Fails if unmatched data remains in the input stream. .coIP @(eol) Explicitly match the end of line. Fails if the current position is not the end of a line. Also fails if no data remains (there is no current line). .coIP @(next) Continue matching in another file or other data source. .coIP @(block) Groups together a sequence of directives into a logical name block, which can be explicitly terminated from within using the .code @(accept) and .code @(fail) directives. Blocks are discussed in the section BLOCKS below. .coIP @(skip) Treat the remaining query as a subquery unit, and search the lines (or characters) of the input file until that subquery matches somewhere. A skip is also an anonymous block. .coIP @(trailer) Treat the remaining query or subquery as a match for a trailing context. That is to say, if the remainder matches, the data position is not advanced. .coIP @(freeform) Treat the remainder of the input as one big string, and apply the following query line to that string. The newline characters (or custom separators) appear explicitly in that string. .coIP @(fuzz) The fuzz directive, inspired by the patch utility, specifies a partial match for some lines. .ccIP @ @(line) and @ @(chr) These directives match a variable or expression against the current line number or character position. .coIP @(some) Multiple clauses are each applied to the same input. Succeeds if at least one of the clauses matches the input. The bindings established by earlier successful clauses are visible to the later clauses. .coIP @(all) Multiple clauses are applied to the same input. Succeeds if and only if each one of the clauses matches. The clauses are applied in sequence, and evaluation stops on the first failure. The bindings established by earlier successful clauses are visible to the later clauses. .coIP @(none) Multiple clauses are applied to the same input. Succeeds if and only if none of them match. The clauses are applied in sequence, and evaluation stops on the first success. No bindings are ever produced by this construct. .coIP @(maybe) Multiple clauses are applied to the same input. No failure occurs if none of them match. The bindings established by earlier successful clauses are visible to the later clauses. .coIP @(cases) Multiple clauses are applied to the same input. Evaluation stops on the first successful clause. .coIP @(require) The require directive is similar to the do directive: it evaluates one or more \*(TL expressions. If the result of the rightmost expression is nil, then require triggers a match failure. See the TXR LISP section far below. .ccIP @, @(if) @, @(elif) and @ @(else) The .code if directive with optional .code elif and .code else clauses is a syntactic sugar which translates to a combination of .code @(cases) and .codn @(require) . .coIP @(choose) Multiple clauses are applied to the same input. The one whose effect persists is the one which maximizes or minimizes the length of a particular variable. .coIP @(empty) The .code @(empty) directive matches the empty string. It is useful in certain situations, such as expressing an empty match in a directive that doesn't accept an empty clause. The .code @(empty) syntax has another meaning in .code @(output) clauses, in conjunction with .codn @(repeat) . .meIP @(define < name >> ( args ...)) Introduces a function. Functions are discussed in the FUNCTIONS section below. .coIP @(gather) Searches text for matches for multiple clauses which may occur in arbitrary order. For convenience, lines of the first clause are treated as separate clauses. .coIP @(collect) Search the data for multiple matches of a clause. Collect the bindings in the clause into lists, which are output as array variables. The .code @(collect) directive is line oriented. It works with a multi-line pattern and scans line by line. A similar directive called .code @(coll) works within one line. A collect is an anonymous block. .coIP @(and) Separator of clauses for .codn @(some) , .codn @(all) , .codn @(none) , .code @(maybe) and .codn @(cases) . Equivalent to .codn @(or) . The choice is stylistic. .coIP @(or) Separator of clauses for .codn @(some) , .codn @(all) , .codn @(none) , .code @(maybe) and .codn @(cases) . Equivalent to .codn @(and) . The choice is stylistic. .coIP @(end) Required terminator for .codn @(some) , .codn @(all) , .codn @(none) , .codn @(maybe) , .codn @(cases) , .codn @(if) , .codn @(collect) , .codn @(coll) , .codn @(output) , .codn @(repeat) , .codn @(rep) , .codn @(try) , .code @(block) and .codn @(define) . .coIP @(fail) Terminate the processing of a block, as if it were a failed match. Blocks are discussed in the section BLOCKS below. .coIP @(accept) Terminate the processing of a block, as if it were a successful match. What bindings emerge may depend on the kind of block: collect has special semantics. Blocks are discussed in the section BLOCKS below. .coIP @(try) Indicates the start of a try block, which is related to exception handling, discussed in the EXCEPTIONS section below. .ccIP @ @(catch) and @ @(finally) Special clauses within .codn @(try) . See EXCEPTIONS below. .ccIP @ @(defex) and @ @(throw) Define custom exception types; throw an exception. See EXCEPTIONS below. .coIP @(assert) The assert directive requires the following material to match, otherwise it throws an exception. It is useful for catching mistakes or omissions in parts of a query that are sure-fire matches. .coIP @(flatten) Normalizes a set of specified variables to one-dimensional lists. Those variables which have scalar value are reduced to lists of that value. Those which are lists of lists (to an arbitrary level of nesting) are converted to flat lists of their leaf values. .coIP @(merge) Binds a new variable which is the result of merging two or more other variables. Merging has somewhat complicated semantics. .coIP @(cat) Decimates a list (any number of dimensions) to a string, by catenating its constituent strings, with an optional separator string between all of the values. .coIP @(bind) Binds one or more variables against a value using a structural pattern match. A limited form of unification takes place which can cause a match to fail. .coIP @(set) Destructively assigns one or more existing variables using a structural pattern, using syntax similar to bind. Assignment to unbound variables triggers an error. .coIP @(rebind) Evaluates an expression in the current binding environment, and then creates new bindings for the variables in the structural pattern. Useful for temporarily overriding variable values in a scope. .coIP @(forget) Removes variable bindings. .coIP @(local) Synonym of .codn @(forget) . .coIP @(output) A directive which encloses an output clause in the query. An output section does not match text, but produces text. The directives above are not understood in an output clause. .coIP @(repeat) A directive understood within an @(output) section, for repeating multi-line text, with successive substitutions pulled from lists. The directive .code @(rep) produces iteration over lists horizontally within one line. .coIP @(deffilter) This directive is used for defining named filters, which are useful for filtering variable substitutions in output blocks. Filters are useful when data must be translated between different representations that have different special characters or other syntax, requiring escaping or similar treatment. Note that it is also possible to use a function as a filter. See Function Filters below. .coIP @(filter) The filter directive passes one or more variables through a given filter or chain or filters, updating them with the filtered values. .ccIP @ @(load) and @ @(include) These directives allow \*(TX programs to be modularized. They bring in code from a file, in two different ways. .coIP @(do) The do directive is used to evaluate \*(TL expressions, discarding their result values. See the TXR LISP section far below. .PP .SH* INPUT SCANNING AND DATA MANIPULATION .dir next The .code next directive indicates that the remaining directives in the current block are to be applied against a new input source. It can only occur by itself as the only element in a query line, and takes various arguments, according to these possibilities: .cblk .mets @(next) .mets @(next << source ) .mets @(next < source :nothrow) .mets @(next :args) .mets @(next :env) .mets @(next :list << expr ) .mets @(next :string << expr ) .cble The lone .code @(next) without arguments specifies that subsequent directives will match inside the next file in the argument list which was passed to \*(TX on the command line. If .meta source is given, it must be string-valued expression which denotes an input source; it may be a string literal, quasiliteral or a string-valued variable. For instance, if variable .code A contains the text .strn "data" , then .code @(next A) means switch to the file called .strn "data" , and .code @(next `@A.txt`) means to switch to the file .strn "data.txt" . If the input source cannot be opened for whatever reason, \*(TX throws an exception (see EXCEPTIONS below). An unhandled exception will terminate the program. Often, such a drastic measure is inconvenient; if .code @(next) is invoked with the .code :nothrow keyword, then if the input source cannot be opened, the situation is treated as a simple match failure. The variant .code @(next :args) means that the remaining command line arguments are to be treated as a data source. For this purpose, each argument is considered to be a line of text. The argument list does include that argument which specifies the file that is currently being processed or was most recently processed. As the arguments are matched, they are consumed. This means that if a .code @(next) directive without arguments is executed in the scope of .codn "@(next :args)" , it opens the file named by the first unconsumed argument. To process arguments, and then continue with the original file and argument list, wrap the argument processing in a .codn @(block) . When the block terminates, the input source and argument list are restored to what they were before the block. The variant .code @(next :env) means that the list of process environment variables is treated as a source of data. It looks like a text file stream consisting of lines of the form .strn "name=value" . If this feature is not available on a given platform, an exception is thrown. The syntax .cblk .meti @(next :list << expr ) .cble treats expression .meta expr as a source of text. The value of .meta expr is flattened to a simple list in a way similar to the .code @(flatten) directive. The resulting list is treated as if it were the lines of a text file: each element of the list must be a string, which represents a line. If the strings happen contain embedded newline characters, they are a visible constituent of the line, and do not act as line separators. The syntax .cblk .meti @(next :string << expr ) .cble treats expression .meta expr as a source of text. The value of the expression must be a string. Newlines in the string are interpreted as line terminators. A string which is not terminated by a newline is tolerated, so that: .cblk @(next :string "abc") @a .cble binds .code a to .strn "abc" . Likewise, this is also the case with input files and other streams whose last line is not terminated by a newline. However, watch out for empty strings, which are analogous to a correctly formed empty file which contains no lines: .cblk @(next :string "") @a .cble This will not bind .code a to .strn "" ; it is a matching failure. The behavior of .code :list is different. The query .cblk @(next :list "") @a .cble binds .code a to .strn "" . The reason is that under .code :list the string .str "" is flattened to the list .cblk ("") .cble which is not an empty input stream, but a stream consisting of one empty line. Note that the .code @(next) directive only redirect the source of input over the scope of subquery in which the next directive appears, not necessarily all remaining directives. For example, the following query looks for the line starting with .str "xyz" at the top of the file .strn "foo.txt" , within a some directive. After the .code @(end) which terminates the .codn @(some) , the .str "abc" is matched in the previous input stream which was in effect before the .code @(next) directive: .cblk @(some) @(next "foo.txt") xyz@suffix @(end) abc .cble However, if the .code @(some) subquery successfully matched .str "xyz@suffix" within the file .codn foo.text , there is now a binding for the .code suffix variable, which is visible to the remainder of the entire query. The variable bindings survive beyond the clause, but the data stream does not. The .code @(next) directive supports the file name conventions as the command line. The name .code - means standard input. Text which starts with a .code ! is interpreted as a shell command whose output is read like a file. These interpretations are applied after variable substitution. If the file is specified as .codn @a , but the variable a expands to .strn "!echo foo" , then the output of the .str "echo foo" command will be processed. .dir skip The .code skip directive considers the remainder of the query as a search pattern. The remainder is no longer required to strictly match at the current line in the current input stream. Rather, the current stream is searched, starting with the current line, for the first line where the entire remainder of the query will successfully match. If no such line is found, the .code skip directive fails. If a matching position is found, the remainder of the query is processed from that point. Of course, the remainder of the query can itself contain skip directives. Each such directive performs a recursive subsearch. Skip comes in vertical and horizontal flavors. For instance, skip and match the last line: .cblk @(skip) @last @(eof) .cble Skip and match the last character of the line: .cblk @(skip)@{last 1}@(eol) .cble The skip directive has two optional arguments. If the first argument is a number, its value limits the range of lines scanned for a match. Judicious use of this feature can improve the performance of queries. Example: scan until .str "size: @SIZE" matches, which must happen within the next 15 lines: .cblk @(skip 15) size: @SIZE .cble Without the range limitation skip will keep searching until it consumes the entire input source. In a horizontal .codn skip , the range-limiting numeric argument is expressed in characters, so that .cblk abc@(skip 5)def .cble means: there must be a match for .str "abc" at the start of the line, and then within the next five characters, there must be a match for .strn "def" . Sometimes a skip is nested within a .codn collect , or following another skip. For instance, consider: .cblk @(collect) begin @BEG_SYMBOL @(skip) end @BEG_SYMBOL @(end) .cble The above .code collect iterates over the entire input. But, potentially, so does the embedded .codn skip . Suppose that .str "begin x" is matched, but the data has no matching .strn "end x" . The skip will search in vain all the way to the end of the data, and then the collect will try another iteration back at the beginning, just one line down from the original starting point. If it is a reasonable expectation that an .code "end x" occurs 15 lines of a .strn "begin x" , this can be specified instead: .cblk @(collect) begin @BEG_SYMBOL @(skip 15) end @BEG_SYMBOL @(end) .cble If the symbol .code nil is used in place of a number, it means to scan an unlimited range of lines; thus, .code @(skip nil) is equivalent to .codn @(skip) . If the symbol .code :greedy is used, it changes the semantics of the skip to longest match semantics. For instance, match the last three space-separated tokens of the line: .cblk @(skip :greedy) @a @b @c .cble Without .codn :greedy , the variable .code @c will can match multiple tokens, and end up with spaces in it, because nothing follows .code @c and so it matches from any position which follows a space to the end of the line. Also note the space in front of .codn @a . Without this space, .code @a will get an empty string. A line oriented example of greedy skip: match the last line without using .codn @eof : .cblk @(skip :greedy) @last_line .cble There may be a second numeric argument. This specifies a minimum number of lines to skip before looking for a match. For instance, skip 15 lines and then search indefinitely for .codn "begin ..." : .cblk @(skip nil 15) begin @BEG_SYMBOL .cble The two arguments may be used together. For instance, the following matches if, and only if, the 15th line of input starts with .codn "begin " : .cblk @(skip 1 15) begin @BEG_SYMBOL .cble Essentially, .cblk .meti @(skip 1 << n ) .cble means "hard skip by .meta n lines". .code @(skip 1 0) is the same as .codn "@(skip 1)" , which is a noop, because it means: "the remainder of the query must match starting on the very next line", or, more briefly, "skip exactly zero lines", which is the behavior if the skip directive is omitted altogether. Here is one trick for grabbing the fourth line from the bottom of the input: .cblk @(skip) @fourth_from_bottom @(skip 1 3) @(eof) .cble Or using greedy skip: .cblk @(skip :greedy) @fourth_from_bottom @(skip 1 3) .cble Nongreedy skip with the .code @(eof) has a slight advantage because the greedy skip will keep scanning even though it has found the correct match, then backtrack to the last good match once it runs out of data. The regular skip with explicit .code @(eof) will stop when the .code @(eof) matches. .SS* Reducing Backtracking with Blocks .code skip can consume considerable CPU time when multiple skips are nested. Consider: .cblk @(skip) A @(skip) B @(skip) C .cble This is actually nesting: the second a third skips occur within the body of the first one, and thus this creates nested iteration. \*(TX is searching for the combination of skips which find match the pattern of lines .codn A , .code B and .codn C , with backtracking behavior. The outermost skip marches through the data until it finds .codn A , followed by a pattern match for the second skip. The second skip iterates within to find .codn B , followed by the third skip, and the third skip iterates to find .codn C . If there is only one line .codn A , and one .codn B , then this is reasonably fast. But suppose there are many lines matching .code A and .codn B , giving rise to a large number combinations of skips which match .code A and .codn B , and yet do not find a match for .codn C , triggering backtracking. The nested stepping which tries the combinations of .code A and .code B can give rise to a considerable running time. One way to deal with the problem is to unravel the nesting with the help of blocks. For example: .cblk @(block) @ (skip) A @(end) @(block) @ (skip) B @(end) @(skip) C .cble Now the scope of each skip is just the remainder of the block in which it occurs. The first skip finds .codn A , and then the block ends. Control passes to the next block, and backtracking will not take place to a block which completed (unless all these blocks are enclosed in some larger construct which backtracks, causing the blocks to be re-executed. Of course, this rewrite is not equivalent, and cannot be used for instance in backreferencing situations such as: .cblk @; @; Find three lines anywhere in the input which are identical. @; @(skip) @line @(skip) @line @(skip) @line .cble This example depends on the nested search-within-search semantics. .dir trailer The .code trailer directive introduces a trailing portion of a query or subquery which matches input material normally, but in the event of a successful match, does not advance the current position. This can be used, for instance, to cause .code @(collect) to match partially overlapping regions. Example: .cblk @(collect) @line @(trailer) @(skip) @line @(end) .cble This script collects each line which has a duplicate somewhere later in the input. Without the .code @(trailer) directive, this does not work properly for inputs like: .cblk 111 222 111 222 .cble Without .codn @(trailer) , the first duplicate pair constitutes a match which spans over the .codn 222 . After that pair is found, the matching continues after the second .codn 111 . With the .code @(trailer) directive in place, the collect body, on each iteration, only consumes the lines matched prior to .codn @(trailer) . .dir freeform The .code freeform directive provides a useful alternative to \*(TX's line-oriented matching discipline. The freeform directive treats all remaining input from the current input source as one big line. The query line which immediately follows freeform is applied to that line. The syntax variations are: .cblk @(freeform) ... query line .. .mets @(freeform << number ) ... query line .. .mets @(freeform << string ) ... query line .. .mets @(freeform < number << string ) ... query line .. .cble If .meta number and .meta string are both present, they may be given in either order. If a numeric argument is given, it limits the range of lines which are combined together. For instance .code @(freeform 5) means to only consider the next five lines to to be one big line. Without a numeric argument, freeform is "bottomless". It can match the entire file, which creates the risk of allocating a large amount of memory. If a string argument is given, it specifies a custom line terminator. The default terminator is .strn "\en" . The terminator does not have to be one character long. Freeform does not convert the entire remainder of the input into one big line all at once, but does so in a dynamic, lazy fashion, which takes place as the data is accessed. So at any time, only some prefix of the data exists as a flat line in which newlines are replaced by the terminator string, and the remainder of the data still remains as a list of lines. After the subquery is applied to the virtual line, the unmatched remainder of that line is broken up into multiple lines again, by looking for and removing all occurrences of the terminator string within the flattened portion. Care must be taken if the terminator is other than the default .strn "\en" . All occurrences of the terminator string are treated as line terminators in the flattened portion of the data, so extra line breaks may be introduced. Likewise, in the yet unflattened portion, no breaking takes place, even if the text contains occurrences of the terminator string. The extent of data which is flattened, and the amount of it which remains, depends entirely on the query line underneath .codn @(flatten) . In the following example, lines of data are flattened using $ as the line terminator. .IP code: .cblk \ @(freeform "$") @a$@b: @c @d .cble .IP data: .cblk \ 1 2:3 4 .cble .IP "output (\f[4]-B\f[]):" .cblk \ a="1" b="2" c="3" d="4" .cble .PP The data is turned into the virtual line .codn 1$2:3$4$ . The .code @a$@b: subquery matches the .code 1$2: portion, binding .code a to .strn 1 , and .code b to .strn 2 . The remaining portion .code 3$4$ is then split into separate lines again according to the line terminator .codn $i : .cblk 3 4 .cble Thus the remainder of the query .cblk @c @d .cble faces these lines, binding .code c to .code 3 and .code d to .codn 4 . Note that since the data does not contain dollar signs, there is no ambiguity; the meaning may be understood in terms of the entire data being flattened and split again. In the following example, .code freeform is used to solve a tokenizing problem. The Unix password file has fields separated by colons. Some fields may be empty. Using freeform, we can join the password file using .str ":" as a terminator. By restricting freeform to one line, we can obtain each line of the password file with a terminating .strn ":" , allowing for a simple tokenization, because now the fields are colon-terminated rather than colon-separated. Example: .cblk @(next "/etc/passwd") @(collect) @(freeform 1 ":") @(coll)@{token /[^:]*/}:@(end) @(end) .cble .dir fuzz The .code fuzz directive allows for an imperfect match spanning a set number of lines. It takes two arguments, both expressions that should evaluate to integers: .cblk @(fuzz m n) ... .cble This expresses that over the next n query lines, the matching strictness is relaxed a little bit. Only m out of those n lines have to match. Afterward, the rest of the query follows normal, strict processing. In the degenerate situation that there are fewer than n query lines following the .code fuzz directive, then m of them must succeed nevertheless. (If there are fewer than m, then this is impossible.) .dirs line chr The .code line and .code chr directives perform binding between the current input line number or character position within a line, against an expression or variable: .cblk @(line 42) @(line x) abc@(chr 3)def@(chr y) .cble The directive .code @(line 42) means "match the current input line number against the integer 42". If the current line is 42, then the directive matches, otherwise it fails. .code line is a vertical directive which doesn't consume a line of input. Thus, the following matches at the beginning of an input stream, and .code x ends up bound to the first line of input: .cblk @(line 1) @(line 1) @(line 1) @x .cble The directive .code @(line x) binds variable .code x to the current input line number, if .code x is an unbound variable. If .code x is already bound, then the value of .code x must match the current line number, otherwise the directive fails. The .code chr directive is similar to .code line except that it's a horizontal directive, and matches the character position rather than the line position. Character positions are measured from zero, rather than one. .code chr does not consume a character. Hence the two occurrences of .code chr in the following example both match, and .code x takes the entire line of input: .cblk @(chr 0)@(chr 0)@x .cble The argument of .code line or .code chr may be a .codn @ -delimited Lisp expression. This is useful for matching computed lines or character positions: .cblk @(line @(+ a (* b c))) .cble .dirs some all none maybe cases choose These directives, called the parallel directives, combine multiple subqueries, which are applied at the same input position, rather than to consecutive input. They come in vertical (line mode) and horizontal (character mode) flavors. In horizontal mode, the current position is understood to be a character position in the line being processed. The clauses advance this character position by moving it to the right. In vertical mode, the current position is understood to be a line of text within the stream. A clause advances the position by some whole number of lines. The syntax of these parallel directives follows this example: .cblk @(some) subquery1 . . . @(and) subquery2 . . . @(and) subquery3 . . . @(end) .cble And in horizontal mode: .cblk @(some)subquery1...@(and)subquery2...@(and)subquery3...@(end) .cble Long horizontal lines can be broken up with line continuations, allowing the above example to be written like this, which is considered a single logical line: .cblk @(some)@\e subquery1...@\e @(and)@\e subquery2...@\e @(and)@\e subquery3...@\e @(end) .cble The .codn @(some) , .codn @(all) , .codn @(none) , .codn @(maybe) , .code @(cases) or .code @(choose) must be followed by at least one subquery clause, and be terminated by .codn @(end) . If there are two or more subqueries, these additional clauses are indicated by .code @(and) or .codn @(or) , which are interchangeable. The separator and terminator directives also must appear as the only element in a query line. The .code choose directive requires keyword arguments. See below. The syntax supports arbitrary nesting. For example: .cblk QUERY: SYNTAX TREE: @(all) all -+ @ (skip) +- skip -+ @ (some) | +- some -+ it | | +- TEXT @ (and) | | +- and @ (none) | | +- none -+ was | | | +- TEXT @ (end) | | | +- end @ (end) | | +- end a dark | +- TEXT @(end) *- end .cble nesting can be indicated using whitespace between .code @ and the directive expression. Thus, the above is an .code @(all) query containing a .code @(skip) clause which applies to a .code @(some) that is followed by the text line .strn "a dark" . The .code @(some) clause combines the text line .strn it , and a .code @(none) clause which contains just one clause consisting of the line .strn was . The semantics of the parallel directives is: .coIP @(all) Each of the clauses is matched at the current position. If any of the clauses fails to match, the directive fails (and thus does not produce any variable bindings). Clauses following the failed directive are not evaluated. Bindings extracted by a successful clause are visible to the clauses which follow, and if the directive succeeds, all of the combined bindings emerge. .meIP @(some [ :resolve >> ( var ...) ]) Each of the clauses is matched at the current position. If any of the clauses succeed, the directive succeeds, retaining the bindings accumulated by the successfully matching clauses. Evaluation does not stop on the first successful clause. Bindings extracted by a successful clause are visible to the clauses which follow. The .code :resolve parameter is for situations when the .code @(some) directive has multiple clauses that need to bind some common variables to different values: for instance, output parameters in functions. Resolve takes a list of variable name symbols as an argument. This is called the resolve set. If the clauses of .code @(some) bind variables in the resolve set, those bindings are not visible to later clauses. However, those bindings do emerge out of the .code @(some) directive as a whole. This creates a conflict: what if two or more clauses introduce different bindings for a variable in the resolve set? This is why it is called the resolve set: conflicts for variables in the resolve set are automatically resolved in favor of later directives. Example: .cblk @(some :resolve (x)) @ (bind a "a") @ (bind x "x1") @(or) @ (bind b "b") @ (bind x "x2") @(end) .cble Here, the two clauses both introduce a binding for .codn x . Without the .code :resolve parameter, this would mean that the second clause fails, because .code x comes in with the value .strn x1 , which does not bind with .strn x2 . But because .code x is placed into the resolve set, the second clause does not see the .str x1 binding. Both clauses establish their bindings independently creating a conflict over .codn x . The conflict is resolved in favor of the second clause, and so the bindings which emerge from the directive are: .cblk a="a" b="b" x="x2" .cble .coIP @(none) Each of the clauses is matched at the current position. The directive succeeds only if all of the clauses fail. If any clause succeeds, the directive fails, and subsequent clauses are not evaluated. Thus, this directive never produces variable bindings, only matching success or failure. .coIP @(maybe) Each of the clauses is matched at the current position. The directive always succeeds, even if all of the clauses fail. Whatever bindings are found in any of the clauses are retained. Bindings extracted by any successful clause are visible to the clauses which follow. .coIP @(cases) Each of the clauses is matched at the current position. The clauses are matched, in order, at the current position. If any clause matches, the matching stops and the bindings collected from that clause are retained. Any remaining clauses after that one are not processed. If no clause matches, the directive fails, and produces no bindings. .meIP @(choose [ :longest < var | :shortest < var ]) Each of the clauses is matched at the current position in order. In this construct, bindings established by an earlier clause are not visible to later clauses. Although any or all of the clauses can potentially match, the clause which succeeds is the one which maximizes or minimizes the length of the text bound to the specified variable. The other clauses have no effect. For all of the parallel directives other than .code @(none) and .codn @(choose) , the query advances the input position by the greatest number of lines that match in any of the successfully matching subclauses that are evaluated. The .code @(none) directive does not advance the input position. For instance if there are two subclauses, and one of them matches three lines, but the other one matches five lines, then the overall clause is considered to have made a five line match at its position. If more directives follow, they begin matching five lines down from that position. .dir require The syntax of .code @(require) is: .cblk .mets @(require << lisp-expression ) .cble The require directive evaluates a \*(TL expression. (See TXR LISP far below.) If the expression yields a true value, then it succeeds, and matching continues with the directives which follow. Otherwise the directive fails. In the context of the .code require directive, the expression should not be introduced by the .code @ symbol; it is expected to be a Lisp expression. Example: .cblk @; require that 4 is greater than 3 @; This succeeds; therefore, @a is processed @(require (> (+ 2 2) 3)) @a .cble .dir if The syntax of the .code if directive can be exemplified as follows: .cblk .mets @(if << lisp-expr ) . . . .mets @(elif << lisp-expr ) . . . .mets @(elif << lisp-expr ) . . . @(else) . . . @(end) .cble The .code @(elif) and .code @(else) clauses are all optional. If .code @(else) is present, it must be last, before .codn @(end) , after any .code @(elif) clauses. Any of the clauses may be empty. See the \*(TL section about \*(TL expressions. In this directive, \*(TL expressions are not introduced by the .code @ symbol, just like in the .code require directive. For example: .cblk @(if (> (length str) 42)) foo: @a @b @(else) {@c} @(end) .cble In this example, if the length of the variable .code str is greater than .codn 42 , then matching continues with .strn "foo: @a b" , otherwise it proceeds with .codn {@c} . The .code if directive is actually a syntactic sugar which is translated to .code @(cases) and .codn @(require) . That is to say, the following pattern: .cblk @(cases) .mets @(require << lisp-expr-1 ) A @(or) .mets @(require << lisp-expr-2 ) B @(or) C @(end) .cble corresponds to the somewhat shorter and clearer: .cblk .mets @(if << lisp-expr-1 ) A .mets @(elsif << lisp-expr-2 ) B @(else) C @(end) .cble .dir gather Sometimes text is structured as items that can appear in an arbitrary order. When multiple matches need to be extracted, there is a combinatorial explosion of possible orders, making it impractical to write pattern matches for all the possible orders. The .code gather directive is for these situations. It specifies multiple clauses which all have to match somewhere in the data, but in any order. For further convenience, the lines of the first clause of the .code gather directive are implicitly treated as separate clauses. The syntax follows this pattern .cblk @(gather) one-line-query1 one-line-query2 . . . one-line-queryN @(and) multi line query1 . . . @(and) multi line query2 . . . @(end) .cble Of course the multi-line clauses are optional. The .code gather directive takes keyword parameters, see below. .coNP The @ until / @ last clause in @ gather Similarly to .codn collect , .code gather has an optional .cod3 until / last clause: .cblk @(gather) ... @(until) ... @(end) .cble How gather works is that the text is searched for matches for the single line and multi-line queries. The clauses are applied in the order in which they appear. Whenever one of the clauses matches, any bindings it produces are retained and it is removed from further consideration. Multiple clauses can match at the same text position. The position advances by the longest match from among the clauses which matched. If no clauses match, the position advances by one line. The search stops when all clauses are eliminated, and then the cumulative bindings are produced. If the data runs out, but unmatched clauses remain, the directive fails. Example: extract several environment variables, which do not appear in a particular order: .cblk @(next :env) @(gather) USER=@USER HOME=@HOME SHELL=@SHELL @(end) .cble If the until or last clause is present and a match occurs, then the matches from the other clauses are discarded and the gather terminates. The difference between .cod3 until / last is that any bindings bindings established in last are retained, and the input position is advanced past the matched material. The .cod3 until / last clause has visibility to bindings established in the previous clauses in that same iteration, even though those bindings end up thrown away. .coNP Keyword parameters in @ gather The .code gather directive accepts the keyword parameter .codn :vars . The argument to vars is a list of required and optional variables. Optional variables are denoted by the specification of a default value. Example: .cblk @(gather :vars (a b c (d "foo"))) ... @(end) .cble Here, .codn a , .codn b , .code c and .code e are required variables, and .code d is optional. Variable .code e is required because its default value is the empty list .codn () , same as the symbol .codn nil . The presence of .code :vars changes the behavior in three ways. Firstly, even if all the clauses in the gather match successfully and are eliminated, the directive will fail if the required variables do not have bindings. It doesn't matter whether the bindings are existing, or whether they are established by the gather. Secondly, if some of the clauses of the gather did not match, but all of the required variables have bindings, then the directive succeeds. Without the presence of .codn :vars , it would fail in this situation. Thirdly, if .code gather succeeds (all required variables have bindings), then all of the optional variables which do not have bindings are given bindings to their default values. .dir collect The syntax of the .code collect directive is: .cblk @(collect) ... lines of subquery @(end) .cble or with an until or last clause: .cblk @(collect) ... lines of subquery: main clause @(until) ... lines of subquery: until clause @(end) @(collect) ... lines of subquery: main clause @(last) ... lines of subquery: last clause @(end) .cble The .code repeat symbol may be specified instead of .codn collect , which changes the meaning, see below: .cblk @(repeat) ... lines of subquery @(end) .cble The subquery is matched repeatedly, starting at the current line. If it fails to match, it is tried starting at the subsequent line. If it matches successfully, it is tried at the line following the entire extent of matched data, if there is one. Thus, the collected regions do not overlap. (Overlapping behavior can be obtained: see the .code @(trailer) directive). Unless certain keywords are specified, or unless the collection is explicitly failed with .codn @(fail) , it always succeeds, even if it collects nothing, and even if the .cod3 until / last clause never finds a match. If no .cod3 until / last last clause is specified, and the collect is not limited using parameters, the collection is unbounded: it consumes the entire data file. If any query material follows such the .code collect clause, it will fail if it tries to match anything in the current file; but of course, it is possible to continue matching in another file by means of .codn @(next) . .coNP The @ until / @ last clause in @ collect If an .cod3 until / last last clause is specified, the collection stops when that clause matches at the current position. If an .code until clause terminates collect, no bindings are collected at that position, even if the main clause matches at that position also. Moreover, the position is not advanced. The remainder of the query begins matching at that position. If a last clause terminates collect, the behavior is different. Any bindings captured by the main clause are thrown away, just like with the until clause. However, the bindings in the last clause itself survive, and the position is advanced to skip over that material. Example: .IP code: .cblk \ @(collect) @a @(until) 42 @b @(end) @c .cble .IP data: .cblk \ 1 2 3 42 5 6 .cble .IP result: .cblk \ a[0]="1" a[1]="2" a[2]="3" c="42" .cble .PP The line .code 42 is not collected, even though it matches .con @a . Furthermore, the .code @(until) does not advance the position, so variable .code c takes .codn 42 . If the .code @(until) is changed to .code @(last) the output will be different: .IP result: .cblk \ a[0]="1" a[1]="2" a[2]="3" b="5" c="6" .cble .PP The .code 42 is not collected into the a list, just like before. But now the binding captured by .code @b emerges. Furthermore, the position advances so variable now takes .codn 6 . The binding variables within the clause of a collect are treated specially. The multiple matches for each variable are collected into lists, which then appear as array variables in the final output. Example: .IP code: .cblk \ @(collect) @a:@b:@c @(end) .cble .IP data: .cblk \ John:Doe:101 Mary:Jane:202 Bob:Coder:313 .cble .IP result: .cblk \ a[0]="John" a[1]="Mary" a[2]="Bob" b[0]="Doe" b[1]="Jane" b[2]="Coder" c[0]="101" c[1]="202" c[2]="313" .cble .PP The query matches the data in three places, so each variable becomes a list of three elements, reported as an array. Variables with list bindings may be referenced in a query. They denote a multiple match. The .code -D command line option can establish a one-dimensional list binding. The clauses of .code collect may be nested. Variable matches collated into lists in an inner collect, are again collated into nested lists in the outer collect. Thus an unbound variable wrapped in N nestings of .code @(collect) will be an N-dimensional list. A one dimensional list is a list of strings; a two dimensional list is a list of lists of strings, etc. It is important to note that the variables which are bound within the main clause of a collect. That is, the variables which are subject to collection appear, within the collect, as normal one-value bindings. The collation into lists happens outside of the collect. So for instance in the query: .cblk @(collect) @x=@x @(end) .cble The left .code @x establishes a binding for some material preceding an equal sign. The right .code @x refers to that binding. The value of .cod @x is different in each iteration, and these values are collected. What finally comes out of the collect clause is a single variable called .code x which holds a list containing each value that was ever instantiated under that name within the collect clause. Also note that the until clause has visibility over the bindings established in the main clause. This is true even in the terminating case when the until clause matches, and the bindings of the main clause are discarded. .coNP Keyword Parameters in @ collect By default, .code collect searches the rest of the input indefinitely, or until the .cod3 until / last clause matches. It skips arbitrary amounts of nonmatching material before the first match, and between matches. Within the .code @(collect) syntax, it is possible to specify some useful keyword parameters for additional control of the behavior. For instance .cblk @(collect :maxgap 5) .cble means that the collect will terminate if it does not find a match within five lines of the starting position, or if more than five lines are skipped since any successful match. A .code :maxgap of .code 0 means that the collected regions must be adjacent. For instance: .cblk @(collect :maxgap 0) M @a @(end) .cble means: from here, collect consecutive lines of the form .strn "M ..." . This will not search for the first such line, nor will it skip lines which do not match this form. Other keywords are .codn :mingap , and .codn :gap . The .code :mingap keyword specifies a minimum gap between matches, but has no effect on the distance to the first match. The .code :gap keyword effectively specifies .code :mingap and .code :maxgap at the same time, and can only be used if these other two are not used. Thus: .cblk @(collect :gap 1) @a @(end) .cble means collect every other line starting with the current line. Several other supported keywords are .codn :times , .codn :mintimes , .code :maxtimes and .codn :lines . The shorthand .code :times N means the same thing as .codn ":mintimes N :maxtimes N" . These specify how many matches should be collected. If there are fewer than .code :mintimes matches, the collect fails. If .code :maxtimes matches are collected, collect stops collecting immediately. Example: .cblk @(collect :times 3) @a @b @(end) .cble This will collect a match for .str @a @b exactly three times. If three matches are not found, it will fail. The .code :lines parameter specifies the upper bound on how many lines should be scanned by collect, measuring from the starting position. The extent of the collect body is not counted. Example: .cblk @(collect :lines 2) foo: @a bar: @b baz: @c @(end) .cble The above .code collect will look for a match only twice: at the current position, and one line down. There is one more keyword, .codn :vars , discussed in the following section. .coNP Specifying Variables in @ collect Normally, any variable for which a new binding occurs in a .code collect block is collected. A collect clause may be "sloppy": it can neglect to collect some variables on some iterations, or bind some variables which are intended to behave like local temporaries, but end up collated into lists. Another issue is that the collect clause might not match anything at all, and then none of the variables are bound. The .code :vars keyword allows the query writer to add discipline the .code collect body. The argument to .code :vars is a list of variable specs. A variable spec is either a symbol, or a .cblk .meti >> ( symbol << expression ) .cble pair, where the expression specifies a default value. When a .code :vars list is specified, it means that only the given variables can emerge from the successful collect. Any newly introduced bindings for other variables do not propagate. Furthermore, for any variable which is not specified with a default value, the collect body, whenever it matches successfully, must bind that variable. If it neglects to bind the variable, an exception of type query-error is thrown. (If a .code collect body matches successfully, but produces no new bindings, then this error is suppressed.) For any variable which does have a default value, if the .code collect body neglects to bind that variable, the behavior is as if .code collect did bind that variable to that default value. The default values are expressions, and so can be quasiliterals. Lastly, if in the event that .code collect does not match anything, the variables specified in vars (whether or not they have a default value) are all bound to empty lists. (These bindings are established after the processing of the .cod3 until / last last clause, if present.) Example: .cblk @(collect :vars (a b (c "foo"))) @a @c @(end) .cble Here, if the body .str @a @c matches, an error will be thrown because one of the mandatory variables is .codn b , and the body neglects to produce a binding for .codn b . Example: .cblk @(collect :vars (a (c "foo"))) @a @b @(end) .cble Here, if .str @a @b matches, only .code a will be collected, but not .codn b , because .code b is not in the variable list. Furthermore, because there is no binding for .code c in the body, a binding is created with the value .strn foo , exactly as if .code c matched such a piece of text. In the following example, the assumption is that .code THIS NEVER MATCHES is not found anywhere in the input but the line .code THIS DOES MATCH is found and has a successor which is bound to .codn a . Because the body did not match, the .code :vars .code a and .cod b should be bound to empty lists. But .code a is bound by the last clause to some text, so this takes precedence. Only .code b is bound to an empty list. .cblk @(collect :vars (a b)) THIS NEVER MATCHES @(last) THIS DOES MATCH @a @(end) .cble The following means: do not allow any variables to propagate out of any iteration of the collect and therefore collect nothing: .cblk @(collect :vars nil) ... @(end) .cble Instead of writing .codn "@(collect :vars nil)" , it is possible to write .codn @(repeat) . .code @(repeat) takes all collect keywords, except for .codn :vars . There is a .code @(repeat) directive used in .code @(output) clauses; that is a different directive. .dir coll The .code coll directive is the horizontal version of .codn collect . Whereas .code collect works with multi-line clauses on line-oriented material, .code coll works within a single line. With .codn coll , it is possible to recognize repeating regularities within a line and collect lists. Regular-expression based Positive Match variables work well with coll. Example: collect a comma-separated list, terminated by a space. .IP code: .cblk \ @(coll)@{A /[^, ]+/}@(until) @(end)@B .cble .IP data: .cblk \ foo,bar,xyzzy blorch .cble .IP result: .cblk \ A[0]="foo" A[1]="bar" A[2]="xyzzy" B=blorch .cble .PP Here, the variable .code A is bound to tokens which match the regular expression .code /[^, ]+/: non-empty sequence of characters other than commas or spaces. Like .codn collect , .code coll searches for matches. If no match occurs at the current character position, it tries at the next character position. Whenever a match occurs, it continues at the character position which follows the last character of the match, if such a position exists. If not bounded by an until clause, it will exhaust the entire line. If the until clause matches, then the collection stops at that position, and any bindings from that iteration are discarded. Like collect, coll also supports an .cod3 until / last clause, which propagates variable bindings and advances the position. .code coll clauses nest, and variables bound within a coll are available to clauses within the rest of the .code coll clause, including the .cod3 until / last clause, and appear as single values. The final list aggregation is only visible after the .code coll clause. The behavior of .code coll leads to difficulties when a delimited variable are used to match material which is delimiter separated rather than terminated. For instance, entries in a comma-separated files usually do not appear as .str a,b,c, but rather .strn a,b,c . So for instance, the following result is not satisfactory: .IP code: .cblk \ @(coll)@a @(end) .cble .IP data: .cblk \ 1 2 3 4 5 .cble .IP result: .cblk \ a[0]="1" a[1]="2" a[2]="3" a[3]="4" .cble .PP The .code 5 is missing because it isn't followed by a space, which the text-delimited variable match .str "@a " looks for. After matching "4 ", coll continues to look for matches, and doesn't find any. It is tempting to try to fix it like this: .IP code: .cblk \ @(coll)@a@/ ?/@(end) .cble .IP data: .cblk \ 1 2 3 4 5 .cble .IP result: .cblk \ a[0]="" a[1]="" a[2]="" a[3]="" a[4]="" a[5]="" a[6]="" a[7]="" a[8]="" .cble .PP The problem now is that the regular expression .code / ?/ (match either a space or nothing), matches at any position. So when it is used as a variable delimiter, it matches at the current position, which binds the empty string to the variable, the extent of the match being zero. In this situation, the .code coll directive proceeds character by character. The solution is to use positive matching: specify the regular expression which matches the item, rather than a trying to match whatever follows. The collect directive will recognize all items which match the regular expression: .IP code: .cblk \ @(coll)@{a /[^ ]+/}@(end) .cble .IP data: .cblk \ 1 2 3 4 5 .cble .IP result: .cblk \ a[0]="1" a[1]="2" a[2]="3" a[3]="4" a[4]="5" .cble .PP The .code until clause can specify a pattern which, when recognized, terminates the collection. So for instance, suppose that the list of items may or may not be terminated by a semicolon. We must exclude the semicolon from being a valid character inside an item, and add an until clause which recognizes a semicolon: .IP code: .cblk \ @(coll)@{a /[^ ;]+/}@(until);@(end); .cble .IP data: .cblk \ 1 2 3 4 5; .cble .IP result: .cblk \ a[0]="1" a[1]="2" a[2]="3" a[3]="4" a[4]="5" .cble .PP Whether followed by the semicolon or not, the items are collected properly. Note that the .code @(end) is followed by a semicolon. That's because when the .code @(until) clause meets a match, the matching material is not consumed. This repetition can, of course, be avoided by using .code @(last) instead of .code @(until) since .code @(last) consumes the terminating material. Instead of the above regular-expression-based approach, this extraction problem can also be solved with .codn cases : .IP code: .cblk \ @(coll)@(cases)@a @(or)@a@(end)@(end) .cble .IP data: .cblk \ 1 2 3 4 5 .cble .IP result: .cblk \ a[0]="1" a[1]="2" a[2]="3" a[3]="4" a[4]="5" .cble .PP .coNP Keyword parameters in @ coll The .code @(coll) directive takes most of the same parameters as .codn @(collect) . See the section Collect Keyword Parameters above. So for instance .code @(coll :gap 0) means that the collects must be consecutive, and .code @(coll :maxtimes 2) means that at most two matches will be collected. The .code :lines keyword does not exist, but there is an analogous .code :chars keyword. .dir flatten The .code flatten directive can be used to convert variables to one dimensional lists. Variables which have a scalar value are converted to lists containing that value. Variables which are multidimensional lists are flattened to one-dimensional lists. Example (without .codn @(flatten) ) .IP code: .cblk \ @b @(collect) @(collect) @a @(end) @(end) .cble .IP data: .cblk \ 0 1 2 3 4 5 .cble .IP result: .cblk \ b="0" a_0[0]="1" a_1[0]="2" a_2[0]="3" a_3[0]="4" a_4[0]="5" .cble .PP Example (with .codn @(flatten) ): .IP code: .cblk \ @b @(collect) @(collect) @a @(end) @(end) @(flatten a b) .cble .IP data: .cblk \ 0 1 2 3 4 5 .cble .IP result: .cblk \ b="0" a[0]="1" a[1]="2" a[2]="3" a[3]="4" a[4]="5" .cble .PP .dir merge The .code merge directive provides a way of combining two or more variables in a somewhat complicated but very useful way. To understand what merge does we first have to define a property called depth. The depth of an atom such as a string is defined as .codn 1 . The depth of an empty list is .codn 0 . The depth of a nonempty list is one plus the depth of its deepest element. So for instance .str foo has depth 1, .cblk ("foo") .cble has depth 2, and .cblk ("foo" ("bar")) .cble has depth three. We can now define the binary (two argument) merge operation as follows. .IP 1 .code (merge A B) first normalizes the values .code A and .code B such that they have equal depth. .IP 2 A value which has depth zero is put into a one element list. .IP 3 If either value has a smaller depth than the other, it is wrapped in a list as many times as needed to give it equal depth. Finally, the values are appended together. .PP Merge takes more than two arguments. These are merged by a left reduction. The leftmost two values are merged, and then this result is merged with the third value, and so on. Merge is useful for combining the results from collects at different levels of nesting such that elements are at the appropriate depth. .dir cat The .code cat directive converts a list variable into a single piece of text. The syntax is: .cblk .mets @(cat < var <> [ sep ]) .cble The .meta sep argument specifies a separating piece of text. If no separator is specified, then a single space is used. Example: .IP code: .cblk \ @(coll)@{a /[^ ]+/}@(end) @(cat a ":") .cble .IP data: .cblk \ 1 2 3 4 5 .cble .IP result: .cblk \ a="1:2:3:4:5" .cble .PP .dir bind The syntax of the .code bind directive is: .cblk .mets @(bind < pattern < expression >> { keyword << value }*) .cble The .code bind directive is a kind of pattern match, which matches one or more variables on the left hand side pattern to the value of a variable on the right hand side. The right hand side variable must have a binding, or else the directive fails. Any variables on the left hand side which are unbound receive a matching piece of the right hand side value. Any variables on the left which are already bound must match their corresponding value, or the bind fails. Any variables which are already bound and which do match their corresponding value remain unchanged (the match can be inexact). The simplest bind is of one variable against itself, for instance bind .code A against .codn A : .cblk @(bind A A) .cble This will fail if .code A is not bound, (and complain loudly). If .code A is bound, it succeeds, since .code A matches .codn A . The next simplest bind binds one variable to another: .cblk @(bind A B) .cble Here, if .code A is unbound, it takes on the same value as .codn B . If .code A is bound, it has to match .codn B , or the bind fails. Matching means that either .IP - .code A and . code B are the same text .IP - .code A is text, .code B is a list, and .code A occurs within .codn B . .IP - vice versa: .code B is text, .code A is a list, and .code B occurs within .codn A . .IP - .code A and .code B are lists and are either identical, or one is found as substructure within the other. .PP The right hand side does not have to be a variable. It may be some other object, like a string, quasiliteral, regexp, or list of strings, et cetera. For instance .cblk @(bind A "ab\etc") .cble will bind the string .str ab\etc to the variable .code A if .code A is unbound. If .code A is bound, this will fail unless .code A already contains an identical string. However, the right hand side of a bind cannot be an unbound variable, nor a complex expression that contains unbound variables. The left hand side of .code bind can be a nested list pattern containing variables. The last item of a list at any nesting level can be preceded by a .code . (dot), which means that the variable matches the rest of the list from that position. Example: suppose that the list A contains .cblk ("now" "now" "brown" "cow"). .cble Then the directive .codn "@(bind (H N . C) A)" , assuming that .codn H , .code N and .code C are unbound variables, will bind .code H to .strn how , code N to .strn now , and .code C to the remainder of the list .cblk ("brown" "cow"). .cble Example: suppose that the list .code A is nested to two dimensions and contains .cblk (("how" "now") ("brown" "cow")). .cble Then .code @(bind ((H N) (B C)) A) binds .code H to .strn how , .code N to .strn now , .code B to .str brown and .code C to .strn cow . The dot notation may be used at any nesting level. it must be followed by an item. The forms .code (.) and .code (X .) are invalid, but .code (. X) is valid and equivalent to .codn X . The number of items in a left pattern match must match the number of items in the corresponding right side object. So the pattern .code () only matches an empty list. The notations .code () and .code nil mean exactly the same thing. The symbols .codn nil , .code t and keyword symbols may be used on either side. They represent themselves. For example .code @(bind :foo :bar) fails, but .code @(bind :foo :foo) succeeds since the two sides denote the same keyword symbol object. .coNP Keywords in the @ bind directive The .code bind directive accepts these keywords: .coIP :lfilt The argument to .code :lfilt is a filter specification. When the left side pattern contains a binding which is therefore matched against its counterpart from the right side expression, the left side is filtered through the filter specified by .code :lfilt for the purposes of the comparison. For example: .cblk @(bind "a" "A" :lfilt :upcase) .cble produces a match, since the left side is the same as the right after filtering through the :upcase filter. .coIP :rfilt The argument to .code :rfilt is a filter specification. The specified filter is applied to the right hand side material prior to matching it against the left side. The filter is not applied if the left side is a variable with no binding. It is only applied to determine a match. Binding takes place the unmodified right hand side object. For example, the following produces a match: .cblk @(bind "A" "a" :rfilt :upcase) .cble .coIP :filter This keyword is a shorthand to specify both filters to the same value. For instance .code :filter :upcase is equivalent to .codn ":lfilt :upcase :rfilt :upcase" . For a description of filters, see Output Filtering below. Of course, compound filters like .code (:from_html :upcase) are supported with all these keywords. The filters apply across arbitrary patterns and nested data. Example: .cblk @(bind (a b c) ("A" "B" "C")) @(bind (a b c) (("z" "a") "b" "c") :rfilt :upcase) .cble Here, the first bind establishes the values for .codn a , .code b and .codn c , and the second bind succeeds, because the value of a matches the second element of the list .code ("z" "a") if it is upcased, and likewise .code b matches .str "b" and .code c matches .str c if these are upcased. .coNP Lisp forms in the @ bind directive \*(TL forms, introduced by .code @ may be used on either side of .codn bind . Example: .cblk @(bind a @(+ 2 2)) @(bind @(+ 2 2) @(* 2 2)) .cble Here, .code a is bound to the integer .codn 4 . The second .code bind then succeeds because the forms .code (+ 2 2) and .code (* 2 2) produce equal values. .dir set The .code set directive syntactically resembles .codn bind , but is not a pattern match. It overwrites the previous values of variables with new values from the right hand side. Each variable that is assigned must have an existing binding: .code set will not induce binding. Examples follow. Store the value of .code A back into .codn A , an operation with no effect: .cblk @(set A A) .cble Exchange the values of .code A and .codn B : .cblk @(set (A B) (B A)) .cble Store a string into .codn A : .cblk @(set A "text") .cble Store a list into .codn A : .cblk @(set A ("line1" "line2")) .cble Destructuring assignment. .code A ends up with .strn A , .code B ends up with .cblk ("B1" "B2") .cble and .code C binds to .cblk ("C1" "C2"). .cble .cblk @(bind D ("A" ("B1" "B2") "C1" "C2")) @(bind (A B C) (() () ())) @(set (A B . C) D) .cble Note that .code set does not support a \*(TL expression on the left side, so the following are invalid syntax: .cblk @(set @(+ 1 1) @(* 2 2)) @(set @b @(list "a")) .cble The second one is erroneous even though there is a variable on the left. Because it is preceded by the .code @ escape, it is a Lisp variable, and not a pattern variable. .dir rebind The .code rebind directive resembles .code set but it is not an assignment. It combines the semantics of .codn local , .code bind and .codn set . The expression on the right hand side is evaluated in the current environment. Then the variables in the pattern on the left are introduced as new bindings, whose values come from the pattern. .code rebind makes it easy to create temporary bindings based on existing bindings. .cblk @(define pattern-function (arg)) @;; inside a pattern function: @(rebind recursion-level @(+ recursion-level 1)) @;; ... @(end) .cble When the function terminates, the previous value of recursion-level is restored. The effect is like the following, but much easier to write and faster to execute: .cblk @(define pattern-function (arg)) @;; inside a pattern function: @(local temp) @(set temp recursion-level) @(local recursion-level) @(set recursion-level @(+ temp 1)) @;; ... @(end) .cble .dir forget The .code forget has two spellings: .code @(forget) and .codn @(local). The arguments are one or more symbols, for example: .cblk @(forget a) @(local a b c) .cble this can be written .cblk @(local a) @(local a b c) .cble Directives which follow the forget or local directive no longer see any bindings for the symbols mentioned in that directive, and can establish new bindings. It is not an error if the bindings do not exist. It is strongly recommended to use the .code @(local) spelling in functions, because the forgetting action simulates local variables: for the given symbols, the machine forgets any earlier variables from outside of the function, and consequently, any new bindings for those variables belong to the function. (Furthermore, functions suppress the propagation of variables that are not in their parameter list, so these locals will be automatically forgotten when the function terminates.) .dir do The syntax of .code @(do) is: .cblk .mets @(do << lisp-expression ) .cble The do directive evaluates a \*(TL expression. (See TXR LISP far below.) The value of the expression is ignored, and matching continues continues with the directives which follow the .code do directive, if any. In the context of the .code do directive, the expression should not be introduced by the .code @ symbol; it is expected to be a Lisp expression. Example: .cblk @; match text into variables a and b, then insert into hash table h @(bind h (hash :equal-based)) @a:@b @(do (set [h a] b)) .cble .SH* BLOCKS .SS* Overview Blocks are sections of a query which are either denoted by a name, or are anonymous. They may nest: blocks can occur within blocks and other constructs. Blocks are useful for terminating parts of a pattern matching search prematurely, and escaping to a higher level. This makes blocks not only useful for simplifying the semantics of certain pattern matches, but also an optimization tool. Judicious use of blocks and escapes can reduce or eliminate the amount of backtracking that \*(TX performs. .dir block The .cblk .meti @(block << name ) .cble directive introduces a named block, except when .meta name is the symbol .codn nil . The .code @(block) directive introduces an unnamed block, equivalent to .codn "@(block nil)" . The .code @(skip) and .code @(collect) directives introduce implicit anonymous blocks, as do function bodies. .SS* Block Scope The names of blocks are in a distinct namespace from the variable binding space. So .code @(block foo) is unrelated to the variable .codn @foo . A block extends from the .code @(block ...) directive which introduces it, until the matching .codn @(end) , and may be empty. For instance: .cblk @(some) abc @(block foo) xyz @(end) @(end) .cble Here, the block foo occurs in a .code @(some) clause, and so it extends to the .code @(end) which terminates the block. After that .codn @(end) , the name foo is not associated with a block (is not "in scope"). The second .code @(end) terminates the .code @(some) block. The implicit anonymous block introduced by .code @(skip) has the same scope as the .codn @(skip) : it extends over all of the material which follows the skip, to the end of the containing subquery. .SS* Block Nesting Blocks may nest, and nested blocks may have the same names as blocks in which they are nested. For instance: .cblk @(block) @(block) ... @(end) @(end) .cble is a nesting of two anonymous blocks, and .cblk @(block foo) @(block foo) @(end) @(end) .cble is a nesting of two named blocks which happen to have the same name. When a nested block has the same name as an outer block, it creates a block scope in which the outer block is "shadowed"; that is to say, directives which refer to that block name within the nested block refer to the inner block, and not to the outer one. .SS* Block Semantics A block normally does nothing. The query material in the block is evaluated normally. However, a block serves as a termination point for .code @(fail) and .code @(accept) directives which are in scope of that block and refer to it. The precise meaning of these directives is: .meIP @(fail << name ) Immediately terminate the enclosing query block called .metn name , as if that block failed to match anything. If more than one block by that name encloses the directive, the inner-most block is terminated. No bindings emerge from a failed block. .coIP @(fail) Immediately terminate the innermost enclosing anonymous block, as if that block failed to match. If the implicit block introduced by .code @(skip) is terminated in this manner, this has the effect of causing .code skip itself to fail. I.e. the behavior is as if skip search did not find a match for the trailing material, except that it takes place prematurely (before the end of the available data source is reached). If the implicit block associated with a .code @(collect) is terminated this way, then the entire .code collect fails. This is a special behavior, because a collect normally does not fail, even if it matches nothing and collects nothing! To prematurely terminate a collect by means of its anonymous block, without failing it, use .codn @(accept) . .meIP @(accept << name ) Immediately terminate the enclosing query block called .metn name , as if that block successfully matched. If more than one block by that name encloses the directive, the inner-most block is terminated. Any bindings established within that block until this point emerge from that block. .coIP @(accept) Immediately terminate the innermost enclosing anonymous block, as if that block successfully matched. Any bindings established within that block until this point emerge from that block. If the implicit block introduced by .code @(skip) is terminated in this manner, this has the effect of causing the skip itself to succeed, as if all of the trailing material had successfully matched. If the implicit block associated with a .code @(collect) is terminated this way, then the collection stops. All bindings collected in the current iteration of the collect are discarded. Bindings collected in previous iterations are retained, and collated into lists in accordance with the semantics of collect. Example: alternative way to achieve .code @(until) termination: .cblk @(collect) @ (maybe) --- @ (accept) @ (end) @LINE @(end) .cble This query will collect entire lines into a list called .codn LINE . However, if the line .code --- is matched (by the embedded .codn @(maybe)), the collection is terminated. Only the lines up to, and not including the .code --- line, are collected. The effect is identical to: .cblk @(collect) @LINE @(until) --- @(end) .cble The difference (not relevant in these examples) is that the until clause has visibility into the bindings set up by the main clause. However, the following example has a different meaning: .cblk @(collect) @LINE @ (maybe) --- @ (accept) @ (end) @(end) .cble Now, lines are collected until the end of the data source, or until a line is found which is followed by a .code --- line. If such a line is found, the collection stops, and that line is not included in the collection! The .code @(accept) terminates the process of the collect body, and so the action of collecting the last .code @LINE binding into the list is not performed. .PP .SS* Data Extent of Terminated Blocks A query block may have matched some material prior to being terminated by .codn accept . In that case, it is deemed to have only matched that material, and not any material which follows. This may matter, depending on the context in which the block occurs. Example: .IP code: .cblk \ @(some) @(block foo) @first @(accept foo) @ignored @(end) @second .cble .IP data: .cblk \ 1 2 3 .cble .IP result: .cblk \ first="1" second="2" .cble .PP At the point where the .code accept occurs, the foo block has matched the first line, bound the text .str 1 to the variable .codn @first . The block is then terminated. Not only does the .code @first binding emerge from this terminated block, but what also emerges is that the block advanced the data past the first line to the second line. Next, the .code @(some) directive ends, and propagates the bindings and position. Thus the .code @second which follows then matches the second line and takes the text .strn 2 . In the following query, the foo block occurs inside a maybe clause. Inside the foo block there is a .code @(some) clause. Its first subclause matches variable .code @first and then terminates block foo. Since block foo is outside of the .code @(some) directive, this has the effect of terminating the .code @(some) clause: .IP code: .cblk \ @(maybe) @(block foo) @ (some) @first @ (accept foo) @ (or) @one @two @three @four @ (end) @(end) @second .cble .IP data: .cblk \ 1 2 3 4 5 .cble .IP result: .cblk \ first="1" second="2" .cble .PP The second clause of the .code @(some) directive, namely: .cblk @one @two @three @four .cble is never processed. The reason is that subclauses are processed in top to bottom order, but the processing was aborted within the first clause the .codn "@(accept foo)" . The .code @(some) construct never gets the opportunity to match four lines. If the .code @(accept foo) line is removed from the above query, the output is different: .IP code: .cblk \ @(maybe) @(block foo) @ (some) @first @# <-- @(accept foo) removed from here!!! @ (or) @one @two @three @four @ (end) @(end) @second .cble .IP data: .cblk \ 1 2 3 4 5 .cble .IP result: .cblk \ first="1" one="1" two="2" three="3" four="4" second="5" .cble .PP Now, all clauses of the .code @(some) directive have the opportunity to match. The second clause grabs four lines, which is the longest match. And so, the next line of input available for matching is .codn 5 , which goes to the .code @second variable. .coSS Interaction Between the @ trailer and @ accept Directives If one of the clauses which follow a .code @(trailer) requests a successful termination to an outer block via .codn @(accept) , then .code @(trailer) intercepts the escape and adjusts the data extent to the position that it was given. Example: .IP code: .cblk \ @(block) @(trailer) @line1 @line2 @(accept) @(end) @line3 .cble .IP data: .cblk \ 1 2 3 .cble .IP result: .cblk \ line1="1" line2="2" line3="1" .cble .PP The variable .code line3 is bound to .str 1 because although .code @(accept) yields a data position which has advanced to the third line, this is intercepted by .code @(trailer) and adjusted back to the first line. Neglecting to do this adjustment would violate the semantics of .codn trailer . Directives other than .code @(trailer) have no such special interaction with accept. .SH* FUNCTIONS .SS* Overview \*(TX functions allow a query to be structured to avoid repetition. On a theoretical note, because \*(TX functions support recursion, functions enable \*(TX to match some kinds of patterns which exhibit self-embedding, or nesting, and thus cannot be matched by a regular language. Functions in \*(TX are not exactly like functions in mathematics or functional languages, and are not like procedures in imperative programming languages. They are not exactly like macros either. What it means for a \*(TX function to take arguments and produce a result is different from the conventional notion of a function. A \*(TX function may have one or more parameters. When such a function is invoked, an argument must be specified for each parameter. However, a special behavior is at play here. Namely, some or all of the argument expressions may be unbound variables. In that case, the corresponding parameters behave like unbound variables also. Thus \*(TX function calls can transmit the "unbound" state from argument to parameter. It should be mentioned that functions have access to all bindings that are visible in the caller; functions may refer to variables which are not mentioned in their parameter list. With regard to returning, \*(TX functions are also unconventional. If the function fails, then the function call is considered to have failed. The function call behaves like a kind of match; if the function fails, then the call is like a failed match. When a function call succeeds, then the bindings emanating from that function are processed specially. Firstly, any bindings for variables which do not correspond to one of the function's parameters are thrown away. Functions may internally bind arbitrary variables in order to get their job done, but only those variables which are named in the function argument list may propagate out of the function call. Thus, a function with no arguments can only indicate matching success or failure, but not produce any bindings. Secondly, variables do not propagate out of the function directly, but undergo a renaming. For each parameter which went into the function as an unbound variable (because its corresponding argument was an unbound variable), if that parameter now has a value, that value is bound onto the corresponding argument. Example: .cblk @(define collect-words (list)) @(coll)@{list /[^ \et]+/}@(end) @(end) .cble The above function .code collect-words contains a query which collects words from a line (sequences of characters other than space or tab), into the list variable called .codn list . This variable is named in the parameter list of the function, therefore, its value, if it has one, is permitted to escape from the function call. Suppose the input data is: .cblk Fine summer day .cble and the function is called like this: .cblk @(collect-words wordlist) .cble The result (with .codn "txr -B" ) is: .cblk wordlist[0]=Fine wordlist[1]=summer wordlist[1]=day .cble How it works is that in the function call .codn "@(collect-words wordlist)" , .code wordlist is an unbound variable. The parameter corresponding to that unbound variable is the parameter .codn list . Therefore, that parameter is unbound over the body of the function. The function body collects the words of .str Fine summer day into the variable .codn list, and then yields the that binding. Then the function call completes by noticing that the function parameter .code list now has a binding, and that the corresponding argument .code wordlist has no binding. The binding is thus transferred to the .code wordlist variable. After that, the bindings produced by the function are thrown away. The only enduring effects are: .IP - the function matched and consumed some input; and .IP - the function succeeded; and .IP - the .code wordlist variable now has a binding. .PP Another way to understand the parameter behavior is that function parameters behave like proxies which represent their arguments. If an argument is an established value, such as a character string or bound variable, the parameter is a proxy for that value and behaves just like that value. If an argument is an unbound variable, the function parameter acts as a proxy representing that unbound variable. The effect of binding the proxy is that the variable becomes bound, an effect which is settled when the function goes out of scope. Within the function, both the original variable and the proxy are visible simultaneously, and are independent. What if a function binds both of them? Suppose a function has a parameter called .codn P , which is called with an argument .codn A , which is an unbound variable, and then, in the function, both .code A and .code P bound. This is permitted, and they can even be bound to different values. However, when the function terminates, the local binding of A simply disappears (because the symbol .code A is not among the parameters of the function). Only the value bound to .code P emerges, and is bound to .codn A , which still appears unbound at that point. The .code P binding disappears also, and the net effect is that .code A is now bound. The "proxy" binding of .code A through the parameter .code P "wins" the conflict with the direct binding. .SS* Definition Syntax Function definition syntax comes in two flavors: vertical and horizontal. Horizontal definitions actually come in two forms, the distinction between which is hardly noticeable, and the need for which is made clear below. A function definition begins with a .code @(define ...) directive. For vertical functions, this is the only element in a line. The .code define symbol must be followed by a symbol, which is the name of the function being defined. After the symbol, there is a parenthesized optional argument list. If there is no such list, or if the list is specified as .code () or the symbol .code nil then the function has no parameters. Examples of valid .code define syntax are: .cblk @(define foo) @(define bar ()) @(define match (a b c)) .cble If the define directive is followed by more material on the same line, then it defines a horizontal function: .cblk @(define match-x)x@(end) .cble If the define is the sole element in a line, then it is a vertical function, and the function definition continues below: .cblk @(define match-x) x @(end) .cble The difference between the two is that a horizontal function matches characters within a line, whereas a vertical function matches lines within a stream. The former .code match-x matches the character .codn x , advancing to the next character position. The latter .code match-x matches a line consisting of the character .codn x , advancing to the next line. Material between .code @(define) and .code @(end) is the function body. The define directive may be followed directly by the .code @(end) directive, in which case the function has an empty body. Functions may be nested within function bodies. Such local functions have dynamic scope. They are visible in the function body in which they are defined, and in any functions invoked from that body. The body of a function is an anonymous block. (See BLOCKS above). .SS* Two Forms of The Horizontal Function If a horizontal function is defined as the only element of a line, it may not be followed by additional material. The following construct is erroneous: .cblk @(define horiz (x))@foo:@bar@(end)lalala .cble This kind of definition is actually considered to be in the vertical context, and like other directives that have special effects and that do not match anything, it does not consume a line of input. If the above syntax were allowed, it would mean that the line would not only define a function but also match .codn "lalala" . This would, in turn, would mean that the .code @(define)...@(end) is actually in horizontal mode, and so it matches a span of zero characters within a line (which means that is would require a line of input to match: a surprising behavior for a non-matching directive!) A horizontal function can be defined in an actual horizontal context. This occurs if its is in a line where it is preceded by other material. For instance: .cblk X@(define fun)...@(end)Y .cble This is a query line which must match the text .codn XY . It also defines the function .codn fun . The main use of this form is for nested horizontal functions: .cblk @(define fun)@(define local_fun)...@(end)@(end) .cble .SS* Vertical-Horizontal Overloading A function of the same name may be defined as both vertical and horizontal. Both functions are available at the same time. Which one is used by a call is resolved by context. See the section Vertical Versus Horizontal Calls below. .SS* Call Syntax A function is invoked by compound directive whose first symbol is the name of that function. Additional elements in the directive are the arguments. Arguments may be symbols, or other objects like string and character literals, quasiliterals ore regular expressions. Example: .IP code: .cblk \ @(define pair (a b)) @a @b @(end) @(pair first second) @(pair "ice" cream) .cble .IP data: .cblk \ one two ice milk .cble .IP result: .cblk \ first="one" second="two" cream="milk" .cble .PP The first call to the function takes the line .strn "one two" . The parameter .code a takes .str one and parameter .code b takes .strn two . These are rebound to the arguments .code first and .codn second . The second call to the function binds the a parameter to the word .strn "ice" , and the .code b is unbound, because the corresponding argument .code cream is unbound. Thus inside the function, .code a is forced to match .codn "ice" . Then a space is matched and .code b collects the text .strn milk . When the function returns, the unbound .str cream variable gets this value. If a symbol occurs multiple times in the argument list, it constrains both parameters to bind to the same value. That is to say, all parameters which, in the body of the function, bind a value, and which are all derived from the same argument symbol must bind to the same value. This is settled when the function terminates, not while it is matching. Example: .IP code: .cblk \ @(define pair (a b)) @a @b @(end) @(pair same same) .cble .IP data: .cblk \ one two .cble .IP result: .cblk \ [query fails] .cble .PP Here the query fails because .code a and .code b are effectively proxies for the same unbound variable .code same and are bound to different values, creating a conflict which constitutes a match failure. .SS* Vertical Versus Horizontal Calls A function call which is the only element of the query line in which it occurs is ambiguous. It can go either to a vertical function or to the horizontal one. If both are defined, then it goes to the vertical one. Example: .IP code: .cblk \ @(define which (x))@(bind x "horizontal")@(end) @(define which (x)) @(bind x "vertical") @(end) @(which fun) .cble .IP result: .cblk \ fun="vertical" .cble .PP Not only does this call go to the vertical function, but it is in a vertical context. If only a horizontal function is defined, then that is the one which is called, even if the call is the only element in the line. This takes place in a horizontal character-matching context, which requires a line of input which can be traversed: Example: .IP code: .cblk \ @(define which (x))@(bind x "horizontal")@(end) @(which fun) .cble .IP data: .cblk \ ABC .cble .IP result: .cblk \ [query fails] .cble .PP The query fails because since .code @(which fun) is in horizontal mode, it matches characters in a line. Since the function body consists only of .code @(bind ...) which doesn't match any characters, the function call requires an empty line to match. The line .code ABC is not empty, and so there is a matching failure. The following example corrects this: Example: .IP code: .cblk \ @(define which (x))@(bind x "horizontal")@(end) @(which fun) .cble .IP data: .cblk \ [empty line] .cble .IP result: .cblk \ fun="horizontal" .cble .PP A call made in a clearly horizontal context will prefer the horizontal function, and only fall back on the vertical one if the horizontal one doesn't exist. (In this fall-back case, the vertical function is called with empty data; it is useful for calling vertical functions which process arguments and produce values.) In the next example, the call is followed by trailing material, placing it in a horizontal context. Leading material will do the same thing: Example: .IP code: .cblk \ @(define which (x))@(bind x "horizontal")@(end) @(define which (x)) @(bind x "vertical") @(end) @(which fun)B .cble .IP data: .cblk \ B .cble .IP result: .cblk \ fun="horizontal" .cble .PP .SS* Local Variables As described earlier, variables bound in a function body which are not parameters of the function are discarded when the function returns. However, that, by itself, doesn't make these variables local, because pattern functions have visibility to all variables in their calling environment. If a variable .code x exists already when a function is called, then an attempt to bind it inside a function may result in a failure. The .code local directive must be used in a pattern function to list which variables are local. Example: .cblk @(define path (path))@\e @(local x y)@\e @(cases)@\e (@(path x))@(path y)@(bind path `(@x)@y`)@\e @(or)@\e @{x /[.,;'!?][^ \et\ef\ev]/}@(path y)@(bind path `@x@y`)@\e @(or)@\e @{x /[^ .,;'!?()\et\ef\ev]/}@(path y)@(bind path `@x@y`)@\e @(or)@\e @(bind path "")@\e @(end)@\e @(end) .cble This is a horizontal function which matches a path, which lands into four recursive cases. A path can be parenthesized path followed by a path; it can be a certain character followed by a path, or it can be empty This function ensures that the variables it uses internally, .code x and .codn y , do not have anything to do with any inherited bindings for .code x and .codn y . Note that the function is recursive, which cannot work without .code x and .code y being local, even if no such bindings exist prior to the top-level invocation of the function. The invocation .code @(path x) causes .code x to be bound, which is visible inside the invocation .codn "@(path y)" , but that invocation needs to have its own binding of .code x for local use. .SS* Nested Functions Function definitions may appear in a function. Such definitions are visible in all functions which are invoked from the body (and not necessarily enclosed in the body). In other words, the scope is dynamic, not lexical. Inner definitions shadow outer definitions. This means that a caller can redirect the function calls that take place in a callee, by defining local functions which capture the references. Example: .IP code: .cblk \ @(define which) @ (fun) @(end) @(define fun) @ (output) toplevel fun! @ (end) @(end) @(define callee) @ (define fun) @ (output) local fun! @ (end) @ (end) @ (which) @(end) @(callee) @(which) .cble .IP output: .cblk \ local fun! toplevel fun! .cble .PP Here, the function .code which is defined which calls .codn fun . A toplevel definition of .code fun is introduced which outputs .strn "toplevel fun!" . The function .code callee provides its own local definition of .code fun which outputs .str "local fun!" before calling .codn which . When .code callee is invoked, it calls .codn which , whose .code @(fun) call is routed to callee's local definition. When .code which is called directly from the top level, its .code fun call goes to the toplevel definition. .SH* MODULARIZATION .dirs load include The syntax of the .code load and .code include directives is: .cblk .mets @(load << expr ) .mets @(include << expr ) .cble Where .meta expr evaluates to a string giving the path of the file to load. Unless the path is absolute, it is interpreted relative to the directory of the source file from which the .code @(load) syntax was read. If there was no such source file (for instance, the script was read from standard input), then it is resolved relative to the current working directory. If the file cannot be opened, then the .code .txr suffix is added and another attempt is made. Thus load expressions need not refer to the suffix. In the future, additional suffixes may be searched (compiled versions of a file). The two directives differ as follows. The action of .code load is not performed immediately but at evaluation time. Evaluation time occurs after a \*(TX program is read from beginning to end and parsed. The action of .code include is performed immediately, as the code is being scanned and parsed. That is to say, as the \*(TX parser encounters .code @(include) it processes it immediately. The included material is read and parsed, and its syntax tree is substituted in place of the .code include directive. The parser then continues processing the original file after the .code include directive. Note: the .code include directive is useful for loading \*(TX files which contain Lisp macros which are needed by the parent program. The parent program cannot use .code load to bring in macros because macros are required during expansion, which takes place prior to evaluation time, whereas .code load doesn't execute until evaluation time. See also: the .code *self-path* and .code stdlib variables in \*(TL. .SH* OUTPUT .SS* Introduction A \*(TX query may perform custom output. Output is performed by .code output clauses, which may be embedded anywhere in the query, or placed at the end. Output occurs as a side effect of producing a part of a query which contains an .code @(output) directive, and is executed even if that part of the query ultimately fails to find a match. Thus output can be useful for debugging. An .code output clause specifies that its output goes to a file, pipe, or (by default) standard output. If any output clause is executed whose destination is standard output, \*(TX makes a note of this, and later, just prior to termination, suppresses the usual printing of the variable bindings or the word false. .dir output The syntax of the .code @(output) directive is: .cblk .mets @(output [ < destination ] { < bool-keyword | < keyword < value }* ) . . one or more output directives or lines . @(end) .cble The optional .meta destination is a string which gives the path name of a file to open for output. If the name is .code - it instead denotes standard output, and if it begins with .code ! then the rest of the shell is treated as a shell command to which the output is piped. The destination may be specified as a variable which holds text, as a string literal or as a quasiliteral The keyword list consists of a mixture of boolean keywords which do not have an argument, or keywords with arguments. The following boolean keywords are supported: .coIP :nothrow The output directive throws an exception if the output destination cannot be opened, unless the .code :nothrow keyword is present, in which case the situation is treated as a match failure. Note that since command pipes are processes that report errors asynchronously, a failing command will not throw an immediate exception that can be suppressed with .codn :nothrow . This is for synchronous errors, like trying to open a destination file, but not having permissions, etc. .coIP :append This keyword is meaningful for files, specifying append mode: the output is to be added to the end of the file rather than overwriting the file. The following value keywords are supported: .coIP :filter The argument can be a symbol, which specifies a filter to be applied to the variable substitutions occurring within the .code output clause. The argument can also be a list of filter symbols, which specifies that multiple filters are to be applied, in left to right order. See the later sections Output Filtering below, and The Deffilter Directive. .coIP :into The argument of .code :into is a symbol which denotes a variable. The output will go into that variable. If the variable is unbound, it will be created. Otherwise, its contents are overwritten unless the .code :append keyword is used. If .code :append is used, then the new content will be appended to the previous content of the variable, after flattening the content to a list, as if by the .code flatten directive. .coIP :named The argument of .code :named is a symbol which denotes a variable. The file or pipe stream which is opened for the output is stored in this variable, and is not closed at the end of the output block. This allows a subsequent output block to continue output on the same stream, which is possible using the next two keywords, .code :continue or .codn :finish . A new binding is established for the variable, even if it already has an existing binding. .coIP :continue A destination should not be specified if .code :continue is used. The argument of .code :continue is an expression, such as a variable name, that evaluates to a stream object. That stream object is used for the output block. At the end of the output block, the stream is flushed, but not closed. A usage example is given in the documentation for the Close Directive below. .coIP :finish A destination should not be specified if .code :finish is used. The argument of .code :finish is an expression, such as a variable name, that evaluates to a stream object. That stream object is used for the output block. At the end of the output block, the stream is closed. An example is given in the documentation for the Close Directive below. .SS* Output Text Text in an output clause is not matched against anything, but is output verbatim to the destination file, device or command pipe. .SS* Output Variables Variables occurring in an output clause do not match anything; instead their contents are output. A variable being output can be any object. If it is of a type other than a list or string, it will be converted to a string as if by the .code tostring function in \*(TL. A list is converted to a string in a special way: the elements are individually converted to a string and then they are catenated together. The default separator string is a single space: an alternate separation can be specified as an argument in the brace substitution syntax. Empty lists turn into an empty string. Lists may be output within .code @(repeat) or .code @(rep) clauses. Each nesting of these constructs removes one level of nesting from the list variables that it contains. In an output clause, the .cblk .meti >> @{ name << number } .cble variable syntax generates fixed-width field, which contains the variable's text. The absolute value of the number specifies the field width. For instance .code -20 and .code 20 both specify a field width of twenty. If the text is longer than the field, then it overflows the field. If the text is shorter than the field, then it is left-adjusted within that field, if the width is specified as a positive number, and right-adjusted if the width is specified as negative. An output variable may specify a filter which overrides any filter established for the output clause. The syntax for this is .cblk .meti @{NAME :filter << filterspec }. .cble The filter specification syntax is the same as in the output clause. See Output Filtering below. .SS* Output Variables: Indexing Additional syntax is supported in output variables that does not appear in pattern matching variables. A square bracket index notation may be used to extract elements or ranges from a variable, which works with strings, vectors and lists. Elements are indexed from zero. This notation is only available in brace-enclosed syntax, and looks like this: .meIP <> @{name[ expr ]} Extract the element at the position given by .metn expr . .meIP <> @{name[ expr1..expr2 ]} Extract a range of elements from the position given by .metn expr1 , up to one position less than the position given by .metn expr2 . If the variable is a list, it is treated as a list substitution, exactly as if it were the value of an unsubscripted list variable. The elements of the list are converted to strings and catenated together wit ha separator string between them, the default one being a single space. An alternate character may be given as a string argument in the brace notation. .PP Example: .cblk @(bind a ("a" "b" "c" "d")) @(output) @{a[1..3] "," 10} @(end) .cble The above produces the text .str b,c in a field .code 10 spaces wide. The .code [1..3] argument extracts a range of .codn a ; the .str "," argument specifies an alternate separator string, and .code 10 specifies the field width. .SS* Output Substitutions The brace syntax has another syntactic and semantic extension in .code output clauses. In place of the symbol, an expression may appear. The value of that expression is substituted. Example: .cblk @(bind a "foo") @(output) @{`@a:` -10} .cble Here, the quasiliteral expression .code `@a:` is evaluated, producing the string .strn foo: . This string is printed right-adjusted in a .code 10 character field. .dir repeat The .code repeat directive generates repeated text from a "boilerplate", by taking successive elements from lists. The syntax of repeat is like this: .cblk @(repeat) . . main clause material, required . . special clauses, optional . . @(end) .cble .code repeat has four types of special clauses, any of which may be specified with empty contents, or omitted entirely. They are described below. .code repeat takes arguments, also described below. All of the material in the main clause and optional clauses is examined for the presence of variables. If none of the variables hold lists which contain at least one item, then no output is performed, (unless the repeat specifies an .code @(empty) clause, see below). Otherwise, among those variables which contain non-empty lists, repeat finds the length of the longest list. This length of this list determines the number of repetitions, R. If the .code repeat contains only a main clause, then the lines of this clause is output R times. Over the first repetition, all of the variables which, outside of the repeat, contain lists are locally rebound to just their first item. Over the second repetition, all of the list variables are bound to their second item, and so forth. Any variables which hold shorter lists than the longest list eventually end up with empty values over some repetitions. Example: if the list .code A holds .strn 1 , .str 2 and .strn 3 ; the list .code B holds .strn A , .strn B ; and the variable .code C holds .strn X , then .cblk @(repeat) >> @C >> @A @B @(end) .cble will produce three repetitions (since there are two lists, the longest of which has three items). The output is: .cblk >> X >> 1 A >> X >> 2 B >> X >> 3 .cble The last line has a trailing space, since it is produced by .strn "@A @B" , where .code B has an empty value. Since .code C is not a list variable, it produces the same value in each repetition. The special clauses are: .coIP @(single) If the .code repeat produces exactly one repetition, then the contents of this clause are processed for that one and only repetition, instead of the main clause or any other clause which would otherwise be processed. .coIP @(first) The body of this clause specifies an alternative body to be used for the first repetition, instead of the material from the main clause. .coIP @(last) The body of this clause is used instead of the main clause for the last repetition. .coIP @(empty) If the repeat produces no repetitions, then the body of this clause is output. If this clause is absent or empty, the repeat produces no output. .coIP "@(mod n m)" The forms .code n and .code m are expressions that evaluate to integers. The value of .code m should be nonzero. The clause denoted this way is active if the repetition modulo .code m is equal to .codn n . The first repetition is numbered zero. For instance the clause headed by .code @(mod 0 2) will be used on repetitions 0, 2, 4, 6, ... and .code @(mod 1 2) will be used on repetitions 1, 3, 5, 7, ... .coIP "@(modlast n m)" The meaning of .code n and .code m is the same as in .codn "@(mod n m)" , but one more condition is imposed. This clause is used if the repetition modulo .code m is equal to .codn n , and if it is the last repetition. .PP The precedence among the clauses which take an iteration is: .codn "single > first > mod > modlast > last > main" . That is if two or more of these clauses can apply to a repetition, then the leftmost one in this precedence list applies. For instance, if there is just a single repetition, then any of these special clause types can apply to that repetition, since it is the only repetition, as well as the first and last one. In this situation, if there is a .code @(single) clause present, then the repetition is processed using that clause. Otherwise, if there is a .code @(first) clause present, that clause is used. Failing that, .code @(mod) is used if there is such a clause and its numeric conditions are satisfied. If there isn't, then .code @(modlast) clauses are considered, and if there are none, or none of them activate, then .code @(last) is considered. Finally if none of all these clauses are present or apply, then the repetition is processed using the main clause. Repeat supports arguments. .cblk .mets @(repeat [:counter << symbol ] [:vars <> ( symbol *)]) .cble The .code :counter argument designates a symbol which will behave as an integer variable over the scope of the clauses inside the repeat. The variable provides access to the repetition count, starting at zero, incrementing with each repetition. The .code :vars argument specifies a list of variables. The repeat directive will pick out from this list those variables which have bindings. It will assume that all these variables occur in the repeat block and are to be iterated. This syntax is needed for situations in which .code @(repeat) is not able to deduce the existence of a variable in the block. It does not dig very deeply to discover variables, and does not "see" variables that are referenced via embedded \*(TL expressions. For instance, the following produces no output: .cblk @(bind list ("a" "b" "c")) @(output) @(repeat) @(format nil "<~a>" list) @(end) @(end) .cble Although the list variable appears in the repeat block, it is embedded in a \*(TL construct. That construct will never be evaluated because no repetitions take place: the repeat construct doesn't find any variables and so doesn't iterate. The remedy is to provide a little help via the :vars parameter: .cblk @(bind list ("a" "b" "c")) @(output) @(repeat :vars (list)) @(format nil "<~a>" list) @(end) @(end) .cble Now the repeat block iterates over list and the output is: .cblk .cble .coSS Nested @ repeat directives If a .code repeat clause encloses variables which hold multidimensional lists, those lists require additional nesting levels of repeat (or rep). It is an error to attempt to output a list variable which has not been decimated into primary elements via a repeat construct. Suppose that a variable .code X is two-dimensional (contains a list of lists). .code X must be twice nested in a .codn repeat . The outer repeat will traverse the lists contained in .codn X . The inner repeat will traverse the elements of each of these lists. A nested repeat may be embedded in any of the clauses of a repeat, not only the main clause. .dir rep The .code rep directive is similar to .codn repeat . Whereas .code repeat is line oriented, .code rep generates material within a line. It has all the same clauses, but everything is specified within one line: .cblk @(rep)... main material ... .... special clauses ...@(end) .cble More than one .code @(rep) can occur within a line, mixed with other material. A .code @(rep) can be nested within a .code @(repeat) or within another .codn @(rep) . Also, .code @(rep) accepts the same .code :counter and .code :vars arguments. .coSS @ repeat and @ rep Examples Example 1: show the list .code L in parentheses, with spaces between the elements, or the word .code EMPTY if the list is empty: .cblk @(output) @(rep)@L @(single)(@L)@(first)(@L @(last)@L)@(empty)EMPTY@(end) @(end) .cble Here, the .code @(empty) clause specifies .codn EMPTY . So if there are no repetitions, the text .code EMPTY is produced. If there is a single item in the list .codn L , then .code @(single)(@L) produces that item between parentheses. Otherwise if there are two or more items, the first item is produced with a leading parenthesis followed by a space by .code @(first)(@L and the last item is produced with a closing parenthesis: .codn @(last)@L) . All items in between are emitted with a trailing space by the main clause: .codn @(rep)@L . Example 2: show the list L like Example 1 above, but the empty list is .codn () . .cblk @(output) (@(rep)@L @(last)@L@(end)) @(end) .cble This is simpler. The parentheses are part of the text which surrounds the .code @(rep) construct, produced unconditionally. If the list .code L is empty, then .code @(rep) produces no output, resulting in .codn () . If the list .code L has one or more items, then they are produced with spaces each one, except the last which has no space. If the list has exactly one item, then the .code @(last) applies to it instead of the main clause: it is produced with no trailing space. .dir close The syntax of the .code close directive is: .cblk .mets @(close << expr ) .cble Where .meta expr evaluates to a stream. The .code close directive can be used to explicitly close streams created using .cblk .meti @(output ... :named << var ) .cble syntax, as an alternative to .cblk .meti @(output :finish << expr ). .cble Examples: Write two lines to .str foo.txt over two output blocks using a single stream: .cblk @(output "foo.txt" :named foo) Hello, @(end) @(output :continue foo) world! @(end) @(close foo) .cble The same as above, using .code :finish rather than .code :continue so that the stream is closed at the end of the second block: .cblk @(output "foo.txt" :named foo) Hello, @(end) @(output :finish foo) world! @(end) .cble .SS* Output Filtering Often it is necessary to transform the output to preserve its meaning under the convention of a given data format. For instance, if a piece of text contains the characters .code < or .codn > , then if that text is being substituted into HTML, these should be replaced by .code < and .codn > . This is what filtering is for. Filtering is applied to the contents of output variables, not to any template text. \*(TX implements named filters. Built-in filters are named by keywords, given below. User-defined filters are possible, however. See notes on the deffilter directive below. Instead of a filter name, the syntax .cblk .meti (fun << name ) .cble can be used. This denotes that the function called .meta name is to be used as a filter. This is discussed in the next section Function Filters below. Built-in filters named by keywords: .coIP :to_html Filter text to HTML, representing special characters using HTML ampersand sequences. For instance .code > is replaced by .codn > . .coIP :from_html Filter text with HTML codes into text in which the codes are replaced by the corresponding characters. For instance .code > is replaced by .codn > . .coIP :upcase Convert the 26 lower case letters of the English alphabet to upper case. .coIP :downcase Convert the 26 upper case letters of the English alphabet to lower case. .coIP :frompercent Decode percent-encoded text. Character triplets consisting of the .code % character followed by a pair of hexadecimal digits (case insensitive) are are converted to bytes having the value represented by the hexadecimal digits (most significant nybble first). Sequences of one or more such bytes are treated as UTF-8 data and decoded to characters. .coIP :topercent Convert to percent encoding according to RFC 3986. The text is first converted to UTF-8 bytes. The bytes are then converted back to text as follows. Bytes in the range 0 to 32, and 127 to 255 (note: including the ASCII DEL), bytes whose values correspond to ASCII characters which are listed by RFC 3986 as being in the "reserved set", and the byte value corresponding to the ASCII .code % character are encoded as a three-character sequence consisting of the .code % character followed by two hexadecimal digits derived from the byte value (most significant nybble first, upper case). All other bytes are converted directly to characters of the same value without any such encoding. .coIP :fromurl Decode from URL encoding, which is like percent encoding, except that if the unencoded .code + character occurs, it is decoded to a space character. Of course .code %20 still decodes to space, and .code %2B to the .code + character. .coIP :tourl Encode to URL encoding, which is like percent encoding except that a space maps to .code + rather than .codn %20 . The .code + character, being in the reserved set, encodes to .codn %2B . .coIP :tonumber Converts strings to numbers. Strings that contain a period, .code e or .code E are converted to floating point as if by the Lisp function .codn flo-str . Otherwise they are converted to integer as if using .code int-str with a radix of 10. Non-numeric junk results in the object .codn nil . .coIP :tointeger Converts strings to integers as if using .code int-str with a radix of 10. Non-numeric junk results in the object .codn nil . .coIP :tofloat Converts strings to floating-point values as if using the function .codn flo-str . Non-numeric junk results in the object .codn nil. .coIP :hextoint Converts strings to integers as if using .code int-str with a radix of 16. Non-numeric junk results in the object .codn nil. .PP Examples: To escape HTML characters in all variable substitutions occurring in an output clause, specify .code :filter :to_html in the directive: .cblk @(output :filter :to_html) ... @(end) .cble To filter an individual variable, add the syntax to the variable spec: .cblk @(output) @{x :filter :to_html} @(end) .cble Multiple filters can be applied at the same time. For instance: .cblk @(output) @{x :filter (:upcase :to_html)} @(end) .cble This will fold the contents of .code x to upper case, and then encode any special characters into HTML. Beware of combinations that do not make sense. For instance, suppose the original text is HTML, containing codes like .codn " . The compound filter .code (:upcase :from_html) will not work because .code " will turn to .code " which no longer be recognized by the .code :from_html filter, since the entity names in HTML codes are case-sensitive. Capture some numeric variables and convert to numbers: .cblk @date @time @temperature @pressure @(filter :tofloat temperature pressure) @;; temperature and pressure can now be used in calculations .cble .SS* Function Filters A function can be used as a filter. For this to be possible, the function must conform to certain rules: .IP 1. The function must take two special arguments, which may be followed by additional arguments. .IP 2. When the function is called, the first argument will be bound to a string, and the second argument will be unbound. The function must produce a value by binding it to the second argument. If the filter is to be used as the final filter in a chain, it must produce a string. For instance, the following is a valid filter function: .cblk @(define foo_to_bar (in out)) @ (next :string in) @ (cases) foo @ (bind out "bar") @ (or) @ (bind out in) @ (end) @(end) .cble This function binds the .code out parameter to .str bar if the in parameter is .strn foo , otherwise it binds the .code out parameter to a copy of the .code in parameter. This is a simple filter. To use the filter, use the syntax .code (:fun foo_to_bar) in place of a filter name. For instance in the bind directive: .cblk @(bind "foo" "bar" :lfilt (:fun foo_to_bar)) .cble The above should succeed since the left side is filtered from .str foo to .strn bar , so that there is a match. Of course, function filters can be used in a chain: .cblk @(output :filter (:downcase (:fun foo_to_bar) :upcase)) ... @(end) .cble Here is a split function which takes an extra argument which specifies the separator: .cblk @(define split (in out sep)) @ (next :list in) @ (coll)@(maybe)@token@sep@(or)@token@(end)@(end) @ (bind out token) @(end) .cble Furthermore, note that it produces a list rather than a string. This function separates the argument in into tokens according to the separator text carried in the variable .codn sep . Here is another function, .codn join , which catenates a list: .cblk @(define join (in out sep)) @ (output :into out) @ (rep)@in@sep@(last)@in@(end) @ (end) @(end) .cble Now here is these two being used in a chain: .cblk @(bind text "how,are,you") @(output :filter (:fun split ",") (:fun join "-")) @text @(end) .cble Output: .cblk how-are-you .cble When the filter invokes a function, it generates the first two arguments internally to pass in the input value and capture the output. The remaining arguments from the .code (:fun ...) construct are also passed to the function. Thus the string objects .str "," and .str "-" are passed as the .code sep argument to .code split and .codn join . Note that .code split puts out a list, which .code join accepts. So the overall filter chain operates on a string: a string goes into split, and a string comes out of join. .dir deffilter The .code deffilter directive allows a query to define a custom filter, which can then be used in .code output clauses to transform substituted data. This directive's syntax is illustrated in this example: .IP code: .cblk \ @(deffilter rot13 ("a" "n") ("b" "o") ("c" "p") ("d" "q") ("e" "r") ("f" "s") ("g" "t") ("h" "u") ("i" "v") ("j" "w") ("k" "x") ("l" "y") ("m" "z") ("n" "a") ("o" "b") ("p" "c") ("q" "d") ("r" "e") ("s" "f") ("t" "g") ("u" "h") ("v" "i") ("w" "j") ("x" "k") ("y" "l") ("z" "m")) @(collect) @line @(end) @(output :filter rot13) @(repeat) @line @(end) @(end) .cble .IP data: .cblk \ hey there! .cble .IP output: .cblk \ url gurer! .cble .PP The .code deffilter symbol must be followed by the name of the filter to be defined, followed by forms which evaluate to lists of strings. Each list must be at least two elements long and specifies one or more texts which are mapped to a replacement text. For instance, the following specifies a telephone keypad mapping from upper case letters to digits. .cblk @(deffilter alpha_to_phone ("E" "0") ("J" "N" "Q" "1") ("R" "W" "X" "2") ("D" "S" "Y" "3") ("F" "T" "4") ("A" "M" "5") ("C" "I" "V" "6") ("B" "K" "U" "7") ("L" "O" "P" "8") ("G" "H" "Z" "9")) @(deffilter foo (`@a` `@b`) ("c" `->@d`)) @(bind x ("from" "to")) @(bind y ("---" "+++")) @(deffilter sub x y) .cble The last deffilter above equivalent to .codn "@(deffilter sub ("from" "to") ("---" "+++"))" . Filtering works using a longest match algorithm. The input is scanned from left to right, and the longest piece of text is identified at every character position which matches a string on the left hand side, and that text is replaced with its associated replacement text. The scanning then continues at the first character after the matched text. If none of the strings matches at a given character position, then that character is passed through the filter untranslated, and the scan continues at the next character in the input. Filtering is not in-place but rather instantiates a new text, and so replacement text is not re-scanned for more replacements. If a filter definition accidentally contains two or more repetitions of the same left hand string with different right hand translations, the later ones take precedence. No warning is issued. .dir filter The syntax of the .code filter directive is: .cblk @(filter FILTER { VAR }+ ) .cble A filter is specified, followed by one or more variables whose values are filtered and stored back into each variable. Example: convert .codn a , .codn b , and .code c to upper case and HTML encode: .cblk @(filter (:upcase :to_html) a b c) .cble .SH* EXCEPTIONS .SS* Introduction The exceptions mechanism in \*(TX is another disciplined form of non-local transfer, in addition to the blocks mechanism (see BLOCKS above). Like blocks, exceptions provide a construct which serves as the target for a dynamic exit. Both blocks and exceptions can be used to bail out of deep nesting when some condition occurs. However, exceptions provide more complexity. Exceptions are useful for error handling, and \*(TX in fact maps certain error situations to exception control transfers. However, exceptions are not inherently an error-handling mechanism; they are a structured dynamic control transfer mechanism, one of whose applications is error handling. An exception control transfer (simply called an exception) is always identified by a symbol, which is its type. Types are organized in a subtype-supertype hierarchy. For instance, the .code file-error exception type is a subtype of the .code error type. This means that a file error is a kind of error. An exception handling block which catches exceptions of type .code error will catch exceptions of type .codn file-error , but a block which catches .code file-error will not catch all exceptions of type .codn error . A .code query-error is a kind of error, but not a kind of .codn file-error . The symbol .code t is the supertype of every type: every exception type is considered to be a kind of .codn t . (Mnemonic: .code t stands for type, as in any type). Exceptions are handled using .code @(catch) clauses within a .code @(try) directive. In addition to being useful for exception handling, the .code @(try) directive also provides unwind protection by means of a .code @(finally) clause, which specifies query material to be executed unconditionally when the try clause terminates, no matter how it terminates. .dir try The general syntax of the .code try directive is .cblk @(try) ... main clause, required ... ... optional catch clauses ... ... optional finally clause @(end) .cble A .code catch clause looks like: .cblk @(catch TYPE [ PARAMETERS ]) . . . .cble and also this simple form: .cblk @(catch) . . . .cble which catches all exceptions, and is equivalent to .codn "@(catch t)" . A .code finally clause looks like: .cblk @(finally) ... . . .cble The main clause may not be empty, but the catch and finally may be. A try clause is surrounded by an implicit anonymous block (see BLOCKS section above). So for instance, the following is a no-op (an operation with no effect, other than successful execution): .cblk @(try) @(accept) @(end) .cble The .code @(accept) causes a successful termination of the implicit anonymous block. Execution resumes with query lines or directives which follow, if any. .code try clauses and blocks interact. For instance, an .code accept from within a try clause invokes a .codn finally . .IP code: .cblk \ @(block foo) @ (try) @ (accept foo) @ (finally) @ (output) bye! @ (end) @ (end) .cble .IP output: .cblk \ bye! .cble .PP How this works: the .code try block's main clause is .codn "@(accept foo)" . This causes the enclosing block named .code foo to terminate, as a successful match. Since the .cod try is nested within this block, it too must terminate in order for the block to terminate. But the try has a .code finally clause, which executes unconditionally, no matter how the try block terminates. The .code finally clause performs some output, which is seen. .coSS The @ finally clause A .code try directive can terminate in one of three ways. The main clause may match successfully, and possibly yield some new variable bindings. The main clause may fail to match. Or the main clause may be terminated by a non-local control transfer, like an exception being thrown or a block return (like the block foo example in the previous section). No matter how the .code try clause terminates, the .code finally clause is processed. The .code finally clause is itself a query which binds variables, which leads to questions: what happens to such variables? What if the .code finally block fails as a query? As well as: what if a .code finally clause itself initiates a control transfer? Answers follow. Firstly, a .code finally clause will contribute variable bindings only if the main clause terminates normally (either as a successful or failed match). If the main clause of the .code try block successfully matches, then the .code finally block continues matching at the next position in the data, and contributes bindings. If the main clause fails, then the .code finally block tries to match at the same position where the main clause failed. The overall .code try directive succeeds as a match if either the main clause or the .code finally clause succeed. If both fail, then the try directive is a failed match. Example: .IP code: .cblk \ @(try) @a @(finally) @b @(end) @c .cble .IP data: .cblk \ 1 2 3 .cble .IP result: .cblk \ a="1" b="2" c="3" .cble .PP In this example, the main clause of the .code try captures line .str 1 of the data as variable .codn a , then the finally clause captures .str 2 as .codn b , and then the query continues with the .code @c line after try block, so that .code c captures .strn "3" . Example: .IP code: .cblk \ @(try) hello @a @(finally) @b @(end) @c .cble .IP data: .cblk \ 1 2 .cble .IP result: .cblk \ b="1" c="2" .cble .PP In this example, the main clause of the .code try fails to match, because the input is not prefixed with .srn "hello " . However, the .code finally clause matches, binding .code b to .strn "1" . This means that the try block is a successful match, and so processing continues with .code @c which captures .strn "2" . When .code finally clauses are processed during a non-local return, they have no externally visible effect if they do not bind variables. However, their execution makes itself known if they perform side effects, such as output. A .code finally clause guards only the main clause and the .code catch clauses. It does not guard itself. Once the finally clause is executing, the .code try block is no longer guarded. This means if a nonlocal transfer, such as a block accept or exception, is initiated within the finally clause, it will not re-execute the .code finally clause. The .code finally clause is simply abandoned. The disestablishment of blocks and .code try clauses is properly interleaved with the execution of .code finally clauses. This means that all surrounding exit points are visible in a .code finally clause, even if the .code finally clause is being invoked as part of a transfer to a distant exit point. The finally clause can make a control transfer to an exit point which is more near than the original one, thereby "hijacking" the control transfer. Also, the anonymous block established by the .code try directive is visible in the .code finally clause. Example: .cblk @(try) @ (try) @ (next "nonexistent-file") @ (finally) @ (accept) @ (end) @(catch file-error) @ (output) file error caught @ (end) @(end) .cble In this example, the .code @(next) directive throws an exception of type .codn file-error , because the given file does not exist. The exit point for this exception is the .code @(catch file-error) clause in the outer-most .code try block. The inner block is not eligible because it contains no catch clauses at all. However, the inner try block has a finally clause, and so during the processing of this exception which is headed for .codn "@(catch file-error)" , the .code finally clause performs an anonymous .codn accept . The exit point for that .code accept is the anonymous block surrounding the inner .codn try . So the original transfer to the .code catch clause is thereby abandoned. The inner .code try terminates successfully due to the .codn accept , and since it constitutes the main clause of the outer try, that also terminates successfully. The .str "file error caught" message is never printed. .c1SS catch clauses .code catch clauses establish their associated .code try blocks as potential exit points for exception-induced control transfers (called "throws"). A .code catch clause specifies an optional list of symbols which represent the exception types which it catches. The .code catch clause will catch exceptions which are a subtype of any one of those exception types. If a .code try block has more than one .code catch clause which can match a given exception, the first one will be invoked. When a .code catch is invoked, it is of course understood that the main clause did not terminate normally, and so the main clause could not have produced any bindings. .code catch clauses are processed prior to .codn finally . If a .code catch clause itself throws an exception, that exception cannot be caught by that same clause or its siblings in the same try block. The .code catch clauses of that block are no longer visible at that point. Nevertheless, the .code catch clauses are still protected by the finally block. If a catch clause throws, or otherwise terminates, the .code finally block is still processed. If a .code finally block throws an exception, then it is simply aborted; the remaining directives in that block are not processed. So the success or failure of the .code try block depends on the behavior of the .code catch clause or the .ocde finally clause, if there is one. If either of them succeed, then the try block is considered a successful match. Example: .IP code: .cblk \ @(try) @ (next "nonexistent-file") @ x @ (catch file-error) @a @(finally) @b @(end) @c .cble .IP data: .cblk \ 1 2 3 .cble .IP result: .cblk \ a="1" b="2" c="3" .cble .PP Here, the .code try block's main clause is terminated abruptly by a .code file-error exception from the .code @(next) directive. This is handled by the .code catch clause, which binds variable .code a to the input line .strn 1 . Then the .code finally clause executes, binding .code b to .strn 2 . The .code try block then terminates successfully, and so .code @c takes .strn "3" . .coSS @ catch Clauses with Parameters A .code catch clause may have parameters following the type name, like this: .cblk @(catch pair (a b)) .cble To write a catch-all with parameters, explicitly write the master supertype t: .cblk @(catch t (arg ...)) .cble Parameters are useful in conjunction with .codn throw . The built-in .code error exceptions carry one argument, which is a string containing the error message. Using .codn throw , arbitrary parameters can be passed from the throw site to the catch site. .dir throw The .code throw directive generates an exception. A type must be specified, followed by optional arguments. For example, .cblk @(throw pair "a" `@file.txt`) .cble throws an exception of type .codn pair , with two arguments, being .str a and the expansion of the quasiliteral .codn `@file.txt` . The selection of the target .code catch is performed purely using the type name; the parameters are not involved in the selection. Binding takes place between the arguments given in .code throw and the target .codn catch . If any .code catch parameter, for which a .code throw argument is given, is a bound variable, it has to be identical to the argument, otherwise the catch fails. (Control still passes to the .codn catch , but the catch is a failed match). .IP code: .cblk \ @(bind a "apple") @(try) @(throw e "banana") @(catch e (a)) @(end) .cble .IP result: .cblk \ [query fails] .cble .PP If any argument is an unbound variable, the corresponding parameter in the .code catch is left alone: if it is an unbound variable, it remains unbound, and if it is bound, it stays as is. .IP code: .cblk \ @(try) @(trow e "honda" unbound) @(catch e (car1 car2)) @car1 @car2 @(end) .cble .IP data: .cblk \ honda toyota .cble .IP result: .cblk \ car1="honda" car2="toyota" .cble .PP If a .code catch has fewer parameters than there are throw arguments, the excess arguments are ignored: .IP code: .cblk \ @(try) @(throw e "banana" "apple" "pear") @(catch e (fruit)) @(end) .cble .IP result: .cblk \ fruit="banana" .cble .PP If a .code catch has more parameters than there are throw arguments, the excess parameters are left alone. They may be bound or unbound variables. .IP code: .cblk \ @(try) @(trow e "honda") @(catch e (car1 car2)) @car1 @car2 @(end) .cble .IP data: .cblk \ honda toyota .cble .IP result: .cblk \ car1="honda" car2="toyota" .cble .PP A .code throw argument passing a value to a .code catch parameter which is unbound causes that parameter to be bound to that value. .code throw arguments are evaluated in the context of the .codn throw , and the bindings which are available there. Consideration of what parameters are bound is done in the context of the catch. .IP code: .cblk \ @(bind c "c") @(try) @(forget c) @(bind (a c) ("a" "lc")) @(throw e a c) @(catch e (b a)) @(end) .cble .IP result: .cblk \ c="c" b="a" a="lc" .cble .PP In the above example, .code c has a toplevel binding to the string .strn "c" , but then becomes unbound via .code forget within the .code try construct, and rebound to the value .strn lc . Since the .code try construct is terminated by a .codn throw , these modifications of the binding environment are discarded. Hence, at the end of the query, variable .code c ends up bound to the original value .strn c . The .code throw still takes place within the scope of the bindings set up by the .code try clause, so the values of .code a and .code c that are thrown are .str a and .strn lc . However, at the .code catch site, variable .code a does not have a binding. At that point, the binding to .str a established in the .code try has disappeared already. Being unbound, the .code catch parameter .code a can take whatever value the corresponding throw argument provides, so it ends up with .strn lc . .dir defex The .code defex directive allows the query writer to invent custom exception types, which are arranged in a type hierarchy (meaning that some exception types are considered subtypes of other types). Subtyping means that if an exception type .code B is a subtype of .codn A , then every exception of type .code B is also considered to be of type .codn A . So a catch for type .code A will also catch exceptions of type .codn B . Every type is a supertype of itself: an .code A is a kind of .codn A . This of course implies that every type is a subtype of itself also. Furthermore, every type is a subtype of the type .codn t , which has no supertype other than itself. Type .code nil is is a subtype of every type, including itself. The subtyping relationship is transitive also. If .code A is a subtype of .codn B , and .code B is a subtype of .codn C , then .code A is a subtype of .codn C . .code defex may be invoked with no arguments, in which case it does nothing: .cblk @(defex) .cble It may be invoked with one argument, which must be a symbol. This introduces a new exception type. Strictly speaking, such an introduction is not necessary; any symbol may be used as an exception type without being introduced by .codn @(defex) : .cblk @(defex a) .cble Therefore, this also does nothing, other than document the intent to use a as an exception. If two or more argument symbols are given, the symbols are all introduced as types, engaged in a subtype-supertype relationship from left to right. That is to say, the first (leftmost) symbol is a subtype of the next one, which is a subtype of the next one and so on. The last symbol, if it had not been already defined as a subtype of some type, becomes a direct subtype of the master supertype .codn t . Example: .cblk @(defex d e) @(defex a b c d) .cble The first directive defines .code d as a subtype of .codn e , and .code e as a subtype of .codn t . The second defines .code a as a subtype of .codn b , .code b as a subtype of .codn c , and .code c as a subtype of .codn d , which is already defined as a subtype of .codn e . Thus .code a is now a subtype of .codn e . The the above can be condensed to: .cblk @(defex a b c d e) .cble Example: .IP code: .cblk \ @(defex gorilla ape primate) @(defex monkey primate) @(defex human primate) @(collect) @(try) @(skip) @(cases) gorilla @name @(throw gorilla name) @(or) monkey @name @(throw monkey name) @(or) human @name @(throw human name) @(end)@#cases @(catch primate (name)) @kind @name @(output) we have a primate @name of kind @kind @(end)@#output @(end)@#try @(end)@#collect .cble .IP data: .cblk \ gorilla joe human bob monkey alice .cble .IP output: .cblk \ we have a primate joe of kind gorilla we have a primate bob of kind human we have a primate alice of kind monkey .cble .PP Exception types have a pervasive scope. Once a type relationship is introduced, it is visible everywhere. Moreover, the .code defex directive is destructive, meaning that the supertype of a type can be redefined. This is necessary so that something like the following works right: .cblk @(defex gorilla ape) @(defex ape primate) .cble These directives are evaluated in sequence. So after the first one, the .code ape type has the type .code t as its immediate supertype. But in the second directive, .code ape appears again, and is assigned the .code primate supertype, while retaining .code gorilla as a subtype. This situation could be diagnosed as an error, forcing the programmer to reorder the statements, but instead \*(TX obliges. However, there are limitations. It is an error to define a subtype-supertype relationship between two types if they are already connected by such a relationship, directly or transitively. So the following definitions are in error: .cblk @(defex a b) @(defex b c) @(defex a c)@# error: a is already a subtype of c, through b @(defex x y) @(defex y x)@# error: circularity; y is already a supertype of x. .cble .dir assert The .code assert directive requires the remaining query or sub-query which follows it to match. If the remainder fails to match, the assert directive throws an exception. If the directive is simply .cblk @(assert) .cble Then it throws an assertion of type assert, which is a subtype of error. The assert directive also takes arguments similar to the throw directive. The following assert directive, if it triggers, will throw an exception of type .codn foo , with arguments .code 1 and .strn 2 : .cblk @(assert foo 1 "2") .cble Example: .cblk @(collect) Important Header ---------------- @(assert) Foo: @a, @b @(end) .cble Without the assertion in places, if the .code Foo: @a, @b part does not match, then the entire interior of the .code @(collect) clause fails, and the collect continues searching for another match. With the assertion in place, if the text .str "Important Header" and its underline match, then the remainder of the collect body must match, otherwise an exception is thrown. Now the program will not silently skip over any Important Header sections due to a problem in its matching logic. This is particularly useful when the matching is varied with numerous cases, and they must all be handled. There is a horizontal .code assert directive also. For instance: .cblk abc@(assert)d@x .cble asserts that if the prefix .str abc is matched, then it must be followed by a successful match for .strn "d@x" , or else an exception is thrown. .SH* TXR LISP The \*(TX language contains an embedded Lisp dialect called \*(TL. This language is exposed in \*(TX in several ways. Firstly, in any situation that calls for an expression, a Lisp expression can be used, if it is preceded by the .code @ character. The Lisp expression is evaluated and its value becomes the value of that expression. Thus, \*(TX directives are embedded in literal text using .codn @ , and Lisp expressions are embedded in directives using .code @ also. Secondly, the .code @(do) directive can be used for evaluating one or more Lisp forms, such that their value is thrown away. This is useful for evaluating some Lisp code for the sake of its side effect, such as defining a variable, updating a hash table, et cetera. Thirdly, the .code @(require) directive can be used to evaluate Lisp expressions as part of the matching logic of the \*(TX pattern language. The return value of the rightmost expression is examined. If it is nil, then the .code @(require) directive triggers a match failure. Otherwise, matching proceeds. Fourth, \*(TL code can be placed into files. On the command line, \*(TX treats files with a .str ".tl" suffix as \*(TL code, and the .code @(load) directive does also. Lastly, \*(TL expressions can be evaluated via the command line, using the .code -e and .code -p options. .B Examples: Bind variable .code a to the integer 4: .cblk @(bind a @(+ 2 2)) .cble Bind variable .code b to the standard input stream: .cblk @(bind a @*stdin*) .cble Define several Lisp functions inside .codn @(do) : .cblk @(do (defun add (x y) (+ x y)) (defun occurs (item list) (cond ((null list) nil) ((atom list) (eql item list)) (t (or (eq (first list) item) (occurs item (rest list))))))) .cble Trigger a failure unless previously bound variable .code answer is greater than 42: .cblk @(require (> (int-str answer) 42) .cble .SS* Overview \*(TL is a small and simple dialect, like Scheme, but much more similar to Common Lisp than Scheme. It has separate value and function binding namespaces, like Common Lisp (and thus is a Lisp-2 type dialect), and represents boolean .B true and .B false with the symbols .code t and .code nil (note the case sensitivity of identifiers denoting symbols!) Furthermore, the symbol .code nil is also the empty list, which terminates nonempty lists. \*(TL has lexically scoped local variables and dynamic global variables, similarly to Common Lisp, including the convention that .code defvar marks symbols for dynamic binding in local scopes. Lexical closures are supported. Functions are lexically scoped in \*(TL; they can be defined in pervasive global environment using .code defun or in local scopes using .code flet and .codn labels . .SS* Additional Syntax Much of the \*(TL syntax has been introduced in the previous sections of the manual, since directive forms are based on it. There is some additional syntax that is useful in \*(TL programming. .SS* Symbol Tokens The symbol tokens in \*(TL, called a .meta lident (Lisp identifier) has a similar syntax to the .meta bident (braced identifier) in the \*(TX pattern language. It may consist of all the same characters, as well as the .code / (slash) character which may not be used in a .metn bident . Thus a .meta lident may consist of these characters, in addition to letters and numbers: .cblk ! $ % & * + - < = > ? \e _ ~ / .cble and of course, may not look like a number. A lone .code / is a symbol in \*(TL. The token .code /abc/ is also a symbol, and not a regular expression, like it is in the braced variable syntax. Within \*(TL, regular expressions are written with a leading .codn # . .SS* Consing Dot Unlike other major Lisp dialects, \*(TL allows a consing dot with no forms preceding it. This construct simply denotes the form which follows the dot. That is to say, the parser implements the following transformation: .cblk (. expr) -> expr .cble This is convenient in writing function argument lists that only take variable arguments. Instead of the syntax: .cblk (defun fun args ...) .cble the following syntax can be used: .cblk (defun fun (. args) ...) .cble When a .code lambda form is printed, it is printed in the following style. .cblk (lambda nil ...) -> (lambda () ...) (lambda sym ...) -> (lambda (. sym) ...) (lambda (sym) ...) -> (lambda (sym) ...) .cble In no other circumstances is .code nil printed as .codn () , or an atom .code sym as .codn "(. sym)" . .SS* Quote and Quasiquote .meIP >> ' expr The quote character in front of an expression is used for suppressing evaluation, which is useful for forms that evaluate to something other than themselves. For instance if .code '(+ 2 2) is evaluated, the value is the three-element list .codn "(+ 2 2)" , whereas if .code (+ 2 2) is evaluated, the value is .codn 4 . Similarly, the value of .code 'a is the symbol .code a itself, whereas the value of .code a is the contents of the variable .codn a . .meIP >> ^ qq-template The caret in front of an expression is a quasiquote. A quasiquote is like a quote, but with the possibility of substitution of material. Under a quasiquote, form is considered to be a quasiquote template. The template is considered to be a literal structure, except that it may contain the notations .cblk .meti >> , expr .cble and .cblk .meti >> ,* expr .cble which denote non-constant parts. A quasiquote gets translated into code which, when evaluated, constructs the structure implied by .metn qq-template , taking into account the unquotes and splices. A quasiquote also processes nested quasiquotes specially. If .meta qq-template does not contain any unquotes or splices (which match its level of nesting), or is simply an atom, then .cblk .meti >> ^ qq-template .cble is equivalent to .cblk .meti >> ' qq-template . .cble in other words, it is like an ordinary quote. For instance .code ^(a b ^(c ,d)) is equivalent to .codn '(a b ^(c ,d)) . Although there is an unquote ,d it belongs to the inner quasiquote .codn ^(c ,d) , and the outer quasiquote does not have any unquotes of its own, making it equivalent to a quote. Dialect note: in Common Lisp and Scheme, .code ^form is written .codn `form , and quasiquotes are also informally known as backquotes. In \*(TX, the backquote character .code ` used for quasi string literals. .meIP >> , expr The comma character is used within a .meta qq-template to denote an unquote. Whereas the quote suppresses evaluation, the comma introduces an exception: an element of a form which is evaluated. For example, list .code ^(a b c ,(+ 2 2) (+ 2 2)) is the list .codn (a b c 4 (+ 2 2)) . Everything in the quasiquote stands for itself, except for the .code ,(+ 2 2) which is evaluated. Note: if a variable is called .codn *x* , then the syntax .code ,*x* means .codn ,* x* : splice the value of .codn x* . In this situation, whitespace between the comma and the variable name should be used: .codn , *x* . .meIP >> ,* expr The comma-star operator is used within quasiquote list to denote a splicing unquote. The form which follows .code ,* must evaluate to a list. That list is spliced into the structure which the quasiquote denotes. For example: .code '(a b c ,*(list (+ 3 3) (+ 4 4) d)) evaluates to .codn "(a b c 6 8 d)" . The expression .code (list (+ 3 3) (+ 4 4)) is evaluated to produce the list .codn "(6 8)" , and this list is spliced into the quoted template. Dialect note: in other Lisp dialects, the equivalent syntax is usually .code ,@ (comma at). The .code @ character already has an assigned meaning, so .code * is used. .SS* Quasiquoting non-List Objects Quasiquoting is supported over hash table and vector literals (see Vectors and Hashes below). A hash table or vector literal can be quoted, like any object, for instance: .cblk '#(1 2 3) .cble The .code #(1 2 3) literal is turned into a vector atom right in the \*(TX parser, and this atom is being quoted: this is .cblk .meti (quote << atom ) .cble syntactically, which evaluates to .metn atom . When a vector is quasi-quoted, this is a case of .cblk .meti >> ^ atom .cble which evaluates to .metn atom . A vector can be quasiquoted, for example: .cblk ^#(1 2 3) .cble Of course, unquotes can occur within it. .cblk (let ((a 42)) ^#(1 ,a 3)) ; value is #(1 42 3) .cble In this situation, the .code ^#(...) notation produces code which constructs a vector. The vector in the following example is also a quasivector. It contains unquotes, and though the quasiquote is not directly applied to it, it is embedded in a quasiquote: .cblk (let ((a 42)) ^(a b c #(d ,a))) ; value is (a b c #(d 42)) .cble Hash table literals have two parts: the list of hash construction arguments and the key-value pairs. For instance: .cblk #H((:equal-based) (a 1) (b 2)) .cble where .code (:equal-based) is the list of construction arguments and the pairs .code (a 1) and .code (b 2) are the key/value entries. Hash literals may be quasiquoted. In quasiquoting, the arguments and pairs are treated as separate syntax; it is not one big list. So the following is not a possible way to express the above hash: .cblk ;; not supported: splicing across the entire syntax (let ((hash-syntax '((:equal-based) (a 1) (b 2)))) ^#H(,*hash-syntax)) .cble This is correct: .cblk ;; fine: splicing hash arguments and contents separately (let ((hash-args '(:equal-based)) (hash-contents '((a 1) (b 2)))) ^#H(,hash-args ,*hash-contents)) .cble .SS* Quasiquoting combined with Quasiliterals When a quasiliteral is embedded in a quasiquote, it is possible to use splicing to insert material into the quasiliteral. Example: .cblk (eval (let ((a 3)) ^`abc @,a @{,a} @{(list 1 2 ,a)}`)) -> "abc 3 3 1 2 3" .cble .SS* Vectors .coIP "#(...)" A hash token followed by a list denotes a vector. For example .code #(1 2 a) is a three-element vector containing the numbers .code 1 and .codn 2 , and the symbol .codn a . .SS* Hashes .meIP <> #H(( hash-argument *) >> ( key << value )*) The notation .code #H followed by a nested list syntax denotes a hash table literal. The first item in the syntax is a list of keywords. These are the same keywords as are used when calling the function hash to construct a hash table. Allowed keywords are: .codn :equal-based , .code :weak-keys and .codn :weak-values . An empty list can be specified as .code nil or .codn () , which defaults to a hash table based on the .code eql function, with no weak semantics. .coSS The @ .. notation In \*(TL, there is a special "dotdot" notation consisting of a pair of dots. This can be written between successive atoms or compound expressions, and is a shorthand for cons. That is to say, .code A .. B translates to .codn "(cons A B)" , and so for instance .code (a b .. (c d) e .. f . g) means .codn "(a (cons b (c d)) (cons e f) . g)" . This is a syntactic sugar useful in certain situations in which a cons is used to represent a pair of numbers or other objects. For instance, if .code L is a list, then .code [L 1 .. 3] computes a sublist of .code L consisting of elements 1 through 2 (counting from zero). Restrictions: The notation must be enclosed in a list. For instance .code a..b is not, by itself, an expression, but .code (a..b) is. This is important if Lisp data is being parsed from a string or stream using the read function. If the data .str "a..b" is parsed, the symbol .code a will be extracted, leaving .strn ..a , which, if parsed, produces a syntax error since it consists of a "dotdot" token followed by a symbol, which is not valid syntax. The notation cannot occur in the dot position; that is, the syntax .code (a . b .. c) is invalid. The dotdot operator can only be used between the non-dot-position elements of a list. .SS* The DWIM Brackets \*(TL has a square bracket notation. The syntax .code [...] is a shorthand way of writing .codn "(dwim ...)" . The .code [] syntax is useful for situations where the expressive style of a Lisp-1 dialect is useful. For instance if .code foo is a variable which holds a function object, then .code [foo 3] can be used to call it, instead of .codn "(call foo 3)" . If foo is a vector, then .code [foo 3] retrieves the fourth element, like .codn "(vecref foo 3)" . Indexing over lists, strings and hash tables is possible, and the notation is assignable. Furthermore, any arguments enclosed in .code [] which are symbols are treated according to a modified namespace lookup rule. More details are given in the documentation for the .code dwim operator. .SS* Compound Forms In \*(TL, there are two types of compound forms: the Lisp-2 style compound forms, denoted by ordinary lists that are expressed with parentheses. There are Lisp-1 style compound forms denoted by the DWIM Brackets, discussed in the previous section. The first position of an ordinary Lisp-2 style compound form, is expected to have a function or operator name. Then arguments follow. There may also be an expression in the dotted position, if the form is a function call. If the form is a function call then the arguments are evaluated. If any of the arguments are symbols, they are treated according to Lisp-2 namespacing rules. Additionally, if there is an expression in the dotted position, it is also evaluated. It should evaluate to a sequence: a list, vector or string. The elements of the sequence generate additional arguments for the function call. Note, however, that a compound form cannot be used in the dot position, for obvious reasons, namely that .code (a b c . (foo z)) does not mean that there is a compound form in the dot position, but denotes an alternate spelling for .codn "(a b c foo z)" , where foo behaves as a variable. The DWIM brackets are similar, except that the first position is an arbitrary expression which is evaluated according to the same rules as the remaining positions. The first expression must evaluate to a function, or else to some other object for which the DWIM syntax is defined, such as a vector, string, list or hash. Operators are not supported. The dotted syntax for application of additional arguments from a list or vector is supported in the DWIM brackets just like in the parentheses. Examples: .cblk ;; a contains 3 ;; b contains 4 ;; c contains #(5 6 7) ;; s contains "xyz" (foo a b . c) ;; calls (foo 3 4 5 6 7) (foo a) ;; calls (foo 3) (foo . s) ;; calls (foo #\ex #\ey #\ez) [foo a b . c] ;; calls (foo 3 4 5 6 7) [c 1] ;; indexes into vector #(5 6 7) to yield 6 .cble Dialect Note: In some other Lisp dialects, the improper list syntax is not supported; a function called apply (or similar) must be used for application even if the expression which gives the trailing arguments is a symbol. Moreover, applying sequences other than lists is not supported. .SS* Regular Expressions In \*(TL, the .code / character can occur in symbol names, and the .code / token is a symbol. Therefore the .code /regex/ syntax is not used for denoting regular expressions; rather, the .code #/regex/ syntax is used. .coSS Generalization of List Accessors In ancient Lisp in the 1960's, it was not possible to apply the operations .code car and .code cdr to the .code nil symbol (empty list), because it is not a .code cons cell. In the InterLisp dialect, this restriction was lifted: these operations were extended to accept .code nil (and return .codn nil ). The convention was adopted in other Lisp dialects such as MacLisp and eventually in Common Lisp. Thus there exists an object which is not a cons, yet which takes .code car and .codn cdr . In \*(TL, this relaxation is extended further. For the sake of convenience, the operations .code car and .codn cdr , are made to work with strings and vectors: .cblk (cdr "") -> nil (car "") -> nil (car "abc") -> #\ea (cdr "abc") -> "bc" (cdr #(1 2 3)) -> #(2 3) (car #(1 2 3)) -> 1 .cble The .code ldiff function is also extended in a special way. When the right parameter is a string or vector, then it uses the equal equality test rather than eq for detecting the tail of the list. .cblk (ldiff "abcd" "cd") -> (#\ea #\eb) .cble The .code ldiff operation starts with .str "abcd" and repeatedly applies .code cdr to produce .str "bcd" and .strn "cd" , until the suffix is equal to the second argument: .cblk (equal "cd" "cd") .cble yields true. Operations based on .codn car , .code cdr and .codn ldiff , such as .code keep-if and .code remq extend to strings and vectors. Most derived list processing operations such as .code remq or .code mapcar obey the following rule: the returned object follows the type of the leftmost input list object. For instance, if one or more sequences are processed by .codn mapcar , and the leftmost one is a character string, the function is expected to return characters, which are converted to a character string. However, in the event that the objects produced cannot be assembled into that type of sequence, a list is returned instead. For example .cblk [mapcar list "ab" "12"] .cble returns .codn "((#\ea #\eb) (#\e1 #\e2))" , because a string cannot hold lists of characters. However .cblk [mappend list "ab" "12"] .cble returns .strn "a1b2" . The lazy versions of these functions such as .code mapcar* do not have this behavior; they produce lazy lists. .SS* Callable Objects In \*(TL, sequences (strings, vectors and lists) and hashes can be used as functions everywhere, not just with the DWIM brackets. Sequences work as one or two-argument functions. With a single argument, an element is selected by position and returned. With two arguments, a range is extracted and returned. Hashes also work as one or two argument functions, corresponding to the arguments of the gethash function. Moreover, when a sequence is used as a function of one argument, and the argument is a cons cell rather than an integer, then the call becomes a two-argument call in which the car and cdr of the cell are passed as separate arguments. This allows for syntax like .cblk (call "abc" 0..1). .cble .B Example 1: .cblk (mapcar "abc" '(2 0 1)) -> (#\ec #\ea #\eb) .cble Here, .code mapcar treats the string .str abc as a function of one argument (since there is one list argument). This function maps the indices .codn 0 , .code 1 and .code 2 to the corresponding characters of string .strn abc . Through this function, the list of integer indices .code (2 0 1) is taken to the list of characters .codn (#\ec #\ea #\eb) . .B Example 2: .cblk (call '(1 2 3 4) 1..3) -> (2 3) .cble Here, the shorthand .code 1 .. 3 denotes .codn (cons 1 3) . A cons cell as an argument to a sequence performs range extraction: taking a slice starting at index 1, up to and not including index 3, as if by the call .codn (sub '(1 2 3 4) 1 3) . .B Example 3: .cblk (call '(1 2 3 4) '(0 2)) -> (1 2) .cble A list of indices applied to a sequence is equivalent to using the select function, as if .code (select '(1 2 3 4) '(0 2)) were called. .SS* Special Variables Similarly to Common Lisp, \*(TL is lexically scoped by default, but also has dynamically scoped (a.k.a "special") variables. When a variable is defined with .codn defvar , a binding for the symbol is introduced in the global name space, regardless of in what scope the .code defvar form occurs. Furthermore, at the time the defvar form is evaluated, the symbol which names the variable is tagged as special. When a symbol is tagged as special, it behaves differently when it is used in a lexical binding construct like .codn let , and all other such constructs such as function parameter lists. Such a binding is not the usual lexical binding, but a "rebinding" of the global variable. Over the dynamic scope of the form, the global variable takes on the value given to it by the rebinding. When the form terminates, the prior value of the variable is restored. (This is true no matter how the form terminates; even if by an exception.) Because of this "pervasive special" behavior of a symbol that has been used as the name of a global variable, a good practice is to make global variables have visually distinct names via the "earmuffs" convention: beginning and ending the name with an asterisk. Certain variables in \*(TX's library break this convention; however, they at least have distinct prefixes, examples being example s-ifmt, log-emerg and sig-hup. .TP* "Example:" .cblk (defvar *x* 42) ;; *x* has a value of 42 (defun print-x () (format t "~a\en" *x*)) (let ((*x* "abc")) ;; this overrides *x* (print-x)) ;; *x* is now "abc" and so that is printed (print-x) ;; *x* is 42 again and so "42" is printed .cble .TP* "Dialect Note:" The terms .I bind and .I binding are used differently in \*(TL compared to ANSI Common Lisp. In \*(TL binding is an association between a symbol and an abstract storage location. The association is registered in some namespace, such as the global namespace or a lexical scope. That storage location, in turn, contains a value. In ANSI Lisp, a binding of a dynamic variable is the association between the symbol and a value. It is possible for a dynamic variable to exist, and not have a value. A value can be assigned, which creates a binding. In \*(TL, an assignment is an operation which transfers a value into a binding, not one which creates a binding. In ANSI Lisp, a dynamic variable can exist which has no value. Accessing the value signals a condition, but storing a value is permitted; doing so creates a binding. By contrast, in \*(TL a global variable cannot exist without a value. If a .code defvar form doesn't specify a value, and the variable doesn't exist, it is created with a value of .codn nil . .SS* Syntactic Places and Accessors The \*(TL feature known as .meta syntactic places allows programs to use the syntax of a form which is used to .I access a value from an environment or object, as an expression which denotes a .I place where a value may be .I stored. They are almost exactly the same concept as "generalized references" in Common Lisp, and are related to "lvalues" in languages in the C family, or "designators" in Pascal. .NP* Symbolic Places A symbol is a is a syntactic place if it names a variable. If .code a is a variable, then it may be assigned using the .code set operator: the form .code (set a 42) causes .code a to have the integer value 42. .NP* Compound Places A compound expression can be a syntactic place, if its leftmost constituent is as symbol which is specially registered, and if the form has the correct syntax for that kind of place, and suitable semantics. Such an expression is a compound place. An example of a compound place is a .code car form. If .code c is an expression denoting a .code cons cell, then .code (car c) is not only an expression which retrieves the value of the .code car field of the cell. It is also a syntactic place which denotes that field as a storage location. Consequently, the expression .cblk (set (car c) "abc") .cble stores the character string .str "abc" in that location. Although the same effect can be obtained with .cblk (rplaca c "abc") .cble the syntactic place frees the programmer from having to remember different update functions for different kinds of places. There are various other advantages. \*(TL provides a plethora of operators for modifying a place in addition to .codn set . Subject to certain usage restrictions, these operators work uniformly on all places. For instance, the expression .code (rotate (car x) [str 3] y) causes three different kinds of places to exchange contents, while the three expressions denoting those places are evaluated only once. New kinds of place update macros like .code rotate are quite easily defined, as are new kinds of compound places. .NP* Accessor Functions When a function call form such as the above .code (car x) is a syntactic place, then the function is called an .IR accessor . This term is used throughout this document to denote functions which have associated syntactic places. .NP* Macro Call Syntactic Places Syntactic places can be macros (global and lexical), including symbol macros. So for instance in .code (set x 42) the .code x place can actually be a symbolic macro which expands to, say, .codn (cdr y) . This means that the assignment is effectively .codn (set (cdr y) 42) . .NP* User-Defined Syntactic Places and Place Operators Syntactic places, as well as operators upon syntactic places, are both open-ended. Code can be written quite easily in \*(TL to introduce new kinds of places, as well as new place-mutating operators. New places can be introduced with the help of the .code defplace macro, or possibly the .code define-place-macro macro in simple cases when a new syntactic place can be expressed as a transformation to the syntax of an existing place. New place update macros (place operators) are written using the ordinary macro definer .codn defmacro , with the help of special utility macros called .codn with-update-expander , .codn with-clobber-expander , and .codn with-delete-expander . Simple update macros similar to .code inc and .code push can be written compactly using .codn define-modify-macro . .NP* Deletable Places Unlike generalized references in Common Lisp, \*(TL syntactic places support the concept of deletion. Some kinds of places can be deleted, which is an action distinct from (but does not preclude) being overwritten with a value. What exactly it means for a place to be deleted, or whether that is even permitted, depends on the kind of place. For instance a place which denotes a lexical variable may not be deleted, whereas a global variable may be. A place which denotes a hash table entry may be deleted, and results in the entry being removed from the hash table. Deleting a place in a list causes the trailing items, if any, or else the terminating atom, to move in to close the gap. Users may, of course, define new kinds of places which support deletion semantics. .NP* Evaluation of Places To bring about their effect, place operators must evaluate one or more places. Moreover, some of them evaluate additional forms which are not places. Which arguments of a place operator form are places and which are ordinary forms depends on its specific syntax. For all the built-in place operators, the position of an argument in the syntax determines whether it is treated as (and consequently required to be) a syntactic place, or whether it is an ordinary form. All built-in place operators perform the evaluation of place and non-place argument forms in strict left to right order. Place forms are evaluated not in order to compute a value, but in order to determine the storage location. In addition to determining a storage location, the evaluation of a place form may possibly give rise to side effects. Once a place is fully evaluated, the storage location can then be accessed. Access to the storage location is not considered part of the evaluation of a place. To determine a storage location means to compute some hidden referential object which provides subsequent access to that location without the need for a re-evaluation of the original place form. (The subsequent access to the place through this referential object may still require a multi-step traversal of a data structure; minimizing such steps is a matter of optimization.) Place forms may themselves be compounds, which contain subexpressions that must be evaluated. All such evaluation for the built-in places takes place in left to right order. Certain place operators, such as .code shift and .codn rotate , exhibit an unspecified behavior with regard to the timing of the access of the prior value of a place, relative to the evaluation of places which occur later in the same place operator form. Access to the prior values may be delayed until the entire form is evaluated, or it may be interleaved into the evaluation of the form. For example, in the form .codn (shift a b c 1) , the prior value of .code a can be accessed and saved as soon as .code a is evaluated, prior to the evaluation of .codn b . Alternatively, .code a may be accessed and saved later, after the evaluation of .code b or after the evaluation of all the forms. This issue affects the behavior of place-modifying forms whose subforms contain side effects. It is recommended that such forms not be used in programs. .NP* Nested Places Certain place forms are required to have one or more arguments which are themselves places. The prime example of this, and the only example from among built-in syntactic places, are DWIM forms. A DWIM form has the syntax .cblk .mets (dwim < obj-place < index <> [ alt ]) .cble and of course the square-bracket-notation equivalent: .cblk .mets >> [ obj-place < index <> [ alt ]] .cble Note that not only is the entire form a place, denoting some element or element range of .metn obj-place , but there is the added constraint that .meta obj-place must also itself be a syntactic place. This requirement is necessary, because it supports the behavior that when the element or element range is updated, then .meta obj-place is also potentially updated. After the assignment .cblk (set [obj 0..3] '("forty" "two")) .cble not only is the range of places denoted by .code [obj 0..3] replaced by the list of strings .cblk ("forty" "two") .cble but .code obj may also be overwritten with a new value. This behavior is necessary because the DWIM brackets notation maintains the illusion of an encapsulated array-like container over several dis-similar types, including Lisp lists. But Lisp lists do not behave as fully encapsulated containers. Some mutations on Lisp lists return new objects, which then have to stored (or otherwise accepted) in place of the original objects in order to maintain the array-like container illusion. .NP* Built-In Syntactic Places The following is a summary of the built-in place forms, in addition to symbolic places denoting variables. Of course, new syntactic place forms can be defined by \*(TX programs. .cblk .mets (car << object ) .mets (first << object ) .mets (cdr << object ) .mets (rest << object ) .mets (vecref < vec << idx ) .mets (chr-str < str << idx ) .mets (gethash < hash < key <> [ alt ]) .mets (dwim < obj-place < index <> [ alt ]) .mets >> [ obj-place < index <> [ alt ]] ;; equivalent to dwim .mets (symbol-value << symbol ) .mets (symbol-function << symbol ) .mets (fun << function-name ) .mets (force << promise ) .mets (errno) .cble .NP* Built-In Place-Mutating Operators The following is a summary of the built-in place mutating macros. They are described in detail in their own sections. .meIP (set >> { place << new-value }*) Assigns the values of expressions to places, performing assignments in left to right order, returning the value assigned to the rightmost place. .meIP (pset >> { place << new-value }*) Assigns the values of expressions to places, performing the determination of places and evaluation of the expressions left to right, but the assignment in parallel. Returns the value assigned to the rightmost place. .meIP (zap < place <> [ new-value ]) Assigns .meta new-value to place, defaulting to .codn nil , and returns the prior value. .meIP (flip << place ) Logically toggles the boolean value of .metn place , and returns the new value. .meIP (inc < place <> [ delta ]) Increments .meta place by .metn delta , which defaults to 1, and returns the new value. .meIP (dec < place <> [ delta ]) Decrements .meta place by .metn delta , which defaults to 1, and returns the new value. .meIP (swap < left-place << right-place ) Exchanges the values of .meta left-place and .metn right-place . .meIP (push < item << place ) Pushes .meta item into the list stored in .code place and returns .codn item . .meIP (pop << place ) Pop the list stored in .meta place and returns the popped value. .meIP (shift << place + << shift-in-value) Treats one or more places as a "multi-place shift register". Values are shifted to the left among the places. The rightmost place receives .metn shift-in-value , and the value of the leftmost place emerges as the return value. .meIP (rotate << place *) Treats zero or more places as a "multi-place rotate register". The places exchange values among themselves, by a rotation by one place to the left. The value of the leftmost place goes to the rightmost place, and that value is returned. .meIP (del << place ) Deletes a place which supports deletion, and returns the value which existed in that place prior to deletion. .PP .SH* TXR LISP OPERATOR AND FUNCTION LIBRARY A compound expression with a symbol as its first element, if intended to be evaluated, denotes either an operator invocation or a function call. This depends on whether the symbol names an operator or a function. When the form is an operator invocation, the interpretation of the meaning of that form is under the complete control of that operator. If the compound form is a function call, the remaining forms, if any, denote argument expressions to the function. They are evaluated in left to right order to produce the argument values, which are passed to the function. An exception is thrown if there are not enough arguments, or too many. Programs can define named functions with the defun operator Some operators are macros. There exist predefined macros in the library, and macro operators can also be user-defined using the macro-defining operator .codn defmacro . Operators that are not macros are called special operators. Macro operators work as functions which are given the source code of the form. They analyze the form, and translate it to another form which is substituted in their place. This happens during a code walking phase called the expansion phase, which is applied to Lisp code prior to evaluation. All macros are expanded in the expansion phase, resulting in code which contains only function calls and the executable forms of the operators. (Special operators can also perform code transformations during the expansion phase, but that is not considered macroexpansion, but rather an adjustment of the representation of the operator into an required executable form.) The following sections list all of the special operators, macros and functions in \*(TL. In these sections Syntax is indicated using these conventions: .ie n \{\ . coIP A symbol in angle brackets .\} .el \{\ . coIP \f[5]word\f[] A symbol in .meta fixed-width-italic font .\} denotes some syntactic unit: it may be a symbol or compound form. The syntactic unit is explained in the Description section. .ie n \{\ .coIP {syntax}* * .\} .el \{\ .coIP {syntax}* \f[5]word\f[]* .\} This indicates a repetition of zero or more of the given syntax enclosed in the braces or syntactic unit. .ie n \{\ .coIP {syntax}+ + .\} .el \{\ .coIP {syntax}+ \f[5]word\f[]+ .\} This indicates a repetition of one or more of the given syntax enclosed in the braces or syntactic unit. .ie n \{\ .coIP [syntax] [] .\} .el \{\ .coIP [syntax] [\f[5]word\f[]] .\} Square brackets indicate optional syntax. .coIP alternative1 | alternative2 | ... | alternativeN Multiple syntactic variations allowed in one place are indicated as bar-separated items. .SS* Control Flow and Sequencing .coNP Operators @ progn and @ prog1 .synb .mets (progn << form *) .mets (prog1 << form *) .syne .desc The .code progn operator evaluates forms in order, and returns the value of the last form. The return value of the form .code (progn) is .codn nil . The .code prog1 operator evaluates forms in order, and returns the value of the first form. The return value of the form .code (prog1) is .codn nil . Various other operators such as .code let also arrange for the evaluation of a body of forms, the value of the last of which is returned. These operators are said to feature an implicit .codn progn . .coNP Operator @ cond .synb .mets (cond >> {( test << form *)}*) .syne .desc The .code cond operator provides a multi-branching conditional evaluation of forms. Enclosed in the cond form are groups of forms expressed as lists. Each group must be a list of at least one form. The forms are processed from left to right as follows: the first form, .metn test , in each group is evaluated. If it evaluates true, then the remaining forms in that group, if any, are also evaluated. Processing then terminates and the result of the last form in the group is taken as the result of cond. If .meta test is the only form in the group, then result of .meta test is taken as the result of .codn cond . If the first form of a group yields .codn nil , then processing continues with the next group, if any. If all form groups yield .codn nil , then the cond form yields .codn nil . This holds in the case that the syntax is empty: .code (cond) yields .codn nil . .coNP Macros @, caseq @ caseql and @ casequal .synb .mets (caseq < test-form << normal-clause * <> [ else-clause ]) .mets (caseql < test-form << normal-clause * <> [ else-clause ]) .mets (casequal < test-form << normal-clause * <> [ else-clause ]) .syne .desc These three macros arrange for the evaluation of of .metn test-form , whose value is then compared against the key or keys in each .meta normal-clause in turn. When the value matches a key, then the remaining forms of .meta normal-clause are evaluated, and the value of the last form is returned; subsequent clauses are not evaluated. When the value doesn't match any of the keys of a .meta normal-clause then the next .meta normal-clause is tested. If all these clauses are exhausted, and there is no .metn else-clause , then the value nil is returned. Otherwise, the forms in the .meta else-clause are evaluated, and the value of the last one is returned. The syntax of a .meta normal-clause takes on these two forms: .cblk .mets >> ( key << form *) .cble where .meta key may be an atom which denotes a single key, or else a list of keys. There is a restriction that the symbol .code t may not be used as .metn key . The form .code (t) may be used as a key to match that symbol. The syntax of an .meta else-clause is: .cblk .mets (t << form *) .cble which resembles a form that is often used as the final clause in the .code cond syntax. The three forms of the case construct differ from what type of test they apply between the value of .meta test-form and the keys. The .code caseq macro generates code which uses the .code eq function's equality. The .code caseql macro uses .codn eql , and .code casequal uses .codn equal . .TP* Example .cblk (let ((command-symbol (casequal command-string (("q" "quit") 'quit) (("a" "add") 'add) (("d" "del" "delete") 'delete) (t 'unknown)))) ...) .cble .coNP Operator/function @ if .synb .mets (if < cond < t-form <> [ e-form ]) .mets [if < cond < then <> [ else ]] .syne .desc There exist both an .code if operator and an .code if function. A list form with the symbol .code if in the fist position is interpreted as an invocation of the .code if operator. The function can be accessed using the DWIM bracket notation and in other ways. The .code if operator provides a simple two-way-selective evaluation control. The .meta cond form is evaluated. If it yields true then .meta t-form is evaluated, and that form's return value becomes the return value of the .codn if . If .meta cond yields false, then .meta e-form is evaluated and its return value is taken to be that of .codn if . If .meta e-form is omitted, then the behavior is as if .meta e-form were specified as .codn nil . The .code if function provides no evaluation control. All of arguments are evaluated from left to right. If the .meta cond argument is true, then it returns the .meta then argument, otherwise it returns the value of the .meta else argument if present, otherwise it returns .codn nil . .coNP Operator/function @ and .synb .mets (and << form *) .mets [and << arg *] .syne .desc There exist both an .code and operator and an .code and function. A list form with the symbol .code and in the fist position is interpreted as an invocation of the operator. The function can be accessed using the DWIM bracket notation and in other ways. The .code and operator provides three functionalities in one. It computes the logical "and" function over several forms. It controls evaluation (a.k.a. "short-circuiting"). It also provides an idiom for the convenient substitution of a value in place of .code nil when some other values are all true. The .code and operator evaluates as follows. First, a return value is established and initialized to the value .codn t . The .metn form s, if any, are evaluated from left to right. The return value is overwritten with the result of each form. Evaluation stops when all forms are exhausted, or when .code nil is stored in the return value. When evaluation stops, the operator yields the return value. The .code and function provides no evaluation control; it receives all of its arguments fully evaluated. If it is given no arguments, it returns .codn t . If it is given one or more arguments, and any of them are .codn nil , it returns .codn nil . Otherwise it returns the value of the last argument. .TP* Examples: .cblk (and) -> t (and (> 10 5) (stringp "foo")) -> t (and 1 2 3) -> 3 ;; short-hand for (if (and 1 2) 3). .cble .coNP Operator/function @ or .synb .mets (or << form *) .mets [or << arg *] .syne .desc There exist both an .code or operator and an .code or function. A list form with the symbol .code or in the fist position is interpreted as an invocation of the operator. The function can be accessed using the DWIM bracket notation and in other ways. The or operator provides three functionalities in one. It computes the logical "or" function over several forms. It controls evaluation (a.k.a. "short-circuiting"). The behavior of .code or also provides an idiom for the selection of the first non-nil value from a sequence of forms. The .code or operator evaluates as follows. First, a return value is established and initialized to the value .codn nil . The .metn form s, if any, are evaluated from left to right. The return value is overwritten with the result of each .metn form . Evaluation stops when all forms are exhausted, or when a true value is stored into the return value. When evaluation stops, the operator yields the return value. The .code or function provides no evaluation control; it receives all of its arguments fully evaluated. If it is given no arguments, it returns .codn nil . If all of its arguments are .codn nil , it also returns .codn nil . Otherwise, it returns the value of the first argument which isn't .codn nil . .TP* Examples: .cblk (or) -> nil (or 1 2) -> 1 (or nil 2) -> 2 (or (> 10 20) (stringp "foo")) -> t .cble .coNP Macros @ when and @ unless .synb .mets (when < expression << form *) .mets (unless < expression << form *) .syne .desc The when macro operator evaluates .metn expression . If .meta expression yields true, and there are additional forms, then each .meta form is evaluated. The value of the last form is becomes the result value of the when form. If there are no forms, then the result is .codn nil . The .code unless operator is similar to when, except that it reverses the logic of the test. The forms, if any, are evaluated if, and only if .meta expression is false. .coNP Macros @ while and @ until .synb .mets (while < expression << form *) .mets (until < expression << form *) .syne .desc The .code while macro operator provides a looping construct. It evaluates .metn expression . If .meta expression yields .codn nil , then the evaluation of the .code while form terminates, producing the value .codn nil . Otherwise, if there are additional forms, then each .meta form is evaluated. Next, evaluation returns to .metn expression , repeating all of the previous steps. The .code until macro operator is similar to while, except that the until form terminates when .meta expression evaluates true, rather than false. These operators arrange for the evaluation of all their enclosed forms in an anonymous block. Any of the .metn form s, or .metn expression , may use the .code return operator to terminate the loop, and optionally to specify a result value for the form. The only way these forms can yield a value other than .code nil is if the .code return operator is used to terminate the implicit anonymous block, and is given an argument, which becomes the result value. .coNP Macros @ while* and @ until* .synb .mets (while* < expression << form *) .mets (until* < expression << form *) .syne .desc The .code while* and .code until* macros are similar, respectively, to the macros .code while and .codn until . They differ in one respect: they begin by evaluating the .metn form -s one time unconditionally, without first evaluating .metn expression . After this evaluation, the subsequent behavior is like that of .code while or .codn until . Another way to regard the behavior is that that these forms execute one iteration unconditionally, without evaluating the termination test prior to the first iteration. Yet another view is that these constructs relocate the test from the "top of the loop" to the "bottom of the loop". .coNP Macro @ whilet .synb .mets (whilet >> ({ sym | >> ( sym << init-form )}+) .mets \ \ << body-form *) .syne .desc The .code whilet macro provides a construct which combines iteration with variable binding. The evaluation of the form takes place as follows. First, fresh bindings are established for .metn sym -s as if by the .code let* operator. It is an error for the list of variable bindings to be empty. After the establishment of the bindings, the the value of the .meta sym is tested. If the value is .codn nil , then .code whilet terminates. Otherwise, .metn body-form -s are evaluated in the scope of the variable bindings, and then .code whilet iterates from the beginning, again establishing fresh bindings for the .metn sym -s, and testing the value of the last .metn sym . All evaluation takes place in an anonymous block, which can be terminated with the .code return operator. Doing so terminates the loop. If the .code whilet loop is thus terminated by an explicit .codn return , a return value can be specified. Under normal termination, the return value is .codn nil . .TP* Examples: .cblk ;; read lines of text from *std-input* and print them, ;; until the end-of-stream condition: (whilet ((line (get-line))) (put-line line)) ;; read lines of text from *std-input* and print them, ;; until the end-of-stream condition occurs or ;; a line is identical to the character string "end". (whilet ((line (get-line)) (more (and line (not (equal line "end"))))) (put-line line)) .cble .coNP Macros @ iflet and @ whenlet .synb .mets (iflet >> ({ sym | >> ( sym << init-form )}+) .mets \ \ < then-form <> [ else-form ]) .mets (whenlet >> ({ sym | >> ( sym << init-form )}+) .mets \ \ << body-form *]) .syne .desc The .code iflet and .code whenlet macros combine the variable binding of .code let* with conditional evaluation of .code if and .codn when , respectively. The evaluation of these forms takes place as follows. First, fresh bindings are established for .metn sym -s as if by the .code let* operator. It is an error for the list of variable bindings to be empty. Then, the last variable's value is tested. If it is not .code nil then the test is true, otherwise false. In the case of the .code iflet operator, if the test is true, the operator evaluates .meta then-form and yields its value. Otherwise the test is false, and if the optional .meta else-form is present, that is evaluated instead and its value is returned. If this form is missing, then .code nil is returned. In the case of the .code whenlet operator, if the test is true, then the .metn body-form -s, if any, are evaluated. The value of the last one is returned, otherwise .code nil if the forms are missing. If the test is false, then evaluation of .metn body-form -s is skipped, and .code nil is returned. .TP* Examples: .cblk ;; dispose of foo-resource if present (whenlet ((foo-res (get-foo-resource obj))) (foo-shutdown foo-res) (set-foo-resource obj nil)) ;; Contrast with: above, using when and let (let ((foo-res (get-foo-resource obj))) (when foo-res (foo-shutdown foo-res) (set-foo-resource obj nil))) ;; print frobosity value if it exceeds 150 (whenlet ((fv (get-frobosity-value)) (exceeds-p (> fv 150))) (format t "frobosity value ~a exceeds 150\en" fv)) .cble .coNP Macro @ ifa .synb .mets (ifa < cond < then <> [ else ]) .syne .desc The .code ifa macro provides a anaphoric conditional operator resembling the .code if operator. Around the evaluation of the .meta then and .meta else forms, the symbol .code it is implicitly bound to a subexpression of .metn cond , providing a reference to that value, similar to the word "it" in the English language, and similar anaphoric pronouns in other languages. If .code it is bound to a place form, the binding is established as if using the .code placelet operator. Otherwise, .code it is bound as an ordinary lexical variable to the form's value. The .code ifa macro imposes several restrictions on the .meta cond expression. Firstly, the .meta cond expression must be either an atom, or a function call form: a compound expression with a symbol in the leftmost position which resolves to a function. Otherwise the .code ifa macro invocation is ill-formed. Secondly, if the .meta cond expression is a function call with two or more arguments, at most one of them may be an it-candidate: an expression viable for having its value or storage location bound to the .code it symbol. If there are two or more it-candidates, the .code ifa expression is ill-formed. If .meta cond is an atom, or a function call expression with no arguments, then the .code it symbol is not bound. Effectively, .code ifa macro behaves like the ordinary .code if operator. If .meta cond is an invocation of the functions .code not or .codn null , If .meta cond is a function call with exactly one argument, then the .code it variable is bound to the value of that argument, except when the function being called is .codn not , .code null , or .codn false . That special situation is rewritten according to the following pattern: .cblk .mets (ifa (not << expr ) < then << else ) -> (ifa < expr < else << then ) .cble Note the reversal of .meta then and .metn else . If .meta cond is a function call with two or more arguments, then it is only well-formed if at most one of those arguments is an it-candidates. If there is one such argument, then the .code it variable is bound to it. Otherwise the variable is bound to the leftmost argument expression, regardless of whether that argument expression is an it-candidate. An it-candidate is any expression which is not a constant expression according to the .code constantp function, and not a symbol. In all other regards, the .code ifa macro behaves similarly to .codn if . The .meta cond expression is evaluated, and, if applicable, the value of, or storage location denoted by the appropriate argument is captured and bound to the variable .code it whose scope extends over the .meta then form, as well as over .metn else , if present. If .meta cond yields a true value, then .meta then is evaluated and the resulting value is returned, otherwise .meta else is evaluated if present and its value is returned. A missing .meta else is treated as if it were the .code nil form. .TP* Examples: .cblk (ifa t 1 0) -> 1 (let ((x 6) (y 49)) (ifa (> y (* x x)) ;; it binds to (* x x) (list it))) -> (36) (ifa (evenp 4) (list it)) -> (4) (ifa (not (oddp 4)) (list it)) -> (4) .cble .coNP Macro @ conda .synb .mets (conda >> {( test << form *)}*) .syne .desc The .code conda operator provides a multi-branching conditional evaluation of forms, similarly to the .code cond operator. Enclosed in the cond form are groups of forms expressed as lists. Each group must be a list of at least one form. The .code conda operator is anaphoric: it expands into a nested structure of zero or more .code ifa invocations, according to these patterns: .cblk (conda) -> nil (conda (x y ...) ...) -> (ifa x (progn y ...) (conda z ...)) .cble Thus, .code conda inherits all the restrictions on the .meta test expressions from .codn ifa , as well as the anaphoric .code it variable feature. .coNP Macro @ dotimes .synb .mets (dotimes >> ( var < count-form <> [ result-form ]) << body-form *) .syne .desc The .code dotimes macro implements a simple counting loop. .meta var is established as a variable, and initialized to zero. .meta count-form is evaluated one time to produce a limiting value, which should be a number. Then, if the value of .meta var is less than the limiting value, the .metn body-form -s are evaluated, .meta var is incremented by one, and the process repeats with a new comparison of .meta var against the limiting value possibly leading to another evaluation of the forms. If .meta var is found to equal or exceed the limiting value, then the loop terminates. When the loop terminates, its return value is .code nil unless a .meta result-form is present, in which case the value of that form specifies the return value. .metn body-form -s as well as .meta result-form are evaluated in the scope in which the binding of .meta var is visible. .coNP Operator @ unwind-protect .synb .mets (unwind-protect < protected-form << cleanup-form *) .syne .desc The .cod unwind-protect operator evaluates .meta protected-form in such a way that no matter how the execution of .meta protected-form terminates, the .metn cleanup-form s will be executed. The .metn cleanup-form s, however, are not protected. If a .meta cleanup-form terminates via some non-local jump, the subsequent .metn cleanup-form s are not evaluated. .metn cleanup-form s themselves can "hijack" a non-local control transfer such as an exception. If a .meta cleanup-form is evaluated during the processing of a dynamic control transfer such as an exception, and that .meta cleanup-form initiates its own dynamic control transfer, the original control transfer is aborted and replaced with the new one. .TP* Example: .cblk (block foo (unwind-protect (progn (return-from foo 42) (format t "not reached!\en")) (format t "cleanup!\en"))) .cble In this example, the protected .code progn form terminates by returning from block .codn foo . Therefore the form does not complete and so the output .str not reached! is not produced. However, the cleanup form executes, producing the output .strn cleanup! . .coNP Operator @ block .synb .mets (block < name << body-form *) .syne .desc The .code block operator introduces a named block around the execution of some forms. The .meta name argument must be a symbol. Since a block name is not a variable binding, keyword symbols are permitted, and so are the symbols .code t and .codn nil . A block named by the symbol nil is slightly special: it is understood to be an anonymous block. Blocks in \*(TL have dynamic scope. This means that the following situation is allowed: .cblk (defun func () (return-from foo 42)) (block foo (func)) .cble The function can return from the .code foo block even though the .code foo block does not lexically surround .codn foo . Thus blocks in \*(TL provide dynamic non-local returns, as well as returns out of lexical nesting. .TP* "Dialect Note:" In Common Lisp, blocks are lexical. A separate mechanism consisting of catch and throw operators performs non-local transfer based on symbols. The \*(TL example: .cblk (defun func () (return-from foo 42)) (block foo (func)) .cble is not allowed in Common Lisp, but can be transliterated to: .cblk (defun func () (throw 'foo 42)) (catch 'foo (func)) .cble Note that foo is quoted in CL. This underscores the dynamic nature of the construct. .code throw itself is a function and not an operator. .coNP Operators @ return and @ return-from .synb .mets (return <> [ value ]) .mets (return-from < name <> [ value ]) .syne .desc The .code return operator must be dynamically enclosed within an anonymous block (a block named by the symbol .codn nil ). It immediately terminates the evaluation of the innermost anonymous block which encloses it, causing it to return the specified value. If the value is omitted, the anonymous block returns .codn nil . The .code return-from operator must be dynamically enclosed within a named block whose name matches the .meta name argument. It immediately terminates the evaluation of the innermost such block, causing it to return the specified value. If the value is omitted, that block returns .codn nil . .TP* Example: .cblk (block foo (let ((a "abc\en") (b "def\en")) (pprint a *stdout*) (return-from foo 42) (pprint b *stdout*))) .cble Here, the output produced is .strn "abc" . The value of .code b is not printed because. .code return-from terminates block .codn foo , and so the second pprint form is not evaluated. .SS* Evaluation .coNP Operator @ dwim .synb .mets (dwim << argument *) .mets <> [ argument *] .mets (set (dwim < obj-place < index <> [ alt ]) << new-value ) .mets (set >> [ obj-place < index <> [ alt ]] << new-value ) .syne .desc The .code dwim operator's name is an acronym: DWIM may be taken to mean "Do What I Mean", or alternatively, "Dispatch, in a Way that is Intelligent and Meaningful". The notation .code [...] is a shorthand equivalent to .code (dwim ...) and is usually the preferred way for writing .code dwim expressions. The .code dwim operator takes a variable number of arguments, which are all evaluated in the same way: the first argument is not evaluated differently from the remaining arguments. This means that the first argument isn't a function name, but an ordinary expression which can simply compute a function object (or, more generally, a callable object). Furthermore, for those arguments of .code dwim which are symbols (after all macro-expansion is performed on the arguments), the evaluation rules are altered. For the purposes of resolving symbols to values, the function and variable binding namespaces are considered to be merged into a single space, creating a situation that is very similar to a Lisp-1 style dialect. This special Lisp-1 evaluation is not recursively applied. All arguments of .code dwim which, after macro expansion, are not symbols are evaluated using the normal Lisp-2 evaluation rules. Thus, the DWIM operator must be used in every expression where the Lisp-1 rules for reducing symbols to values are desired. After macro expansion, the first argument of .code dwim may not be an operator such as .codn let , or the name of a macro. Prior to macroexpansion, any argument of .code dwim may be a symbol macro. If a symbol has bindings both in the variable and function namespace in scope, and is referenced by a dwim argument, this constitutes a conflict which is resolved according to two rules. When nested scopes are concerned, then an inner binding shadows an outer binding, regardless of their kind. An inner variable binding for a symbol shadows an outer or global function binding, and vice versa. If a symbol is bound to both a function and variable in the global namespace, then the variable binding is favored. Macros do not participate in the special scope conflation, with one exception. What this means is that the space of symbol macros is not folded together with the space of operator macros. An argument of .code dwim that is a symbol might be symbol macro, variable or function, but it cannot be interpreted as the name of a operator macro. The exception is this: from the perspective of a .code dwim form, function bindings can shadow symbol macros. If a function binding is defined in an inner scope relative to a symbol macro for the same symbol, using .code flet or .codn labels , the function hides the symbol macro. In other words, when macro expansion processes an argument of a .code dwim form, and that argument is a symbol, it is treated specially in order to provide a consistent name lookup behavior. If the innermost binding for that symbol is a function binding, it refers to that function binding, even if a more outer symbol macro binding exists, and so the symbol is not expanded using the symbol macro. By contrast, in an ordinary form, a symbolic argument never resolves to a function binding. The symbol refers to either a symbol macro or a variable, whichever is nested closer. How many arguments are required by the .code dwim operator depends on the type of object to which the first argument expression evaluates. The possibilities are: .RS .meIP >> [ function << argument *] Call the given function object with the given arguments. .meIP >> [ symbol << argument *] If the first expression evaluates to a symbol, that symbol is resolved in the function namespace, and then the resulting function, if found, is called with the given arguments. .meIP >> [ sequence << index ] Retrieve an element from .metn sequence , at the specified .metn index , which is a nonnegative integer. This form is also a place if the .meta sequence subform is a place. If a value is stored to this place, it replaces the element. The place may also be deleted, which has the effect of removing the element from the sequence, shifting the elements at higher indices, if any, down one element position, and shortening the sequence by one. .meIP >> [ sequence << from-index..to-below-index ] Retrieve the specified range of elements. The range of elements is specified in the .code car and .code cdr fields of a cons cell, for which the .code .. (dotdot) syntactic sugar is useful. See the section on Range Indexing below. This form is also a syntactic place, if the .meta sequence subform is a place. Storing a value in this place has the effect of replacing the subsequence with a new subsequence. Deleting the place has the effect of removing the specified subsequence from .metn sequence . The .meta new-value argument in a range assignment can be a string, vector or list, regardless of whether the target is a string, vector or list. If the target is a string, the replacement sequence must be a string, or a list or vector of characters. .meIP >> [ sequence << index-list ] Elements specified by .metn index-list , which may be a list or vector, are extracted from .meta sequence and returned as a sequence of the same kind as .metn sequence . This form is equivalent to .cblk .meti (select < sequence << where-index ) .cble except when the target of an assignment operation. This form is a syntactic place if .meta sequence is one. If a sequence is assigned to this place, then elements of the sequence are distributed to the specified locations. The following equivalences hold between index-list-based indexing and the .code select and .code replace functions, except that .code set always returns the value assigned, whereas .code replace returns its first argument: .cblk [seq idx-list] <--> (select seq idx-list) (set [seq idx-list] new) <--> (replace seq new idx-list) .cble Note that unlike the select function, this does not support .cblk .meti >> [ hash << index-list ] .cble because since hash keys may be lists, that syntax is indistinguishable from a simple hash lookup where .meta index-list is the key. .meIP >> [ hash < key <> [ alt ]] Retrieve a value from the hash table corresponding to .metn key , or else return .meta alt if there is no such entry. The expression .meta alt is always evaluated, whether or not its value is used. .RE .PP .TP* "Range Indexing:" Vector and list range indexing is based from zero, meaning that the first element is numbered zero, the second one and so on. zero. Negative values are allowed; the value .code -1 refers to the last element of the vector or list, and .code -2 to the second last and so forth. Thus the range .code 1 .. -2 means "everything except for the first element and the last two". The symbol .code t represents the position one past the end of the vector, string or list, so .code 0 .. t denotes the entire list or vector, and the range .code t .. t represents the empty range just beyond the last element. It is possible to assign to .codn t .. t . For instance: .cblk (defvar list '(1 2 3)) (set [list t .. t] '(4)) ;; list is now (1 2 3 4) .cble The value zero has a "floating" behavior when used as the end of a range. If the start of the range is a negative value, and the end of the range is zero, the zero is interpreted as being the position past the end of the sequence, rather than the first element. For instance the range .code -1..0 means the same thing as .codn -1..t . Zero at the start of a range always means the first element, so that .code 0..-1 refers to all the elements except for the last one. .TP* Notes: The dwim operator allows for a Lisp-1 flavor of programming in \*(TL, which is principally a Lisp-2 dialect. A Lisp-1 dialect is one in which an expression like .code (a b) treats both a and b as expressions subject to the same evaluation rules\(emat least, when .code a isn't an operator or an operator macro. This means that the symbols .code a and .code b are resolved to values in the same namespace. The form denotes a function call if the value of variable .code a is a function object. Thus in a Lisp-1, named functions do not exist as such: they are just variable bindings. In a Lisp-1 .code (car 1 2) means that there is a variable called .codn car , which holds a function. In a Lisp-2 .code (car 1 2) means that there is a function called .codn car , and so .code (car car car) is possible, because there can be also a variable called .code car which holds a cons cell object, rather than the .code car function. The Lisp-1 approach is useful for functional programming, because it eliminates cluttering occurrences of the call and fun operators. For instance: .cblk ;; regular notation (call foo (fun second) '((1 a) (2 b))) ;; [] notation [foo second '((1 a) (2 b))] .cble Lisp-1 dialects can also provide useful extensions by giving a meaning to objects other than functions in the first position of a form, and the .code dwim/[...] syntax does exactly this. \*(TL is a Lisp-2 because Lisp-2 also has advantages. Lisp-2 programs which use macros naturally achieve hygiene because lexical variables do not interfere with the function namespace. If a Lisp-2 program has a local variable called .codn list , this does not interfere with the hidden use of the function .code list in a macro expansion in the same block of code. Lisp-1 dialects have to provide hygienic macro systems to attack this problem. Furthermore, even when not using macros, Lisp-1 programmers have to avoid using the names of functions as lexical variable names, if the enclosing code might use them. The two namespaces of a Lisp-2 also naturally accommodate symbol macros and operator macros. Whereas functions and variables can be represented in a single namespace readily, because functions are data objects, this is not so with symbol macros and operator macros, the latter of which are distinguished syntactically by their position in a form. In a Lisp-1 dialect, given .codn (foo bar) , either of the two symbols could be a symbol macro, but only .code foo can possibly be an operator macro. Yet, having only a single namespace, a Lisp-1 doesn't permit .codn (foo foo) , where .code foo is simultaneously a symbol macro and an operator macro, though the situation is unambiguous by syntax even in Lisp-1. In other words, Lisp-1 dialects do not entirely remove the special syntactic recognition given to the leftmost position of a compound form, yet at the same time they prohibit the user from taking full advantage of it by providing only one namespace. \*(TL provides the "best of both worlds": the DWIM brackets notation provides a model of Lisp-1 computation that is purer than Lisp-1 dialects (since the leftmost argument is not given any special syntactic treatment at all) while the Lisp-2 foundation provides a traditional Lisp environment with its "natural hygiene". .coNP Function @ identity .synb .mets (identity << value ) .syne .desc The .code identity function returns its argument. .TP* Notes: The .code identity function is useful as a functional argument, when a transformation function is required, but no transformation is actually desired. .coNP Function @ eval .synb .mets (eval < form <> [ env ]) .syne .desc The .code eval function treats the .meta form object as a Lisp expression, which is evaluated. The side effects implied by the form are performed, and the value which it produces is returned. The optional .meta env object specifies an environment for resolving the function and variable references encountered in the expression. If this argument is omitted .code nil then evaluation takes place in the global environment. See also: the .code make-env function. .coNP Function @ make-env .synb .mets (make-env >> [ variable-bindings >> [ function-bindings <> [ next-env ]]]) .syne .desc The .code make-env function creates an environment object suitable as the .code env parameter. The .meta variable-bindings and .meta function-bindings parameters, if specified, should be association lists, mapping symbols to objects. The objects in .meta function-bindings should be functions, or objects callable as functions. The .meta next-env argument, if specified, should be an environment. Note: bindings can also be added to an environment using the .code env-vbind and .code env-fbind functions. .coNP Functions @ env-vbind and @ env-fbind .synb .mets (env-vbind < env < symbol << value ) .mets (env-fbind < env < symbol << value ) .syne .desc These functions bind a symbol to a value in either the function or variable space of environment .codn env . Values established in the function space should be functions or objects that can be used as functions such as lists, strings, arrays or hashes. If .meta symbol already exists in the environment, in the given space, then its value is updated with .codn value . .coNP Function @ constantp .synb .mets (constantp < form >> [ env ]) .syne .desc The .code constantp function determines whether .mode form is a constant form, with respect to environment .mode env . If .mode env is absent, the global environment is used. The .mode env argument is used for macro-expanding .modn form . Currently, .code constantp returns true for any form, which, after macro-expansion is a compound form with the symbol .code quote in its first position, a non-symbolic atom, or one of the symbols which evaluate to themselves and cannot be bound as variables. These symbols are the keyword symbols, and the symbols .code t and .codn nil . In the future, .code constantp will be able to recognize more constant forms, such as calls to certain functions whose arguments are constant forms. .SS* Mutation of Syntactic Places .coNP Macro @ set .synb .mets (set >> { place << new-value }*) .syne .desc The .code set operator stores the values of expressions in places. It must be given an even number of arguments. If there are no arguments, then .code set does nothing and returns .codn nil . If there are two arguments, .meta place and .metn new-value , then .meta place is evaluated to determine its storage location, then .meta new-value is evaluated to determine the value to be stored there, and then the value is stored in that location. Finally, the value is also returned as the result value. If there are more than two arguments, then .code set performs multiple assignments in left to right order. Effectively, .code (set v1 e1 v2 e2 ... vn en) is precisely equivalent to .codn (progn (set v1 e1) (set v2 e2) ... (set vn en)) . .coNP Macro @ pset .synb .mets (pset >> { place << new-value }*) .syne .desc The syntax of .code pset is similar to that of .codn set , and the semantics is similar also in that zero or more places are assigned zero or more values. In fact, if there are no arguments, or if there is exactly one pair of arguments, .code pset is equivalent to .codn set . If there are two or more argument pairs, then all of the arguments are evaluated first, in left-to-right order. No store takes place until after every .meta place is determined, and every .meta new-value is calculated. During the calculation, the values to be stored are retained in hidden, temporary locations. Finally, these values are moved into the determined places. The rightmost value is returned as the form's value. The assignments thus appear to take place in parallel, and .code pset is capable of exchanging the values of a pair of places, or rotating the values among three or more places. (However, there are more convenient operators for this, namely .code rotate and .codn swap ). .TP* Example: .cblk ;; exchange x and y (pset x y y x) ;; exchange elements 0 and 1; and 2 and 3 of vector v: (let ((v (vec 0 10 20 30)) (i -1)) (pset [vec (inc i)] [vec (inc i)] [vec (inc i)] [vec (inc i)]) vec) -> #(10 0 30 20) .cble .coNP Macro @ zap .synb .mets (zap < place <> [ new-value ]) .syne .desc The .code zap macro assigns .meta new-value to .meta place and returns the previous value of .metn place . If .meta new-value is missing, then .code nil is used. In more detail, first .code place is evaluated to determine the storage location. Then, the location is accessed to retrieve the previous value. Then, the .code new-value expression is evaluated, and that value is placed into the storage location. Finally, the previously retrieved value is returned. .coNP Macro @ flip .synb .mets (flip << place ) .syne .desc The .code flip macro toggles the boolean value stored in .metn place . If .meta place previously held .codn nil , it is set to .codn t , and if it previously held a value other than .codn nil , it is set to .codn nil . .coNP Macros @ inc and @ dec .synb .mets (inc < place <> [ delta ]) .mets (dec < place <> [ delta ]) .syne .desc The .code inc macro increments .meta place by adding .meta delta to its value. If .meta delta is missing, the value used in its place the integer 1. First the .meta place argument is evaluated as a syntactic place to determine the location. Then, the value currently stored in that location is retrieved. Next, the .meta delta expression is evaluated. Its value is added to the previously retrieved value as if by the .code + function. The resulting value is stored in the place, and returned. The macro .code dec works exactly like .code inc except that addition is replaced by subtraction. The similarly defaulted .meta delta value is subtracted from the previous value of the place. .coNP Macro @ swap .synb .mets (swap < left-place << right-place ) .syne .desc The .code swap macro exchanges the values of .meta left-place and .meta right-place and returns the value which is thereby transferred to .metn right-place . First, .meta left-place and .meta right-place are evaluated, in that order, to determine their locations. Then the prior values are retrieved, exchanged and stored back. The value stored in .meta right-place is also returned. .coNP Macro @ push .synb .mets (push < item << place ) .syne .desc The .code push macro places .meta item at the head of the list stored in .meta place and returns the updated list which is stored back in .metn place . First, the expression .meta item is evaluated to produce the push value. Then, .meta place is evaluated to determine its storage location. Next, the storage location is accessed to retrieve the list value which is stored there. A new object is produced as if by invoking .code cons function on the push value and list value. This object is stored into the location, and returned. .coNP Macro @ pop .synb .mets (pop << place ) .syne The .code pop macro removes an element from the list stored in .meta place and returns it. First, .meta place is evaluated to determine the place. The place is accessed to retrieve the original value. Then a new value is calculated, as if by applying the .code cdr function to the old value. This new value is stored. Finally, a return value is calculated and returned, as if by applying the .code car function to the original value. .coNP Macro @ pushnew .synb .mets (pushnew < item < place >> [ testfun <> [ keyfun ]]) .syne .desc The .code pushnew macro inspects the list stored in .metn place . If the list already contains the item, then it returns the list. Otherwise it creates a new list with the item at the front and stores it back into .metn place , and returns it. First, the expression .meta item is evaluated to produce the push value. Then, .meta place is evaluated to determine its storage location. Next, the storage location is accessed to retrieve the list value which is stored there. The list is inspected to check whether it already contains the push value, as if using the .code member function. If that is the case, the list is returned and the operation finishes. Otherwise, a new object is produced as if by invoking .code cons function on the push value and list value. This object is stored into the location and returned. .coNP Macro @ shift .synb .mets (shift << place + << shift-in-value) .syne .desc The .code shift macro treats one or more places as a "multi-place shift register". The values of the places are shifted one place to the left. The first (leftmost) place receives the value of the second place, the second receives that of the third, and so on. The last (rightmost) place receives .meta shift-in-value (which is not treated as a place, even if it is a syntactic place form). The previous value of the first place is returned. More precisely, all of the argument forms are evaluated left to right, in the process of which the storage locations of the places are determined, .meta shift-in-value is reduced to its value. The values stored in the places are sampled and saved. Note that it is not specified whether the places are sampled in a separate pass after the evaluation of the argument forms, or whether the sampling is interleaved into the argument evaluation. This affects the behavior in situations in which the evaluation of any of the .meta place forms, or of .metn shift-in-value , has the side effect of modifying later places. Next, the places are updated by storing the saved value of the second place into the first place, the third place into the second and so forth, and the value of .meta shift-in-value into the last place. Finally, the saved original value of the first place is returned. .coNP Macro @ rotate .synb .mets (rotate << place *) .syne .desc Treats zero or more places as a "multi-place rotate register". If there are no arguments, there is no effect and .code nil is returned. Otherwise, the last (rightmost) place receives the value of the first (leftmost) place. The leftmost place receives the value of the second place, and so on. If there are two arguments, this equivalent to .codn swap . The prior value of the first place, which is the the value rotated into the last place, is returned. More precisely, the .meta place arguments are evaluated left to right, and the storage locations are thereby determined. The storage locations are sampled, and then the sampled values are stored back into the locations, but rotated by one place as described above. The saved original value of the leftmost .meta place is returned. It is not specified whether the sampling of the original values is a separate pass which takes place after the arguments are evaluated, or whether this sampling it is interleaved into argument evaluation. This affects the behavior in situations in which the evaluation of any of the .meta place forms has the side effect of modifying the value stored in a later .meta place form. .coNP Macro @ del .synb .mets (del << place ) .syne .desc The .code del macro requests the deletion of .codn place . If .code place doesn't support deletion, an exception is thrown. First .code place is evaluated, thereby determining its location. Then the place is accessed to retrieve its value. The place is then subject to deletion. Finally, the previously retrieved value is returned. Precisely what deletion means depends on the kind of place. The built-in places in \*(TL have deletion semantics which are intended to be unsurprising to the programmer familiar with the data structure which holds the place. Generally, if a place denotes the element of a sequence, then deletion of the place implies deletion of the element, and deletion of the element implies that the gap produced by the element is closed. The deleted element is effectively replaced by its successor, that successor by its successor and so on. If a place denotes a value stored in a dynamic data set such as a hash table, then deletion of that place implies deletion of the entry which holds that value. If the entry is identified by a key, that key is also removed. .SS* Binding and Iteration .coNP Operators @ defvar and @ defparm .synb .mets (defvar < sym <> [ value ]) .mets (defparm < sym << value ) .syne .desc The .code defvar operator binds a name in the variable namespace of the global environment. Binding a name means creating a binding: recording, in some namespace of some environment, an association between a name and some named entity. In the case of a variable binding, that entity is a storage location for a value. The value of a variable is that which has most recently been written into the storage location, and is also said to be a value of the binding, or stored in the binding. If the variable named .meta sym already exists in the global environment, the form has no effect; the .meta value form is not evaluated, and the value of the variable is unchanged. If the variable does not exist, then a new binding is introduced, with a value given by evaluating the .meta value form. If the form is absent, the variable is initialized to .codn nil . The .meta value form is evaluated in the environment in which the .code defvar form occurs, not necessarily in the global environment. The symbols .code t and .code nil may not be used as variables, and neither can be keyword symbols: symbols denoted by a leading colon. In addition to creating a binding, the .code defvar operator also marks .meta sym as the name of a special variable. This changes what it means to bind that symbol in a lexical binding construct such as the .code let operator, or a function parameter list. See the section "Special Variables" far above. The .code defparm operator behaves like .code defvar when a variable named .meta sym doesn't already exist. If .meta sym already denotes a variable binding in the global namespace, .code defparm evaluates the .meta value form and assigns the resulting value to the variable. The following equivalence holds: .cblk (defparm x y) <--> (prog1 (defvar x) (set x y)) .cble The .code defvar and .code defparm operators return .metn sym . .coNP Operators @ let and @ let* .synb .mets (let >> ({ sym | >> ( sym << init-form )}*) << body-form *) .mets (let* >> ({ sym | >> ( sym << init-form )}*) << body-form *) .syne .desc The .code let and .code let* operators introduce a new scope with variables and evaluate forms in that scope. The operator symbol, either .code let or .codn let* , is followed by a list which can contain any mixture of variable name symbols, or .cblk .meti >> ( sym << init-form ) .cble pairs. A symbol denotes the name of variable to be instantiated and initialized to the value .codn nil . A symbol specified with an init-form denotes a variable which is initialized from the value of the .metn init-form . The symbols .code t and .code nil may not be used as variables, and neither can be keyword symbols: symbols denoted by a leading colon. The difference between .code let and .code let* is that in .codn let* , later .codn init-form s have visibility over the variables established by earlier variables in the same let* construct. In plain .codn let , the variables are not visible to any of the .metn init-form s. When the variables are established, then the .metn body-form s are evaluated in order. The value of the last .meta body-form becomes the return value of the .codn let . If there are no .metn body-form s, then the return value .code nil is produced. The list of variables may be empty. .TP* Examples: .cblk (let ((a 1) (b 2)) (list a b)) -> (1 2) (let* ((a 1) (b (+ a 1))) (list a b (+ a b))) -> (1 2 3) (let ()) -> nil (let (:a nil)) -> error, :a and nil can't be used as variables .cble .coNP Operators @ for and @ for* .synb .mets ({for | for*} >> ({ sym | >> ( sym << init-form )}*) .mets \ \ \ \ \ \ \ \ \ \ \ \ \ >> ([ test-form << result-form *]) .mets \ \ \ \ \ \ \ \ \ \ \ \ \ <> ( inc-form *) .mets \ \ << body-form *) .syne .desc The .code for and .code for* operators combine variable binding with loop iteration. The first argument is a list of variables with optional initializers, exactly the same as in the .code let and .code let* operators. Furthermore, the difference between .code for and .code for* is like that between .code let and .code let* with regard to this list of variables. The .code for and .code for* operators execute these steps: .RS .IP 1. Establish bindings for the specified variables similarly to .code let and .codn let* . The variable bindings are visible over the .metn test-form , each .metn result-form , each .meta inc-form and each .metn body-form . .IP 2. Establish an anonymous block over the remaining forms, allowing the .code return operator to be used to terminate the loop. .IP 3. Evaluate .metn test-form . If .meta test-form yields .codn nil , then the loop terminates. Each .meta result-form is evaluated, and the value of the last of these forms is is the result value of the loop. If there are no .metn result-form s then the result value is .codn nil . If the .meta test-form is omitted, then the test is taken to be true, and the loop does not terminate. .IP 4. Otherwise, if .meta test-form yields true, then each .meta body-form is evaluated in turn. Then, each .code inc-form is evaluated in turn and processing resumes at step 2. .RE .IP Furthermore, the .code for and .code for* operators establish an anonymous block, allowing the .code return operator to be used to terminate at any point. .coNP Operators @, each @, each* @, collect-each @, collect-each* @ append-each and @ append-each* .synb .mets (each >> ({( sym << init-form )}*) << body-form *) .mets (each* >> ({( sym << init-form )}*) << body-form *) .mets (collect-each >> ({( sym << init-form )}*) << body-form *) .mets (collect-each* >> ({( sym << init-form )}*) << body-form *) .mets (append-each >> ({( sym << init-form )}*) << body-form *) .mets (append-each* >> ({( sym << init-form )}*) << body-form *) .syne .desc These operators establish a loop for iterating over the elements of one or more lists. Each .meta init-form must evaluate to a list. The lists are then iterated in parallel over repeated evaluations of the .metn body-form s, with each .meta sym variable being assigned to successive elements of its list. The shortest list determines the number of iterations, so if any of the .metn init-form s evaluate to an empty list, the body is not executed. The body forms are enclosed in an anonymous block, allowing the .code return operator to terminate the loop prematurely and optionally specify the return value. The .code collect-each and .code collect-each* variants are like .code each and .codn each* , except that for each iteration, the resulting value of the body is collected into a list. When the iteration terminates, the return value of the .code collect-each or .code collect-each* operator is this collection. The .code append-each and .code append-each* variants are like .code each and .codn each* , except that for each iteration other than the last, the resulting value of the body must be a list. The last iteration may produce either an atom or a list. The objects produced by the iterations are combined together as if they were arguments to the append function, and the resulting value is the value of the .code append-each or .code append-each* operator. The alternate forms denoted by the adorned symbols .codn each* , .code collect-each* and .codn append-each* , differ from .codn each , .code collect-each and .code append-each* in the following way. The plain forms evaluate the .metn init-form s in an environment in which none of the .code sym variables are yet visible. By contrast, the alternate forms evaluate each .meta init-form in an environment in which bindings for the previous .meta sym variables are visible. In this phase of evaluation, .meta sym variables are list-valued: one by one they are each bound to the list object emanating from their corresponding .metn init-form . Just before the first loop iteration, however, the .meta sym variables are assigned the first item from each of their lists. .TP* Example: .cblk ;; print numbers from 1 to 10 and whether they are even or odd (each* ((n (range 1 10)) ;; n list a list here! (even (collect-each ((n m)) (evenp m)))) ;; n is an item here! (format t "~s is ~s\en" n (if even "even" "odd"))) .cble .TP* Output: .cblk 1 is odd 2 is even 3 is odd 4 is even 5 is odd 6 is even 7 is odd 8 is even 9 is odd 10 is even .cble .SS* Function Objects and Named Functions .coNP Operator @ defun .synb .mets (defun < name <> ( param * [: << opt-param *] [. << rest-param ]) .mets \ \ << body-form ) .syne .desc The .code defun operator introduces a new function in the global function namespace. The function is similar to a lambda, and has the same parameter syntax and semantics as the .code lambda operator. Unlike in .codn lambda , the .metn body-form s of a .code defun are surrounded by a block. The name of this block is the same as the name of the function, making it possible to terminate the function and return a value using .cblk .meti (return-from < name << value ). .cble For more information, see the definition of the block operator. A function may call itself by name, allowing for recursion. .coNP Operator @ lambda .synb .mets (lambda <> ( param * [: << opt-param *] [. << rest-param ]) .mets \ \ << body-form ) .mets (lambda < rest-param .mets \ \ << body-form ) .syne .desc The .code lambda operator produces a value which is a function. Like in most other Lisps, functions are objects in \*(TL. They can be passed to functions as arguments, returned from functions, aggregated into lists, stored in variables, et cetera. The first argument of .code lambda is the list of parameters for the function. It may be empty, and it may also be an improper list (dot notation) where the terminating atom is a symbol other than .codn nil . It can also be a single symbol. The second and subsequent arguments are the forms making up the function body. The body may be empty. When a function is called, the parameters are instantiated as variables that are visible to the body forms. The variables are initialized from the values of the argument expressions appearing in the function call. The dotted notation can be used to write a function that accepts a variable number of arguments. There are two ways write a function that accepts only a variable argument list and no required arguments: .cblk .mets (lambda (. << rest-param ) ...) .mets (lambda < rest-param ...) .cble (These notations are syntactically equivalent because the list notation .code (. X) actually denotes the object .code X which isn't wrapped in any list). The keyword symbol .code : (colon) can appear in the parameter list. This is the symbol in the keyword package whose name is the empty string. This symbol is treated specially: it serves as a separator between required parameters and optional parameters. Furthermore, the .code : symbol has a role to play in function calls: it can be specified as an argument value to an optional parameter by which the caller indicates that the optional argument is not being specified. It will be processed exactly that way. An optional parameter can also be written in the form .cblk .meti >> ( name < expr <> [ sym ]). .cble In this situation, if the call does not specify a value for the parameter (or specifies a value as the keyword .code : (colon)) then the parameter takes on the value of the expression .metn expr . If .meta sym is specified, then .meta sym will be introduced as an additional binding with a boolean value which indicates whether or not the optional parameter had been specified by the caller. The initializer expressions are evaluated an environment in which all of the previous parameters are visible, in addition to the surrounding environment of the lambda. For instance: .cblk (let ((default 0)) (lambda (str : (end (length str)) (counter default)) (list str end counter))) .cble In this .codn lambda , the initializing expression for the optional parameter end is .codn (length str) , and the .code str variable it refers to is the previous argument. The initializer for the optional variable counter is the expression default, and it refers to the binding established by the surrounding let. This reference is captured as part of the .codn lambda 's lexical closure. .TP* Examples: .IP "Counting function:" This function, which takes no arguments, captures the variable .codn counter . Whenever this object is called, it increments .code counter by .code 1 and returns the incremented value. .cblk (let ((counter 0)) (lambda () (inc counter))) .cble .IP "Function that takes two or more arguments:" The third and subsequent arguments are aggregated into a list passed as the single parameter .codn z : .cblk (lambda (x y . z) (list 'my-arguments-are x y z)) .cble .IP "Variadic function:" .cblk (lambda args (list 'my-list-of-arguments args)) .cble .IP "Optional arguments:" .cblk [(lambda (x : y) (list x y)) 1] -> (1 nil) [(lambda (x : y) (list x y)) 1 2] -> (1 2) .cble .coNP Function @ call .synb .mets (call < function << argument *) .syne .desc The .code call function invokes .metn function , passing it the given arguments, if any. .TP* Examples: Apply arguments .code 1 2 to a .code lambda which adds them to produce .codn 3 : .cblk (call (lambda (a b) (+ a b)) 1 2) .cble Useless use of .code call on a named function; equivalent to .codn (list 1 2) : .cblk (call (fun list) 1 2) .cble .coNP Operator @ fun .synb .mets (fun << function-name ) .syne .desc The .code fun operator retrieves the function object corresponding to a named function in the current lexical environment. The .meta function-name is a symbol denoting a named function: a built in function, or one defined by .codn defun . Note: the .code fun operator does not see macro bindings. It is possible to retrieve a global macro expander using .codn symbol-function . .TP* "Dialect Note:" A lambda expression is not a function name in \*(TL. The syntax .code (fun (lambda ...)) is invalid. .coNP Accessors @ symbol-function and @ symbol-value .synb .mets (symbol-function << symbol ) .mets (symbol-value << symbol ) .mets (set (symbol-function << symbol ) << new-value ) .mets (set (symbol-value << symbol ) << new-value ) .syne .desc The .code symbol-function function retrieves the value of the global function binding of the given .meta symbol if it has one: that is, the function object bound to the .metn symbol . If .meta symbol has no global function binding, then the value of the global macro binding is returned. If that doesn't exist, then the value of a global special operator binding is returned, and if that doesn't exist, then .code nil is returned. The value of a macro binding isn't a function object, but a list of the following form: .cblk .mets (# < macro-parameter-list << body-form *) .cble The value of a special operator binding is a "C pointer" object, whose printed representation looks like: .cblk # .cble These details may change in future version of \*(TX. The .code symbol-value function retrieves the value of a either a global variable or a global symbol macro, whichever exists. Otherwise it returns .codn nil . The value of a symbol macro binding is simply the replacement form. .TP* "Dialect note:" Forms which call .code symbol-function or .code symbol-value are currently not assignable places. Only the .code defun operator defines functions. A .code symbol-function or .code symbol-value form denotes a place. If .meta symbol has a binding in, respectively, the function or variable namespace of the global environment, then a value may be stored in the place, otherwise an error is thrown. Deleting a place doesn't require a binding; it takes place as if using the .code fmakunbound or .code makunbound functions. If a nonexistent place is deleted, the prior value yielded by the deletion is deemed to be .codn nil . .coNP Functions @ boundp and @ fboundp .synb .mets (boundp << symbol ) .mets (fboundp << symbol ) .syne .desc .code boundp returns .code t if the symbol is bound as a variable or symbol macro in the global environment, otherwise .codn nil . .code fboundp returns .code t if the symbol has a function or macro binding in the global environment, or if it is an operator, otherwise .codn nil . .coNP Functions @ makunbound and @ fmakunbound .synb .mets (makunbound << symbol ) .mets (fmakunbound << symbol ) .syne .desc The function .code makunbound removes any binding for .meta symbol from the variable namespace of the global environment. If .meta symbol has no such binding, it does nothing. In either case, it returns .metn symbol . Both variables and symbol macros are bindings in the variable namespace. Additionally, if .meta symbol was previously marked as special, for instance by .codn defvar , this marking is removed by .codn makunbound . The function .code fmakunbound removes any binding for .meta symbol from the function namespace of the global environment. If .meta symbol has no such binding, it does nothing. In either case, it returns .metn symbol . Both functions and macros are bindings in the function namespace. .TP* "Dialect Note:" The behavior of these functions differs from ANSI Common Lisp. The .code makunbound function in ANSI Lisp only removes a value from a dynamic variable. The dynamic variable does not cease to exist, it only ceases to have a value (because a binding is a value). The special property from a symbol is also not removed. In \*(TL, the variable ceases to exist. The binding of a variable isn't its value, it is the variable itself: the association between a name and an abstract storage location, in some environment. If the binding is undone, the variable disappears. .coNP Function @ func-get-form .synb .mets (func-get-form << func ) .syne .desc The .code func-get-form function retrieves a source code form of .metn func , which must be an interpreted function. The source code form has the syntax .cblk .meti >> ( name < arglist << body-form *) . .cble .coNP Function @ func-get-env .synb .mets (func-get-env << func ) .syne .desc The .code func-get-env function retrieves the environment object associated with function .metn func . The environment object holds the captured bindings of a lexical closure. .coNP Function @ functionp .synb .mets (functionp << obj ) .syne .desc The .code functionp function returns .code t if .meta obj is a function, otherwise it returns .codn nil . .coNP Function @ interp-fun-p .synb .mets (interp-fun-p << obj ) .syne .desc The .code interp-fun-p function returns .code t if .meta obj is an interpreted function, otherwise it returns .codn nil . .coNP Macros @ flet and @ labels .synb .mets (flet >> ({( name < param-list << function-body-form *)}*) .mets \ \ << body-form *) .mets (labels >> ({( name < param-list << function-body-form *)}*) .mets \ \ << body-form *) .syne .desc The .code flet and .code labels macros bind local, named functions in the lexical scope. The difference between .code flet and .code labels is that a function defined by .code labels can see itself, and therefore recurse directly by name. Moreover, if multiple functions are defined by the same labels construct, they all see each other. By contrast, a .codn flet -defined function does not have itself in scope and cannot recurse. Multiple functions in the same .code flet do not have each other's names in their scopes. More formally, the .metn function-body-form -s and .meta param-list of the functions defined by .code labels are in a scope in which all of the function names being defined by that same .code labels construct are visible. Under both .code labels and .codn flet , the local functions that are defined are lexically visible to the main .metn body-form -s. Note that .code labels and .code flet are properly scoped with regard to macros. During macro expansion, function bindings introduced by these macro operators shadow macros defined by .code macrolet and .codn defmacro . Furthermore, function bindings introduced by .code labels and .code flet also shadow symbol macros defined by .codn symacrolet , when those symbol macros occur as arguments of a .code dwim form. See also: the .code macrolet operator. .TP* Examples: .cblk ;; Wastefully slow algorithm for determining evenness. ;; Note: ;; - mutual recursion between labels-defined functions ;; - inner is-even bound by labels shadows the outer ;; one bound by defun so the (is-even n) call goes ;; to the local function. (defun is-even (n) (labels ((is-even (n) (if (zerop n) t (is-odd (- n 1)))) (is-odd (n) (if (zerop n) nil (is-even (- n 1))))) (is-even n))) .cblk .SS* Object Type And Equivalence .coNP Function @ typeof .synb .mets (typeof << value ) .syne .desc The .code typeof function returns a symbol representing the type of .metn value . The core types are identified by the following symbols: .coIP cons Cons cell. .coIP str String. .coIP lit Literal string embedded in the \*(TX executable image. .coIP chr Character. .coIP fixnum Fixnum integer: an integer that fits into the value word, not having to be heap allocated. .coIP sym Symbol. .coIP pkg Symbol package. .coIP fun Function. .coIP vec Vector. .coIP lcons Lazy cons. .coIP lstr Lazy string. .coIP env Function/variable binding environment. .coIP bignum A bignum integer: arbitrary precision integer that is heap-allocated. .PP There are additional kinds of objects, such as streams. .coNP Functions @, null @, not and @ false .synb .mets (null << value ) .mets (not << value ) .mets (false << value ) .syne .desc The .codn null , .code not and .code false functions are synonyms. They tests whether .meta value is the object .codn nil . They return .code t if this is the case, .code nil otherwise. .TP* Examples: .cblk (null '()) -> t (null nil) -> t (null ()) -> t (false t) -> nil (if (null x) (format t "x is nil!")) (let ((list '(b c d))) (if (not (memq 'a list)) (format t "list ~s does not contain the symbol a\en"))) .cble .coNP Functions @ true and @ have .synb .mets (true << value ) .mets (have << value ) .syne .desc The .code true function is the complement of the .codn null , .code not and .code false functions. The .code have function is a synonym for .codn true . It return .code t if the .meta value is any object other than .codn nil . If .meta value is .codn nil , it returns .codn nil . Note: programs should avoid explicitly testing values with true. For instance .code (if x ...) should be favored over .codn (if (true x) ...) . However, the latter is useful with the .code ifa macro because .cblk .meti (ifa (true << expr ) ...) .cble binds the .code it variable to the value of .metn expr , no matter what kind of form .meta expr is, which is not true in the .cblk .meti (ifa < expr ...) .cble form. .TP* Example: .cblk ;; Compute indices where the list '(1 nil 2 nil 3) ;; has true values: [where '(1 nil 2 nil 3) true] -> (1 3) .cble .coNP Functions @, eq @ eql and @ equal .synb .mets (eq < left-obj << right-obj ) .mets (eql < left-obj << right-obj ) .mets (equal < left-obj << right-obj ) .syne .desc The principal equality test functions .codn eq , .code eql and .code equal test whether two objects are equivalent, using different criteria. They return .code t if the objects are equivalent, and .code nil otherwise. The .code eq function uses the strictest equivalence test, called implementation equality. The eq function returns .code t if, and only if, .meta left-obj and .meta right-obj are actually the same object. The .code eq test is is implemented by comparing the raw bit pattern of the value, whether or not it is an immediate value or a pointer to a heaped object. Two character values are .code eq if they are the same character, and two fixnum integers are .code eq if they have the same value. All other object representations are actually pointers, and are .code eq if, and only, if they point to the same object in memory. So, for instance, two bignum integers might not be .code eq even if they have the same numeric value, two lists might not be .code eq even if all their corresponding elements are .code eq and two strings might not be eq even if they hold identical text. The .code eql function is slightly less strict than .codn eq . The difference between .code eql and .code eq is that if .meta left-obj and .meta right-obj are numbers which are of the same kind and have the same numeric value, .code eql returns .metn t , even if they are different objects. Note that an integers and a floating-point number are not .code eql even if one has a value which converts to the other: thus, .code (eql 0.0 0) yields .codn nil ; the comparison operation which finds these numbers equal is the .codn (= 0.0 0) . For all other object types, .code eql behaves like .codn eq . The .code equal function is less strict still than .codn eql . In general, it recurses into some kinds of aggregate objects to perform a structural equivalence check. Firstly, if .meta left-obj and .meta right-obj are .code eql then they are also .codn equal , though of course the converse isn't necessarily the case. If two objects are both cons cells, then they are equal if their .code car fields are .code equal and their .code cdr fields are .codn equal . If two objects are vectors, they are .code equal if they have the same length, and their corresponding elements are .codn equal . If two objects are strings, they are equal if they are textually identical. If two objects are functions, they are .code equal if they have .code equal environments, and if they have the same code. Two compiled functions are considered to have the same code if and only if they are pointers to the same function. Two interpreted functions are considered to have the same code if their list structure is .codn equal . Two hashes are .code equal if they use the same equality (both are .codn :equal-based , or both are the default .codn :eql-based ), if their associated user data elements are equal (see the function .codn get-hash-userdata ), if their sets of keys are identical, and if the data items associated with corresponding keys from each respective hash are .code equal objects. For some aggregate objects, there is no special semantics. Two arguments which are symbols, packages, or streams are .code equal if and only if they are the same object. Certain object types have a custom .code equal function. .SS* Basic List Library .coNP Function @ cons .synb .mets (cons < car-value << cdr-value ) .syne .desc The .code cons function allocates, initializes and returns a single cons cell. A cons cell has two fields called .code car and .codn cdr , which are accessed by functions of the same name, or by the functions .code first and .codn rest , which are synonyms for these. Lists are made up of conses. A (proper) list is either the symbol .code nil denoting an empty list, or a cons cell which holds the first item of the list in its .codn car , and the list of the remaining items in .codn cdr . The expression .code (cons 1 nil) allocates and returns a single cons cell which denotes the one-element list .codn (1) . The .code cdr is .codn nil , so there are no additional items. A cons cell whose .code cdr is an atom other than .code nil is printed with the dotted pair notation. For example the cell produced by .code (cons 1 2) is denoted .codn (1 . 2) . The notation .code (1 . nil) is perfectly valid as input, but the cell which it denotes will print back as .codn (1) . The notations are equivalent. The dotted pair notation can be used regardless of what type of object is the cons cell's .codn cdr . so that for instance .code (a . (b c)) denotes the cons cell whose .code car is the symbol a .code a and whose .code cdr is the list .codn (b c) . This is exactly the same thing as .codn (a b c) . In other words .code (a b ... l m . (n o ... w . (x y z))) is exactly the same as .codn (a b ... l m n o ... w x y z). Every list, and more generally cons cell tree structure, can be written in a "fully dotted" notation, such that there are as many dots as there are cells. For instance the cons structure of the nested list .code (1 (2) (3 4 (5))) can be made more explicit using .codn (1 . ((2 . nil) . ((3 . (4 . ((5 . nil) . nil))) . nil)))) . The structure contains eight conses, and so there are eight dots in the fully dotted notation. The number of conses in a linear list like .code (1 2 3) is simply the number of items, so that list in particular is made of three conses. Additional nestings require additional conses, so for instance .code (1 2 (3)) requires four conses. A visual way to count the conses from the printed representation is to count the atoms, then add the count of open parentheses, and finally subtract one. A list terminated by an atom other than .code nil is called an improper list, and the dot notation is extended to cover improper lists. For instance .code (1 2 . 3) is an improper list of two elements, terminated by .codn 3 , and can be constructed using .codn (cons 1 (cons 2 3)) . The fully dotted notation for this list is .codn (1 . (2 . 3)) . .coNP Function @ atom .synb .mets (atom << value ) .syne .desc The .code atom function tests whether .meta value is an atom. It returns .code t if this is the case, .code nil otherwise. All values which are not cons cells are atoms. .code (atom x) is equivalent to .codn (not (consp x)) . .TP* Examples: .cblk (atom 3) -> t (atom (cons 1 2)) -> nil (atom "abc") -> t (atom '(3)) -> nil .cble .coNP Function @ consp .synb .mets (consp << value ) .syne .desc The .code consp function tests whether .meta value is a cons. It returns .code t if this is the case, .code nil otherwise. .code (consp x) is equivalent to .codn (not (atom x)) . Non-empty lists test positive under .code consp because a list is represented as a reference to the first cons in a chain of one or more conses. Note that a lazy cons is a cons and satisfies the .code consp test. See the function .code make-lazy-cons and the macro .codn lcons . .TP* Examples: .cblk (consp 3) -> nil (consp (cons 1 2)) -> t (consp "abc") -> nil (consp '(3)) -> t .cble .coNP Accessors @ car and @ first .synb .mets (car << object ) .mets (first << object ) .mets (set (car << object ) << new-value ) .mets (set (first << object ) << new-value ) .syne .desc The functions .code car and .code first are synonyms. If .meta object is a cons cell, these functions retrieve the .code car field of that cons cell. .code (car (cons 1 2)) yields .codn 1 . For programming convenience, .meta object may be of several other kinds in addition to conses. .code (car nil) is allowed, and returns .codn nil . .meta object may also be a vector or a string. If it is an empty vector or string, then .code nil is returned. Otherwise the first character of the string or first element of the vector is returned. A .code car form denotes a valid place when .meta object is accessible via .codn car , isn't the object .codn nil , and is modifiable. A .code car form supports deletion. The following equivalence then applies: .cblk (del (car place)) <--> (pop place) .cble This implies that deletion requires the argument of the .code car form to be a place, rather than the whole form itself. In this situation, the argument place may have a value which is .codn nil , because .code pop is defined on an empty list. The abstract concept behind deleting a .code car is that physically deleting this field from a cons, thereby breaking it in half, would result in just the .code cdr remaining. Though fragmenting a cons in this manner is impossible, deletion simulates it by replacing the place which previously held the cons, with that cons' .code cdr field. This semantics happens to coincide with deleting the first element of a list by a .code pop operation. .coNP Accessors @ cdr and @ rest .synb .mets (cdr << object ) .mets (rest << object ) .mets (set (cdr << object ) << new-value ) .mets (set (rest << object ) << new-value ) .syne .desc The functions .code cdr and .code rest are synonyms. If .meta object is a cons cell, these functions retrieve the .code cdr field of that cons cell. .code (cdr (cons 1 2)) yields .codn 2 . For programming convenience, .meta object may be of several other kinds in addition to conses. .code (cdr nil) is allowed, and returns .codn nil . .meta object may also be a vector or a string. If it is a non-empty string or vector containing at least two items, then the remaining part of the object is returned, with the first element removed. For example .code (cdr "abc") yields .codn "bc" . If .meta object is is a one-element vector or string, or an empty vector or string, then .code nil is returned. Thus .code (cdr "a") and .code (cdr "") both result in .codn nil . The invocation syntax of a .code cdr or .code rest form is a syntactic place. The place is semantically valid when .meta object is accessible via .codn cdr , isn't the object .codn nil , and is modifiable. A .code cdr place supports deletion, according to the following near equivalence: .cblk (del (cdr place)) <--> (prog1 (cdr place) (set place (car place))) .cble Of course, .code place is evaluated only once. Note that this is symmetric with the delete semantics of .code car in that the cons stored in .code place goes away, as does the .code cdr field, leaving just the .codn car , which takes the place of the original cons. .TP* Example: Walk every element of the list .code (1 2 3) using a .code for loop: .cblk (for ((i '(1 2 3))) (i) ((set i (cdr i))) (print (car i) *stdout*) (print #\enewline *stdout*)) .cble The variable .code i marches over the cons cells which make up the "backbone" of the list. The elements are retrieved using the .code car function. Advancing to the next cell is achieved using .codn (cdr i) . If .code i is the last cell in a (proper) list, .code (cdr i) yields .code nil and so .code i becomes .codn nil , the loop guard expression .code i fails and the loop terminates. .coNP Functions @ rplaca and @ rplacd .synb .mets (rplaca < cons << new-car-value ) .mets (rplacd < cons << new-cdr-value ) .syne .desc The .code rplaca and .code rplacd functions assign new values into the .code car and .code cdr fields of the cell .metn cons . Note that, except for the difference in return value, .code (rplaca x y) is the same as the more generic .codn (set (car x) y) , and likewise .code (rplacd x y) can be written as .codn (set (cdr x) y) . It is an error if .meta cons is not a cons or lazy cons. In particular, whereas .code (car nil) is correct, .code (rplaca nil ...) is erroneous. The .code rplaca and .code rplacd functions return .metn cons . Note: \*(TX versions 89 and earlier, these functions returned the new value. The behavior was undocumented. The .meta cons argument does not have to be a cons cell. Both functions support meaningful semantics for vectors and strings. If .meta cons is a string, it must be modifiable. The .code rplaca function replaces the first element of a vector or first character of a string. The vector or string must be at least one element long. The .code rplacd function replaces the suffix of a vector or string after the first element with a new suffix. The .meta new-cdr-value must be a sequence, and if the suffix of a string is being replaced, it must be a sequence of characters. The suffix here refers to the portion of the vector or string after the first element. It is permissible to use .code rplacd on an empty string or vector. In this case, .meta new-cdr-value specifies the contents of the entire string or vector, as if the operation were done on a non-empty vector or string, followed by the deletion of the first element. .coNP Accessors @, second @, third @, fourth @, fifth @, sixth @, seventh @, eighth @ ninth and @ tenth .synb .mets (first << object ) .mets (second << object ) .mets (third << object ) .mets (fourth << object ) .mets (fifth << object ) .mets (sixth << object ) .mets (seventh << object ) .mets (eighth << object ) .mets (ninth << object ) .mets (tenth << object ) .mets (set (first << object ) << new-value ) .mets (set (second << object ) << new-value ) .mets ... .mets (set (tenth << object ) << new-value ) .syne .desc Used as functions, these accessors retrieve the elements of a sequence by position. If the sequence is shorter than implied by the position, these functions return .codn nil . When used as syntactic places, these accessors denote the storage locations by position. The location must exist, otherwise an error exception results. The places support deletion. .TP* Examples: .cblk (third '(1 2)) -> nil (second "ab") -> #\eb (third '(1 2 . 3)) -> **error, improper list* (let ((x (copy "abcd"))) (inc (third x)) x) -> "abce" .cble .coNP Functions @, append @ nconc and @ append* .synb .mets (append <> [ list * << last-arg ]) .mets (nconc <> [ list * << last-arg ]) .mets (append* <> [ list * << last-arg ]) .syne .desc The .code append function creates a new list which is a catenation of the .meta list arguments. All arguments are optional; .code (append) produces the empty list. If a single argument is specified, then .code append simply returns the value of that argument. It may be any kind of object. If N arguments are specified, where N > 1, then the first N-1 arguments must be proper lists. Copies of these lists are catenated together. The last argument N, shown in the above syntax as .metn last-arg , may be any kind of object. It is installed into the .code cdr field of the last cons cell of the resulting list. Thus, if argument N is also a list, it is catenated onto the resulting list, but without being copied. Argument N may be an atom other than .codn nil ; in that case .code append produces an improper list. The .code nconc function works like .codn append , but avoids consing. It destructively manipulates (that is to say, mutates) incoming lists to catenate them, and so must be used with care. The .code append* function works like .codn append , but returns a lazy list which produces the catenation of the lists on demand. If some of the arguments are themselves lazy lists which are infinite, then .code append* can return immediately, whereas append will get caught in an infinite loop trying to produce a catenation and eventually exhaust available memory. (However, the last argument to append may be an infinite lazy list, because append does not traverse the last argument.) .TP* Examples: .cblk ;; An atom is returned. (append 3) -> 3 ;; A list is also just returned: no copying takes place. ;; The eq function can verify that the same object emerges ;; from append that went in. (let ((list '(1 2 3))) (eq (append list) list)) -> t (append '(1 2 3) '(4 5 6) 7) -> '(1 2 3 4 5 6 . 7)) ;; the (4 5 6) tail of the resulting list is the original ;; (4 5 6) object, shared with that list. (append '(1 2 3) '(4 5 6)) -> '(1 2 3 4 5 6) (append nil) -> nil ;; (1 2 3) is copied: it is not the last argument (append '(1 2 3) nil) -> (1 2 3) ;; empty lists disappear (append nil '(1 2 3) nil '(4 5 6)) -> (1 2 3 4 5 6) (append nil nil nil) -> nil ;; atoms and improper lists other than in the last position ;; are erroneous (append '(a . b) 3 '(1 2 3)) -> **error** .cble .coNP Function @ list .synb .mets (list << value *) .syne .desc The .code list function creates a new list, whose elements are the argument values. .TP* Examples: .cblk (list) -> nil (list 1) -> (1) (list 'a 'b) -> (a b) .cble .coNP Function @ list* .synb .mets >> ( list* << value *) .syne .desc The .code list* function is a generalization of cons. If called with exactly two arguments, it behaves exactly like cons: .code (list* x y) is identical to .codn (cons x y) . If three or more arguments are specified, the leading arguments specify additional atoms to be consed to the front of the list. So for instance .code (list* 1 2 3) is the same as .code (cons 1 (cons 2 3)) and produces the improper list .codn (1 2 . 3) . Generalizing in the other direction, .code list* can be called with just one argument, in which case it returns that argument, and can also be called with no arguments in which case it returns .codn nil . .TP* Examples: .cblk (list*) -> nil (list* 1) -> 1 (list* 'a 'b) -> (a . b) (list* 'a 'b 'c) -> (a b . c) .cble .TP* "Dialect Note:" Note that unlike in some other Lisp dialects, the effect of .code (list* 1 2 x) can also be obtained using .codn (list 1 2 . x) . However, .code (list* 1 2 (func 3)) cannot be rewritten as .code (list 1 2 . (func 3)) because the latter is equivalent to .codn (list 1 2 func 3) . .coNP Function @ sub-list .synb .mets (sub-list < list >> [ from <> [ to ]]) .syne .desc This function is like the .code sub function, except that it operates strictly on lists. For a description of the arguments and semantics, refer to the .code sub function. .coNP Function @ replace-list .synb .mets (replace-list < list < item-sequence >> [ from <> [ to ]]) .syne .desc The .code replace-list function is like the replace function, except that the first argument must be a list. For a description of the arguments, semantics and return value, refer to the .code replace function. .coNP Functions @ listp and @ proper-listp .synb .mets (listp << value ) .mets (proper-listp << value ) .syne .desc The .code listp and .code proper-listp functions test, respectively, whether .meta value is a list, or a proper list, and return .code t or .code nil accordingly. The .code listp test is weaker, and executes without having to traverse the object. .code (listp x) is equivalent to .codn (or (null x) (consp x)) . The empty list .code nil is a list, and a cons cell is a list. The .code proper-listp function returns .code t only for proper lists. A proper list is either .codn nil , or a cons whose .code cdr is a proper list. .code proper-listp traverses the list, and its execution will not terminate if the list is circular. .coNP Function @ length-list .synb .mets (length-list << list ) .syne .desc The .code length-list function returns the length of .metn list , which may be a proper or improper list. The length of a list is the number of conses in that list. .coNP Function @ copy-list .synb .mets (copy-list << list ) .syne .desc The .code copy-list function which returns a list similar to .metn list , but with a newly allocated cons cell structure. If .meta list is an atom, it is simply returned. Otherwise, .meta list is a cons cell, and .code copy-list returns the same object as the expression .cblk .meti (cons (car << list ) (copy-list (cdr << list ))). .cble Note that the object .cblk .meti (car << list ) .cble is not deeply copied, but only propagated by reference into the new list. .code copy-list produces a new list structure out of the same items that are in .metn list . .TP* "Dialect Note:" Common Lisp does not allow the argument to be an atom, except for the empty list .codn nil . .coNP Function @ copy-cons .synb .mets (copy-cons << cons ) .syne .desc This function creates a fresh cons cell, whose .code car and .code cdr fields are copied from .codn cons . .coNP Functions @ reverse and @ nreverse .synb .mets (reverse << list ) .mets (nreverse << list ) .syne .desc Description: The functions .code reverse and .code nreverse produce an object which contains the same items as proper list .metn list , but in reverse order. If .meta list is .codn nil , then both functions return .codn nil . The .code reverse function is non-destructive: it creates a new list. The .code nreverse function creates the structure of the reversed list out of the cons cells of the input list, thereby destructively altering it (if it contains more than one element). How .code nreverse uses the material from the original list is unspecified. It may rearrange the cons cells into a reverse order, or it may keep the structure intact, but transfer the .code car values among cons cells into reverse order. Other approaches are possible. .coNP Function @ ldiff .synb .mets (ldiff < list << sublist ) .syne .desc The values .meta list and .meta sublist are proper lists. The .code ldiff function determines whether .meta sublist is a structural suffix of .meta list (meaning that it actually is a suffix, and is not merely equal to one). This is true if .meta list and .meta sublist are the same object, or else, recursively, if .meta sublist is a suffix of .cblk .meti (cdr << list ). .cble The object .code nil is the sublist of every list, including itself. The ldiff function returns a new list consisting of the elements of the prefix of .meta list which come before the .meta sublist suffix. The elements are in the same order as in .metn list . If .meta sublist is not a suffix of .metn list , then a copy of .meta list is returned. These functions also work more generally on sequences. The .meta list and .meta sublist arguments may be strings or vectors. In this case, the suffixing matching behavior is relaxed to one of structural equivalence. See the relevant examples below. .TP* Examples: .cblk ;;; unspecified: the compiler could make ;;; '(2 3) a suffix of '(1 2 3), ;;; or they could be separate objects. (ldiff '(1 2 3) '(2 3)) -> either (1) or (1 2 3) ;; b is the (1 2) suffix of a, so the ldiff is (1) (let ((a '(1 2 3)) (b (cdr a))) (ldiff a b)) -> (1) ;; string and vector behavior (ldiff "abc" "bc") -> "a" (ldiff "abc" nil) -> "abc" (ldiff #(1 2 3) #(3)) -> #(1 2) ;; mixtures do not have above behavior (ldiff #(1 2 3) '(3)) -> #(1 2 3) (ldiff '(1 2 3) #(3)) -> #(1 2 3) (ldiff "abc" #(#\eb #\ec)) -> "abc" .cble .coNP Function @ last .synb .mets (last << list ) .syne .desc If .meta list is a nonempty proper or improper list, the .code last function returns the last cons cell in the list: that cons cell whose .code cdr field is a terminating atom. If .meta list is .codn nil , then .code nil is returned. .coNP Accessor @ nthcdr .synb .mets (nthcdr < index << list ) .mets (set (nthcdr < index << list ) << new-value ) .syne .desc The .code nthcdr function retrieves the n-th cons cell of a list, indexed from zero. The .meta index parameter must be a non-negative integer. If .meta index specifies a nonexistent cons beyond the end of the list, then .code nthcdr yields nil. The following equivalences hold: .cblk (nthcdr 0 list) <--> list (nthcdr 1 list) <--> (cdr list) (nthcdr 2 list) <--> (cddr list) .cble An .code nthcdr place designates the storage location which holds the n-th cell, as indicated by the value of .metn index . Indices beyond the last cell of .meta list do not designate a valid place. If .meta list is itself a place, then the zeroth index is permitted and the resulting place denotes .metn list . Storing a value to .cblk .meti (nthcdr < 0 << list) .cble overwrites .metn list . Otherwise if .meta list isn't a syntactic place, then the zeroth index does not designate a valid place; .meta index must have a positive value. A .code nthcdr place does not support deletion. .TP* "Dialect Note:" In Common Lisp, .code nthcdr is only a function, not an accessor; .code nthcdr forms do not denote places. .coNP Accessors @, caar @, cadr @, cdar @, cddr ... @ cdddddr .synb .mets (caar << object ) .mets (cadr << object ) .mets (cdar << object ) .mets (cddr << object ) .mets ... .mets (cdddr << object ) .mets (set (caar << object ) << new-value ) .mets (set (cadr << object ) << new-value ) .mets ... .syne .desc The .I a-d accessors provide a shorthand notation for accessing two to five levels deep into a cons-cell-based tree structure. For instance, the the equivalent of the nested function call expression .cblk .meti (car (car (cdr << object ))) .cble can be achieved using the single function call .cblk .meti (caadr << object ). The symbol names of the a-d accessors are a generalization of the words "car" and "cdr". They encodes the pattern of .code car and .code cdr traversal of the structure using a sequence of the the letters .code a and .code d placed between .code c and .codn r . The traversal is encoded in right-to-left order, so that .code cadr indicates a traversal of the .code cdr link, followed by the .codn car . This order corresponds to the nested function call notation, which also encodes the traversal right-to-left. The following diagram illustrates the straightforward relationship: .cblk (cdr (car (cdr x))) ^ ^ ^ | / | | / / | / ____/ || / (cdadr x) .cble \*(TL provides all possible a-d accessors up to five levels deep, from .code caar all the way through .codn cdddddr . Expressions involving a-d accessors are places. For example, .code (caddr x) denotes the same place as .codn (car (cddr x)) , and .code (cdadr x) denotes the same place as .codn (cdr (cadr x)) . The a-d accessor places support deletion, with semantics derived from the deletion semantics of the .code car and .code cdr places. For example, .code (del (caddr x)) means the same as .code (del (car (cddr x))) . .coNP Functions @ flatten and @ flatten* .synb .mets (flatten << list ) .mets (flatten* << list ) .syne .desc The .code flatten function produces a list whose elements are all of the .cod2 non- nil atoms contained in the structure of .metn list . The .code flatten* function works like .code flatten except that .code flatten creates and returns a complete flattened list, whereas .code flatten* produces a lazy list which is instantiated on demand. This is particularly useful when the input structure is itself lazy. .TP* Examples: .cblk (flatten '(1 2 () (3 4))) -> (1 2 3 4) ;; equivalent to previous, since ;; nil is the same thing as () (flatten '(1 2 nil (3 4))) -> (1 2 3 4) (flatten nil) -> nil (flatten '(((()) ()))) -> nil .cble .coNP Functions @, memq @ memql and @ memqual .synb .mets (memq < object << list ) .mets (memql < object << list ) .mets (memqual < object << list ) .syne .desc The .codn memq , .code memql and .code memqual functions search .meta list for a member which is, respectively, .codn eq , .code eql or .coe equal to .metn object . (See the .codn eq , .code eql and .code equal functions above.) If no such element found, .code nil is returned. Otherwise, that suffix of .meta list is returned whose first element is the matching object. .coNP Functions @ member and @ member-if .synb .mets (member < key < sequence >> [ testfun <> [ keyfun ]]) .mets (member-if < predfun < sequence <> [ keyfun ]) .syne .desc The .code member and .code member-if functions search through .meta sequence for an item which matches a key, or satisfies a predicate function, respectively. The .meta keyfun argument specifies a function which is applied to the elements of the sequence to produce the comparison key. If this argument is omitted, then the untransformed elements of the sequence themselves are examined. The .code member function's .meta testfun argument specifies the test function which is used to compare the comparison keys taken from the sequence to the search key. If this argument is omitted, then the .code equal function is used. If .code member does not find a matching element, it returns .codn nil . Otherwise it returns the suffix of .meta sequence which begins with the matching element. The .code member-if function's .meta predfun argument specifies a predicate function which is applied to the successive comparison keys pulled from the sequence by applying the key function to successive elements. If no match is found, then .code nil is returned, otherwise what is returned is the suffix of .meta sequence which begins with the matching element. .coNP Functions @, remq @ remql and @ remqual .synb .mets (remq < object << list ) .mets (remql < object << list ) .mets (remqual < object << list ) .syne .desc The .codn remq , .code remql and .code remqual functions produce a new list based on .metn list , removing the items which are .codn eq , .code eql or .code equal to .metn object . The input .meta list is unmodified, but the returned list may share substructure with it. If no items are removed, it is possible that the return value is .meta list itself. .coNP Functions @, remq @ remql* and @ remqual* .synb .mets (remq* < object << list ) .mets (remql* < object << list ) .mets (remqual* < object << list ) .syne .desc The .codn remq* , .code remql* and .code remqual* functions are lazy versions of .codn remq , .code remql and .codn remqual . Rather than computing the entire new list prior to returning, these functions return a lazy list. Caution: these functions can still get into infinite looping behavior. For instance, in .codn (remql* 0 (repeat '(0))) , .code remql will keep consuming the .code 0 values coming out of the infinite list, looking for the first item that does not have to be deleted, in order to instantiate the first lazy value. .TP* Examples: .cblk ;; Return a list of all the natural numbers, excluding 13, ;; then take the first 100 of these. ;; If remql is used, it will loop until memory is exhausted, ;; because (range 1) is an infinite list. [(remql* 13 (range 1)) 0..100] .cble .coNP Functions @, countqual @ countql and @ countq .synb .mets (countq < object << list ) .mets (countql < object << list ) .mets (countqual < object << list ) .syne .desc The .codn countq , .code countql and .code countqual functions count the number of objects in .meta list which are .codn eq , .code eql or .code equal to .metn object , and return the count. .SS* Applicative List Processing .coNP Functions @, remove-if @, keep-if @ remove-if* and @ keep-if* .synb .mets (remove-if < predicate-function < list <> [ key-function ]) .mets (keep-if < predicate-function < list <> [ key-function ]) .mets (remove-if* < predicate-function < list <> [ key-function ]) .mets (keep-if* < predicate-function < list <> [ key-function ]) .syne .desc The .code remove-if function produces a list whose contents are those of .meta list but with those elements removed which satisfy .metn predicate-function . Those elements which are not removed appear in the same order. The result list may share substructure with the input list, and may even be the same list object if no items are removed. The optional .meta key-function specifies how each element from the .meta list is transformed to an argument to .metn predicate-function . If this argument is omitted then the predicate function is applied to the elements directly, a behavior which is identical to .meta key-function being .codn (fun identity) . The .code keep-if function is exactly like .codn remove-if , except the sense of the predicate is inverted. The function .code keep-if retains those items which .code remove-if will delete, and removes those that .code remove-if will preserve. The .code remove-if* and .code keep-if* functions are like .code remove-if and .codn keep-if , but produce lazy lists. .TP* Examples: .cblk ;; remove any element numerically equal to 3. (remove-if (op = 3) '(1 2 3 4 3.0 5)) -> (1 2 4 5) ;; remove those pairs whose first element begins with "abc" [remove-if (op equal [@1 0..3] "abc") '(("abcd" 4) ("defg" 5)) car] -> (("defg" 5)) ;; equivalent, without test function (remove-if (op equal [(car @1) 0..3] "abc") '(("abcd" 4) ("defg" 5))) -> (("defg" 5)) .cble .coNP Function @ count-if .synb .mets (count-if < predicate-function < list <> [ key-function ]) .syne .desc The .code count-if function counts the number of elements of .meta list which satisfy .meta predicate-function and returns the count. The optional .meta key-function specifies how each element from the .meta list is transformed to an argument to .metn predicate-function . If this argument is omitted then the predicate function is applied to the elements directly, a behavior which is identical to .meta key-function being .codn (fun identity) . .coNP Functions @, posqual @ posql and @ posq .synb .mets (posq < object << list ) .mets (posql < object << list ) .mets (posqual < object << list ) .syne .desc The .codn posq , .code posql and .code posqual functions return the zero-based position of the first item in .meta list which is, respectively, .codn eq , .code eql or .code equal to .metn object . .coNP Functions @ pos and @ pos-if .synb .mets (pos < key < list >> [ testfun <> [ keyfun ]]) .mets (pos-if < predfun < list <> [ keyfun ]) .syne .desc The .code pos and .code pos-if functions search through .meta list for an item which matches .metn key , or satisfies predicate function .metn predfun , respectively. They return the zero-based position of the matching item. The .meta keyfun argument specifies a function which is applied to the elements of .meta list to produce the comparison key. If this argument is omitted, then the untransformed elements of .meta list are examined. The .code pos function's .meta testfun argument specifies the test function which is used to compare the comparison keys from .meta list to .metn key . If this argument is omitted, then the .code equal function is used. The position of the first element .meta list whose comparison key (as retrieved by .metn keyfun ) matches the search (under .metn testfun ) is returned. If no such element is found, .code nil is returned. The .code pos-if function's .meta predfun argument specifies a predicate function which is applied to the successive comparison keys taken from .meta list by applying .meta keyfun to successive elements. The position of the first element for which .meta predfun yields true is returned. If no such element is found, .code nil is returned. .coNP Functions @ pos-max and @ pos-min .synb .mets (pos-max < sequence >> [ testfun <> [ keyfun ]]) .mets (pos-min < sequence >> [ testfun <> [ keyfun ]]) .syne .desc The .code pos-min and .code pos-max functions implement exactly the same algorithm; they differ only in their defaulting behavior with regard to the .meta testfun argument. If .meta testfun is not given, then the pos-max function defaults .meta testfun to the .code greater function, whereas .code pos-min defaults it to the .code less function. If .meta sequence is empty, both functions return .codn nil . Without a .meta testfun argument, the .code pos-max function finds the zero-based position index of the numerically maximum value occurring in .metn sequence , whereas .code pos-min without a .meta testfun argument finds the index of the minimum value. If a .meta testfun argument is given, the two functions are equivalent. The .meta testfun function must be callable with two arguments. If .meta testfun behaves like a greater-than comparison, then .code pos-max and .code pos-min return the index of the maximum element. If .meta testfun behaves like a .code less-than comparison, then the functions return the index of the minimum element. The .meta keyfun argument defaults to the .code identity function. Each element from .meta sequence is passed through this one-argument function, and the resulting value is used in its place. .coNP Function @ where .synb .mets (where < function << object ) .syne .desc If .meta object is a sequence, the .code where function returns a list of the numeric indices of those of its elements which satisfy .metn function . The numeric indices appear in increasing order. If .meta object is a hash, the .code where function returns an unordered list of keys which have values which satisfy .metn function . .meta function must be a function that can be called with one argument. For each element of .metn object , .meta function is called with that element as an argument. If a .cod2 non- nil value is returned, then the zero-based index of that element is added to a list. Finally, the list is returned. .coNP Function @ select .synb .mets (select < object >> { index-list <> | function }) .syne .desc The .code select function returns an object, of the same kind as .metn object , which consists of those elements of .meta object which are identified by the indices in .metn index-list , which may be a list or a vector. If .meta function is given instead of .metn index-list , then .meta function is invoked with .meta object as its argument. The return value is then taken as if it were the .meta index-list argument . If .meta object is a sequence, then .meta index-list consists of numeric indices. The .code select function stops processing .meta object upon encountering an index inside .meta index-list which is out of range. (Rationale: without this strict behavior, .code select would not be able to terminate if .meta index-list is infinite.) If .meta object is a list, then .meta index-list must contain monotonically increasing numeric values, even if no value is out of range, since the .code select function makes a single pass through the list based on the assumption that indices are ordered. (Rationale: optimization.) If .meta object is a hash, then .meta index-list is a list of keys. A new hash is returned which contains those elements of .meta object whose keys appear in .metn index-list . All of .meta index-list is processed, even if it contains keys which are not in .metn object . .coNP Function @ in .synb .mets (in < sequence < key >> [ testfun <> [ keyfun ]]) .mets (in < hash << key ) .syne .desc The .code in function tests whether .meta key is found inside .meta sequence or .metn hash . If the .meta testfun argument is specified, it specifies the function which is used to comparison keys from the sequence to .metn key . Otherwise the .code equal function is used. If the .meta keyfun argument is specified, it specifies a function which is applied to the elements of .meta sequence to produce the comparison keys. Without this argument, the elements themselves are taken as the comparison keys. If the object being searched is a hash, then the .meta keyfun and .meta testfun arguments are ignored. The .code in function returns .code t if it finds .meta key in .meta sequence or .metn hash, otherwise .codn nil . .coNP Function @ partition .synb .mets (partition < sequence >> { index-list >> | index <> | function }) .syne .desc If .meta sequence is empty, then .code partition returns an empty list, and the second argument is ignored; if it is .metn function , it is not called. Otherwise, .code partition returns a lazy list of partitions of .metn sequence . Partitions are consecutive, non-overlapping, non-empty sub-strings of .metn sequence , of the same kind as .metn sequence , such that if these sub-strings are catenated together in their order of appearance, a sequence .code equal to the original is produced. If the second argument is of the form .metn index-list , it shall be a sequence of strictly non-decreasing, integers. First, any leading negative or zero values in this sequence are dropped. The .code partition function then divides .meta sequence according to the indices in index list. The first partition begins with the first element of .metn sequence . The second partition begins at the first position in .metn index-list , and so on. Indices beyond the length of the sequence are ignored. If .meta index-list is empty then a one-element list containing the entire .meta sequence is returned. If the second argument is a function, then this function is applied to .metn sequence , and the return value of this call is then used in place of the second argument, which must be an .meta index or .metn index-list . If the second argument is an atom other than a function, it is assumed to be an integer index, and is turned into an .meta index-list of one element. .TP* Examples: .cblk (partition '(1 2 3) 1) -> ((1) (2 3)) ;; split the string where there is a "b" (partition "abcbcbd" (op where (op eql #\eb))) -> ("a" "bc" "bc" "bd") .cble .coNP Function @ split .synb .mets (split < sequence >> { index-list >> | index <> | function }) .syne .desc If .meta sequence is empty, then .code split returns an empty list, and the second argument is ignored; if it is .metn function , it is not called. Otherwise, .code split returns a lazy list of pieces of .metn sequence : consecutive, non-overlapping, possibly empty sub-strings of .metn sequence , of the same kind as .metn sequence . A catenation of these pieces in the order they appear would produce a sequence that is .code equal to the original sequence. If the second argument is of the form .metn index-list , it shall be a sequence of increasing integers. The .code split function divides .meta sequence according to the indices in index list. The first piece always begins with the first element of .metn sequence . Each subsequent piece begins with the position indicated by an element of .metn index-list . Negative indices are ignored. Repeated values give rise to empty pieces. If .meta index-list includes index zero, then an empty first piece is generated. If .meta index-list includes an index greater than or equal to the length of .meta sequence (equivalently, an index beyond the last element of the sequence) then an additional empty last piece is generated. If .meta index-list is empty then a one-element list containing the entire .meta sequence is returned. If the second argument is a function, then this function is applied to .metn sequence , and the return value of this call is then used in place of the second argument, which must be an .meta index or .metn index-list . If the second argument is an atom other than a function, it is assumed to be an integer index, and is turned into an .meta index-list of one element. .TP* Examples: .cblk (split '(1 2 3) 1) -> ((1) (2 3)) (split "abc" 0) -> ("" "abc") (split "abc" 3) -> ("abc" "") (split "abc" 1) -> ("a" "bc") (split "abc" 0 1 2 3) -> ("" "a" "b" "c" "") (split "abc" 1 2) -> ("a" "b" "c") (split "abc" -1 1 2 15) -> ("a" "b" "c") ;; triple split at makes two additional empty pieces (split "abc" '(1 1 1)) -> ("a" "" "" "bc") .cble .coNP Function @ partition* .synb .mets (partition* < sequence >> { index-list >> | index <> | function }) .syne .desc If .meta sequence is empty, then .code partition* returns an empty list, and the second argument is ignored; if it is .metn function , it is not called. If the second argument is of the form .metn index-list , which is a sequence of strictly increasing non-negative integers, then .code partition* produces a lazy list of pieces taken from .metn sequence . The pieces are formed by deleting from .meta sequence the elements at the positions given in .metn index-list . The pieces are the non-empty sub-strings between the deleted elements. If .meta index-list is empty then a one-element list containing the entire .meta sequence is returned. If the second argument is a function, then this function is applied to .metn sequence , and the return value of this call is then used in place of the second argument, which must be an .meta index or .metn index-list . If the second argument is an atom other than a function, it is assumed to be an integer index, and is turned into an .meta index-list of one element. .TP* Examples: .cblk (partition* '(1 2 3 4 5) '(0 2 4)) -> ((1) (3) (5)) (partition* "abcd" '(0 3)) -> "bc" (partition* "abcd" '(0 1 2 3)) -> nil .cble .coNP Function @ tree-find .synb .mets (tree-find < obj < tree << test-function ) .syne .desc The .code tree-find function searches .meta tree for an occurrence of .metn obj . Tree can be any atom, or a cons. If .meta tree it is a cons, it is understood to be a proper list whose elements are also trees. The equivalence test is performed by .meta test-function which must take two arguments, and has conventions similar to .codn eq , .code eql or .codn equal . .code tree-find works as follows. If .meta tree is equivalent to .meta obj under .metn test-function , then .code t is returned to announce a successful finding. If this test fails, and .meta tree is an atom, .code nil is returned immediately to indicate that the find failed. Otherwise, .meta tree is taken to be a proper list, and tree-find is recursively applied to each element of the list in turn, using the same .meta obj and .meta test-function arguments, stopping at the first element which returns a .cod2 non- nil value. .coNP Functions @ find and @ find-if .synb .mets (find < key < sequence >> [ testfun <> [ keyfun ]]) .mets (find-if < predfun < sequence <> [ keyfun ]) .syne .desc The .code find and .code find-if functions search through a sequence for an item which matches a key, or satisfies a predicate function, respectively. The .meta keyfun argument specifies a function which is applied to the elements of .meta sequence to produce the comparison key. If this argument is omitted, then the untransformed elements of the .meta sequence are searched. The .code find function's .meta testfun argument specifies the test function which is used to compare the comparison keys from .meta sequence to the search key. If this argument is omitted, then the .code equal function is used. The first element from the list whose comparison key (as retrieved by .metn keyfun ) matches the search (under .metn testfun ) is returned. If no such element is found, .code nil is returned. The .code find-if function's .meta predfun argument specifies a predicate function which is applied to the successive comparison keys pulled from the list by applying .meta keyfun to successive elements. The first element for which .meta predfun yields true is returned. If no such element is found, .code nil is returned. .coNP Functions @ find-max and @ find-min .synb .mets (find-max < sequence >> [ testfun <> [ keyfun ]]) .mets (find-min < sequence >> [ testfun <> [ keyfun ]]) .syne .desc The .code find-min and .code find-max function implement exactly the same algorithm; they differ only in their defaulting behavior with regard to the .meta testfun argument. If .meta testfun is not given, then the find-max function defaults it to the .code greater function, whereas .code find-min defaults it to the .code less function. Without a .meta testfun argument, the .code find-max function finds the numerically maximum value occurring in .metn sequence , whereas .code pos-min without a .meta testfun argument finds the minimum value. If a .meta testfun argument is given, the two functions are equivalent. The .meta testfun function must be callable with two arguments. If .meta testfun behaves like a greater-than comparison, then .code find-max and .code find-min both return the maximum element. If .meta testfun behaves like a less-than comparison, then the functions return the minimum element. The .meta keyfun argument defaults to the .code identity function. Each element from .meta sequence is passed through this one-argument function, and the resulting value is used in its place for the purposes of the comparison. However, the original element is returned. .coNP Function @ set-diff .synb .mets (set-diff < seq1 < seq2 >> [ testfun <> [ keyfun ]]) .syne .desc The .code set-diff function treats the sequences .meta seq1 and .meta seq2 as if they were sets and computes the set difference: a sequence which contains those elements in .meta seq1 which do not occur in .metn seq2 . .code set-diff returns a sequence of the same kind as .metn seq1 . Element equivalence is determined by a combination of .meta testfun and .metn keyfun . Elements are compared pairwise, and each element of a pair is passed through .meta keyfun function to produce a comparison value. The comparison values are compared using .metn testfun . If .meta keyfun is omitted, then the untransformed elements themselves are compared, and if .meta testfun is omitted, then the .code equal function is used. If .meta seq1 contains duplicate elements which do not occur in .meta seq2 (and thus are preserved in the set difference) then these duplicates appear in the resulting sequence. Furthermore, the order of the items from .meta seq1 is preserved. .coNP Functions @, mapcar @ mappend @ mapcar* and @ mappend* .synb .mets (mapcar < function << sequence *) .mets (mappend < function << sequence *) .mets (mapcar* < function << sequence *) .mets (mappend* < function << sequence *) .syne .desc When given only one argument, the .code mapcar function returns .codn nil . .meta function is never called. When given two arguments, the .code mapcar function applies .meta function to each elements of .meta sequence and returns a sequence of the resulting values in the same order as the original values. The returned sequence is the same kind as .metn sequence , if possible. If the accumulated values cannot be elements of that type of sequence, then a list is returned. When additional sequences are given as arguments, this filtering behavior is generalized in the following way: .code mapcar traverses the sequences in parallel, taking a value from each sequence as an argument to the function. If there are two lists, .meta function is called with two arguments and so forth. The traversal is limited by the length of the shortest sequence. The return values of the function are collected into a new sequence which is returned. The returned sequence is of the same kind as the leftmost input sequence, unless the accumulated values cannot be elements of that type of sequence, in which case a list is returned. The .code mappend function works like .codn mapcar , with the following difference. Rather than accumulating the values returned by the function into a sequence, mappend expects the items returned by the function to be sequences which are catenated with .codn append , and the resulting sequence is returned. The returned sequence is of the same kind as the leftmost input sequence, unless the values cannot be elements of that type of sequence, in which case a list is returned. The .code mapcar* and .code mappend* functions work like .code mapcar and .codn mappend , respectively. However, they return lazy lists rather than generating the entire output list prior to returning. .TP* Caveats: Like .codn mappend , .code mappend* must "consume" empty lists. For instance, if the function being mapped puts out a sequence of .codn nil s, then the result must be the empty list .codn nil , because .code (append nil nil nil nil ...) is .codn nil . But suppose that .code mappend* is used on inputs which are infinite lazy lists, such that the function returns .code nil values indefinitely. For instance: .cblk ;; Danger: infinite loop!!! (mappend* (fun identity) (repeat '(nil))) .cble The .code mappend* function is caught in a loop trying to consume and squash an infinite stream of .codn nil s, and so doesn't return. .TP* Examples: .cblk ;; multiply every element by two (mapcar (lambda (item) (* 2 item)) '(1 2 3)) -> (4 6 8) ;; "zipper" two lists together (mapcar (lambda (le ri) (list le ri)) '(1 2 3) '(a b c)) '((1 a) (2 b) (3 c))) ;; like append, mappend allows a lone atom or a trailing atom: (mappend (fun identity) 3) -> (3) (mappend (fun identity) '((1) 2)) -> (1 . 2) ;; take just the even numbers (mappend (lambda (item) (if (evenp x) (list x))) '(1 2 3 4 5)) -> (2 4) .cble .coNP Function @ mapdo .synb .mets (mapdo < function << sequence *) .syne .desc The .code mapdo function is similar to .codn mapcar , but always returns .codn nil . It is useful when .meta function performs some kind of side effect, hence the "do" in the name, which is a mnemonic for the execution of imperative actions. When only the .meta function argument is given, .meta function is never called, and .code nil is returned. If a single .meta sequence argument is given, then .code mapdo iterates over .metn sequence , invoking .meta function on each element. If two or more .meta sequence arguments are given, then .code mapdo iterates over the sequences in parallel, extracting parallel tuples of items. These tuples are passed as arguments to .metn function, which must accept as many arguments as there are sequences. .coNP Functions @ transpose and @ zip .synb .mets (transpose << sequence ) .mets (zip << sequence *) .syne .desc The .code transpose function performs a transposition on .metn sequence . This means that the elements of .meta sequence must be sequences. These sequences are understood to be columns; transpose exchanges rows and columns, returning a sequence of the rows which make up the columns. The returned sequence is of the same kind as .metn sequence , and the rows are also the same kind of sequence as the first column of the original sequence. The number of rows returned is limited by the shortest column among the sequences. All of the input sequences (the elements of .metn sequence) must have elements which are compatible with the first sequence. This means that if the first element of .meta sequence is a string, then the remaining sequences must be strings, or else sequences of characters, or of strings. The .code zip function takes variable arguments, and is equivalent to calling .code transpose on a list of the arguments. The following equivalences hold: .synb (zip . x) <--> (transpose x) [apply zip x] <--> (transpose x) .syne .TP* Examples: .cblk ;; transpose list of lists (transpose '((a b c) (c d e))) -> ((a c) (b d) (c e)) ;; transpose vector of strings: ;; - string columns become string rows ;; - vector input becomes vector output (transpose #("abc" "def" "ghij")) -> #("adg" "beh" "cfi") ;; error: transpose wants to make a list of strings ;; but 1 is not a character (transpose #("abc" "def" '(1 2 3))) ;; error! ;; String elements are catenated: (transpose #("abc" "def" ("UV" "XY" "WZ"))) -> #("adUV" "beXY" "cfWZ") (zip '(a b c) '(c d e)) -> ((a c) (b d) (c e)) .cble .coNP Function @ interpose .synb .mets (interpose < sep << sequence ) .syne .desc The .code interpose function returns a sequence of the same type as .metn sequence , in which the elements from .meta sequence appear with the .meta sep value inserted between them. If .meta sequence is an empty sequence or a sequence of length 1, then a sequence identical to .meta sequence is returned. It may be a copy of .meta sequence or it may be .meta sequence itself. If .meta sequence is a character string, then the value .meta sep must be a character. It is permissible for .metn sequence , or for a suffix of .meta sequence to be a lazy list, in which case interpose returns a lazy list, or a list with a lazy suffix. .TP* Examples: .cblk (interpose #\e- "xyz") -> "x-y-z" (interpose t nil) -> nil (interpose t #()) -> #() (interpose #\ea "") -> "" (interpose t (range 0 0)) -> (0) (interpose t (range 0 1)) -> (0 t 1) (interpose t (range 0 2)) -> (0 t 1 t 2) .cble .coNP Functions @ conses and @ conses* .synb .mets (conses << list ) .mets (conses* << list ) .syne .desc These functions return a list whose elements are the conses which make up .metn list . The .code conses* function does this in a lazy way, avoiding the computation of the entire list: it returns a lazy list of the conses of .metn list . The .code conses function computes the entire list before returning. The input .meta list may be proper or improper. The first cons of .meta list is that .meta list itself. The second cons is the rest of the list, or .cblk .meti (cdr << list ). .cble The third cons is .cblk .meti (cdr (cdr << list )) .cble and so on. .TP* Example: .cblk (conses '(1 2 3)) -> ((1 2 3) (2 3) (3)) .cble .TP* "Dialect Note:" These functions are useful for simulating the .code maplist function found in other dialects like Common Lisp. \*(TL's .code (conses x) can be expressed in Common Lisp as .codn (maplist #'identity x) . Conversely, the Common Lisp operation .code (maplist function list) can be computed in \*(TL as .codn (mapcar function (conses list)) . More generally, the Common Lisp operation .cblk (maplist function list0 list1 ... listn) .cble can be expressed as: .cblk (mapcar function (conses list0) (conses list1) ... (conses listn)) .cble .coNP Functions @ apply and @ iapply .synb .mets (apply < function <> [ arg * << trailing-args ]) .mets (iapply < function <> [ arg * << trailing-args ]) .syne .desc The .code apply function invokes .metn function , optionally passing to it an argument list. The return value of the .code apply call is that of .metn function . If no arguments are present after .metn function , then .meta function is invoked without arguments. If one argument is present after .metn function , then it is interpreted as .metn trailing-args . If this is a sequence (a list, vector or string), then the elements of the sequence are passed as individual arguments to .metn function . If .meta trailing-args is not a sequence, then .meta function is invoked with an improper argument list, terminated by the .meta trailing-args atom. If two or more arguments are present after .metn function , then the last of these arguments is interpreted as .metn trailing-args . The previous arguments represent leading arguments which are applied to .metn function , prior to the arguments taken from .metn trailing-args . The .code iapply function ("improper apply") is similar to .codn apply , except with regard to the treatment of .metn trailing-args . Firstly, under .codn iapply , if .meta trailing-args is an atom other than .code nil (possibly a sequence, such as a vector or string), then it is treated as an ordinary argument: .meta function is invoked with a proper argument list, whose last element is .metn trailing-args . Secondly, if .meta trailing-args is a list, but an improper list, then the terminating atom of .meta trailing-args becomes an ordinary argument. Thus, in all possible cases, .code iapply treats an extra .cod2 non- nil atom as an argument, and never calls .meta function with an improper argument list. .TP* Examples: .cblk ;; '(1 2 3) becomes arguments to list, thus (list 1 2 3). (apply (fun list) '(1 2 3)) -> (1 2 3) ;; this effectively invokes (list 1 2 3 4) (apply (fun list) 1 2 '(3 4)) -> (1 2 3) ;; this effectively invokes (list 1 2 . 3) (apply (fun list) 1 2 3)) -> (1 2 . 3) ;; "abc" is separated into characters which become arguments of list (apply (fun list) "abc") -> (#\ea #\eb #\ec) .cble .TP* "Dialect Note:" Note that some uses of this function that are necessary in other Lisp dialects are not necessary in \*(TL. The reason is that in \*(TL, improper list syntax is accepted as a compound form, and performs application: .cblk (foo a b . x) .cble Here, the variables .code a and .code b supply the first two arguments for .codn foo . In the dotted position, .code x must evaluate to a list or vector. The list or vector's elements are pulled out and treated as additional arguments for .codn foo . Of course, this syntax can only be used if .code x is a symbolic form or an atom. It cannot be a compound form, because .code (foo a b . (x)) and .code (foo a b x) are equivalent structures. .coNP Functions @ reduce-left and @ reduce-right .synb .mets (reduce-left < binary-function < list .mets \ \ \ \ \ \ \ \ \ \ \ \ >> [ init-value <> [ key-function ]]) .mets (reduce-right < binary-function < list .mets \ \ \ \ \ \ \ \ \ \ \ \ \ >> [ init-value <> [ key-function ]]) .syne .desc The .code reduce-left and .code reduce-right functions reduce lists of operands specified by .meta list and .meta init-value to a single value by the repeated application of .metn binary-function . An effective list of operands is formed by combining .meta list and .metn init-value . If .meta key-function is specified, then the items of .meta list are mapped to a new values through .metn key-function . If .meta init-value is supplied, then in the case of .codn reduce-left , the effective list of operands is formed by prepending .meta init-value to .metn list . In the case of .codn reduce-right , the effective operand list is produced by appending .meta init-value to .metn list . The production of the effective list can be expressed like this, though this is not to be understood as the actual implementation: .cblk (append (if init-value-present (list init-value)) [mapcar (or key-function identity) list])))) .cble In the .code reduce-right case, the arguments to .code append are reversed. If the effective list of operands is empty, then .meta binary-function is called with no arguments at all, and its value is returned. This is the only case in which .meta binary-function is called with no arguments; in all remaining cases, it is called with two arguments. If the effective list contains one item, then that item is returned. Otherwise, the effective list contains two or more items, and is decimated as follows. Note that an .meta init-value specified as .code nil is not the same as a missing .metn init-value ; this means that the initial value is the object .codn nil . Omitting .meta init-value is the same as specifying a value of .code : (the colon symbol). It is possible to specify .meta key-function while omitting an .meta init-value argument. This is achieved by explicitly specifying .code : as the .meta init-value argument. Under .codn reduce-left , the leftmost pair of operands is removed from the list and passed as arguments to .metn binary-function , in the same order that they appear in the list, and the resulting value initializes an accumulator. Then, for each remaining item in the list, .meta binary-function is invoked on two arguments: the current accumulator value, and the next element from the list. After each call, the accumulator is updated with the return value of .metn binary-function . The final value of the accumulator is returned. Under .codn reduce-right , the list is processed right to left. The rightmost pair of elements in the effective list is removed, and passed as arguments to .metn binary-function , in the same order that they appear in the list. The resulting value initializes an accumulator. Then, for each remaining item in the list, .meta binary-function is invoked on two arguments: the next element from the list, in right to left order, and the current accumulator value. After each call, the accumulator is updated with the return value of .metn binary-function . The final value of the accumulator is returned. .TP* Examples: .cblk ;;; effective list is (1) so 1 is returned (reduce-left (fun +) () 1 nil) -> 1 ;;; computes (- (- (- 0 1) 2) 3) (reduce-left (fun -) '(1 2 3) 0 nil) -> -6 ;;; computes (- 1 (- 2 (- 3 0))) (reduce-right (fun -) '(1 2 3) 0 nil) -> 2 ;;; computes (* 1 2 3) (reduce-left (fun *) '((1) (2) (3)) nil (fun first)) -> 6 ;;; computes 1 because the effective list is empty ;;; and so * is called with no arguments, which yields 1. (reduce-left (fun *) nil) .cble .coNP Function @, some @ all and @ none .synb .mets (some < sequence >> [ predicate-fun <> [ key-fun ]]) .mets (all < sequence >> [ predicate-fun <> [ key-fun ]]) .mets (none < sequence >> [ predicate-fun <> [ key-fun ]]) .syne .desc The .codn some , .code all and .code none functions apply a predicate test function .meta predicate-fun over a list of elements. If the argument .meta key-fun is specified, then elements of .meta sequence are passed into .metn key-fun , and .meta predicate-fun is applied to the resulting values. If .meta key-fun is omitted, the behavior is as if .meta key-fun is the identity function. If .meta predicate-fun is omitted, the behavior is as if .meta predicate-fun is the identity function. These functions have short-circuiting semantics and return conventions similar to the and and or operators. The some function applies .meta predicate-fun to successive values produced by retrieving elements of .meta list and processing them through .metn key-fun . If the list is empty, it returns .codn nil . Otherwise it returns the first .cod2 non- nil return value returned by a call to .meta predicate-fun and stops evaluating more elements. If .meta predicate-fun returns .code nil for all elements, it returns .metn nil . The .code all function applies .meta predicate-fun to successive values produced by retrieving elements of .meta list and processing them through .metn key-fun . If the list is empty, it returns .codn t . Otherwise, if .meta predicate-fun yields .code nil for any value, the .cod all function immediately returns without invoking .meta predicate-fun on any more elements. If all the elements are processed, then the all function returns the value which .meta predicate-fun yielded for the last element. The .code none function applies .meta predicate-fun to successive values produced by retrieving elements of .meta list and processing them through .metn key-fun . If the list is empty, it returns .codn t . Otherwise, if .meta predicate-fun yields .cod2 non- nil for any value, the none function immediately returns nil. If .meta predicate-fun yields nil for all values, the none function returns .codn t . .TP* Examples: .cblk ;; some of the integers are odd [some '(2 4 6 9) oddp] -> t ;; none of the integers are even [none '(1 3 4 7) evenp] -> t .cble .coNP Function @ multi .synb .mets (multi < function << list *) .syne .desc The .code multi function distributes an arbitrary list processing function .meta multi over multiple lists given by the .meta list arguments. The .meta list arguments are first transposed into a single list of tuples. Each successive element of this transposed list consists of a tuple of the successive items from the lists. The length of the transposed list is that of the shortest .meta list argument. The transposed list is then passed to .meta function as an argument. The .meta function is expected to produce a list of tuples, which are transposed again to produce a list of lists which is then returned. Conceptually, the input lists are columns and .meta function is invoked on a list of the rows formed from these columns. The output of .meta function is a transformed list of rows which is reconstituted into a list of columns. .TP* Example: .cblk ;; Take three lists in parallel, and remove from all of them ;; them the element at all positions where the third list ;; has an element of 20. (multi (op remove-if (op eql 20) @1 third) '(1 2 3) '(a b c) '(10 20 30)) -> ((1 3) (a c) (10 30)) ;; The (2 b 20) "row" is gone from the three "columns". ;; Note that the (op remove if (op eql 20) @1 third) ;; expression can be simplified using the ap operator: ;; ;; (op remove-if (ap eql @3 20)) .cble .SS* Association Lists Association lists are ordinary lists formed according to a special convention. Firstly, any empty list is a valid association list. A non-empty association list contains only cons cells as the key elements. These cons cells are understood to represent key/value associations, hence the name "association list". .coNP Function @ assoc .synb .mets (assoc < key << alist ) .syne .desc The .code assoc function searches an association list .meta alist for a cons cell whose car field is equivalent to .meta key (with equality determined by the equal function). The first such cons is returned. If no such cons is found, .code nil is returned. .coNP Function @ assql .synb .mets (assql < key << alist ) .syne .desc The .code assql function is just like .codn assoc , except that the equality test is determined using the .code eql function rather than .codn equal . .coNP Function @ acons .synb .mets (acons < car < cdr << alist ) .syne .desc The .code acons function constructs a new alist by consing a new cons to the front of .metn alist . The following equivalence holds: .cblk (acons car cdr alist) <--> (cons (cons car cdr) alist) .cble .coNP Function @ acons-new .synb .mets (acons-new < car < cdr << alist ) .syne .desc The .code acons-new function searches .metn alist , as if using the assoc function, for an existing cell which matches the key provided by the car argument. If such a cell exists, then its cdr field is overwritten with the .meta cdr argument, and then the .meta alist is returned. If no such cell exists, then a new list is returned by adding a new cell to the input list consisting of the .meta car and .meta cdr values, as if by the .code acons function. .coNP Function @ aconsql-new .synb .mets (aconsql-new < car < cdr << alist ) .syne .desc This function is like .codn acons-new , except that the .code eql function is used for equality testing. Thus, the list is searched for an existing cell as if using the .code assql function rather than .codn assoc . .coNP Function @ alist-remove .synb .mets (alist-remove < alist << keys ) .syne .desc The .code alist-remove function takes association list .meta alist and produces a duplicate from which cells matching the specified keys have been removed. The .meta keys argument is a list of the keys not to appear in the output list. .coNP Function @ alist-nremove .synb .mets (alist-nremove < alist << keys ) .syne .desc The .code alist-nremove function is like .codn alist-remove , but potentially destructive. The input list .meta alist may be destroyed and its structural material re-used to form the output list. The application should not retain references to the input list. .coNP Function @ copy-alist .synb .mets (copy-alist << alist ) .syne .desc The .code copy-alist function duplicates .codn alist . Unlike .codn copy-list , which only duplicates list structure, .code copy-alist also duplicates each cons cell of the input alist. That is to say, each element of the output list is produced as if by the .code copy-cons function applied to the corresponding element of the input list. .SS* Property Lists .coNP Function @ prop .synb .mets (prop < plist << key ) .syne .desc A property list a flat list of even length consisting of interleaved pairs of property names (usually symbols) and their values (arbitrary objects). An example property list is (:a 1 :b "two") which contains two properties, :a having value 1, and :b having value "two". The .code prop function searches property list .meta plist for key .metn key . If the key is found, then the value next to it is returned. Otherwise .code nil is returned. It is ambiguous whether .code nil is returned due to the property not being found, or due to the property being present with a .code nil value. .SS* List Sorting .coNP Function @ merge .synb .mets (merge < seq1 < seq2 >> [ lessfun <> [ keyfun ]]) .syne .desc The .code merge function merges two sorted sequences .meta seq1 and .meta seq2 into a single sorted sequence. The semantics and defaulting behavior of the .meta lessfun and .meta keyfun arguments are the same as those of the sort function. The sequence which is returned is of the same kind as .metn seq1 . This function is destructive of any inputs that are lists. If the output is a list, it is formed out of the structure of the input lists. .coNP Function @ multi-sort .synb .mets (multi-sort < columns < less-funcs <> [ key-funcs ]) .syne .desc The .code multi-sort function regards a list of lists to be the columns of a database. The corresponding elements from each list constitute a record. These records are to be sorted, producing a new list of lists. The .meta columns argument supplies the list of lists which comprise the columns of the database. The lists should ideally be of the same length. If the lists are of different lengths, then the shortest list is taken to be the length of the database. Excess elements in the longer lists are ignored, and do not appear in the sorted output. The .meta less-funcs argument supplies a list of comparison functions which are applied to the columns. Successive functions correspond to successive columns. If .meta less-funcs is an empty list, then the sorted database will emerge in the original order. If .meta less-funcs contains exactly one function, then the rows of the database is sorted according to the first column. The remaining columns simply follow their row. If .meta less-funcs contains more than one function, then additional columns are taken into consideration if the items in the previous columns compare .codn equal . For instance if two elements from column one compare .codn equal , then the corresponding second column elements are compared using the second column comparison function. The optional .meta key-funcs argument supplies transformation functions through which column entries are converted to comparison keys, similarly to the single key function used in the sort function and others. If there are more key functions than less functions, the excess key functions are ignored. .SS* Lazy Lists and Lazy Evaluation .coNP Function @ make-lazy-cons .synb .mets (make-lazy-cons << function ) .syne .desc The function .code make-lazy-cons makes a special kind of cons cell called a lazy cons, or lcons. Lazy conses are useful for implementing lazy lists. Lazy lists are lists which are not allocated all at once. Rather, their elements materialize when they are accessed, like magic stepping stones appearing under one's feet out of thin air. A lazy cons has .code car and .code cdr fields like a regular cons, and those fields are initialized to .code nil when the lazy cons is created. A lazy cons also has an update function, the one which is provided as the .meta function argument to .codn make-lazy-cons . When either the .code car and .code cdr fields of a cons are accessed for the first time, the function is automatically invoked first. That function has the opportunity to initialize the .code car and .code cdr fields. Once the function is called, it is removed from the lazy cons: the lazy cons no longer has an update function. To continue a lazy list, the function can make another call to .code make-lazy-cons and install the resulting cons as the .code cdr of the lazy cons. .TP* Example: .cblk ;;; lazy list of integers between min and max (defun integer-range (min max) (let ((counter min)) ;; min is greater than max; just return empty list, ;; otherwise return a lazy list (if (> min max) nil (make-lazy-cons (lambda (lcons) ;; install next number into car (rplaca lcons counter) ;; now deal wit cdr field (cond ;; max reached, terminate list with nil! ((eql counter max) (rplacd lcons nil)) ;; max not reached: increment counter ;; and extend with another lazy cons (t (inc counter) (rplacd lcons (make-lazy-cons (lcons-fun lcons)))))))))) .cble .coNP Function @ lconsp .synb .mets (lconsp << value ) .syne .desc The .code lconsp function returns .code t if .meta value is a lazy cons cell. Otherwise it returns .codn nil , even if .meta value is an ordinary cons cell. .coNP Function @ lcons-fun .synb .mets (lcons-fun << lazy-cons ) .syne .desc The .code lcons-fun function retrieves the update function of a lazy cons. Once a lazy cons has been accessed, it no longer has an update function and .code lcons-fun returns .codn nil . While the update function of a lazy cons is executing, it is still accessible. This allows the update function to retrieve a reference to itself and propagate itself into another lazy cons (as in the example under .codn make-lazy-cons ). .coNP Macro @ lcons .synb .mets (lcons < car-expression << cdr-expression ) .syne .desc The .code lcons macro simplifies the construction of structures based on lazy conses. Syntactically, it resembles the .code cons function. However, the arguments are expressions rather than values. The macro generates code which, when evaluated, immediately produces a lazy cons. The expressions .meta car-expression and .meta cdr-expression are not immediately evaluated. Rather, when either the .code car or .code cdr field of the lazy cons cell is accessed, these expressions are both evaluated at that time, in the order that they appear in the .code lcons expression, and in the original lexical scope in which that expression was evaluated. The return values of these expressions are used, respectively, to initialize the corresponding fields of the lazy cons. Note: the .code lcons macro may be understood in terms of the following reference implementation, as a syntactic sugar combining the .code make-lazy-cons constructor with a lexical closure provided by a .code lambda function: .cblk (defmacro lcons (car-form cdr-form) (let ((lc (gensym))) ^(make-lazy-cons (lambda (,lc) (rplaca ,lc ,car-form) (rplacd ,lc ,cdr-form))))) .cble .TP* Example: .cblk ;; Given the following function ... (defun fib-generator (a b) (lcons a (fib-generator b (+ a b)))) ;; ... the following function call generates the Fibonacci ;; sequence as an infinite lazy list. (fib-generator 1 1) -> (1 1 2 3 5 8 13 ...) .cble .coNP Functions @ lazy-stream-cons and @ get-lines .synb .mets (lazy-stream-cons << stream ) .mets (get-lines <> [ stream ]) .syne .desc The .code lazy-stream-cons and .code get-lines functions are synonyms, except that the .meta stream argument is optional in .code get-lines and defaults to .codn *stdin* . Thus, the following description of .code lazy-stream-cons also applies to .codn get-lines . The .code lazy-stream-cons returns a lazy cons which generates a lazy list based on reading lines of text from input stream .metn stream , which form the elements of the list. The .code get-line function is called on demand to add elements to the list. The .code lazy-stream-cons function itself makes the first call to .code get-line on the stream. If this returns .codn nil , then the stream is closed and .code nil is returned. Otherwise, a lazy cons is returned whose update function will install that line into the .code car field of the lazy cons, and continue the lazy list by making another call to .codn lazy-stream-cons , installing the result into the .code cdr field. .code lazy-stream-cons inspects the real-time property of a stream as if by the .code real-time-stream-p function. This determines which of two styles of lazy list are returned. For an ordinary (non-real-time) stream, the lazy list treats the end-of-file condition accurately: an empty file turns into the empty list .codn nil , a one line file into a one-element list which contains that line and so on. This accuracy requires one line of lookahead which is not acceptable in real-time streams, and so a different type of lazy list is used, which generates an extra .code nil item after the last line. Under this type of lazy list, an empty input stream translates to the list .codn (nil) ; a one-line stream translates to .code ("line" nil) and so forth. .coNP Macro @ delay .synb .mets (delay << expression ) .syne .desc The delay operator arranges for the delayed (or "lazy") evaluation of .metn expression . This means that the expression is not evaluated immediately. Rather, the delay expression produces a promise object. The promise object can later be passed to the .code force function (described later in this document). The force function will trigger the evaluation of the expression and retrieve the value. The expression is evaluated in the original scope, no matter where the .code force takes place. The expression is evaluated at most once, by the first call to .codn force . Additional calls to .code force only retrieve a cached value. .TP* Example: .cblk @(do ;; list is popped only once: the value is computed ;; just once when force is called on a given promise ;; for the first time. (defun get-it (promise) (format t "*list* is ~s\en" *list*) (format t "item is ~s\en" (force promise)) (format t "item is ~s\en" (force promise)) (format t "*list* is ~s\en" *list*)) (defvar *list* '(1 2 3)) (get-it (delay (pop *list*)))) Output: *list* is (1 2 3) item is 1 item is 1 *list* is (2 3) .cble .coNP Accessor @ force .synb .mets (force << promise ) .mets (set (force << promise ) << new-value ) .syne .desc The .code force function accepts a promise object produced by the .code delay macro. The first time .code force is invoked, the .meta expression which was wrapped inside .meta promise by the .code delay macro is evaluated (in its original lexical environment, regardless of where in the program the .code force call takes place). The value of .meta expression is cached inside .meta promise and returned, becoming the return value of the .code force function call. If the .code force function is invoked additional times on the same promise, the cached value is retrieved. A .code force form is a syntactic place, denoting the value cache location within .metn promise . Storing a value in a .code force place causes future accesses to the .meta promise to return that value. If the promise had not yet been forced, then storing a value into it prevents that from ever happening. The delayed .meta expression will never be evaluated. If, while a promise is being forced, the evaluation of .meta expression itself causes an assignment to the promise, it is not specified whether the promise will take on the value of .meta expression or the assigned value. .coNP Macro @ mlet .synb .mets (mlet >> ({ sym | >> ( sym << init-form )}*) << body-form *) .syne .desc The .code mlet macro ("magic let" or "mutual let") implements a variable binding construct similar to .code let and .codn let* . Under .codn mlet , the scope of the bindings of the .meta sym variables extends over the .metn init-form -s, as well as the .metn body-form -s. Unlike the .code let* construct, each .meta init-form has each .meta sym in scope. That is to say, an .metn init-form can refer not only to previous variables, but also to later variables as well as to its own variable. The variables are not initialized until their values are accessed for the first time. Any .meta sym whose value is not accessed is not initialized. Furthermore, the evaluation of each .meta init-form does not take place until the time when its value is needed to initialize the associated .metn sym . This evaluation takes place once. If a given .meta sym is not accessed during the evaluation of the .code mlet construct, then its .meta init-form is never evaluated. The bound variables may be assigned. If, before initialization, a variable is updated in such a way that its prior value is not needed, it is unspecified whether initialization takes place, and thus whether its .meta init-form is evaluated. Direct circular references erroneous and are diagnosed. .TP* Examples: .cblk ;; Dependent calculations in arbitrary order (mlet ((x (+ y 3)) (z (+ x 1)) (y 4)) (+ z 4)) --> 12 ;; Error: circular reference: ;; x depends on y, y on z, but z on x again. (mlet ((x (+ y 1)) (y (+ z 1)) (z (+ x 1))) z) ;; Okay: lazy circular reference because lcons is used (mlet ((list (lcons 1 list))) list) --> (1 1 1 1 1 ...) ;; circular list .cble In the last example, the .code list variable is accessed for the first time in the body of the .code mlet form. This causes the evaluation of the .code lcons form. This form evaluates its arguments lazily, which means that it is not a problem that .code list is not yet initialized. The form produces a lazy cons, which is then used to initialize .code list. When the .code car or .code cdr fields of the lazy cons are accessed, the .code list expression in the .code lcons argument is accessed. By that time, the variable is initialized and holds the lazy cons itself, which creates the circular reference, and a circular list. .coNP Functions @ generate and @ giterate .synb .mets (generate < while-fun << gen-fun ) .mets (giterate < while-fun < gen-fun <> [ value ]) .syne .desc The .code generate function produces a lazy list which dynamically produces items according to the following logic. The arguments to .code generate are functions which do not take any arguments. The return value of generate is a lazy list. When the lazy list is accessed, for instance with the functions car and cdr, it produces items on demand. Prior to producing each item, .meta while-fun is called. If it returns a true boolean value (any value other than .codn nil ), then the .meta gen-fun function is called, and its return value is incorporated as the next item of the lazy list. But if .meta while-fun yields .codn nil , then the lazy list immediately terminates. Prior to returning the lazy list, generate invokes the .meta while-fun one time. If .code while-fun yields .codn nil , then .code generate returns the empty list .code nil instead of a lazy list. Otherwise, it instantiates a lazy list, and invokes the .code gen-func to populate it with the first item. The .code giterate function is similar to .codn generate , except that .meta while-fun and .meta gen-fun are functions of one argument rather than functions of no arguments. The optional .meta value argument defaults to .code nil and is threaded through the function calls. That is to say, the lazy list returned is .cblk .meti >> ( value >> [ gen-fun << value ] >> [ gen-fun >> [ gen-fun << value ]] ...). .cble The lazy list terminates when a value fails to satisfy .metn while-fun . That is to say, prior to generating each value, the lazy list tests the value using .metn while-fun . If that function returns .codn nil , then the item is not added, and the sequence terminates. Note: .code giterate could be written in terms of .code generate like this: .cblk (defun giterate (w g v) (generate (lambda () [w v]) (lambda () (prog1 v (set v [g v]))))) .cble .TP* Example: .cblk (giterate (op > 5) (op + 1) 0) -> (0 1 2 3 4) .cble .coNP Function @ repeat .synb .mets (repeat < list <> [ count ]) .syne .desc If .meta list is empty, then repeat returns an empty list. If .meta count is omitted, the .code repeat function produces an infinite lazy list formed by catenating together copies of .metn list . If .meta count is specified and is zero or negative, then an empty list is returned. Otherwise a list is returned consisting of .meta count repetitions of .meta list catenated together. .coNP Function @ pad .synb .mets (pad < sequence < object <> [ count ]) .syne .desc The .code pad function produces a lazy list which consists of all of the elements of .meta sequence followed by repetitions of .metn object . If .meta count is omitted, then the repetition of .meta object is infinite. Otherwise the specified number of repetitions occur. Note that .meta sequence may be a lazy list which is infinite. In that case, the repetitions of .meta object will never occur. .coNP Function @ weave .synb .mets (weave <> { sequence }*) .syne .desc The .code weave function interleaves elements from the sequences given as arguments. If called with no arguments, it returns the empty list. If called with a single sequence, it returns the elements of that sequence as a new lazy list. When called with two or more sequences, .code weave returns a lazy list which draws elements from the sequences in a round-robin fashion, repeatedly scanning the sequences from left to right, and taking an item from each one, removing it from the sequence. Whenever a sequence runs out of items, it is deleted; the weaving then continues with the remaining sequences. The weaved sequence terminates when all sequences are eliminated. (If at least one of the sequences is an infinite lazy list, then the weaved sequence is infinite.) .TP* Examples: .cblk ;; Weave negative integers with positive ones: (weave (range 1) (range -1 : -1)) -> (1 -1 2 -2 3 -3 ...) (weave "abcd" (range 1 3) '(x x x x x x x)) --> (#\ea 1 x #\eb 2 x #\ec 3 x #\ed x x x x) .cble .coNP Macros @ gen and @ gun .synb .mets (gen < while-expression << produce-item-expression ) .mets (gun << produce-item-expression ) .syne .desc The .code gen macro operator produces a lazy list, in a manner similar to the .code generate function. Whereas the .code generate function takes functional arguments, the .code gen operator takes two expressions, which is often more convenient. The return value of .code gen is a lazy list. When the lazy list is accessed, for instance with the functions .code car and .codn cdr , it produces items on demand. Prior to producing each item, the .meta while-expression is evaluated, in its original lexical scope. If the expression yields a .cod2 non- nil value, then .meta produce-item-expression is evaluated, and its return value is incorporated as the next item of the lazy list. If the expression yields .codn nil , then the lazy list immediately terminates. The .code gen operator itself immediately evaluates .meta while-expression before producing the lazy list. If the expression yields .codn nil , then the operator returns the empty list .codn nil . Otherwise, it instantiates the lazy list and invokes the .meta produce-item-expression to force the first item. The .code gun macro similarly creates a lazy list according to the following rules. Each successive item of the lazy list is obtained as a result of evaluating .metn produce-item-expression . However, when .meta produce-item-expression yields .codn nil , then the list terminates (without adding that .code nil as an item). Note 1: the form .code gun can be implemented as a macro-expanding to an instance of the .code gen operator, like this: .cblk (defmacro gun (expr) (let ((var (gensym))) ^(let (,var) (gen (set ,var ,expr) ,var)))) .cble This exploits the fact that the .code set operator returns the value that is assigned, so the set expression is tested as a condition by .codn gen , while having the side effect of storing the next item temporarily in a hidden variable. In turn, .code gen can be implemented as a macro expanding to some .code lambda functions which are passed to the .code generate function: .cblk (defmacro gen (while-expr produce-expr) ^(generate (lambda () ,while-expr) (lambda () ,produce-expr))) .cble Note 2: .code gen can be considered as an acronym for Generate, testing Expression before Next item, whereas .code gun stands for Generate Until Null. .TP* Example: .cblk @(do ;; Make a lazy list of integers up to 1000 ;; access and print the first three. (let* ((counter 0) (list (gen (< counter 1000) (inc counter)))) (format t "~s ~s ~s\en" (pop list) (pop list) (pop list)))) Output: 1 2 3 .cble .coNP Functions @ range and @ range* .synb .mets (range >> [ from >> [ to <> [ step ]]]) .mets (range* >> [ from >> [ to <> [ step ]]]) .syne .desc The .code range and .code range* functions generate a lazy sequence of integers, with a fixed step between successive values. The difference between .code range and .code range* is that .code range* excludes the endpoint. For instance .code (range 0 3) generates the list .codn (0 1 2 3) , whereas .code (range* 0 3) generates .codn (0 1 2) . All arguments are optional. If the .meta step argument is omitted, then it defaults to .codn 1 : each value in the sequence is greater than the previous one by .codn 1 . Positive or negative step sizes are allowed. There is no check for a step size of zero, or for a step direction which cannot meet the endpoint. The .meta to argument specifies the endpoint value, which, if it occurs in the sequence, is excluded from it by the .code range* function, but included by the range function. If .meta to is missing, or specified as .codn nil , then there is no endpoint, and the sequence which is generated is infinite, regardless of .metn step . If .meta from is omitted, then the sequence begins at zero, otherwise .meta from must be an integer which specifies the initial value. The sequence stops if it reaches the endpoint value (which is included in the case of .codn range , and excluded in the case of .codn range *). However, a sequence with a stepsize greater than .code 1 or less than .code -1 might step over the endpoint value, and therefore never attain it. In this situation, the sequence also stops, and the excess value which surpasses the endpoint is excluded from the sequence. .coNP Function @ perm .synb .mets (perm < seq <> [ len ]) .syne .desc The .code rperm function returns a lazy list which consists of all length .meta len permutations of formed by items taken from .metn seq . The permutations do not use any element of .meta seq more than once. Argument .metn len , if present, must be a positive integer, and .meta seq must be a sequence. If .meta len is not present, then its value defaults to the length of .metn seq : the list of the full permutations of the entire sequence is returned. The permutations in the returned list are sequences of the same kind as .codn seq . If .meta len is zero, then a list containing one permutation is returned, and that permutations is of zero length. If .meta len exceeds the length of .metn seq , then an empty list is returned, since it is impossible to make a single non-repeating permutation that requires more items than are available. The permutations are lexicographically ordered. .coNP Function @ rperm .synb .mets (rperm < seq << len ) .syne .desc The .code rperm function returns a lazy list which consists of all the repeating permutations of length .meta len formed by items taken from .metn seq . "Repeating" means that the items from .meta seq can appear more than once in the permutations. The permutations which are returned are sequences of the same kind as .metn seq . Argument .meta len must be a nonnegative integer, and .meta seq must be a sequence. If .meta len is zero, then a single permutation is returned, of zero length. This is true regardless of whether .meta seq is itself empty. If .meta seq is empty and .meta len is greater than zero, then no permutations are returned, since permutations of a positive length require items, and the sequence has no items. Thus there exist no such permutations. The first permutation consists of .meta le repetitions of the first element of .metn seq . The next repetition, if there is one, differs from the first repetition in that its last element is the second element of .metn seq . That is to say, the permutations are lexicographically ordered. .TP* Examples: .cblk (rperm "01" 4) -> ("000" "001" "010" "011" "100" "101" "110" "111") (rperm #(1) 3) -> (#(1 1 1)) (rperm '(0 1 2) 2) -> ((0 0) (0 1) (0 2) (1 0) (1 1) (1 2) (2 0) (2 1) (2 2)) .cble .coNP Function @ comb .synb .mets (comb < seq << len ) .syne .desc The .code comb function returns a lazy list which consists of all length .meta len non-repeating combinations formed by taking items taken from .metn seq . "Non-repeating combinations" means that the combinations do not use any element of .meta seq more than once. If .meta seq contains no duplicates, then the combinations contain no duplicates. Argument .meta len must be a nonnegative integer, and .meta seq must be a sequence or a hash table. The combinations in the returned list are objects of the same kind as .metn seq . If .meta len is zero, then a list containing one combination is returned, and that permutations is of zero length. If .meta len exceeds the number of elements in .metn seq , then an empty list is returned, since it is impossible to make a single non-repeating combination that requires more items than are available. If .meta seq is a sequence, the returned combinations are lexicographically ordered. This requirement is not applicable when .meta seq is a hash table. .TP* Example: .cblk ;; powerset function, in terms of comb. ;; Yields a lazy list of all subsets of s, ;; expressed as sequences of the same type as s. (defun powerset (s) (mappend* (op comb s) (range 0 (length s)))) .cble .coNP Function @ rcomb .synb .mets (rcomb < seq << len ) .syne .desc The .code comb function returns a lazy list which consists of all length .meta len repeating combinations formed by taking items taken from .metn seq . "Repeating combinations" means that the combinations can use an element of .meta seq more than once. Argument .meta len must be a nonnegative integer, and .meta seq must be a sequence. The combinations in the returned list are sequences of the same kind as .metn seq . If .meta len is zero, then a list containing one combination is returned, and that permutations is of zero length. This is true even if .meta seq is empty. If .meta seq is empty, and .meta len is nonzero, then an empty list is returned. The combinations are lexicographically ordered. .SS* Characters and Strings .coNP Function @ mkstring .synb .mets (mkstring < length << char ) .syne .desc The .code mkstring function constructs a string object of a length specified by the .meta length parameter. Every position in the string is initialized with .metn char , which must be a character value. .coNP Function @ copy-str .synb .mets (copy-str << string ) .syne .desc The .code copy-str function constructs a new string whose contents are identical to .metn string . .coNP Function @ upcase-str .synb .mets (upcase-str << string ) .syne .desc The .code upcase-str function produces a copy of .meta string such that all lower-case characters of the English alphabet are mapped to their upper case counterparts. .coNP Function @ downcase-str .synb .mets (downcase-str << string ) .syne .desc The .code downcase-str function produces a copy of .meta string such that all upper case characters of the English alphabet are mapped to their lower case counterparts. .coNP Function @ string-extend .synb .mets (string-extend < string << tail ) .syne .desc The .code string-extend function destructively increases the length of .metn string , which must be an ordinary dynamic string. It is an error to invoke this function on a literal string or a lazy string. The .meta tail argument can be a character, string or integer. If it is a string or character, it specifies material which is to be added to the end of the string: either a single character or a sequence of characters. If it is an integer, it specifies the number of characters to be added to the string. If .meta tail is an integer, the newly added characters have indeterminate contents. The string appears to be the original one because of an internal terminating null character remains in place, but the characters beyond the terminating zero are indeterminate. .coNP Function @ stringp .synb .mets (stringp << obj ) .syne .desc The .code stringp function returns t if .meta obj is one of the several kinds of strings. Otherwise it returns .codn nil . .coNP Function @ length-str .synb .mets (length-str << string ) .syne .desc The .code length-str function returns the length .meta string in characters. The argument must be a string. .coNP Function @ search-str .synb .mets (search-str < haystack < needle >> [ start <> [ from-end ]]) .syne .desc The .code search-str function finds an occurrence of the string .meta needle inside the .meta haystack string and returns its position. If no such occurrence exists, it returns .codn nil . If a .meta start argument is not specified, it defaults to zero. If it is a non-negative integer, it specifies the starting character position for the search. Negative values of .meta start indicate positions from the end of the string, such that .code -1 is the last character of the string. If the .meta from-end argument is specified and is not .codn nil , it means that the search is conducted right-to-left. If multiple matches are possible, it will find the rightmost one rather than the leftmost one. .coNP Function @ search-str-tree .synb .mets (search-str-tree < haystack < tree >> [ start <> [ from-end ]]) .syne .desc The .code search-str-tree function is similar to .codn search-str , except that instead of searching .meta haystack for the occurrence of a single needle string, it searches for the occurrence of numerous strings at the same time. These search strings are specified, via the .meta tree argument, as an arbitrarily structured tree whose leaves are strings. The function finds the earliest possible match, in the given search direction, from among all of the needle strings. If .meta tree is a single string, the semantics is equivalent to .codn search-str . .coNP Function @ match-str .synb .mets (match-str < bigstring < littlestring <> [ start ]) .syne .desc Without the .meta start argument, the .code match-str function determines whether .meta littlestring is a prefix of .metn bigstring , returning a .code t or .code nil indication. If the .meta start argument is specified, and is a non-negative integer, then the function tests whether .meta littlestring matches a prefix of that portion of .meta bigstring which starts at the given position. If the .meta start argument is a negative integer, then .code match-str determines whether .meta littlestring is a suffix of .metn bigstring , ending on that position of bigstring, where .code -1 denotes the last character of .metn bigstring , .code -2 the second last one and so on. If .meta start is .codn -1 , then this corresponds to testing whether .meta littlestring is a suffix of .metn bigstring . .coNP Function @ match-str-tree .synb .mets (match-str-tree < bigstring < tree <> [ start ]) .syne .desc The .code match-str-tree function is a generalization of match-str which matches multiple test strings against .meta bigstring at the same time. The value reported is the longest match from among any of the strings. The strings are specified as an arbitrarily shaped tree structure which has strings at the leaves. If .meta tree is a single string atom, then the function behaves exactly like match-str. .coNP Function @ sub-str .synb .mets (sub-str < string >> [ from <> [ to ]]) .syne .desc The .code sub-str function is like the more generic function .codn sub , except that it operates only on strings. For a description of the arguments and semantics, refer to the .code sub function. .coNP Function @ replace-str .synb .mets (replace-str < string < item-sequence >> [ from <> [ to ]]) .syne .desc The .code replace-str function is like the .code replace function, except that the first argument must be a string. For a description of the arguments, semantics and return value, refer to the .code replace function. .coNP Function @ cat-str .synb .mets (cat-str < string-list <> [ sep-string ]) .syne .desc The .code cat-str function catenates a list of strings given by .meta string-list into a single string. The optional .meta sep-string argument specifies a separator string which is interposed between the catenated strings. .coNP Function @ split-str .synb .mets (split-str < string << sep ) .syne .desc The .code split-str function breaks the .meta string into pieces, returning a list thereof. The .meta sep argument must be either a string or a regular expression. It specifies the separator character sequence within .metn string . All non-overlapping matches for .meta sep within .meta string are identified in left to right order, and are removed from .metn string . The string is broken into pieces according to the gaps left behind by the removed separators, and a list of the remaining pieces is returned. If .meta sep is the empty string, then the separator pieces removed from the string are considered to be the empty strings between its characters. In this case, if .meta string is of length one or zero, then it is considered to have no such pieces, and a list of one element is returned containing the original string. If a match for .meta sep is not found in the string at all, then the string is not split at all: a list of one element is returned containing the original string. If .meta sep matches the entire string, then a list of two empty strings is returned, except in the case that the original string is empty, in which case a list of one element is returned, containing the empty string. Whenever two adjacent matches for .meta sep occur, they are considered separate cuts with an empty piece between them. This operation is nondestructive: .meta string is not modified in any way. Note: To split a string into pieces of length one such that an empty string produces .code nil rather than .codn ("") , use the .cblk .meti (tok-str < string #/./) .cble pattern. .coNP Function @ split-str-set .synb .mets (split-str-set < string << set ) .syne .desc The .code split-str-set function breaks the .meta string into pieces, returning a list thereof. The .meta set argument must be a string. It specifies a set of characters. All occurrences of any of these characters within .meta string are identified, and are removed from .metn string . The string is broken into pieces according to the gaps left behind by the removed separators. Adjacent occurrences of characters from .meta set within .meta string are considered to be separate gaps which come between empty strings. This operation is nondestructive: .meta string is not modified in any way. .coNP Functions @ tok-str and @ tok-where .synb .mets (tok-str < string < regex <> [ keep-between ]) .mets (tok-where < string << regex ) .syne .desc The .code tok-str function searches .meta string for tokens, which are defined as substrings of .mta string which match the regular expression .meta regex in the longest possible way, and do not overlap. These tokens are extracted from the string and returned as a list. Whenever .meta regex matches an empty string, then an empty token is returned, and the search for another token within .meta string resumes after advancing by one character position. So for instance, .code (tok-str "abc" #/a?/) returns the .cblk ("a" "" "" ""). .cble After the token .str "a" is extracted from a non-empty match for the regex, the regex is considered to match three more times: before the .strn "b" , between .str "b" and .strn "c" , and after the .strn "c" . If the .meta keep-between argument is specified, and is not .codn nil , then the behavior of .code tok-str changes in the following way. The pieces of .meta string which are skipped by the search for tokens are included in the output. If no token is found in .metn string , then a list of one element is returned, containing .metn string . Generally, if N tokens are found, then the returned list consists of 2N + 1 elements. The first element of the list is the (possibly empty) substring which had to be skipped to find the first token. Then the token follows. The next element is the next skipped substring and so on. The last element is the substring of .meta string between the last token and the end. The .code tok-where function works similarly to .codn tok-str , but instead of returning the extracted tokens themselves, it returns a list of the character position ranges within .meta string where matches for .meta regex occur. The ranges are pairs of numbers, represented as cons cells, where the first number of the pair gives the starting character position, and the second number is one position past the end of the match. If a match is empty, then the two numbers are equal. The tok-where function does not support the .meta keep-between parameter. .coNP Function @ list-str .synb .mets (list-str << string ) .syne .desc The .code list-str function converts a string into a list of characters. .coNP Function @ trim-str .synb .mets (trim-str << string ) .syne .desc The .code trim-str function produces a copy of .meta string from which leading and trailing whitespace is removed. Whitespace consists of spaces, tabs, carriage returns, linefeeds, vertical tabs and form feeds. .coNP Function @ chrp .synb .mets (chrp << obj ) .syne .desc Returns .code t if .meta obj is a character, otherwise nil. .coNP Function @ chr-isalnum .synb .mets (chr-isalnum << char ) .syne .desc Returns .code t if .meta char is an alpha-numeric character, otherwise nil. Alpha-numeric means one of the upper or lower case letters of the English alphabet found in ASCII, or an ASCII digit. This function is not affected by locale. .coNP Function @ chr-isalpha .synb .mets (chr-isalpha << char ) .syne .desc Returns .code t if .meta char is an alphabetic character, otherwise .codn nil . Alphabetic means one of the upper or lower case letters of the English alphabet found in ASCII. This function is not affected by locale. .coNP Function @ chr-isascii .synb .mets (chr-isalpha << char ) .syne .desc This function returns .code t if the code of character .meta char is in the range 0 to 127 inclusive. For characters outside of this range, it returns .codn nil . .coNP Function @ chr-iscntrl .synb .mets (chr-iscntrl << char ) .syne .desc This function returns .code t if the character .meta char is a character whose code ranges from 0 to 31, or is 127. In other words, any non-printable ASCII character. For other characters, it returns .codn nil . .coNP Function @ chr-isdigit .synb .mets (chr-isdigit << char ) .syne .desc This function returns .code t if the character .meta char is is an ASCII digit. Otherwise, it returns .codn nil . .coNP Function @ chr-isgraph .synb .mets (chr-isgraph << char ) .syne .desc This function returns .code t if .meta char is a non-space printable ASCII character. It returns nil if it is a space or control character. It also returns nil for non-ASCII characters: Unicode characters with a code above 127. .coNP Function @ chr-islower .synb .mets (chr-islower << char ) .syne .desc This function returns .code t if .meta char is an ASCII lower case letter. Otherwise it returns .codn nil . .coNP Function @ chr-isprint .synb .mets (chr-isprint << char ) .syne .desc This function returns .code t if .meta char is an ASCII character which is not a control character. It also returns .code nil for all non-ASCII characters: Unicode characters with a code above 127. .coNP Function @ chr-ispunct .synb .mets (chr-ispunct << char ) .syne .desc This function returns .code t if .meta char is an ASCII character which is not a control character. It also returns nil for all non-ASCII characters: Unicode characters with a code above 127. .coNP Function @ chr-isspace .synb .mets (chr-isspace << char ) .syne .desc This function returns .code t if .meta char is an ASCII whitespace character: any of the characters in the set .codn #\espace , .codn #\etab , .codn #\elinefeed , .codn #\enewline , .codn #\ereturn , .code #\evtab and .codn #\epage . For all other characters, it returns .codn nil . .coNP Function @ chr-isblank .synb .mets (chr-isblank << char ) .syne .desc This function returns .code t if .meta char is a space or tab: the character .code #\espace or .codn #\etab . For all other characters, it returns .codn nil . .coNP Function @ chr-isunisp .synb .mets (chr-isunisp << char ) .syne .desc This function returns .code t if .meta char is a Unicode whitespace character. This the case for all the characters for which .code chr-isspace returns .codn t. It also returns .code t for these additional characters: .codn #\exa0 , .codn #\ex1680 , .codn #\ex180e , .codn #\ex2000 , .codn #\ex2001 , .codn #\ex2002 , .codn #\ex2003 , .codn #\ex2004 , .codn #\ex2005 , .codn #\ex2006 , .codn #\ex2007 , .codn #\ex2008 , .codn #\ex2009 , .codn #\ex200a , .codn #\ex2028 , .codn #\ex2029 , .codn #\ex205f , and .codn #\ex3000 . For all other characters, it returns .codn nil . .coNP Function @ chr-isupper .synb .mets (chr-isupper < char ) .syne .desc This function returns .code t if .meta char is an ASCII upper case letter. Otherwise it returns .codn nil . .coNP Function @ chr-isxdigit .synb .mets (chr-isxdigit << char ) .syne .desc This function returns .code t if .meta char is a hexadecimal digit. One of the ASCII letters .code A through .codn F , or their lower-case equivalents, or an ASCII digit .code 0 through .codn 9 . .coNP Function @ chr-toupper .synb .mets (chr-toupper << char ) .syne .desc If character .meta char is a lower case ASCII letter character, this function returns the upper case equivalent character. If it is some other character, then it just returns .metn char . .coNP Function @ chr-tolower .synb .mets (chr-tolower << char ) .syne .desc If character .meta char is an upper case ASCII letter character, this function returns the lower case equivalent character. If it is some other character, then it just returns .metn char . .coNP Functions @ num-chr and @ chr-num .synb .mets (num-chr << char ) .mets (chr-num << num ) .syne .desc The argument .meta char must be a character. The .code num-chr function returns that character's Unicode code point value as an integer. The argument .meta num must be a fixnum integer in the range .code 0 to .codn #\ex10FFFF . The argument is taken to be a Unicode code point value and the corresponding character object is returned. .coNP Accessor @ chr-str .synb .mets (chr-str < str << idx ) .mets (set (chr-str < str << idx ) << new-value ) .syne .desc The .code chr-str function performs random access on string .meta str to retrieve the character whose position is given by integer .metn idx , which must be within range of the string. The index value 0 corresponds to the first (leftmost) character of the string and so non-negative values up to one less than the length are possible. Negative index values are also allowed, such that -1 corresponds to the last (rightmost) character of the string, and so negative values down to the additive inverse of the string length are possible. An empty string cannot be indexed. A string of length one supports index 0 and index -1. A string of length two is indexed left to right by the values 0 and 1, and from right to left by -1 and -2. If the element .meta idx of string .meta str exists, and the string is modifiable, then the .code chr-str form denotes a place. A .code chr-str place supports deletion. When a deletion takes place, then the character at .meta idx is removed from the string. Any characters after that position move by one position to close the gap, and the length of the string decreases by one. .TP* Notes: Direct use of .code chr-str is equivalent to the DWIM bracket notation except that .code str must be a string. The following relation holds: .cblk (chr-str s i) --> [s i] .cble since .codn [s i] <--> (ref s i) , this also holds: .cblk (chr-str s i) --> (ref s i) .cble However, note the following difference. When the expression .code [s i] is used as a place, then the subexpression .code s must be a place. When .code (chr-str s i) is used as a place, .code s need not be a place. .coNP Function @ chr-str-set .synb .mets (chr-str-set < str < idx << char ) .syne .desc The .code chr-str function performs random access on string .meta str to overwrite the character whose position is given by integer .metn idx , which must be within range of the string. The character at .meta idx is overwritten with character .metn char . The .meta idx argument works exactly as in .codn chr-str . The .meta str argument must be a modifiable string. .TP* Notes: Direct use of .code chr-str is equivalent to the DWIM bracket notation except that .meta str must be a string. The following relation holds: .cblk (chr-str-set s i c) --> (set [s i] c) .cble since .codn (set [s i] c) <--> (refset s i c) , this also holds: .cblk (chr-str s i) --> (refset s i c) .cble .coNP Function @ span-str .synb .mets (span-str < str << set ) .syne .desc The .code span-str function determines the longest prefix of string .meta str which consists only of the characters in string .metn set , in any combination. .coNP Function @ compl-span-str .synb .mets (compl-span-str < str << set ) .syne .desc The .code compl-span-str function determines the longest prefix of string .meta str which consists only of the characters which do not appear in .metn set , in any combination. .coNP Function @ break-str .synb .mets (break-str < str << set ) .syne .desc The .code break-str function returns an integer which represents the position of the first character in string .meta str which appears in string .metn set . If there is no such character, then .code nil is returned. .SS* Lazy Strings Lazy strings are objects that were developed for the \*(TX pattern matching language, and are exposed via \*(TL. Lazy strings behave much like strings, and can be substituted for strings. However, unlike regular strings, which exist in their entirety, first to last character, from the moment they are created, lazy strings do not exist all at once, but are created on demand. If character at index N of a lazy string is accessed, then characters 0 through N of that string are forced into existence. However, characters at indices beyond N need not necessarily exist. A lazy string dynamically grows by acquiring new text from a list of strings which is attached to that lazy string object. When the lazy string is accessed beyond the end of its hitherto materialized prefix, it takes enough strings from the list in order to materialize the index. If the list doesn't have enough material, then the access fails, just like an access beyond the end of a regular string. A lazy string always takes whole strings from the attached list. Lazy string growth is achieved via the .code lazy-str-force-upto function which forces a string to exist up to a given character position. This function is used internally to handle various situations. The .code lazy-str-force function forces the entire string to materialize. If the string is connected to an infinite lazy list, this will exhaust all memory. Lazy strings are specially recognized in many of the regular string functions, which do the right thing with lazy strings. For instance when .code sub-str is invoked on a lazy string, a special version of the .code sub-str logic is used which handles various lazy string cases, and can potentially return another lazy string. Taking a .code sub-str of a lazy string from a given character position to the end does not force the entire lazy string to exist, and in fact the operation will work on a lazy string that is infinite. Furthermore, special lazy string functions are provided which allow programs to be written carefully to take better advantage of lazy strings. What carefully means is code that avoids unnecessarily forcing the lazy string. For instance, in many situations it is necessary to obtain the length of a string, only to test it for equality or inequality with some number. But it is not necessary to compute the length of a string in order to know that it is greater than some value. .coNP Function @ lazy-str .synb .mets (lazy-str < string-list >> [ terminator <> [ limit-count ]]) .syne .desc The .code lazy-str function constructs a lazy string which draws material from .meta string-list which is a list of strings. If the optional .meta terminator argument is given, then it specifies a string which is appended to every string from .metn string-list , before that string is incorporated into the lazy string. If .meta terminator is not given, then it defaults to the string .strn "\en" , and so the strings from .meta string-list are effectively treated as lines which get terminated by newlines as they accumulate into the growing prefix of the lazy string. To avoid the use of a terminator string, a null string .meta terminator argument must be explicitly passed. In that case, the lazy string grows simply by catenating elements from .metn string-list . If the .meta limit-count argument is specified, it must be a positive integer. It expresses a maximum limit on how many elements will be consumed from .meta string-list in order to feed the lazy string. Once that many elements are drawn, the string ends, even if the list has not been exhausted. .coNP Function @ lazy-stringp .synb .mets (lazy-stringp << obj ) .syne .desc The .code lazy-stringp function returns .code t if .meta obj is a lazy string. Otherwise it returns .codn nil . .coNP Function @ lazy-str-force-upto .synb .mets (lazy-str-force-upto < lazy-str << index ) .syne .desc The .code lazy-str-force-upto function tries to instantiate the lazy string such that the position given by .meta index materializes. The .meta index is a character position, exactly as used in the .code chr-str function. Some positions beyond .meta index may also materialize, as a side effect. If the string is already materialized through to at least .metn index , or if it is possible to materialize the string that far, then the value .code t is returned to indicate success. If there is insufficient material to force the lazy string through to the .meta index position, then nil is returned. It is an error if the .meta lazy-str argument isn't a lazy string. .coNP Function @ lazy-str-force .synb .mets (lazy-str-force << lazy-str ) .syne .desc The .meta lazy-str argument must be a lazy string. The lazy string is forced to fully materialize. The return value is an ordinary, non-lazy string equivalent to the fully materialized lazy string. .coNP Function @ lazy-str-get-trailing-list .synb .mets (lazy-str-get-trailing-list < string << index ) .syne .desc The .code lazy-str-get-trailing-list function can be considered, in some way, an inverse operation to the production of the lazy string from its associated list. First, .meta string is forced up through the position .metn index . That is the only extent to which .meta string is modified by this function. Next, the suffix of the materialized part of the lazy string starting at position .metn index , is split into pieces on occurrences of the terminator character (which had been given as the .meta terminator argument in the .code lazy-str constructor, and defaults to newline). If the .meta index position is beyond the part of the string which can be materialized (in adherence with the lazy string's .meta limit-count constructor parameter), then the list of pieces is considered to be empty. Finally, a list is returned consisting of the pieces produced by the split, to which is appended the remaining list of the string which has not yet been forced to materialize. .coNP Functions @, length-str-> @, length-str->= @ length-str-< and @ length-str-<= .synb .mets (length-str-> < string << len ) .mets (length-str->= < string << len ) .mets (length-str-< < string << len ) .mets (length-str-<= < string << len ) .syne .desc These functions compare the lengths of two strings. The following equivalences hold, as far as the resulting value is concerned: .cblk (length-str-> s l) <--> (> (length-str s) l) (length-str->= s l) <--> (>= (length-str s) l) (length-str-< s l) <--> (< (length-str s) l) (length-str-<= s l) <--> (<= (length-str s) l) .cble The difference between the functions and the equivalent forms is that if the string is lazy, the .code length-str function will fully force it in order to calculate and return its length. These functions only force a string up to position .metn len , so they are not only more efficient, but on infinitely long lazy strings they are usable. .code length-str cannot compute the length of a lazy string with an unbounded length; it will exhaust all memory trying to force the string. These functions can be used to test such as string whether it is longer or shorter than a given length, without forcing the string beyond that length. .coNP Function @ cmp-str .synb .mets (cmp-str < left-string << right-string ) .syne .desc The .code cmp-str function returns a negative integer if .meta left-string is lexicographically prior to .metn right-string , and a positive integer if the reverse situation is the case. Otherwise the strings are equal and zero is returned. If either or both of the strings are lazy, then they are only forced to the minimum extent necessary for the function to reach a conclusion and return the appropriate value, since there is no need to look beyond the first character position in which they differ. The lexicographic ordering is naive, based on the character code point values in Unicode taken as integers, without regard for locale-specific collation orders. .coNP Functions @, str= @, str< @, str> @ str>= and @ str<= .synb .mets (str= < left-string << right-string ) .mets (str< < left-string << right-string ) .mets (str> < left-string << right-string ) .mets (str<= < left-string << right-string ) .mets (str>= < left-string << right-string ) .syne .desc These functions compare .meta left-string and .meta right-string lexicographically, as if by the .code cmp-str function. The .code str= function returns .code t if the two strings are exactly the same, character for character, otherwise it returns .codn nil . The .code str< function returns .code t if .meta left-string is lexicographically before .metn right-string , otherwise nil. The .code str> function returns .code t if .meta left-string is lexicographically after .metn right-string , otherwise .codn nil . The .code str< function returns .code t if .meta left-string is lexicographically before .metn right-string , or if they are exactly the same, otherwise .codn nil . The .code str< function returns .code t if .meta left-string is lexicographically after .metn right-string , or if they are exactly the same, otherwise .codn nil . .coNP Function @ string-lt .synb .mets (string-lt < left-str << right-str ) .syne .desc The .code string-lt is a deprecated alias for .codn str< . .SS* Vectors .coNP Function @ vector .synb .mets (vector < length <> [ initval ]) .syne .desc The .code vector function creates and returns a vector object of the specified length. The elements of the vector are initialized to .metn initval , or to nil if .meta initval is omitted. .coNP Function @ vec .synb .mets (vec << arg *) .syne .desc The .code vec function creates a vector out of its arguments. .coNP Function @ vectorp .synb .mets (vectorp << obj ) .syne .desc The .code vectorp function returns t if .meta obj is a vector, otherwise it returns .codn nil . .coNP Function @ vec-set-length .synb .mets (vec-set-length < vec << len ) .syne .desc The .code vec-set-length modifies the length of .metn vec , making it longer or shorter. If the vector is made longer, then the newly added elements are initialized to nil. The .meta len argument must be nonnegative. The return value is .metn vec . .coNP Accessor @ vecref .synb .mets (vecref < vec << idx ) .mets (set (vecref < vec << idx ) << new-value ) .syne .desc The .code vecref function performs indexing into a vector. It retrieves an element of .meta vec at position .metn idx , counted from zero. The .meta idx value must range from 0 to one less than the length of the vector. The specified element is returned. If the element .meta idx of vector .meta vec exists, then the .code vecref form denotes a place. A .code vecref place supports deletion. When a deletion takes place, then if .meta idx denotes the last element in the vector, the vector's length is decreased by one, so that the vector no longer has that element. Otherwise, if .meta idx isn't the last element, then each elements values at a higher index than .meta idx shifts by one one element position to the adjacent lower index. Then, the length of the vector is decreased by one, so that the last element position disappears. .coNP Function @ vec-push .synb .mets (vec-push < vec << elem ) .syne .desc The .code vec-push function extends the length of a vector .meta vec by one element, and sets the new element to the value .metn elem . The previous length of the vector (which is also the position of .metn elem ) is returned. This function performs similarly to the generic function .codn ref , except that the first argument must be a vector. .coNP Function @ length-vec .synb .mets (length-vec << vec ) .syne .desc The .code length-vec function returns the length of vector .metn vec . It performs similarly to the generic .code length function, except that the argument must be a vector. .coNP Function @ size-vec .synb .mets (size-vec << vec ) .syne .desc The .code size-vec function returns the number of elements for which storage is reserved in the vector .metn vec . .TP* Notes: The .code length of the vector can be extended up to this size without any memory allocation operations having to be performed. .coNP Function @ vector-list .synb .mets (vector-list << list ) .syne .desc This function returns a vector which contains all of the same elements and in the same order as list .metn list . .coNP Function @ list-vector .synb .mets (list-vector << vec ) .syne .desc The .code list-vector function returns a list of the elements of vector .metn vec . .coNP Function @ copy-vec .synb .mets (copy-vec << vec ) .syne .desc The .code copy-vec function returns a new vector object of the same length as .meta vec and containing the same elements in the same order. .coNP Function @ sub-vec .synb .mets (sub-vec < vec >> [ from <> [ to ]]) .syne .desc The .code sub-vec function is like the more generic function .codn sub , except that it operates only on vectors. For a description of the arguments and semantics, refer to the .code sub function. .coNP Function @ replace-vec .synb .mets (replace-vec < vec < item-sequence >> [ from <> [ to ]]) .syne .desc The .code replace-vec is like the .code replace function, except that the first argument must be a vector. For a description of the arguments, semantics and return value, refer to the .code replace function. .coNP Function cat-vec .synb .mets (cat-vec << vec-list ) .syne .desc The .meta vec-list argument is a list of vectors. The .code cat-vec function produces a catenation of the vectors listed in .metn vec-list . It returns a single large vector formed by catenating those vectors together in order. .SS* Sequence Manipulation .coNP Function @ seqp .synb .mets (seqp << object ) .syne .desc The function .code seqp returns .code t if .meta object is a sequence, otherwise .codn nil . A sequence is defined as a list, vector or string. The object .code nil denotes the empty list and so is a sequence. .coNP Function @ length .synb .mets (length << sequence ) .syne .desc The .code length function returns the number of items in .metn sequence , and returns it. .meta sequence may be a hash, in which case .cblk .meti (hash-count << sequence ) .cble is returned. .coNP Function @ empty .synb .mets (empty << sequence ) .syne .desc Returns .code t if .cblk .meti (length << sequence ) .cble is zero, otherwise .codn nil . .coNP Function @ copy .synb .mets (copy << object ) .syne .desc The .code copy function duplicates objects of various supported types: sequences, hashes and random states. If .meta object is .codn nil , it returns .codn nil . If .meta object is a list, it returns .cblk .meti (copy-list << object ). .cble If .meta object is a string, it returns .cblk .meti (copy-str << object ). .cble If .meta object is a vector, it returns .cblk .meti (copy-vec << object ). .cble If .meta object is a hash, it returns .cblk .meti (copy-hash << object ) .cble Lastly, if .meta object is a random state, it returns .cblk .meti (make-random-state << object ). .cble Except in the case when .meta sequence is .codn nil , .code copy returns a value that is distinct from (not .code eq to) .metn sequence . This is different from the behavior of .cblk .meti >> [ sequence 0..t] .cblk or .cblk .meti (sub < sequence 0 t) .cble which recognize that they need not make a copy of .metn sequence , and just return it. Note however, that the elements of the returned sequence may be eq to elements of the original sequence. In other words, copy is a deeper copy than just duplicating the .code sequence value itself, but it is not a deep copy. .coNP Function @ sub .synb .mets (sub < sequence >> [ from <> [ to ]]) .syne .desc The .code sub function extracts a slice from input sequence .metn sequence . The slice is a sequence of the same type as .metn sequence . If the .meta from argument is omitted, it defaults to .codn 0 . If the .meta to parameter is omitted, it defaults to .codn t . Thus .code (sub a) means .codn (sub a 0 t) . The following equivalence holds between the .code sub function and the DWIM-bracket syntax: .cblk ;; from is not a list (sub seq from to) <--> [seq from..to] .cble The description of the .code dwim operator\(emin particular, the section on Range Indexing\(emexplains the semantics of the range specification. If the sequence is a list, the output sequence may share substructure with the input sequence. .coNP Function @ replace .synb .mets (replace < sequence < replacement-sequence >> [ from <> [ to ]]) .mets (replace < sequence < replacement-sequence << index-list ) .syne .desc The .meta replace function modifies .meta sequence in the ways described below. The operation is destructive: it may work "in place" by modifying the original sequence. The caller should retain the return value and stop relying on the original input sequence. The return value of .code replace is the modified version of .metn sequence . This may be the same object as .meta sequence or it may be a newly allocated object. Note that the form: .cblk (set seq (replace seq new fr to)) .cble has the same effect on the variable .code seq as the form: .cblk (set [seq fr..to] new) .cble except that the former .code set form returns the entire modified sequence, whereas the latter returns the value of the .code new argument. The .code replace function has two invocation styles, distinguished by the type of the third argument. If the third argument is a list or vector, then it is deemed to be the .meta index-list parameter of the second form. Otherwise, if the third argument is missing, or is not a list, then it is deemed to be the .meta from argument of the first form. The first form of the replace function replaces a contiguous subsequence of the .meta sequence with .metn replacement-sequence . The replaced subsequence may be empty, in which case an insertion is performed. If .meta replacement-sequence is empty (for example, the empty list .codn nil ), then a deletion is performed. If the .meta from and .meta to arguments are omitted, their values default to .code 0 and .code t respectively. The description of the dwim operator\(emin particular, the section on Range Indexing\(emexplains the semantics of the range specification. The second form of the replace function replaces a subsequence of elements from .meta sequence given by .metn index-list , with their counterparts from .metn replacement-sequence . This form of the replace function does not insert or delete; it simply overwrites elements. If .meta replacement-sequence and .meta index-list are of different lengths, then the shorter of the two determines the maximum number of elements which are overwritten. Furthermore, similar restrictions apply on .meta index-list as under the select function. Namely, the replacement stops when an index value in .meta index-list is encountered which is out of range for .metn sequence . furthermore, if .meta sequence is a list, then .meta index-list must be monotonically increasing. .coNP Function @ search .synb .mets (search < haystack < needle >> [ testfun <> [ keyfun ]) .syne .desc The .code search function determines whether the sequence .meta needle occurs as substring within .metn haystack , under the given comparison function .meta testfun and key function .metn keyfun . If this is the case, then the zero-based position of the leftmost occurrence of .meta key within .meta haystack is returned. Otherwise .code nil is returned to indicate that .meta key does not occur within .metn haystack . If .meta key is empty, then zero is always returned. The arguments .meta haystack and .meta needle are sequences: lists, vectors or strings, in any combination. If .meta needle is not empty, then occurs at some position N within .meta haystack if the first element of .meta needle matches the element at position N of .metn haystack , the second element of .meta needle matches the element at position N+1 of .meta haystack and so forth, for all elements of .metn needle . A match between elements is determined by passing each element through .metn keyfun , and then comparing the resulting values using .metn testfun . If .meta testfun is supplied, it must be a function which can be called with two arguments. If it is not supplied, it defaults to .codn eql . If .meta keyfun is supplied, it must be a function which can be called with one argument. If it is not supplied, it defaults to .codn identity . .TP* Examples: .cblk ;; fails because 3.0 doesn't match 3 ;; under the default eql function [search #(1.0 3.0 4.0 7.0) '(3 4)] -> nil ;; occurrence found at position 1: ;; (3.0 4.0) matches (3 4) under = [search #(1.0 3.0 4.0 7.0) '(3 4) =] -> 1 ;; "even odd odd odd even" pattern ;; matches at position 2 [search #(1 1 2 3 5 7 8) '(2 1 1 1 2) : evenp] -> 2 ;; Case insensitive string search [search "abcd" "CD" : chr-toupper] -> 2 ;; Case insensitive string search ;; using vector of characters as key [search "abcd" #(#\eC #\eD) : chr-toupper] -> 2 .cble .coNP Functions @ ref and @ refset .synb .mets (ref < seq << index ) .mets (refset < seq < index << new-value ) .syne .desc The .code ref and .code refset functions perform array-like indexing into sequences. The .code ref function retrieves an element of .metn seq , whereas .code refset overwrites an element of .meta seq with a new value. The .meta index argument is based from zero, and negative values are permitted, with a special meaning as described in the Range Indexing section under the description of the .code dwim operator. The .code refset function returns the new value. The following equivalences hold between .code ref and .codn refset , and the DWIM bracket syntax: .cblk (ref seq idx) <--> [seq idx] (refset seq idx new) <--> (set [seq idx] new) .cble The difference is that .code ref and .code refset are first class functions which can be used in functional programming as higher order functions, whereas the bracket notation is syntactic sugar, and .code set is an operator, not a function. Therefore the brackets cannot replace all uses of .code ref and .codn refset . .coNP Function @ update .synb .mets (update < sequence-or-hash << function ) .syne .desc The .code update function replaces each elements in a sequence, or each value in a hash table, with the value of .meta function applied to that element or value. The sequence or hash table is returned. .coNP Function @ less .synb .mets (less < left-obj << right-obj ) .mets (less < obj << obj *) .syne .desc The .code less function, when called with two arguments, determines whether .meta left-obj compares less than .meta right-obj in a generic way which handles arguments of various types. The argument syntax of .code less is generalized. It can accept one argument, in which case it unconditionally returns .code t regardless of that argument's value. If more than two arguments are given, then .code less generalizes in a way which can be described by the following equivalence pattern, with the understanding that each argument expression is evaluated exactly once: .cblk (less a b c) <--> (and (less a b) (less b c)) (less a b c d) <--> (and (less a b) (less b c) (less c d)) .cble The .code less function is used as the default for the .meta lessfun argument of the functions .code sort and .codn merge , as well as the .meta testfun argument of the .code pos-min and .codn find-min . The .code less function is capable of comparing numbers, characters, symbols, strings, as well as lists and vectors of these. If both arguments are the same object so that .cblk .meti (eq < left-obj << right-obj ) .cble holds true, then the function returns .code nil regardless of the type of .metn left-obj , even if the function doesn't handle comparing different instances of that type. In other words, no object is less than itself, no matter what it is. If both arguments are numbers or characters, they are compared as if using the .code < function. If both arguments are strings, they are compared as if using the .code string-lt function. If both arguments are symbols, then their names are compared in their place, as if by the .code string-lt function. If both arguments are conses, then they are compared as follows: .RS .IP 1. The .code less function is recursively applied to the .code car fields of both arguments. If it yields true, then .meta left-obj is deemed to be less than .metn right-obj . .IP 2. Otherwise, if the .code car fields are unequal under the .code equal function, .code less returns .codn nil. .IP 3. If the .code car fields are .code equal then .code less is recursively applied to the .code cdr fields of the arguments, and the result of that comparison is returned. .RE .IP This logic performs a lexicographic comparison on ordinary lists such that for instance .code (1 1) is less than .code (1 1 1) but not less than .code (1 0) or .codn (1) . Note that the empty .code nil list nil compared to a cons is handled by type-based precedence, described below. If the arguments are vectors, they are compared lexicographically, similar to strings. Corresponding elements, starting with element 0, of the vectors are compared until an index position is found where the vectors differ. If this differing position is beyond the end of one of the two vectors, then the shorter vector is considered to be lesser. Otherwise, the result of .code less is the outcome of comparing those differing elements themselves with .codn less . If the two arguments are of the above types, but of mutually different types, then .code less resolves the situation based on the following precedence: numbers and characters are less than strings, which are less than symbols, which are less than conses, which are less than vectors. Note that since .code nil is a symbol, it is ranked lower than a cons. This interpretation ensures correct behavior when .code nil is regarded as an empty list, since the empty list is lexicographically prior to a nonempty list. Finally, if either of the arguments has a type other than the above discussed types, the situation is an error. .coNP Function @ greater .synb .mets (greater < left-obj << right-obj ) .mets (greater < obj << obj *) .syne .desc The .code greater function is equivalent to .code less with the arguments reversed. That is to say, the following equivalences hold: .cblk (greater a <--> (less a) <--> t (greater a b) <--> (less b a) (greater a b c ...) <--> (less ... c b a) .cble The .code greater function is used as the default for the .meta testfun argument of the .code pos-max and .code find-max functions. .coNP Functions @ lequal and @ gequal .synb .mets (lequal < obj << obj *) .mets (gequal < obj << obj *) .syne .desc The functions .code lequal and .code gequal are similar to .code less and .code greater respectively, but differ in the following respect: when called with two arguments which compare true under the .code equal function, the .code lequal and .code gequal functions return .codn t . When called with only one argument, both functions return .code t and both functions generalize to three or more arguments in the same way as do .code less and .codn greater . .coNP Function @ sort .synb .mets (sort < sequence >> [ lessfun <> [ keyfun ]]) .syne .desc The .code sort function destructively sorts .metn sequence , producing a sequence which is sorted according to the .meta lessfun and .meta keyfun arguments. The .meta keyfun argument specifies a function which is applied to elements of the sequence to obtain the key values which are then compared using the lessfun. If .meta keyfun is omitted, the identity function is used by default: the sequence elements themselves are their own sort keys. The .meta lessfun argument specifies the comparison function which determines the sorting order. It must be a binary function which can be invoked on pairs of keys as produced by the key function. It must return a .cod2 non- nil value if the left argument is considered to be lesser than the right argument. For instance, if the numeric function .code < is used on numeric keys, it produces an ascending sorted order. If the function .code > is used, then a descending sort is produced. If .meta lessfun is omitted, then it defaults to the generic .code less function. The .code sort function is stable for sequences which are lists. This means that the original order of items which are considered identical is preserved. For strings and vectors, .code sort is not stable. .coNP Function @ sort-group .synb .mets (sort-group < sequence >> [ keyfun <> [ lessfun ]]) .syne .desc The .code sort-group function sorts .meta sequence according to the .meta keyfun and .meta lessfun arguments, and then breaks the resulting sequence into groups, based on the equivalence of the elements under .metn keyfun . The following equivalence holds: .cblk (sort-group sq lf kf) <--> (partition-by kf (sort (copy sq) kf lf)) .cble Note the reversed order of .meta keyfun and .meta lessfun arguments between .code sort and .codn sort-group . .coNP Function @ uniq .synb .mets (uniq << sequence ) .syne .desc The .code uniq function returns a sequence of the same kind as .metn sequence , but with duplicates removed. Elements of .meta sequence are considered equal under the .code equal function. The first occurrence of each element is retained, and the subsequent duplicates of that element, of any, are suppressed, such that the order of the elements is otherwise preserved. The following equivalence holds between .code uniq and .codn unique : .cblk (uniq s) <--> [unique s : :equal-based] .cble That is, .code uniq is like .code unique with the default .meta keyfun argument (the .code identity function) and an .codn equal -based hash table. .coNP Function @ unique .synb .mets (uniq < sequence >> [ keyfun <> { hash-arg }* ]) .syne .desc The .code unique function is a generalization of .codn uniq . It returns a sequence of the same kind as .metn sequence , but with duplicates removed. If neither .meta keyfun nor .metn hash-arg -s are specified, then elements of sequence are considered equal under the .code eql function. The first occurrence of each element is retained, and the subsequent duplicates of that element, of any, are suppressed, such that the order of the elements is otherwise preserved. If .meta keyfun is specified, then that function is applied to each element, and the resulting values are compared for equality. In other words, the behavior is as if .meta keyfun were the .code identity function. If one or more .metn hash-arg -s are present, these specify the arguments for the construction of the internal hash table used by .codn unique . The arguments are like those of the .code hash function. In particular, the argument .code :equal-based causes .code unique to use .code equal equality. .coNP Function @ tuples .synb .mets (tuples < length < sequence <> [ fill-value ]) .syne .desc The .code tuples function produces a lazy list which represents a reorganization of the elements of .meta sequence into tuples of .metn length , where .meta length must be a positive integer. The length of the sequence might not be evenly divisible by the tuple length. In this case, if a .meta fill-value argument is specified, then the last tuple is padded with enough repetitions of .meta fill-value to make it have .meta length elements. If .meta fill-value is not specified, then the last tuple is left shorter than .metn length . The output of the function is a list, but the tuples themselves are sequences of the same kind as .metn sequence . If .meta sequence is any kind of list, they are lists, and not lazy lists. .TP* Examples: .cblk (tuples 3 #(1 2 3 4 5 6 7 8) 0) -> (#(1 2 3) #(4 5 6) #(7 8 0)) (tuples 3 "abc") -> ("abc") (tuples 3 "abcd") -> ("abc" "d") (tuples 3 "abcd" #\ez) -> ("abc" "dzz") (tuples 3 (list 1 2) #\ez) -> ((1 2 #\ez)) .cble .coNP Function @ partition-by .synb .mets (partition-by < function << sequence ) .syne .desc If .meta sequence is empty, then .code partition-by returns an empty list, and .meta function is never called. Otherwise, .code partition-by returns a lazy list of partitions of the sequence .metn sequence . Partitions are consecutive, non-empty sub-strings of .metn sequence , of the same kind as .metn sequence . The partitioning begins with the first element of .meta sequence being placed into a partition. The subsequent partitioning is done according to .metn function , which is applied to each element of .metn sequence . Whenever, for the next element, the function returns the same value as it returned for the previous element, the element is placed into the same partition. Otherwise, the next element is placed into, and begins, a new partition. The return values of the calls to .meta function are compared using the .code equal function. .TP* Examples: .cblk [partition-by identity '(1 2 3 3 4 4 4 5)] -> ((1) (2) (3 3) (4 4 4) (5)) (partition-by (op = 3) #(1 2 3 4 5 6 7)) -> (#(1 2) #(3) #(4 5 6 7)) .cble .coNP Function @ make-like .synb .mets (make-like < list << ref-sequence ) .syne .desc The .meta list argument must be a list. If .meta ref-sequence is a sequence type, then .meta list is converted to the same type of sequence and returned. Otherwise the original .meta list is returned. Note: the .code make-like function is a helper which supports the development of unoptimized versions of a generic function that accepts any type of sequence as input, and produces a sequence of the same type as output. The implementation of such a function can internally accumulate a list, and then convert the resulting list to the same type as an input value by using .codn make-like . .coNP Function @ nullify .synb .mets (nullify << sequence ) .syne .desc The .code nullify function returns .code nil if .meta sequence is an empty sequence. Otherwise it returns .meta sequence itself. Note: the .code nullify function is a helper to support unoptimized generic programming over sequences. Thanks to the generic behavior of .codn cdr , any sequence can be traversed using .code cdr functions, checking for the .code nil value as a terminator. This, however, breaks for empty sequences which are not lists, because they are not equal to .codn nil : to .code car and .code cdr they look like a one-element sequence containing .codn nil . The .code nullify function reduces all empty sequences to .codn nil , thereby correcting the behavior of code which traverses sequences using .codn cdr , and tests for termination with .codn nil . .SS* Math Library .coNP Functions @ + and @ - .synb .mets (+ << number *) .mets (- < number << number *) .mets (* << number *) .syne .desc The .codn + , .code - and .code * functions perform addition, subtraction and multiplication, respectively. Additionally, the .code - function performs additive inverse. The .code + function requires zero or more arguments. When called with no arguments, it produces 0 (the identity element for addition), otherwise it produces the sum over all of the arguments. Similarly, the .code * function requires zero or more arguments. When called with no arguments, it produces 1 (the identity element for multiplication). Otherwise it produces the product of all the arguments. The semantics of .code - changes from subtraction to additive inverse when there is only one argument. The argument is treated as a subtrahend, against an implicit minuend of zero. When there are two or more argument, the first one is the minuend, and the remaining are subtrahends. When there are three or more operands, these operations are performed as if by binary operations, in a left-associative way. That is to say, .code (+ a b c) means .codn (+ (+ a b) c) . The sum of .code a and .code b is computed first, and then this is added to .codn c . Similarly .code (- a b c) means .codn (- (- a b) c) . First, .code b is subtracted from .codn a , and then .code c is subtracted from that result. The arithmetic inverse is performed as if it were subtraction from integer 0. That is, .code (- x) means the same thing as .codn (- 0 x) . The operands of .codn + , .code - and .code * can be characters, integers (fixnum and bignum), and floats, in nearly any combination. If two operands have different types, then one of them is converted to the type of the one with the higher rank, according to this ranking: character < integer < float. For instance if one operand is integer, and the other float, the integer is converted to a float. .TP* Restrictions: Characters are not considered numbers, and participate in these operations in limited ways. Subtraction can be used to computed the displacement between the Unicode values of characters, and an integer displacement can be added to a character, or subtracted from a character. For instance .codn (- #\e9 #\e0) is 9 . The Unicode value of a character .code C can be found using .codn (- C #\ex0) : the displacement from the NUL character. The rules can be stated as a set of restrictions: .RS .IP 1 Two characters may not be added together. .IP 2 A character may not be subtracted from an integer (which also rules out the possibility of computing the additive inverse of a character). .IP 3 A character operand may not be opposite to a floating point operand in any operation. .IP 4 A character may not be an operand of multiplication. .RE .PP .coNP Functions @, / @ trunc, @ mod and @ trunc-rem .synb .mets (/ <> [ dividend ] << divisor ) .mets (trunc < dividend << divisor ) .mets (mod < dividend << divisor ) .mets (trunc-rem < dividend << divisor ) .syne .desc The arguments to these functions are numbers. Characters are not permitted. The .code / function performs floating-point division. Each operands is first converted to floating-point type, if necessary. If .meta dividend is omitted, then it is taken to be .code 1.0 and the function calculates the reciprocal. The .code trunc function performs a division of .meta dividend by .meta divisor whose result is truncated to integer toward zero. If both operands are integers, then an integer division is performed and the result is an integer. If either operand is a floating point value, a floating point division occurs, and the result is truncated toward zero to a floating-point integral value. The .code mod function performs a modulus operation. Firstly, the absolute value of .meta divisor is taken to be a modulus. Then a residue of .meta dividend with respect to .meta modulus is calculated. The residue's sign follows that of the sign of .metn divisor . That is, it is the smallest magnitude (closest to zero) residue of .meta dividend with respect to the absolute value of .metn divisor , having the same sign as .metn divisor . If the operands are integer, the result is an integer. If either operand is of type float, then the result is a float. The modulus operation is then generalized into the floating point domain. For instance the expression .code (mod 0.75 0.5) yields a residue of 0.25 because 0.5 "goes into" 0.75 only once, with a "remainder" of 0.25. The .code trunc-rem function returns a list of two values: a .meta quotient and a .metn remainder . The .meta quotient is exactly the same value as what .code trunc would return for the same inputs. The .meta remainder obeys the following identity: .cblk .mets (eql < remainder (- < dividend >> (* divisor << quotient ))) .cble .coNP Functions @ wrap and @ wrap* .synb .mets (wrap < start < end << number ) .mets (wrap* < start < end << number ) .syne .desc The .code wrap and .code wrap* functions reduce .meta number into the range specified by .meta start and .metn end . Under .code wrap the range is inclusive of the .meta end value, whereas under .code wrap* it is exclusive. The following equivalence holds .cblk (wrap a b c) <--> (wrap* a (succ b) c) .cble The expression .code (wrap* x0 x1 x) performs the following calculation: .cblk .mets (+ (mod (- x x0) (- x1 x0)) x0) .cble In other words, first .meta start is subtracted from .metn number . Then the result is reduced modulo the displacement between .code start and .codn end . Finally, .meta start is added back to that result, which is returned. .TP* Example: .cblk ;; perform ROT13 on the string "nop" [mapcar (opip (+ 13) (wrap #\ea #\ez)) "nop"] -> "abc" .cble .coNP Functions @ gcd and @ lcm .synb .mets (gcd << number *) .mets (lcm << number *) .syne .desc The .code gcd function computes the greatest common divisor: the largest positive integer which divides each .metn number . The .code lcm function computes the lowest common multiple: the smallest positive integer which is a multiple of each .metn number . Each .meta number must be an integer. Negative integers are replaced by their absolute values, so .code (lcm -3 -4) is .code 12 and .code (gcd -12 -9) yields .codn 3 . The value of .code (gcd) is .code 0 and that of .code (lcm) is 1 . The value of .code (gcd x) and .code (lcm x) is .codn (abs x) . Any arguments of .code gcd which are zero are effectively ignored so that .code (gcd 0) and .code (gcd 0 0 0) are both the same as .code (gcd) and .code (gcd 1 0 2 0 3) is the same as .codn (gcd 1 2 3) . If .code lcm has any argument which is zero, it yields zero. .coNP Function @ abs .synb .mets (abs << number ) .syne .desc The .code abs function computes the absolute value of .metn number . If .meta number is positive, it is returned. If .meta number is negative, its additive inverse is returned: a positive number of the same type with exactly the same magnitude. .coNP Functions @ floor and @ ceil .synb .mets (floor << number ) .mets (ceil << number ) .syne .desc The .code floor function returns the highest integer which does not exceed the value of .metn number . The ceiling function returns the lowest integer which does not exceed the value of .metn number . If .meta number an integer, it is simply returned. If the argument is a float, then the value returned is a float. For instance .code (floor 1.1) returns 1.0 rather than 1. .coNP Functions @, sin @, cos @, tan @, asin @, acos @ atan and @ atan2 .synb .mets (sin << radians ) .mets (cos << radians ) .mets (tan << radians ) .mets (atan << slope ) .mets (atan2 < y << x ) .mets (asin << num ) .mets (acos << num ) .syne .desc These trigonometric functions convert their argument to floating point and return a float result. The .codn sin , .code cos and .code tan functions compute the sine and cosine and tangent of the .meta radians argument which represents an angle expressed in radians. The .codn atan , .code acos and .code asin are their respective inverse functions. The .meta num argument to .code asin and .code acos must be in the range -1.0 to 1.0. The .code atan2 function converts the rectilinear coordinates .meta x and .meta y to an angle in polar coordinates in the range [0, 2\(*p). .coNP Functions @, exp @, log @ log10 and @ log2 .synb .mets (exp << arg ) .mets (log << arg ) .mets (log10 << arg ) .mets (log2 << arg ) .syne .desc The .code exp function calculates the value of the transcendental number e raised to the exponent .metn arg . The .code log function calculates the base e logarithm of .metn arg , which must be a positive value. The .code log10 function calculates the base 10 logarithm of .metn arg , which must be a positive value. The .code log2 function calculates the base 2 logarithm of .metn arg , which must be a positive value. .coNP Functions @, expt @ sqrt and @ isqrt .synb .mets (expt < base << exponent *) .mets (sqrt << arg ) .mets (isqrt << arg ) .syne .desc The .code expt function raises .meta base to zero or more exponents given by the .meta exponent arguments. .code (expt x) is equivalent to .codn (expt x 1) , and yields .code x for all .codn x . For three or more arguments, the operation is right-associative. That is to say, .code (expt x y z) is equivalent to .codn (expt x (expt y z)) , similarly to the way nested exponents work in standard algebraic notation. Exponentiation is done pairwise using a binary operation. If both operands to this binary operation are integers, then the result is an integer. If either operand is a float, then the other operand is converted to a float, and a floating point exponentiation is performed. Exponentiation that would produce a complex number is not supported. The .code sqrt function produces a floating-point square root of .metn arg , which is converted from integer to floating-point if necessary. Negative operands are not supported. The .code isqrt function computes the integer square root of .metn arg , which must be an integer. The integer square root is a value which is the greatest integer that is no greater than the real square root of .metn arg . The input value must be an integer. .coNP Function @ exptmod .synb .mets (exptmod < base < exponent << modulus ) .syne .desc The .code exptmod function performs modular exponentiation and accepts only integer arguments. Furthermore, .meta exponent must be a non-negative and .meta modulus must be positive. The return value is .meta base raised to .metn exponent , and reduced to the least positive residue modulo .metn modulus . .coNP Function @ cum-norm-dist .synb .mets (cum-norm-dist << argument ) .syne .desc The .code cum-norm-dist function calculates an approximation to the cumulative normal distribution function: the integral, of the normal distribution function, from negative infinity to the .metn argument . .coNP Functions @ n-choose-k and @ n-perm-k .synb .mets (n-choose-k < n << k ) .mets (n-perm-k < n << k ) .syne .desc The .code n-choose-k function computes the binomial coefficient nCk which expresses the number of combinations of .meta k items that can be chosen from a set of .metn n , where combinations are subsets. The .code n-perm-k function computes nPk: the number of permutations of size .meta k that can be drawn from a set of .metn n , where permutations are sequences, whose order is significant. The calculations only make sense when .meta n and .meta k are nonnegative integers, and .meta k does not exceed .metn n . The behavior is not specified if these conditions are not met. .coNP Functions @, fixnump @, bignump @, integerp @ floatp and @ numberp .synb .mets (fixnump << object ) .mets (bignump << object ) .mets (integerp << object ) .mets (floatp << object ) .mets (numberp << object ) .syne .desc These functions test the type of .metn object , returning .code t if it is an object of the implied type, .code nil otherwise. The .codn fixnump , .code bignump and .code floatp functions return .code t if the object is of the basic type .codn fixnum , .code bignum or .codn float . The function .code integerp returns true of .meta object is either a .code fixnum or a .codn bignum . The function .code numberp returns .code t if .meta object is either a .codn fixnum , .code bignum or .codn float . .coNP Function @ zerop .synb .mets (zerop << number ) .syne .desc The .code zerop function tests .meta number for equivalence to zero. The argument must be a number or character. It returns .code t for the integer value .code 0 and for the floating-point value .codn 0.0 . For other numbers, it returns .codn nil . It returns .code t for the null character .code #\enul and .code nil for all other characters. .coNP Functions @ plusp and @ minusp .synb .mets (plusp << number ) .mets (minusp << number ) .syne .desc These functions test whether a number is positive or negative, returning .code t or .codn nil , as the case may be. The argument may also be a character. All characters other than the null character .code #\enul are positive. No character is negative. .coNP Functions @ evenp and @ oddp .synb .mets (evenp << integer ) .mets (oddp << integer ) .syne .desc The .code evenp and .code oddp functions require integer arguments. .code evenp returns .code t if .meta integer is even (divisible by two), otherwise it returns .codn nil . .code oddp returns .code t if .meta integer is not divisible by two (odd), otherwise it returns .codn nil . .coNP Functions @, succ @, ssucc @, sssucc @, pred @, ppred @ and pppred .synb .mets (succ << number ) .mets (ssucc << number ) .mets (sssucc << number ) .mets (pred << number ) .mets (ppred << number ) .mets (pppred << number ) .syne .desc The .code succ function adds 1 to its argument and returns the resulting value. If the argument is an integer, then the return value is the successor of that integer, and if it is a character, then the return value is the successor of that character according to Unicode. The .code pred function subtracts 1 from its argument, and under similar considerations as above, the result represents the predecessor. The .code ssucc and .code sssucc functions add 2 and 3, respectively. Similarly, .code ppred and .code pppred subtract 2 and 3 from their argument. .coNP Functions @, > @, < @, >= @ <= and @ = .synb .mets (> < number << number *) .mets (> < number << number *) .mets (>= < number << number *) .mets (<= < number << number *) .mets (= < number << number *) .syne .desc These relational functions compare characters and numbers for numeric equality or inequality. The arguments must be one or more numbers or characters. If just one argument is given, then these functions all return .codn t . If two arguments are given then, they are compared as follows. First, if the numbers do not have the same type, then the one which has the lower ranking type is converted to the type of the other, according to this ranking: character < integer < float. For instance if a character and integer are compared, the character is converted to integer. Then a straightforward numeric comparison is applied. Three or more arguments may be given, in which case the comparison proceeds pairwise from left to right. For instance in .codn (< a b c) , the comparison .code (< a b) is performed in isolation. If the comparison is false, then .code nil is returned, otherwise the comparison .code (< b c) is performed in isolation, and if that is false, .code nil is returned, otherwise .code t is returned. Note that it is possible for .code b to undergo two different conversions. For instance in the .cblk .meti (< < float < character << integer ) .cble comparison, .meta character will first convert to a floating-point representation of its Unicode value so that it can be compared to .metn float , and if that comparison succeeds, then in the second comparison, .meta character will be converted to integer so that it can be compared to .metn integer . .coNP Function @ /= .synb .mets (/= << number *) .syne .desc The arguments to .code /= may be numbers or characters. The .code /= function returns .code t if no two of its arguments are numerically equal. That is to say, if there exist some .code a and .code b which are distinct arguments such that .code (= a b) is true, then the function returns .codn nil . Otherwise it returns .codn t . .coNP Functions @ max and @ min .synb .mets (max < first-arg << arg *) .mets (min < first-arg << args *) .syne .desc The .code max and .code min functions determine and return the highest or lowest value from among their arguments. If only .meta first-arg is given, that value is returned. These functions are type generic, since they compare arguments using the same semantics as the .code less function. If two or more arguments are given, then .code (max a b) is equivalent to .codn (if (less a b) b a) , and .code (min a b) is equivalent to .codn (if (less a b) a b) . If the operands do not have the same type, then one of them is converted to the type of the other; however, the original unconverted values are returned. For instance .code (max 4 3.0) yields the integer .codn 4 , not .codn 4.0 . If three or more arguments are given, .code max and .code min reduce the arguments in a left-associative manner. Thus .code (max a b c) means .codn (max (max a b) c) . .coNP Functions @, int-str @ flo-str and @ num-str .synb .mets (int-str < string <> [ radix ]) .mets (flo-str << string ) .mets (num-str << string ) .syne .desc These functions extract numeric values from character string .metn string . Leading whitespace in .metn string , if any, is skipped. If no digits can be successfully extracted, then .code nil is returned. Trailing material which does not contribute to the number is ignored. The .code int-str function converts a string of digits in the specified radix to an integer value. If the radix isn't specified, it defaults to 10. Otherwise it must be an integer in the range 2 to 36. For radices above 10, letters of the alphabet are used for digits: .code A represent a digit whose value is 10, .code B represents 11 and so forth until .codn Z . For values of radix above 36, the returned value is unspecified. Upper and lower case letters are recognized. Any character which is not a digit of the specified radix is regarded as the start of trailing junk at which the extraction of the digits stops. The .code flo-str function converts a floating-point decimal notation to a nearby floating point value. The material which contributes to the value is the longest match for optional leading space, followed by a mantissa which consists of an optional sign followed by a mixture of at least one digit, and at most one decimal point, optionally followed by an exponent part denoted by the letter .code E or .codn e , an optional sign and one or more optional exponent digits. The .code num-str function converts a decimal notation to either an integer as if by a radix 10 application of .codn int-str , or to a floating point value as if by .codn flo-str . The floating point interpretation is chosen if the possibly empty initial sequence of digits (following any whitespace and optional sign) is followed by a period, or by .code e or .codn E . .coNP Functions @ int-flo and @ flo-int .synb .mets (int-flo << float ) .mets (flo-int << integer ) .syne .desc These functions perform numeric conversion between integer and floating point type. The .code int-flo function returns an integer by truncating toward zero. The .code flo-int function returns an exact floating point value corresponding to .metn integer , if possible, otherwise an approximation using a nearby floating point value. .coNP Functions @ tofloat and @ toint .synb .mets (tofloat << value ) .mets (toint < value <> [ radix ]) .syne .desc These convenience functions convert .meta value to floating-point or integer, respectively. If a floating-point value is passed into tofloat, or an integer value into toint, then the value is simply returned. If .meta value is a character, then it is treated as a string of length one containing that character. If .meta value is a string, then it is converted by .code tofloat as if by the function .metn flo-str , , and by .code toint as if by the function .codn int-str . If .meta value is an integer, then it is converted by .code tofloat as if by the function .codn flo-int . If .meta value is a floating-point number, then it is converted by .code toint as if by the function .codn int-flo . .coNP Variables @, *flo-min* @, *flo-max* and @ *flo-epsilon* .desc These variables hold, respectively: the smallest positive floating-point value; the largest positive floating-point value; and the difference between 1.0 and the smallest representable value greater than 1.0. .code *flo-min* and .code *flo-max* define the floating-point range, which consists of three regions: values from .code (- *flo-max*) to .codn (- *flo-min*) ; the value 0.0, and values from .code *flo-min* to .codn *flo-max* . .coNP Variable @ *flo-dig* .desc This variable holds an integer representing the number of decimal digits in a decimal floating-point number such that this number can be converted to a \*(TX floating-point number, and back to decimal, without a change in any of the digits. This holds regardless of the value of the number, provided that it does not exceed the floating-point range. .coNP Variables @ *pi* and @ *e* .desc These variables hold an approximation of the mathematical constants \(*p and e. To four digits of precision, \(*p is 3.142 and e is 2.718. The .code *pi* and .code *e* approximations are accurate to .code *flo-dig* decimal digits. .SS* Bit Operations In \*(TL, similarly to Common Lisp, bit operations on integers are based on a concept that might be called "infinite two's-complement". Under infinite two's complement, a positive number is regarded as having a binary representation prefixed by an infinite stream of zero digits (for example .code 1 is .codn ...00001 ). A negative number in infinite two's complement is the bitwise negation of its positive counterpart, plus one: it carries an infinite prefix of 1 digits. So for instance the number .code -1 is represented by .codn ...11111111 : an infinite sequence of 1 bits. There is no specific sign bit; any operation which produces such an infinite sequence of 1 digits on the left gives rise to a negative number. For instance, consider the operation of computing the bitwise complement of the number .codn 1 . Since the number .code 1 is represented as .codn ...0000001 , its complement is .codn ...11111110 . Each one of the .cod 0 digits in the infinite sequence is replaced by .codn 1 , And this leading sequence means that the number is negative, in fact corresponding to the two's-complement representation of the value .codn -2 . Hence, the infinite digit concept corresponds to an arithmetic interpretation. In fact \*(TL's bignum integers do not use a two's complement representation internally. Numbers are represented as an array which holds a pure binary number. A separate field indicates the sign: negative, or non-negative. That negative numbers appear as two's-complement under the bit operations is merely a carefully maintained illusion (which makes bit operations on negative numbers more expensive). The .code logtrunc function, as well as a feature of the .code lognot function, allow bit manipulation code to be written which works with positive numbers only, even if complements are required. The trade off is that the application has to manage a limit on the number of bits. .coNP Functions @, logand @, logior and @ logxor .synb .mets (logand << integer *) .mets (logior << integer *) .mets (logxor < int1 << int2 ) .syne .desc These operations perform the familiar bitwise and, inclusive or, and exclusive or operations, respectively. Positive values inputs are treated as pure binary numbers. Negative inputs are treated as infinite-bit two's-complement. For example .code (logand -2 7) produces .codn 6 . This is because .code -2 is .code ...111110 in infinite-bit two's-complement. And-ing this value with .code 7 (or .codn ...000111 ) produces .codn 110 . The .code logand and .code logior functions are variadic, and may be called with zero, one, two, or more input values. If .code logand is called with no arguments, it produces the value -1 (all bits 1). If .code logior is called with no arguments it produces zero. In the one-argument case, the functions just return their argument value. .coNP Function @ logtest .synb .mets (logtest < int1 << int2 ) .syne .desc The .code logtest function returns true if .meta int1 and .meta int2 have bits in common. The following equivalence holds: .cblk (logtest a b) <--> (not (zerop (logand a b))) .cble .coNP Functions @ lognot and @ logtrunc .synb .mets (lognot < value <> [ bits ]) .mets (logtrunc < value << bits ) .syne .desc The .code lognot function performs a bitwise complement of .metn value . When the one-argument form of lognot is used, then if .meta value is nonnegative, then the result is negative, and vice versa, according to the infinite-bit two's complement representation. For instance .code (lognot -2) is .codn 1 , and .code (lognot 1) is .codn -2 . The two-argument form of .code lognot produces a truncated complement. Conceptually, a bitwise complement is first calculated, and then the resulting number is truncated to the number of bits given by .metn bits , which must be a nonnegative integer. The following equivalence holds: .cblk (lognot a b) <--> (logtrunc (lognot a) b) .cble The .code logtrunc function truncates the integer .meta value to the specified number of bits. If .meta value is negative, then the two's-complement representation is truncated. The return value of .code logtrunc is always a non-negative integer. .coNP Function @ sign-extend .synb .mets (sign-extend < value << bits ) .syne .desc The .code sign-extend function first truncates the infinite-bit two's complement representation of the integer .meta value to the specified number of bits, similarly to the .code logtrunc function. Then, this truncated value is regarded as a .meta bits wide two's complement integer. The value of this integer is calculated and returned. .TP* Examples: .cblk (sign-extend 127 8) -> 127 (sign-extend 128 8) -> -128 (sign-extend 129 8) -> -127 (sign-extend 255 8) -> -1 (sign-extend 256 8) -> 0 (sign-extend -1 8) -> -1 (sign-extend -255 8) -> 0 .cble .coNP Function @ ash .synb .mets (ash < value << bits ) .syne .desc The .code ash function shifts .meta value by the specified number of .meta bits producing a new value. If .meta bits is positive, then a left shift takes place. If .meta bits is negative, then a right shift takes place. If .meta bit is zero, then .meta value is returned unaltered. For positive numbers, a left shift by n bits is equivalent to a multiplication by two to the power of n, or .codn (expt 2 n) . A right shift by n bits of a positive integer is equivalent to integer division by .codn (expt 2 n) , with truncation toward zero. For negative numbers, the bit shift is performed as if on the two's-complement representation. Under the infinite two's-complement representation, a right shift does not exhaust the infinite sequence of .code 1 digits which extends to the left. Thus if .code -4 is shifted right it becomes .code -2 because the bitwise representations of these values are .code ...111100 and .codn ...11110 . .coNP Function @ bit .synb .mets (bit < value << bit ) .syne .desc The .code bit function tests whether the integer .meta value has a 1 in bit position .metn bit . The .meta bit argument must be a non-negative integer. A value of zero of .meta bit indicates the least significant bit position of .metn value . The .code bit function has a boolean result, returning the symbol .code t if bit .meta bit of .meta value is set, otherwise .codn nil . If .meta value is negative, it is treated as if it had an infinite-bit two's complement representation. For instance, if value is .codn -2 , then the bit function returns .code nil for a .meta bit value of zero, and .code t for all other values, since the infinite bit two's complement representation of .code -2 is .codn ...11110 . .coNP Function @ mask .synb .mets (mask << integer *) .syne .desc The .code mask function takes zero or more integer arguments, and produces an integer value which corresponds a bitmask made up of the bit positions specified by the integer values. If .code mask is called with no arguments, then the return value is zero. If .code mask is called with a single argument .meta integer then the return value is the same as that of the expression .codn (ash 1 ) : the value 1 shifted left by .meta integer bit positions. If .meta integer is zero, then the result is .codn 1 ; if .meta integer is .codn 1 , the result is .code 2 and so forth. If .meta value is negative, then the result is zero. If .code mask is called with two or more arguments, then the result is a bitwise of the masks individually computed for each of the values. In other words, the following equivalences hold: .cblk (mask) <--> 0 (mask a) <--> (ash 1 a) (mask a b c ...) <--> (logior (mask a) (mask b) (mask c) ...) .cble .coNP Function @ width .synb .mets (width << integer *) .syne .desc A two's complement representation of an integer consists of a sign bit and a mantissa field. The .code width function computes the minimum number of bits required for the mantissa portion of the two's complement representation of the .meta integer argument. For a nonnegative argument, the width also corresponds to the number of bits required for a natural binary representation of that value. Two integer values have a width of zero, namely 0 and -1. This means that these two values can be represented in a one-bit two's complement, consisting of only a sign bit: the one-bit two's complement bitfield 1 denotes -1, and 0 denotes 0. Similarly, two integer values have a width of 1: 1 and -2. The two-bit two's complement bitfield 01 denotes 1, and 10 denotes -2. The argument may be a character. .SS* Exceptions .coNP Functions @, throw @ throwf and @ error .synb .mets (throw < symbol << arg *) .mets (throwf < symbol < format-string << format-arg *) .mets (error < format-string << format-arg *) .syne .desc These functions generate an exception. The .code throw and .code throwf functions generate an exception identified by .metn symbol , whereas .code error throws an exception of type .codn error . The call .code (error ...) can be regarded as a shorthand for .codn (throwf 'error ...) . The .code throw function takes zero or more additional arguments. These arguments become the arguments of a .code catch handler which takes the exception. The handler will have to be capable of accepting that number of arguments. The .code throwf and .code error functions generate an exception which has a single argument: a character string created by a formatted print to a string stream using the .code format string and additional arguments. .coNP Operator @ catch .synb .mets (catch < try-expression .mets \ \ >> {( symbol <> ( arg *) << body-form *)}*) .syne .desc The .code catch operator establishes an exception catching block around the .metan try-expression . The .meta try-expression is followed by zero or more catch clauses. Each catch clause consists of a symbol which denotes an exception type, an argument list, and zero or more body forms. If .meta try-expression terminates normally, then the catch clauses are ignored. The catch itself terminates, and its return value is that of the .metn try-expression . If .meta try-expression throws an exception which is a subtype of one or more of the type symbols given in the exception clauses, then the first (leftmost) such clause becomes the exit point where the exception is handled. The exception is converted into arguments for the clause, and the clause body is executed. When the clause body terminates, the catch terminates, and the return value of the catch is that of the clause body. If .meta try-expression throws an exception which is not a subtype of any of the symbols given in the clauses, then the search for an exit point for the exception continues through the enclosing forms. The catch clauses are not involved in the handling of that exception. When a clause catches an exception, the number of arguments in the catch must match the number of elements in the exception. A catch argument list resembles a function or lambda argument list, and may be dotted. For instance the clause .code (foo (a . b)) catches an exception subtyped from .codn foo , with one or more elements. The first element binds to parameter .codn a , and the rest, if any, bind to parameter .codn b . If there is only one element, .code b takes on the value .codn nil . Also see: the .code unwind-protect operator, and the functions .codn throw , .code throwf and .codn error . .coNP Macro @ ignerr .synb .mets (ignerr << form *) .syne .desc The .code ignerr macro operator evaluates each .meta form similarly to the .code progn operator. If no forms are present, it returns .codn nil . Otherwise it evaluates each .meta form in turn, yielding the value of the last one. If the evaluation of any .meta form is abandoned due to an exception of type .codn error , the code generated by the .code ignerr macro catches this exception. In this situation, the execution of the .code ignerr form terminates without evaluating the remaining forms, and yields .codn nil . .coNP Variable @ *unhandled-hook* The .code *unhandled-hook* variable is initialized with .code nil by default. It may instead be assigned a function which is capable of taking three arguments. When an exception occurs which has no handler, this function is called, with the following arguments: the exception type symbol, the exception object, and a third value which is either .code nil or else the form which was being evaluated the exception was thrown. Otherwise, if the variable is .code nil some informational messages are printed about the exception, and the process exits with a failed termination status. In the same situation, if the variable contains an object which is not a function, the process terminates abnormally as if by a call to the .code abort function. Prior to the function being called, the .code *unhandled-hook* variable is reset to .codn nil . If the function registered in .code *unhandled-hook* returns, the process exits with a failed termination status. Note: the functions .code source-loc or .code source-loc-str may be applied to the third argument of the .code *unhandled-hook* function to obtain more information about the form. .SS* Regular Expression Library .coNP Functions @ search-regex and @ range-regex .synb .mets (search-regex < string < regex >> [ start <> [ from-end ]]) .mets (range-regex < string < regex >> [ start <> [ from-end ]]) .mets (range-regst < string < regex >> [ start <> [ from-end ]]) .syne .desc The .code search-regex function searches through .meta string starting at position .meta start for a match for .metn regex . If .meta start is omitted, the search starts at position 0. If .meta from-end is specified and has a .cod2 non- nil value, the search proceeds in reverse, from the last position in the string, toward .metn start . This function returns .code nil if no match is found, otherwise it returns a cons, whose .code car indicates the position of the match, and whose .code cdr indicates the length of the match. The .code range-regex function is similar to .codn search-regex , except that when a match is found, it returns a position range, rather than a position and length. A cons is returned whose .code car indicates the position of the match, and whose .code cdr indicates the position one element past the last character of the match. If the match is empty, the two integers are equal. The .code search-regst differs from .code search-regex in the representation of the return value in the matching case. Rather than returning the position and length of the match, it returns the matching substring of .metn string . .coNP Functions @ match-regex and @ match-regst .synb .mets (match-regex < string < regex <> [ position ]) .mets (match-regst < string < regex <> [ position ]) .syne .desc The .code match-regex function tests whether .meta regex matches at .meta position in .metn string . If .meta position is not specified, it is taken to be zero. If the regex matches, then the length of the match is returned. If it does not match, then .code nil is returned. The .code match-regst differs from .code match-regex in the representation of the return value in the matching case. Rather than returning the length of the match, it returns matching substring of .metn string . .coNP Functions @ match-regex-right and @ match-regst-right .synb .mets (match-regex-right < string < regex <> [ end-position ]) .mets (match-regst-right < string < regex <> [ end-position ]) .syne .desc The .code match-regex function tests whether .meta string contains a match which ends precisely on the character just before .metn end-position . If .meta end-position is not specified, it defaults to the length of the string, and the function performs a right-anchored regex match. If a match is found, then the length of the match is returned. The match must terminate just before .meta end-position in the sense that additional characters at .meta end-position and beyond can no longer satisfy the regular expression. More formally, the function searches, starting from position zero, for positions where there occurs a match for the regular expression, taking the longest possible match. The length of first such a match which terminates on the character just before .meta end-position is returned. If no such a match is found, then .code nil is returned. The .code match-regst-right differs from .code match-regst-right in the representation of the return value in the matching case. Rather than returning the length of the match, it returns matching substring of .metn string . .TP* Examples: .cblk ;; Return matching portion rather than length thereof. (defun match-regex-right-substring (str reg : end-pos) (set end-pos (or end-pos (length str))) (let ((len (match-regex-right str reg end-pos))) (if len [str (- end-pos len)..end-pos] nil))) (match-regex-right-substring "abc" #/c/) -> "" (match-regex-right-substring "acc" #/c*/) -> "cc" ;; Regex matches starting at multiple positions, but all ;; the matches extend past the limit. (match-regex-right-substring "acc" #/c*/ 2) -> nil ;; If the above behavior is not wanted, then ;; we can extract the string up to the limiting ;; position and do the match on that. (match-regex-right-substring ["acc" 0..2] #/c*/) -> "c" ;; Equivalent of above call (match-regex-right-substring "ac" #/c*/) -> "c" .cble .coNP Function @ regsub .synb .mets (regsub < regex < replacement << string ) .syne .desc The .code regsub function searches .meta string for multiple occurrences of non-overlapping matches for .metn regex . A new string is constructed similar to .meta string but in which each matching region is replaced with using .meta replacement as follows. The .meta replacement object may be a character or a string, in which case it is simply taken to be the replacement for each match of the regular expression. The .meta replacement object may be a function of one argument, in which case for every match which is found, this function is invoked, with the matching piece of text as an argument. The function's return value is then taken to be the replacement text. .TP* Examples: .cblk ;; match every lower case e or o, and replace by filtering ;; through the upcase-str function: [regsub #/[eo]/ upcase-str "Hello world!"] -> "HEllO wOrld!" ;; Replace Hello with Goodbye: (regsub #/Hello/ "Goodbye" "Hello world!") -> "Goodbye world!" .cble .coNP Function @ regexp .synb .mets (regexp << obj ) .syne .desc The .code regexp function returns .code t if .meta obj is a compiled regular expression object. For any other object type, it returns .codn nil . .coNP Function @ regex-compile .synb .mets (regex-compile < form-or-string <> [ error-stream ]) .syne .desc The .code regex-compile function takes the source code of a regular expression, expressed as a Lisp data structure representing an abstract syntax tree, or else a regular expression specified as a character string, and compiles it to a regular expression object. If .meta form-or-string is a character string, it is parsed to an abstract syntax tree first, if by the .code regex-parse function. If the parse is successful (the result is not .codn nil ) then the resulting tree structure is compiled by a recursive call to .codn regex-compile . The optional .meta error-stream argument is passed down to .code regex-parse as well as in the recursive call to .codn regex-compile , if that call takes place. If .meta error-stream is specified, it must be a stream. Any error diagnostics are sent to that stream. .TP* Examples: .cblk ;; the equivalent of #/[a-zA-Z0-9_/ (regex-compile '(set (#\ea . #\ez) (#\eA . #\eZ) (#\e0 . #\e9) #\e_)) ;; the equivalent of #/.*/ and #/.+/ (regex-compile '(0+ wild)) (regex-compile '(1+ wild)) ;; #/a|b|c/ (regex-compile '(or (or #\ea #\eb) #\ec)) ;; string (regex-compile "a|b|c") .cble .coNP Function @ regex-parse .synb .mets (regex-parse < string <> [ error-stream ]) .syne .desc The .code regex-parse function parses a character string which contains a regular expression (without any surrounding / characters) and turns it into a Lisp data structure (the abstract syntax tree representation of the regular expression). The regular expression syntax .code #/RE/ produces the same structure, but as a literal which is processed at the time \*(TX source code is read; the .code regex-parse function performs this parsing at run-time. If there are parse errors, the function returns .codn nil . The optional .meta error-stream argument specifies a stream to which error messages are sent from the parser. By default, diagnostic output goes to the .code *stdnull* stream, which discards it. If .meta error-stream is specified as .codn t , then the diagnostic output goes to the .code *stdout* stream. If .code regex-parse returns a .cod2 non- nil value, that structure is then something which is suitable as input to .codn regex-compile . .SS* Hashing Library .coNP Functions @, make-hash and @ hash .synb .mets (make-hash < weak-keys < weak-vals << equal-based ) .mets (hash { :weak-keys | :weak-vals | :equal-based }*) .syne .desc These functions construct a new hash table. A hash table is an object which retains an association between pairs of objects. Each pair consists of a key and value. Given an object which is similar to a key in the hash table, it is possible to retrieve the corresponding value. Entries in a hash table are not ordered in any way, and lookup is facilitated by hashing: quickly mapping a key object to a numeric value which is then used to index into one of many buckets where the matching key will be found (if such a key is present in the hash table). .code make-hash takes three mandatory boolean arguments. The .meta weak-keys argument specifies whether the hash table shall have weak keys. The .meta weak-vals argument specifies whether it shall have weak values, and .meta equal-based specifies whether it is .codn equal- based. The hash function defaults all three of these properties to false, and allows them to be overridden to true by the presence of keyword arguments. It is an error to attempt to construct an .codn equal -based hash table which has weak keys. The hash function provides an alternative interface. It accepts optional arguments which are keyword symbols. Any combination of the three symbols .codn :weak-keys , .code :weak-vals and .code :equal-based can be specified in any order to turn on the corresponding properties in the newly constructed hash table. If any of the keywords is not specified, the corresponding property defaults to .codn nil . If a hash table has weak keys, this means that from the point of view of garbage collection, that table holds only weak references to the keys stored in it. Similarly, if a hash table has weak values, it means that it holds a weak reference to each value stored. A weak reference is one which does not prevent the reclamation of an object by the garbage collector. That is to say, when the garbage collector discovers that the only references to some object are weak references, then that object is considered garbage, just as if it had no references to it. The object is reclaimed, and the weak references "lapse" in some way, which depends on what kind they are. Hash table weak references lapse by entry removal: if either a key or a value object is reclaimed, then the corresponding key-value entry is erased from the hash table. Important to the operation of a hash table is the criterion by which keys are considered same. By default, this similarity follows the eql function. A hash table will search for a stored key which is .code eql to the given search key. A hash table constructed with the .codn equal -based property compares keys using the .code equal function instead. In addition to storing key-value pairs, a hash table can have a piece of information associated with it, called the user data. A hash table can be traversed to visit all of the keys and data. The order of traversal bears no relation to the order of insertion, or to any properties of the key type. During an open traversal, new keys can be inserted into a hash table or deleted from it while a a traversal is in progress. Insertion of a new key during traversal will not cause any existing key to be visited twice or to be skipped; however, it is not specified whether the new key will be traversed. Similarly, if a key is deleted during traversal, and that key has not yet been visited, it is not specified whether it will be visited during the remainder of the traversal. An open traversal of a hash table is performed by the .code maphash function and the .code dohash operator. The traversal is open because code supplied by the program is evaluated for each entry. The functions .codn hash-keys , .codn hash-values , .codn hash-pairs , and .code hash-alist also perform an open traversal, because they return lazy lists. The traversal isn't complete until the returned lazy list is fully instantiated. In the meanwhile, the \*(TX program can mutate the hash table from which the lazy list is being generated. .coNP Functions @ hash-construct and @ hash-from-pairs .synb .mets (hash-construct < hash-args << key-val-pairs ) .mets (hash-from-pairs < key-val-pairs << hash-arg *) .syne .desc The .code hash-construct function constructs a populated hash in one step. The .meta hash-args argument specifies a list suitable as an argument list in a call to the hash function. The .meta key-val-pairs is a sequence of pairs, which are two-element lists representing key-value pairs. A hash is constructed as if by a call to .cblk .meti (apply hash << hash-args ), .cble then populated with the specified pairs, and returned. The .code hash-from-pairs function is an alternative interface to the same semantics. The .meta key-val-pairs argument is first, and the .meta hash-args are passed as trailing variadic arguments, rather than a single list argument. .coNP Function @ hash-list .synb .mets (hash-list < key-list << hash-arg *) .syne .desc The .code hash-list function constructs a hash as if by a call to .cblk .meti (apply hash << hash-args ), .cble where .meta hash-args is a list of the individual .meta hash-arg variadic arguments. The hash is then populated with keys taken from .meta key-list and returned. The value associated with each key is that key itself. .coNP Function @ hash-update .synb .mets (hash-update < hash << function ) .syne .desc The .code hash-update function replaces each values in .metn hash , with the value of .meta function applied to that value. The return value is .metn hash . .coNP Function @ hash-update-1 .synb .mets (hash-update-1 < hash < key < function <> [ init ]) .syne .desc The .code hash-update-1 function operates on a single entry in the hash table. If .meta key exists in the hash table, then its corresponding value is passed into .metn function , and the return value of .meta function is then installed in place of the key's value. The value is then returned. If .meta key does not exist in the hash table, and no .meta init argument is given, then .code hash-update-1 does nothing and returns .codn nil . If .meta key does not exist in the hash table, and an .meta init argument is given, then .meta function is applied to .metn init , and then .meta key is inserted into .meta hash with the value returned by .meta function as the datum. This value is also returned. .coNP Function @ group-by .synb .mets (group-by < func < sequence << option *) .syne .desc The .code group-by function produces a hash table from .metn sequence , which is a list or vector. Entries of the hash table are not elements of .metn sequence , but lists of elements of .metn sequence . The function .meta func is applied to each element of .meta sequence to compute a key. That key is used to determine which list the item is added to in the hash table. The trailing arguments .cblk .meti << option * .cble if any, consist of the same keywords that are understood by the hash function, and determine the properties of the hash. .TP* Example: Group the integers from 0 to 10 into three buckets keyed on 0, 1 and 2 according to the modulo 3 congruence: .cblk (group-by (op mod @1 3) (range 0 10))) -> #H(() (0 (0 3 6 9)) (1 (1 4 7 10)) (2 (2 5 8))) .cble .coNP Functions @ make-similar-hash and @ copy-hash .synb .mets (make-similar-hash << hash ) .mets (copy-hash << hash ) .syne .desc The .code make-similar-hash and copy-hash functions create a new hash object based on the existing .meta hash object. .code make-similar-hash produces an empty hash table which inherits all of the attributes of .metn hash . It uses the same kind of key equality, the same configuration of weak keys and values, and has the same user data (see the .code set-hash-userdata function). The .code copy-hash function is like .codn make-similar-hash , except that instead of producing an empty hash table, it produces one which has all the same elements as .metn hash : it contains the same key and value objects. .coNP Function @ inhash .synb .mets (inhash < hash < key <> [ init ]) .syne .desc The .code inhash function searches hash table .meta hash for .metn key . If .meta key is found, then it return the hash table's cons cell which represents the association between .meta hash and .metn key . Otherwise, it returns .codn nil . If argument .meta init is specified, then the function will create an entry for .meta key in .meta hash whose value is that of .metn init . The cons cell representing that association is returned. Note: for as long as the .meta key continues to exist inside .metn hash . modifying the .code car field of the returned cons has ramifications for the logical integrity of the hash. Modifying the .code cdr field has the effect of updating the association with a new value. .coNP Accessor @ gethash .synb .mets (gethash < hash < key <> [ alt ]) .mets (set (gethash < hash < key <> [ alt ]) << new-value ) .syne .desc The .code gethash function searches hash table .meta hash for key .metn key . If the key is found then the associated value is returned. Otherwise, if the .meta alt argument was specified, it is returned. If the .meta alt argument was not specified, .code nil is returned. A valid .code gethash form serves as a place. It denotes either an existing value in a hash table or a value that would be created by the evaluation of the form. The .meta alt argument is meaningful when .code gethash is used as a place, and, if present, is always evaluated whenever the place is evaluated. In place update operations, it provides the initial value, which defaults to .code nil if the argument is not specified. For example .code (inc (gethash h k d)) will increment the value stored under key .code k in hash table .code h by one. If the key does not exist in the hash table, then the value .code (+ 1 d) is inserted into the table under that key. The expression .code d is always evaluated, whether or not its value is needed. If a .code gethash place is subject to a deletion, but doesn't exist, it is not an error. The operation does nothing, and .code nil is considered the prior value of the place yielded by the deletion. .coNP Function @ sethash .synb .mets (sethash < hash < key << value ) .syne .desc The .code sethash function places a value into .meta hash table under the given .metn key . If a similar key already exists in the hash table, then that key's value is replaced by .metn value . Otherwise, the .meta key and .meta value pair is newly inserted into .metn hash . The .code sethash function returns the .metn value argument. .coNP Function @ pushhash .synb .mets (pushhash < hash < key << element ) .syne .desc The .code pushhash function is useful when the values stored in a hash table are lists. If the given .meta key does not already exist in .metn hash , then a list of length one is made which contains .metn element , and stored in .meta hash table under .metn key . If the .meta key already exists in the hash table, then the corresponding value must be a list. The .meta element value is added to the front of that list, and the extended list then becomes the new value under .metn key . The return value is boolean. If true, indicates that the hash table entry was newly created. If false, it indicates that the push took place on an existing entry. .coNP Function @ remhash .synb .mets (remhash < hash << key ) .syne .desc The .code remhash function searches .meta hash for a key similar to the .metn key . If that key is found, then that key and its corresponding value are removed from the hash table. If the key is found and removal takes place, then the associated value is returned. Otherwise .code nil is returned. .coNP Function @ hash-count .synb .mets (hash-count << hash ) .syne .desc The .code hash-count function returns an integer representing the number of key-value pairs stored in .metn hash . .coNP Function @ get-hash-userdata .synb .mets (get-hash-userdata << hash ) .syne .desc This function retrieves the user data object associated with .metn hash . The user data object of a newly-created hash table is initialized to .codn nil . .coNP Function @ set-hash-userdata .synb .mets (set-hash-userdata < hash << object ) .syne .desc The .code set-hash-userdata replaces, with the .metn object , the user data object associated with .metn hash . .coNP Function @ hashp .synb .mets (hashp << object ) .syne .desc The .code hashp function returns .code t if the .meta object is a hash table, otherwise it returns .codn nil . .coNP Function @ maphash .synb .mets (maphash < hash << binary-function ) .syne .desc The .code maphash function successively invokes .meta binary-function for each entry stored in .metn hash . Each entry's key and value are passed as arguments to .codn binary-function . The function returns .codn nil . .coNP Functions @ hash-eql and @ hash-equal .synb .mets (hash-eql << object ) .mets (hash-equal << object ) .syne .desc These functions each compute an integer hash value from the internal representation of .metn object , which satisfies the following properties. If two objects .code A and .code B are the same under the .code eql function, then .code (hash-eql A) and .code (hash-eql B) produce the same integer hash value. Similarly, if two objects .code A and .code B are the same under the .code equal function, then .code (hash-equal A) and .code (hash-equal B) each produce the same integer hash value. In all other circumstances, the hash values of two distinct objects are unrelated, and may or may not be the same. .coNP Functions @, hash_keys @, hash_values @ hash_pairs and @ hash_alist .synb .mets (hash-keys << hash ) .mets (hash-values << hash ) .mets (hash-pairs << hash ) .mets (hash-alist << hash ) .syne .desc These functions retrieve the bulk key-value data of hash table .meta hash in various ways. .code hash-keys retrieves a list of the keys. .code hash-values retrieves a list of the values. .code hash-pairs retrieves a list of pairs, which are two-element lists consisting of the key, followed by the value. Finally, .code hash-alist retrieves the key-value pairs as a Lisp association list: a list of cons cells whose .code car fields are keys, and whose .code cdr fields are the values. Note that .code hash-alist returns the actual entries from the hash table, which are conses. Modifying the .code cdr fields of these conses constitutes modifying the hash values in the original hash table. Modifying the .code car fields interferes with the integrity of the hash table. These functions all retrieve the keys and values in the same order. For example, if the keys are retrieved with .codn hash-keys , and the values with .codn hash-values , then the corresponding entries from each list pairwise correspond to the pairs in .metn hash . The list returned by each of these functions is lazy, and hence constitutes an open traversal of the hash table. .coNP Operator @ dohash .synb .mets (dohash >> ( key-var < value-var < hash-form <> [ result-form ]) .mets \ \ << body-form *) .syne .desc The .code dohash operator iterates over a hash table. The .meta hash-form expression must evaluate to an object of hash table type. The .meta key-var and .meta value-var arguments must be symbols suitable for use as variable names. Bindings are established for these variables over the scope of the .metn body-form s and the optional .metn result-form . For each element in the hash table, the .meta key-var and .meta value-var variables are set to the key and value of that entry, respectively, and each .metn body-form , if there are any, is evaluated. When all of the entries of the table are thus processed, the .meta result-form is evaluated, and its return value becomes the return value of the dohash form. If there is no .metn result-form , the return value is .codn nil . The .meta result-form and .metn body-form s are in the scope of an implicit anonymous block, which means that it is possible to terminate the execution of dohash early using .cblk .meti (return << value ) .cble or .codn (return) . .coNP Functions @, hash-uni @ hash-diff and @ hash-isec .synb .mets (hash-uni < hash1 < hash2 <> [ join-func ]) .mets (hash-diff < hash1 << hash2 ) .mets (hash-isec < hash1 < hash2 <> [ join-func ]) .syne .desc These functions perform basic set operations on hash tables in a nondestructive way, returning a new hash table without altering the inputs. The arguments .meta hash1 and .meta hash2 must be compatible hash tables. This means that their keys must use the same kind of equality. The resulting hash table inherits attributes from .metn hash1 , as if created by the .code make-similar-hash function. If .meta hash1 has userdata, the resulting hash table has the same userdata. If .meta hash1 has weak keys, the resulting table has weak keys, and so forth. The .code hash-uni function performs a set union. The resulting hash contains all of the keys from .meta hash1 and all of the keys from .metn hash2 , and their corresponding values. If a key occurs both in .meta hash1 and .metn hash2 , then it occurs only once in the resulting hash. In this case, if the .meta join-func argument is not given, the value associated with this key is the one from .metn hash1 . If .meta join-func is specified then it is called with two arguments: the respective data items from .meta hash1 and .metn hash2 . The return value of this function is used as the value in the union hash. The .code hash-diff function performs a set difference. First, a copy of .meta hash1 is made as if by the .code copy-hash function. Then from this copy, all keys which occur in .code hash2 are deleted. The .code hash-isec function performs a set intersection. The resulting hash contains only those keys which occur both in .meta hash1 and .metn hash2 . If .meta join-func is not specified, the values selected for these common keys are those from .metn hash1 . If .meta join-func is specified, then for each key which occurs in both .meta hash1 and .metn hash2 , it is called with two arguments: the respective data items. The return value is then used as the data item in the intersection hash. .coNP Functions @ hash-subset and @ hash-proper-subset .synb .mets (hash-subset < hash1 << hash2 ) .mets (hash-proper-subset < hash1 << hash2 ) .syne .desc The .code hash-subset function returns .code t if the keys in .meta hash1 are a subset of the keys in .metn hash2 . The .code hash-proper-subset function returns .code t if the keys in .meta hash1 are a proper subset of the keys in .metn hash2 . This means that .meta hash2 has all the keys which are in .meta hash1 and at least one which isn't. Note: the return value may not be mathematically meaningful if .meta hash1 and .meta hash2 use different equality. In any case, the actual behavior may be understood as follows. The implementation of .code hash-subset tests whether each of the keys in .meta hash1 occurs in .meta hash2 using their respective equalities. The implementation of .code hash-proper-subset applies .code hash-subset first, as above. If that is true, and the two hashes have the same number of elements, the result is falsified. .SS* Partial Evaluation and Combinators .coNP Macros @ op and @ do .synb .mets (op << form +) .mets (do << form +) .syne .desc The .code op and .code do macro operators are similar. Like the lambda operator, the .code op operator creates an anonymous function based on its syntax. The difference is that the arguments of the function are implicit, or optionally specified within the function body, rather than as a formal parameter list before the body. Also, the .meta form arguments of .code op are implicitly turned into a DWIM expression, which means that argument evaluation follows Lisp-1 rules. (See the .code dwim operator). The .code do operator is like the .code op operator with the following difference: the .meta form arguments of .code op are not implicitly treated as a DWIM expression, but as an ordinary expression. In particular, this means that operator syntax is permitted. Note that the syntax .code (op @1) makes sense, since the argument can be a function, which will be invoked, but .code (do @1) doesn't make sense because it will produce a Lisp-2 form like .code (#:arg1 ...) referring to nonexistent function .codn #:arg1 . Because it accepts operators, .code do can be used with imperative constructs which are not functions, like set: like set: for instance .code (do set x) produces an anonymous function which, if called with one argument, stores that argument into .codn x . The argument forms are arbitrary expressions, within which a special convention is permitted: .RS .meIP >> @ num A number preceded by a .code @ is a metanumber. This is a special syntax which denotes an argument. For instance .code @2 means that the second argument of the anonymous function is to be substituted in place of the .codn @2 . .code op generates a function which has a number of required arguments equal to the highest value of .meta num appearing in a .cblk .mati >> @ num .cble construct in the body. For instance .code (op car @3) generates a three-argument function (which passes its third argument to .codn car , returning the result, and ignores its first two arguments). There is no way to use .code op to generate functions which have optional arguments. .meIP < @rest If the meta-symbol .meta @rest appears in the .code op syntax, it explicitly denotes the list of trailing arguments, allowing them to be placed anywhere in the expression. .RE .IP Functions generated by .code op are always variadic; they always take additional arguments after any required ones, whether or not the .meta @rest syntax is used. If the body does not contain any .meta @num or .meta @rest syntax, then .code @rest is implicitly inserted. What this means is that, for example, since the form .code (op foo) does not contain any numeric positional arguments like .codn @1 , and does not contain .codn @rest , it is actually a shorthand for .codn (op foo . @rest) : a function which applies all of its arguments to .codn foo . The actions of .code op be understood by these examples, which show how .code op is rewritten to lambda. However, note that the real translator uses generated symbols for the arguments, which are not equal to any symbols in the program. .cblk (op) -> invalid (op +) -> (lambda rest [+ . rest]) (op + foo) -> (lambda rest [+ foo . rest]) (op @1 @2) -> (lambda (arg1 arg2 . rest) [arg1 arg2]) (op @1 . @rest) -> (lambda (arg1 . rest) [arg1 . @rest]) (op @1 @rest) -> (lambda (arg1 . rest) [arg1 @rest]) (op @1 @2) -> (lambda (arg1 arg2 . rest) [arg1 arg2]) (op foo @1 (@2) (bar @3)) -> (lambda (arg1 arg2 arg3 . rest) [foo arg1 (arg2) (bar arg3)]) (op foo @rest @1) -> (lambda (arg1 . rest) [foo rest arg1]) (do + foo) -> (lambda rest (+ foo . rest)) (do @1 @2) -> (lambda (arg1 arg2 . rest) (arg1 arg2)) (do foo @rest @1) -> (lambda (arg1 . rest) (foo rest arg1)) .cble Note that if argument .meta @n appears, it is not necessary for arguments .meta @1 through .meta @n-1 to appear. The function will have .code n arguments: .cblk (op @3) -> (lambda (arg1 arg2 arg3 . rest) [arg3]) .cble The .code op and .code do operators can be nested, in any combination. This raises the question: if a metanumber like .code @1 or .code @rest occurs in an .code op that is nested within an .codn op , what is the meaning? A metanumber always belongs with the inner-most op or do operator. So for instance .code (op (op @1)) means that an .code (op @1) expression is nested within an .code op expression which itself contains no meta-syntax. The .code @1 belongs with the inner op. There is a way for an inner .code op to refer to an outer op metanumber argument. This is expressed by adding an extra .code @ prefix for every level of escape. For example in .code (op (op @@1)) the .code @@1 belongs to the outer .codn op : it is the same as .code @1 appearing in the outer .codn op . That is to say, in the expression .codn (op @1 (op @@1)) , the .code @1 and .code @@1 are the same thing: both are parameter 1 of the lambda function generated by the outer .codn op . By contrast, in the expression .code (op @1 (op @1)) there are two different parameters: the first .code @1 is argument of the outer function, and the second .code @1 is the first argument of the inner function. Of course, if there are three levels of nesting, then three .code @ meta-prefixes are needed to insert a parameter from the outermost .code op into the innermost .codn op . .TP* Examples: .cblk ;; Take a list of pairs and produce a list in which those pairs ;; are reversed. (mapcar (op list @2 @1) '((1 2) (a b))) -> ((2 1) (b a)) .cble The .code op syntax interacts with quasiliterals which are nested within it. The metanumber notation as well as .code @rest are recognized without requiring an additional .code @ escape: .cblk (apply (op list `@1-@rest`) '(1 2 3)) -> "1-2 3" (apply (op list `@@1-@@rest`) '(1 2 3)) -> "1-2 3" .cble This is because the .code op macro traverses the code structure produced by the literal without recognizing it specially, and there imposes its own meaning on these elements. Though they produce the same result, the above two examples differ in that .code @rest embeds a metasymbol into the quasiliteral structure, whereas .code @@rest embeds the Lisp expression .code @rest into the quasiliteral. In ordinary circumstances, the former refers to the variable .codn rest . Contrast the previous example with: .cblk (let ((rest "0")) `rest: @rest`) -> "rest: 0" (let ((rest "0")) `rest: @@rest`) -> ;; error: no such function or operator: sys:var .cblk Under the .code op macro and its relatives, occurrences of .code @rest are replaced with syntax which refers to the trailing arguments of the anonymous function. This happens before the interior of the .code op syntax undergoes expansion. Therefore the quasiliteral expander never sees the .codn @rest . This convenient omission of the .codn @ character isn't supported for reaching the arguments of an outer .code op from a quasiliteral within a nested .codn op : .cblk ;; To reach @@1, @@@1 must be written. ;; @@1 Lisp expression introduced by @. (op ... (op ... `@@@1`)) .cble .coNP Macros @, ap @, ip @ ado and @ ido. .synb .mets (ap << form +) .mets (ip << form +) .mets (ado << form +) .mets (ido << form +) .syne .desc The .code ap macro is based on the .code op macro and has identical argument conventions. The .code ap macro analyzes its arguments and produces a function .metn f , in exactly the same same way as the .code op macro. However, instead of returning .metn f , directly, it returns a different function .metn g , which is a one-argument function which accepts a list, and then applies the list as arguments to .metn f . In other words, the following equivalence holds: .cblk (ap form ...) <--> (apf (op form ...)) .cble The .code ap macro nests properly with .code op and .codn do , in any combination, in regard to the .meta ...@@n notation. The .ode ip macro is very similar to the .code ap macro, except that it is based on the semantics of the function .code iapply rather than .codn apply , according to the following equivalence: .cblk (ip form ...) <--> (ipf (op form ...)) .cble The .code ado and .code ido macros are related to do macro in the same way that .code ap and .code ip are related to .codn op . They produce a one-argument function which works as if by applying its arguments to the function generated by do, according to the following equivalence: .cblk (ado form ...) <--> (apf (do form ...)) (ido form ...) <--> (ipf (do form ...)) .cblk See also: the .code apf and .code ipf functions. .coNP Macros @ opip and @ oand .synb .mets (opip << clause *) .mets (oand << clause *) .syne .desc The .code opip and .code oand operators make it possible to chain together functions which are expressed using the .code op syntax. (See the .code op operator for more information). Both macros perform the same transformation except that .code opip translates its arguments to a call to the .code chain function, whereas .code oand translates its arguments in the same way to a call to the .code chand function. More precisely, these macros perform the following rewrites: .cblk (opip arg1 arg2 ... argn) -> [chain {arg1} {arg2} ... {argn}] (oand arg1 arg2 ... argn) -> [chand {arg1} {arg2} ... {argn}] .cble where the above .code {arg} notation denotes the following transformation applied to each argument: .cblk (function ...) -> (op function ...) (operator ...) -> (do operator ...) (macro ...) -> (do macro ...) (dwim ...) -> (dwim ...) [...] -> [...] atom -> atom .cble In other words, compound forms whose leftmost symbol is a macro or operator are translated to the .code do notation. Compound forms denoting function calls are translated to the .code op notation. Compound forms which are .code dwim invocations, either explicit or via the DWIM brackets notation, are preserved, as are any forms which are atoms. Note: the .code opip and .code oand macros use their macro environment in determining whether a form is a macro call, thereby respecting lexical scoping. .TP* Example: Take each element from the list .code (1 2 3 4) and multiply it by three, then add 1. If the result is odd, collect that into the resulting list: .cblk (mappend (opip (* 3) (+ 1) [iff oddp list]) (range 1 4)) .cble The above is equivalent to: .cblk (mappend (chain (op * 3) (op + 1) [iff oddp list]) (range 1 4)) .cble The .code (* 3) and .code (+ 1) terms are rewritten to .code (op * 3) and .codn (op + 1) , respectively, whereas .code [iff oddp list] is passed through untransformed. .coNP Macro @ ret .synb .mets (ret << form ) .syne .desc The .code ret macro's .meta form argument is treated similarly to the second and subsequent arguments of the .code op operator. The .code ret macro produces a function which takes any number of arguments, and returns the value specified by .metn form . .meta form can contain .code op meta syntax like .code @n and .codn @rest . The following equivalence holds: .cblk (ret x) <--> (op identity x)) .cble Thus the expression .code (ret @2) returns a function similar to .codn (lambda (x y . z) y) , and the expression .code (ret 42) returns a function similar to .codn (lambda (. rest) 42) . .coNP Macro @ aret .synb .mets (aret << form ) .syne .desc The .code ret macro's .meta form argument is treated similarly to the second and subsequent arguments of the .code op operator. The .code aret macro produces a function which takes any number of arguments, and returns the value specified by .metn form . .meta form can contain .code ap meta syntax like .meta @n and .codn @rest . The following equivalence holds: .cblk (aret x) <--> (ap identity x)) .cble Thus the expression .code (aret @2) returns a function similar to .codn (lambda (. rest) (second rest)) , and the expression .code (aret 42) returns a function similar to .codn (lambda (. rest) 42) . .coNP Function @ dup .synb .mets (dup << func ) .syne .desc The .code dup function returns a one-argument function which calls the two-argument function .metn func by duplicating its argument. .TP* Example: .cblk ;; square the elements of a list (mapcar [dup *] '(1 2 3)) -> (1 4 9) .cble .coNP Function @ flipargs .synb .mets (flipargs << func ) .syne .desc The .code flipargs function returns a two-argument function which calls the two-argument function .metn func with reversed arguments. .coNP Functions @ chain and @ chand .synb .mets (chain << func *) .mets (chand << func *) .syne .desc The .code chain function accepts zero or more functions as arguments, and returns a single function, called the chained function, which represents the chained application of those functions, in left to right order. If .code chain is given no arguments, then it returns a variadic function which ignores all of its arguments and returns .codn nil . Otherwise, the first function may accept any number of arguments. The second and subsequent functions, if any, must accept one argument. The chained function can be called with an argument list which is acceptable to the first function. Those arguments are in fact passed to the first function. The return value of that call is then passed to the second function, and the return value of that call is passed to the third function and so on. The final return value is returned to the caller. The .code chand function is similar, except that it combines the functionality of .code andf into chaining. The difference between .code chain and .code chand is that .code chand immediately terminates and returns .code nil whenever any of the functions returns .codn nil , without calling the remaining functions. .TP* Example: .cblk (call [chain + (op * 2)] 3 4) -> 14 .cble In this example, a two-element chain is formed from the .code + function and the function produced by .code (op * 2) which is a one-argument function that returns the value of its argument multiplied by two. (See the definition of the .code op operator). The chained function is invoked using the .code call function, with the arguments .code 3 and .codn 4 . The chained evaluation begins by passing .code 3 and .code 4 to .codn + , which yields .codn 7 . This .code 7 is then passed to the .code (op * 2) doubling function, resulting in .codn 14 . A way to write the above example without the use of the DWIM brackets and the op operator is this: .cblk (call (chain (fun +) (lambda (x) (* 2 x))) 3 4) .cble .coNP Function @ juxt .synb .mets (juxt << func *) .syne .desc The .code juxt function accepts a variable number of arguments which are functions. It combines these into a single function which, when invoked, passes its arguments to each of these functions, and collects the results into a list. Note: the juxt function can be understood in terms of the following reference implementation: .cblk (defun juxt (funcs) (lambda (. args) (mapcar (lambda (fun) (apply fun args)) funcs))) .cble .TP* Example: .cblk ;; separate list (1 2 3 4 5 6) into lists of evens and odds, ;; which end up juxtaposed in the output list: [(op [juxt keep-if remove-if] evenp) '(1 2 3 4 5 6)] -> ((2 4 6) (1 3 5)) ;; call several functions on 1, collecting their results: [[juxt (op + 1) (op - 1) evenp sin cos] 1]' -> (2 0 nil 0.841470984807897 0.54030230586814) .cble .coNP Functions @ andf and @ orf .synb .mets (andf << func *) .mets (orf << func *) .syne .desc The .code andf and .code orf functions are the functional equivalent of the .code and and .code or operators. These functions accept multiple functions and return a new function which represents the logical combination of those functions. The input functions should have the same arity. Failing that, there should exist some common argument arity with which each of these can be invoked. The resulting combined function is then callable with that many arguments. The .code andf function returns a function which combines the input functions with a short-circuiting logical conjunction. The resulting function passes its arguments to the functions successively, in left to right order. As soon as any of the functions returns .codn nil , then nil is returned immediately, and the remaining functions are not called. Otherwise, if none of the functions return .codn nil , then the value returned by the last function is returned. If the list of functions is empty, then .code t is returned. That is, .code (andf) returns a function which accepts any arguments, and returns .codn t . The .code orf function combines the input functions with a short-circuiting logical disjunction. The function produced by .code orf passes its arguments down to the functions successively, in left to right order. As soon as any function returns a .cod2 non- nil value, that value is returned and the remaining functions are not called. If all functions return .codn nil , then .code nil is returned. The expression .code (orf) returns a function which accepts any arguments and returns .codn nil . .coNP Function @ notf .synb .mets (notf << function ) .syne .desc The .code notf function returns a function which is the boolean negation of .metn function . The returned function takes a variable number of arguments. When invoked, it passes all of these arguments to .meta function and then inverts the result as if by application of the .codn not . .coNP Functions @ iff and @ iffi .synb .mets (iff < cond-func >> [ then-func <> [ else-func ]]) .mets (iffi < cond-func < then-func <> [ else-func ]) .syne .desc The .code iff function is the functional equivalent of the .code if operator. It accepts functional arguments and returns a function. The resulting function takes its arguments, if any, and applies them to .metn cond-func . If .meta cond-func yields true, then the arguments are passed to .meta then-func and the resulting value is returned. Otherwise the arguments are passed to .meta else-func and the resulting value is returned. If .meta then-func is omitted then .code identity is used as default. This omission is not permitted by .codn iffi , only .codn iff . If .meta else-func needs to be called, but is omitted, then .code nil is returned. The .code iffi function differs from .code iff only in the defaulting behavior with respect to the .meta else-func argument. If .meta else-func is omitted in a call to .code iffi then the default function is .codn identity . This is useful in situations when one value is to be replaced with another one when the condition is true, otherwise preserved. The following equivalences hold between .code iffi and .codn iff : .cblk (iffi a b c) <--> (iff a b c) (iffi a b) <--> (iff a b identity) (iffi a b false) <--> (iff a b) (iffi a identity false) <--> (iff a) .cble The following equivalence illustrates .code iff with both optional arguments omitted: .cblk (iff a) <---> (iff a identity false) .cble .coNP Functions @ tf and @ nilf .synb .mets (tf << arg *) .mets (nilf << arg *) .syne .desc The .code tf and .code nilf functions take zero or more arguments, and ignore them. The .code tf function returns .codn t , and the .code nilf function returns .codn nil . Note: the following equivalences hold between these functions and the .code ret operator, and .code retf function. .cblk (fun tf) <--> (ret t) <--> (retf t) (fun nilf) <--> (ret nil) <--> (ret) <--> (retf nil) .cble In Lisp-1-style code, .code tf and .code nilf behave like constants which can replace uses of .code (ret t) and .codn (ret nil) : .cblk [mapcar (ret nil) list] <--> [mapcar nilf list] .cble .TP* Example: .cblk ;; tf and nilf are useful when functions are chained together. ;; test whether (trunc n 2) is odd. (defun trunc-n-2-odd (n) [[chain (op trunc @1 2) [iff oddp tf nilf]] n]) .cble In this example, two functions are chained together, and .code n is passed through the chain such that it is first divided by two via the function denoted by .code (op trunc @1 2) and then the result is passed into the function denoted by .codn [iff oddp tf nilf] . The .code iff function passes its argument into .codn oddp , and if .code oddp yields true, it passes the same argument to .codn tf . Here .code tf proves its utility by ignoring that value and returning .codn t . If the argument (the divided value) passed into .code iff is even, then iff passes it into the .code nilf function, which ignores the value and returns .codn nil . .coNP Function retf .synb .mets (retf << value ) .syne .desc The .code retf function returns a function. That function can take zero or more arguments. When called, it ignores its arguments and returns .metn value . See also: the .code ret macro. .TP* Example: .cblk ;; the function returned by (retf 42) ;; ignores 1 2 3 and returns 42. (call (retf 42) 1 2 3) -> 42 .cble .coNP Functions @ apf and @ ipf .synb .mets (apf << function ) .mets (ipf << function ) .syne .desc The .code apf function returns a one-argument function which accepts a list. When the function is called, it treats the list as arguments which are applied to .meta function as if by apply. It returns whatever .meta function returns. The .code ipf function is similar to .codn apf , except that the returned function applies arguments as if by .code iapply rather than .codn apply . See also: the .code ap macro. .TP* Example: .cblk ;; Function returned by [apf +] accepts the ;; (1 2 3) list and applies it to +, as ;; if (+ 1 2 3) were called. (call [apf +] '(1 2 3)) -> 6 .cble .coNP Function @ callf .synb .mets (callf < main-function << arg-function *) .syne .desc The .code callf function returns a function which applies its arguments to each .metn arg-function , juxtaposing the return values of these calls to form arguments which are then passed to .metn main-function . The return value of .meta main-function is returned. The following equivalence holds, except for the order of evaluation of arguments: .cblk (callf fm f0 f1 f2 ...) <--> (chain (juxt f0 f1 f2 ...) (apf fm)) .cble .TP* Example: .cblk ;; Keep those pairs which are two of a kind (keep-if [callf eql first second] '((1 1) (2 3) (4 4) (5 6))) -> ((1 1) (4 4)) .cble .coNP Function @ mapf .synb .mets (mapf < main-function << arg-function *) .syne .desc The .code mapf function returns a function which distributes its arguments into the .metn arg-function -s. That is to say, each successive argument of the returned function is associated with a successive .metn arg-function . Each .metn arg-function is called, passed the corresponding argument. The return values of these functions are then passed as arguments to .meta main function and the resulting value is returned. If the returned function is called with fewer arguments than there are .metn arg-function -s, then only that many functions are used. Conversely, if the function is called with more arguments than there are .metn arg-function -s, then those arguments are ignored. The following equivalence holds: .cblk (mapf fm f0 f1 ...) <--> (lambda (. rest) [apply fm [mapcar call (list f0 f1 ...) rest]]) .cble .TP* Example: .cblk ;; Add the squares of 2 and 3 [[mapf + [dup *] [dup *]] 2 3] -> 13 .cble .SS* Input and Output (Streams) \*(TL supports input and output streams of various kinds, with generic operations that work across the stream types. In general, I/O errors are usually turned into exceptions. When the description of error reporting is omitted from the description of a function, it can be assumed that it throws an error. .coNP Special variables @, *stdout* @, *stddebug* @, *stdin* @ *stderr* and @ *stdnull* .desc These variables hold predefined stream objects. The .codn *stdin* , .code *stdout* and .code *stderr* streams closely correspond to the underlying operating system streams. Various I/O functions require stream objects as arguments. The .code *stddebug* stream goes to the same destination as .codn *stdout* , but is a separate object which can be redirected independently, allowing debugging output to be separated from normal output. The .code *stdnull* stream is a special kind of stream called a null stream. This stream is not connected to any device or file. It is similar to the .code /dev/null device on Unix, but does not involve the operating system. .coNP Function @ format .synb .mets (format < stream-designator < format-string << format-arg *) .syne .desc The .code format function performs output to a stream given by .metn stream-designator , by interpreting the actions implicit in a .metn format-string , incorporating material pulled from additional arguments given by .cblk .meti << format-arg *. .cble Though the function is simple to invoke, there is complexity in format string language, which is documented below. The .meta stream-designator argument can be a stream object, or one of the values .code t or .codn nil . The value .code t serves as a shorthand for .codn *stdout* . The value .code nil means that the function will send output into a newly instantiated string output stream, and then return the resulting string. .TP* "Format string syntax:" Within .metn format-string , most characters represent themselves. Those characters are simply output. The character .code ~ (tilde) introduces formatting directives, which are denoted by a single character, usually a letter. The special sequence .code ~~ (tilde-tilde) encodes a single tilde. Nothing is permitted between the two tildes. The syntax of a directive is generally as follows: .cblk .mets <> ~[ width ] <> [, precision ] < letter .cble In other words, the .code ~ (tilde) character, followed by a .meta width specifier, a .meta precision specifier introduced by a comma, and a .metn letter , such that .meta width and .meta precision are independently optional: either or both may be omitted. No whitespace is allowed between these elements. The .meta letter is a single alphabetic character which determines the general action of the directive. The optional width and precision are specified as follows: .RS .meIP < width The width specifier consists of an optional .code < (left angle bracket) character, followed by an optional width specification. If the leading .code < character is present, then the printing will be left-adjusted within this field. Otherwise it will be right-adjusted by default. The width can be specified as a decimal integer, or as the character .codn * . The .code * notation means that instead of digits, the value of the next argument is consumed, and expected to be an integer which specifies the width. If that integer value is negative, then the field will be left-adjusted. If the value is positive, but the .code < character is present in the width specifier, then the field is left adjusted. .meIP < precision The precision specifier is introduced by a leading comma. If this comma appears immediately after the directive's .code ~ character, then it means that .meta width is being omitted; there is only a precision field. The precision specifier may begin with these optional characters: .RS .coIP 0 (the "leading zero flag"), .coIP + (print a sign for positive values") .IP space (print a space in place of a positive sign). .RE The precision specifier itself is either a decimal integer that does not begin with a zero digit, or the .code * character. The precision field's components have a meaning which depends on the type of object printed and the conversion specifier. For integer arguments, the precision value specifies the minimum number of digits to print. If the precision field has a leading zero flag, then the integer is padded with zeros to the required number of digits, otherwise the number is padded with spaces instead of zeros. If zero or space padding is present, and a leading positive or negative sign must be printed, then it is placed before leading zeros, or after leading spaces, as the case may be. For floating-point values, the meaning of the precision value depends on which specific conversion specifier .cod1 ( f , .codn e , .code a or .codn s ) is used. The details are documented in the description of each of these, below. The leading zero flag is ignored for floating-point values regardless of the conversion specifier. For integer or floating-point arguments, if the precision specifier has a .code + sign among the special characters, then a .code + sign is printed for positive numbers. If the precision specifier has a leading space instead of a .code + sign, then the .ocde + sign is rendered as a space for positive numbers. If there is no leading space or .codn + , then a sign character is omitted for positive numbers. Negative numbers are unconditionally prefixed with a .code - sign. For all other objects, the precision specifies the maximum number of characters to print. The object's printed representation is crudely truncated at that number of characters. .RE .TP* "Format directives:" .RS Format directives are case sensitive, so that for example .code ~x and .code ~X have a different effect, and .code ~A doesn't exist whereas .code ~a does. They are: .coIP a Prints any object in an aesthetic way, as if by the .code pprint function. The aesthetic notation violates read-print consistency: this notation is not necessarily readable if it is implanted in \*(TX source code. The field width specifier is honored, including the left-right adjustment semantics. When this specifier is used for floating-point values, the precision specifies the maximum number of total significant figures, which do not include any digits in the exponent, if one is printed. Numbers are printed in exponential notation if their magnitude is small, or else if their exponent exceeds their precision. (If the precision is not specified, then it defaults to the system-dependent number of digits in a floating point value, derived from the C language .code DBL_DIG constant.) Floating point values which are integers are printed without a trailing .code .0 (point zero). .coIP s Prints any object in a standard way, as if by the print function. Objects for which read-print consistency is possible are printed in a way such that if their notation is implanted in \*(TX source, they are readable. The field width specifier is honored, including the left-right adjustment semantics. The precision field is treated very similarly to the .code ~a format directive, except that non-exponentiated floating point numbers that would be mistaken for integers include a trailing .code .0 for the sake of read-print consistency. Objects truncated by precision may not have read-print consistency. For instance, if a string object is truncated, it loses its trailing closing quote, so that the resulting representation is no longer a properly formed string object. .coIP x Requires an argument of character or integer type. The integer value or character code is printed in hexadecimal, using lower-case letters for the digits .code a through .codn f . Width and precision semantics are as described for the .code a format directive, for integers. .coIP X Like the .code x directive, but the hexadecimal digits .code a through .code f are rendered in upper case. .coIP o Like the .code x directive, but octal is used instead of hexadecimal. .coIP f The .code f directive prints numbers in a fixed point decimal notation, with a fixed number of digits after the decimal point. It requires a numeric argument. (Unlike .codn x , .code X and .codn o , it does not allow an argument of character type). The precision specifier gives the number of digits past the decimal point. The number is rounded off to the specified precision, if necessary. Furthermore, that many digits are always printed, regardless of the actual precision of the number or its type. If it is omitted, then the default value is three: three digits past the decimal point. A precision of zero means no digits pas the decimal point, and in this case the decimal point is suppressed (regardless of whether the numeric argument is floating-point or integer). .coIP e The .code e directive prints numbers in exponential notation. It requires a numeric argument. (Unlike .codn x , .code X and .codn o , it does not allow an argument of character type). The precision specifier gives the number of digits past the decimal point printed in the exponential notation, not counting the digits in the exponent. Exactly that many digits are printed, regardless of the precision of the number. If the precision is omitted, then the number of digits after the decimal point is three. If the precision is zero, then a decimal portion is truncated off entirely, including the decimal point. .coIP p The .code p directive prints a numeric representation in hexadecimal of the bit pattern of the object, which is meaningful to someone familiar with the internals of \*(TX. If the object is a pointer to heaped data, that value has a correspondence to its address. .RE .PP .coNP Functions @, print @, pprint @, prinl @, pprinl @ tostring and @ tostringp .synb .mets (print < obj <> [ stream ]) .mets (pprint < obj <> [ stream ]) .mets (prinl < obj <> [ stream ]) .mets (pprinl < obj <> [ stream ]) .mets (tostring << obj ) .mets (tostringp << obj ) .syne .desc The .code print and .code pprint functions render a printed character representation of the .meta obj argument into .metn stream . If a stream argument is not supplied, then the destination is the stream currently stored in the .code *stdout* variable. The .code print function renders in a way which strives for read-print consistency: an object is printed in a notation which is recognized as a similar object of the same kind when it appears in \*(TX source code. The .code pprint function ("pretty print") does not strive for read-print consistency. For instance it prints a string object simply by dumping its characters, rather than by adding the surrounding quotes and rendering escape syntax for special characters. Both functions return .metn obj . The .code prinl and .code pprinl functions are like .code print and .codn pprint , except that they issue a newline character after printing the object. These functions also return .metn obj . The .code tostring and .code tostringp functions are like .code print and .codn pprint , but they do not accept a stream argument. Instead they print to a freshly instantiated string stream, and return the resulting string. The following equivalences hold between calls to the .code format function and calls to the above functions: .cblk (format stream "~s" obj) <--> (print obj stream) (format t "~s" obj) <--> (print obj) (format t "~s\en" obj) <--> (prinl obj) (format nil "~s" obj) <--> (tostring obj) .cble For .codn pprint , .code tostringp and .codn pprinl , the equivalence is produced by using .code ~a in format rather than .codn ~s . Note: for characters, the print function behaves as follows: most control characters in the Unicode .code C0 and .code C1 range are rendered using the .code #\ex notation, using two hex digits. Codes in the range .code D800 to .codn DFFF , and the codes .code FFFE and .code FFFF are printed in the .code #\exNNNN with four hexadecimal digits, and character above this range are printed using the same notation, but with six hexadecimal digits. Certain characters in the .code C0 range are printed using their names such as .code #\enul and .codn #\ereturn , which are documented in the Character Literals section. The .code DC00 character is printed as .codn #\epnul . All other characters are printed as .cblk .meti >> #\e char .cble where .meta char is the actual character. Caution: read-print consistency is affected by trailing material. If additional digits are printed immediately after a number without intervening whitespace, they extend that number. If hex digits are printed after the character .codn x , which is rendered as .codn #\ex , they look like a hex character code. .coNP Function @ tprint .synb .mets (tprint < obj <> [ stream ]) .syne .desc The .codn tprint function prints a representation of .meta obj on .metn stream . If the stream argument is not supplied, then the destination is the stream currently stored in the .code *stdout* variable. For all object types except lists and vectors, .code tprint behaves like .codn pprinl . If .code obj is a list or vector, then .code tprint recurses: the .code tprint function is applied to each element. An empty list or vector results in no output at all. This effectively means that an arbitrarily nested structure of lists and vectors is printed flattened, with one element on each line. .coNP Function @ streamp .synb .mets (streamp << obj ) .syne .desc The .code streamp function returns .code t if .meta obj is any type of stream. Otherwise it returns .codn nil . .coNP Function @ real-time-stream-p .synb .mets (real-time-stream-p << obj ) .syne .desc The .code real-time-streamp-p function returns .code t if .meta obj is a stream marked as "real-time". If .meta obj is not a stream, or not a stream marked as "real-time", then it returns .codn nil . Only certain kinds of streams accept the real-time attribute: file streams and tail streams. This attribute controls the semantics of the application of .code lazy-stream-cons to the stream. For a real-time stream, .code lazy-stream-cons returns a stream with "naive" semantics which returns data as soon as it is available, at the cost of generating spurious .code nil item when the stream terminates. The application has to recognize and discard that .code nil item. The ordinary lazy streams read ahead by one line and suppress this extra item, so their representation is more accurate. When \*(TX starts up, it automatically marks the .code *std-input* stream as real-time, if it is connected to a TTY device (a device for which the POSIX function .code isatty reports true). This is only supported on platforms that have this function. The behavior is overridden by the .code -n command line option. .coNP Function @ make-string-input-stream .synb .mets (make-string-input-stream << string ) .syne .desc This function produces an input stream object. Character read operations on the stream object read successive characters from .metn string . Output operations and byte operations are not supported. .coNP Function @ make-string-byte-input-stream .synb .mets (make-string-byte-input-stream << string ) .syne .desc This function produces an input stream object. Byte read operations on this stream object read successive byte values obtained by encoding .meta string into UTF-8. Character read operations are not supported, and neither are output operations. .coNP Function @ make-string-output-stream .synb (make-string-output-stream) .syne .desc This function, which takes no arguments, creates a string output stream. Data sent to this stream is accumulated into a string object. String output streams supports both character and byte output operations. Bytes are assumed to represent a UTF-8 encoding, and are decoded in order to form characters which are stored into the string. If an incomplete UTF-8 code is output, and a character output operation then takes place, that code is assumed to be terminated and is decoded as invalid bytes. The UTF-8 decoding machine is reset and ready for the start of a new code. The .code get-string-from-stream function is used to retrieve the accumulated string. If the null character is written to a string output stream, either via a character output operation or as a byte operation, the resulting string will appear to be prematurely terminated. \*(TX strings cannot contain null bytes. .coNP Function @ get-string-from-stream .synb .mets (get-string-from-stream << stream ) .syne .desc The .meta stream argument must be a string output stream. This function finalizes the data sent to the stream and retrieves the accumulated character string. If a partial UTF-8 code has been written to .metn stream , and then this function is called, the byte stream is considered complete and the partial code is decoded as invalid bytes. After this function is called, further output on the stream is not possible. .coNP Function @ make-strlist-output-stream .synb (make-strlist-output-stream) .syne .desc This function is very similar to .codn make-string-output-stream . However, the stream object produced by this function does not produce a string, but a list of strings. The data is broken into multiple strings by newline characters written to the stream. Newline characters do not appear in the string list. Also, byte output operations are not supported. .coNP Function @ get-list-from-stream .synb .mets (get-list-from-stream << stream ) .syne .desc This function returns the string list which has accumulated inside a string output stream given by .metn stream . The string output stream is finalized, so that further output is no longer possible. .coNP Function @ close-stream .synb .mets (close-stream < stream <> [ throw-on-error-p ]) .syne .desc The .code close-stream function performs a close operation on .metn stream , whose meaning is depends on the type of the stream. For some types of streams, such as string streams, it does nothing. For streams which are connected to operating system files or devices, will perform a close of the underlying file descriptor, and dissociate that descriptor from the stream. Any buffered data is flushed first. .code close-stream returns a boolean true value if the close has occurred without errors, otherwise .codn nil . For most streams, "without errors" means that any buffered output data is flushed successfully. For command and process pipes (see open-command and open-process), success also means that the process terminates normally, with a successful error code, or an unsuccessful one. An abnormal termination is considered an error, as as is the inability to retrieve the termination status, as well as the situation that the process continues running in spite of the close attempt. Detecting these situations is platform specific. If the .meta throw-on-error-p argument is specified, and isn't .codn nil , then the function throws an exception if an error occurs during the close operation instead of returning .codn nil . .coNP Functions @, get-error @ get-error-str and @ clear-error .synb .mets (get-error << stream ) .mets (get-error-str << stream ) .mets (clear-error << stream ) .syne .desc When a stream operation fails, the .code get-error and .code get-error-str functions may be used to inquire about a more detailed cause of the error. Not all streams support these functions to the same extent. For instance, string input streams have no persistent state. The only error which occurs is the condition when the string has no more data. The .code get-error inquires .meta stream about its error condition. The function returns .code nil to indicate there is no error condition, .code t to indicate an end-of-data condition, or else a value which is specific to the stream type indicating the specific error type. Note: for some streams, it is possible for the .code t value to be returned even though no operation has failed; that is to say, the streams "know" they are at the end of the data even though no read operation has failed. Code which depends on this will not work with streams which do not thus indicate the end-of-data .I a priori, but by means of a read operation which fails. The .code get-error-str function returns a text representation of the error code. The .code nil error code is represented as the string .codn "no error" ; the .code t error code as .code "eof" and other codes have a stream-specific representation. The .code clear-error function removes the error situation from a stream. On some streams, it does nothing. If an error has occurred on a stream, this function should be called prior to re-trying any I/O or positioning operations. The return value is the previous error code, or .code nil if there was no error, or the operation is not supported on the stream. .coNP Functions @, get-line @ get-char and @ get-byte .synb .mets (get-line <> [ stream ]) .mets (get-char <> [ stream ]) .mets (get-byte <> [ stream ]) .syne .desc These fundamental stream functions perform input. The .meta stream argument is optional. If it is specified, it should be an input stream which supports the given operation. If it is not specified, then the .code *stdin* stream is used. The .code get-char function pulls a character from a stream which supports character input. Streams which support character input also support the .code get-line function which extracts a line of text delimited by the end of the stream or a newline character and returns it as a string. (The newline character does not appear in the string which is returned). Character input from streams based on bytes requires UTF-8 decoding, so that get-char actually may read several bytes from the underlying low level operating system stream. The .code get-byte function bypasses UTF-8 decoding and reads raw bytes from any stream which supports byte input. Bytes are represented as integer values in the range 0 to 255. Note that if a stream supports both byte input and character input, then mixing the two operations will interfere with the UTF-8 decoding. These functions return .code nil when the end of data is reached. Errors are represented as exceptions. See also: .code get-lines .coNP Function @ get-string .synb .mets (get-string >> [ stream >> [ count <> [ close-after-p ]]]) .syne .desc The .code get-string function reads characters from a stream, and assembles them into a string, which is returned. If the .meta stream argument is omitted, then the .code *stdin* stream is used. The stream is closed after extracting the data, unless .meta close-after-p is specified as .codn nil . The default value of this argument is .codn t . If the .meta count argument is missing, then all of the characters from the stream are read and assembled into a string. If present, the .meta count argument should be a positive integer indicating a limit on how many characters to read. The returned string will be no longer than .metn count , but may be shorter. .coNP Functions @ unget-char and @ unget-byte .synb .mets (unget-char < char <> [ stream ]) .mets (unget-byte < byte <> [ stream ]) .syne .desc These functions put back, into a stream, a character or byte which was previously read. The character or byte must match the one which was most recently read. If the .meta stream argument is omitted, then the .code *stdin* stream is used. If the operation succeeds, the byte or character value is returned. A .code nil return indicates that the operation is unsupported. Some streams do not support these operations; some support only one of them. In general, if a stream supports .codn get-char , it supports .codn unget-char , and likewise for .code get-byte and .codn unget-byte . Space is available for only one character or byte of pushback. Pushing both a byte and a character, in either order, is also unsupported. Pushing a byte and then reading a character, or pushing a character and reading a byte, are unsupported mixtures of operations. If the stream is binary, then pushing back a byte decrements its position, except if the position is already zero. At that point, the position becomes indeterminate. .coNP Functions @, put-string @, put-line @ put-char and @ put-byte .synb .mets (put-string < string <> [ stream ]) .mets (put-line >> [ string <> [ stream ]]) .mets (put-char < char <> [ stream ]) .mets (put-byte < byte <> [ stream ]) .syne .desc These functions perform output on an output stream. The .meta stream argument must be an output stream which supports the given operation. If it is omitted, then .code *stdout* is used. The .code put-char function writes a character given by .code char to a stream. If the stream is based on bytes, then the character is encoded into UTF-8 and multiple bytes are written. Streams which support .code put-char also support put-line, and .codn put-string . The .code put-string function writes the characters of a string out to the stream as if by multiple calls to put-char. The .meta string argument may be a symbol, in which case its name is used as the string. The .code put-line function is like .codn put-string , but also writes an additional newline character. The string is optional in .codn put-line , and defaults to the empty string. The .code put-byte function writes a raw byte given by the .meta byte argument to .metn stream , if .meta stream supports a byte write operation. The byte value is specified as an integer value in the range 0 to 255. .coNP Functions @ put-strings and @ put-lines .synb .mets (put-strings < sequence <> [ stream ]]) .mets (put-lines < sequence <> [ stream ]]) .syne .desc These functions assume .meta sequence to be a sequence of strings, or of symbols, or a mixture thereof. These strings are sent to the stream. The .meta stream argument must be an output stream. If it is omitted, then .code *stdout* is used. The .code put-strings function iterates over .meta sequence and writes each element to the stream as if using the .code put-string function. The .code put-lines function iterates over .code sequence and writes each element to the stream as if using the .code put-line function. Both functions return .codn t . .coNP Function @ flush-stream .synb .mets (flush-stream << stream ) .syne .desc This function is meaningful for output streams which accumulate data which is passed on to the operating system in larger transfer units. Calling .code flush-stream causes all accumulated data inside .meta stream to be passed to the operating system. If called on streams for which this function is not meaningful, it does nothing, and returns .codn nil . .coNP Function @ seek-stream .synb .mets (seek-stream < stream < offset << whence ) .syne .desc The .code seek-stream function is meaningful for file streams. It changes the current read/write position within .metn stream . It can also be used to determine the current position: see the notes about the return value below. The .meta offset argument is a positive or negative integer which gives a displacement that is measured from the point identified by the .meta whence argument. Note that for text files, there isn't necessarily a 1:1 correspondence between characters and positions due to line-ending conversions and conversions to and from UTF-8. The .meta whence argument is one of three keywords: :from-start, :from-current and :from-end. These denote the start of the file, the current position and the end of the file. If .meta offset is zero, and .meta whence is .codn :from-current , then .code seek-stream returns the current absolute position within the stream, if it can successfully obtain it. Otherwise, it returns .code t if it is successful. If a character has been successfully put back into a text stream with .code unget-char and is still pending, then the position value is unspecified. If a byte has been put back into a binary stream with .codn unget-byte , and the previous position wasn't zero, then the position is decremented by one. On failure, it throws an exception of type .codn stream-error . .coNP Functions @ stream-get-prop and @ stream-set-prop .synb .mets (stream-get-prop < stream << indicator ) .mets (stream-set-prop < stream < indicator << value ) .syne .desc These functions get and set properties on a stream. Only certain properties are meaningful with certain kinds of streams, and the meaning depends on the stream. If two or more stream types support a property of the same name, it is expected that the property has the same or very similar meaning for both streams to the maximum extent that similarity is possible. The .code stream-set-prop function sets a property on a stream. The .meta indicator argument is a symbol, usually a keyword symbol, denoting the property, and .meta value is the property value. If the stream understands and accepts the property, the function returns .codn t . Otherwise it returns .codn nil . The .code stream-get-prop function inquires about the value of a property on a stream. If the stream understands the property, then it returns its current value. If the stream does not understand a property, nil is returned, which is also returned if the property exists, but its value happens to be .codn nil . Properties are currently used for marking certain streams as "real-time" (see the .code real-time-stream-p function above), and also for setting the priority at which messages are reported to syslog by the .code *stdlog* stream (see .code *stdlog* in the UNIX SYSLOG section). If .meta stream is a catenated stream (see the function .codn make-catenated-stream ) then these functions transparently operate on the current head stream of the catenation. .coNP Functions @ make-catenated-stream and @ cat-streams .synb .mets (make-catenated-stream << stream *) .mets (cat-streams << stream-list ) .syne .desc The .code make-catenated-stream function takes zero or more arguments which are input streams of the same type, and combines them into a single virtual stream called a catenated stream. The .code cat-streams function takes a single list of input streams of the same type, and similarly combines them into a catenated stream. A catenated stream does not support seeking operations or output, regardless of the capabilities of the streams in the list. If the stream list is not empty, then the leftmost element of the list is called the head stream. The .codn get-char , .codn get-byte , .codn get-line , .code unget-char and .code unget-byte functions delegate to the corresponding operations on the head stream, if it exists. If the stream list is empty, they return .code nil to the caller. If the .codn get-char , .code get-byte or .code get-line operation on the head stream yields .codn nil , and there are more lists in the stream, then the stream is closed, removed from the list, and the next stream, if any, becomes the head list. The operation is then tried again. If any of these operations fail on the last list, it is not removed from the list, so that a stream remains in place which can take the .code unget-char or .code unget-byte operations. In this manner, the catenated streams appear to be a single stream. Note that the operations can fail due to being unsupported. It is the caller's responsibility to make sure all of the streams in the list are compatible with the intended operations. If the stream list is empty then an empty catenated stream is produced. Input operations on this stream yield .codn nil , and the .code unget-char and .code unget-byte operations throw an exception. .coNP Function @ catenated-stream-p .synb .mets (catenated-stream-p << obj ) .syne .desc The .code catenated-stream-p function returns .code t if .meta obj is a catenated stream. Otherwise it returns .codn nil . .coNP Function @ catenated-stream-push .synb .mets (catenated-stream-push < new-stream << cat-stream ) .syne .desc The .code catenated-stream-push function pushes .meta new-stream to the front of the stream list inside .metn cat-stream . If an .code unget-byte or .code unget-char operation was successfully performed on .meta cat-stream previously to a call to .codn catenated-stream-push , those operations were forwarded to the front stream. If those bytes or characters are still pending, they are pending inside that stream, and thus are logically preceded by the contents of .metn new-stream . .coNP Functions @ open-files and @ open-files* .synb .mets (open-files < path-list <> [ alternative-stream ]) .mets (open-files* < path-list <> [ alternative-stream ]) .syne .desc The .code open-files and .code open-files* functions create a list of streams by invoking the open-file function on each element of .metn path-list . These streams are turned into a catenated stream as if applied as arguments to .codn make-catenated-stream . The effect is that multiple files appear to be catenated together into a single input stream. If the optional .meta alternative-stream argument is supplied, then if .meta path-list is empty, .meta alternative-stream is returned instead of an empty catenated stream. The difference between .code open-files and .code open-files* is that .code open-files creates all of the streams up-front. So if any of the paths cannot be opened, the operation throws. The .code open-files* variant is lazy: it creates a lazy list of streams out of the path list. The streams are opened as needed: before the second stream is opened, the program has to read the first stream to the end, and so on. .TP* Example: Collect lines from all files that are given as arguments on the command line. If there are no files, then read from standard input: .cblk @(next @(open-files *args* *stdin*)) @(collect) @line @(end) .cble .coNP Function @ abs-path-p .synb .mets (abs-path-p << path ) .syne .desc The .code abs-path-function tests whether the argument .meta path is an absolute path, returning a .code t or .code nil indication. An absolute path is a string which either begins with a slash or backslash character, or which begins with an alphanumeric word, followed by a colon, followed by a slash or backslash. Examples of absolute paths: .cblk /etc c:/tmp ftp://user@server disk0:/home Z:\eUsers .cble Examples of strings which are not absolute paths. .cblk .mets < (the < empty < string) . abc foo:bar/x $:\eabc .cble .coNP Function @ read .synb .mets (read >> [ source >> [ error-stream >> [ error-return-value <> [ name ]]]]) .syne .desc The .code read function converts text denoting \*(TL structure, into the corresponding data structure. The .meta source argument may be either a character string, or a stream. If it is omitted, then .code *stdin* is used as the stream. The source must provide the text representation of one complete \*(TL object. Multiple calls to read on the same stream will extract successive objects from the stream. To parse successive objects from a string, it is necessary to convert it to a string stream. The optional .meta error-stream argument can be used to specify a stream to which parse errors diagnostics are sent. If absent, the diagnostics are suppressed. The optional .meta name argument can be used to specify the file name which is used for reporting errors. If this argument is missing, the name is taken from the name property of the .meta source argument if it is a stream, or else the word .code string is used as the name if .meta source is a string. If there are no parse errors, the function returns the parsed data structure. If there are parse errors, and the .meta error-return-value parameter is present, its value is returned. If the .meta error-return-value parameter is not present, then an exception of type .code syntax-error is thrown. .SS* Filesystem Access .coNP Function @ open-directory .synb .mets (open-directory << path ) .syne .desc The .code open-directory function tries to create a stream which reads the directory given by the string argument .metn path . If a filesystem object exists under the path, is accessible, and is a directory, then the function returns a stream. Otherwise, a file error exception is thrown. The resulting stream supports the get-line operation. Each call to the .code get-line operation retrieves a string representing the next directory entry. The value .code nil is returned when there are no more directory entries. The .code . and .code .. entries in Unix filesystems are not skipped. .coNP Function @ open-file .synb .mets (open-file < path <> [ mode-string ]) .syne .desc The .code open-file function creates a stream connected to the file which is located at the given .metn path , which is a string. The .meta mode-string argument is a string which uses the same conventions as the mode argument of the C language .code fopen function, with some extensions. The mode string determines whether the stream is an input stream or output stream. Note that the .str b mode is passed through to the C library, but has no special meaning to \*(TX. Whether a stream is text or binary depends on which operations are invoked on it. If the .meta mode-string argument is omitted, mode .str r is used. The option letter .str i is supported. If present, it will create a stream which has the real-time property set. .coNP Function @ open-tail .synb .mets (open-tail < path >> [ mode-string <> [ seek-to-end-p ]]) .syne .desc The .code open-tail function creates a tail stream connected to the file which is located at the given .metn path . The .meta mode-string argument is a string which uses the same conventions as the mode argument of the C language .code fopen function. If this argument is omitted, then .str r is used. See the .code open-file function for a discussion of modes. The .code seek-to-end-p argument is a boolean which determines whether the initial read/write position is at the start of the file, or just past the end. It defaults to .codn nil . This argument only makes a difference if the file exists at the time .code open-tail is called. If the file does not exist, and is later created, then the tail stream will follow that file from the beginning. In other words, .meta seek-to-end-p controls whether the tail stream reads all the existing data in the file, if any, or whether it reads only newly added data from approximately the time the stream is created. A tail stream has special semantics with regard to reading at the end of file. A tail stream never reports an end-of-file condition; instead it polls the file until more data is added. Furthermore, if the file is truncated, or replaced with a smaller file, the tail stream follows this change: it automatically opens the smaller file and starts reading from the beginning (the .meta seek-to-end-p flag only applies to the initial open). In this manner, a tail stream can dynamically growing rotating log files. Caveat: since a tail stream can re-open a new file which has the same name as the original file, it behave incorrectly if the program changes the current working directory, and the path name is relative. .coNP Function @ remove-path .synb .mets (remove-path << path ) .syne .desc The .code remove-path function tries to remove the filesystem object named by .metn path , which may be a file, directory or something else. If successful, it returns .codn t . A failure to remove the object results in an exception of type .codn file-error . .coNP Function @ rename-path .synb .mets (rename-path < from-path << to-path ) .syne .desc The .code remove-path function tries to rename filesystem path .metn from-path , which may refer to a file, directory or something else, to the path .metn to-path . If successful, it returns .codn t . A failure results in an exception of type .codn file-error . .SS* Coprocesses .coNP Functions @ open-command and @ open-process .synb .mets (open-command < system-command <> [ mode-string ]) .mets (open-process < program < mode-string <> [ argument-list ]) .syne .desc These functions spawn external programs which execute concurrently with the \*(TX program. Both functions return a unidirectional stream for communicating with these programs: either an output stream, or an input stream, depending on the contents of .metn mode-string . In .codn open-command , the .meta mode-string argument is optional, defaulting to the value .str r if it is missing. See the .code open-file function for a discussion of modes. The .code open-command function accepts, via the .meta system-command string parameter, a system command, which is in a system-dependent syntax. On a POSIX system, this would be in the POSIX Shell Command Language. The .code open-process function specifies a program to invoke via the .meta command argument. This is subject to the operating system's search strategy. On POSIX systems, if it is an absolute or relative path, it is treated as such, but if it is a simple base name, then it is subject to searching via the components of the PATH environment variable. If open-process is not able to find .metn program , or is otherwise unable to execute the program, the child process will exit, using the value of the C variable .code errno as its exit status. This value can be retrieved via .codn close-stream . The .meta mode-string argument follows the convention used by the POSIX .code popen function. The .meta argument-list argument is a list of strings which specifies additional optional arguments to be passed passed to the program. The .meta program argument becomes the first argument, and .meta argument-string become the second and subsequent arguments. If .meta argument-strings is omitted, it defaults to empty. If a coprocess is open for writing .cblk .meti >> ( mode-string .cble is specified as .strn w ), then writing on the returned stream feeds input to that program's standard input file descriptor. Indicating the end of input is performed by closing the stream. If a coprocess is open for reading .cblk .meti >> ( mode-string .cble is specified as .strn r ), then the program's output can be gathered by reading from the returned stream. When the program finishes output, it will close the stream, which can be detected as normal end of data. If a coprocess terminates abnormally or unsuccessfully, an exception is raised. .SS* Symbols and Packages A package is an object which serves as a container of symbols. A symbol which exists inside a package is said to be interned in that package. A symbol can be interned in at most one package at a time. Each symbol has a name, which is a string. It is not necessarily unique: two distinct symbols can have the same name. However, a symbol name is unique within a package, because it serves as the key which associates the symbol with the package. Two symbols cannot be in the same package if they have the same name. Moreover, a symbol cannot exist in more than one package at at time, although it can be relocated from one package to another. A symbols exist which is not entered into any package: such a symbol is called "uninterned". Packages are held in a global list which can be used to search for a package by name. The .code find-package function performs this lookup. A package may be deleted from the list with the .code delete-package function, but it continues to exist until the program loses the last reference to that package. .coNP Special variables @, *user-package* @, *keyword-package* and @ *system-package* .desc These variables hold predefined packages. The .code *user-package* is the one in which symbols are read when a \*(TX program is being scanned. The .code *keyword-package* holds keyword symbols, which are printed with a leading colon. The .code *system-package* is for internal symbols, helping the implementation avoid name clashes with user code in some situations. .coNP Function @ make-sym .synb .mets (make-sym << name ) .syne .desc The .code make-sym function creates and returns a new symbol object. The argument .metn name , which must be a string, specifies the name of the symbol. The symbol does not belong to any package (it is said to be "uninterned"). Note: an uninterned symbol can be interned into a package with the .code rehome-sym function. Also see the .code intern function. .coNP Function @ gensym .synb .mets (gensym <> [ prefix ]) .syne .desc The .code gensym function is similar to make-sym. It creates and returns a new symbol object. If the .meta prefix argument is omitted, it defaults to .strn g . Otherwise it must be a string. The difference between .code gensym and .code make-sym is that .code gensym creates the name by combining the prefix with a numeric suffix. The numeric suffix is a decimal digit string, taken from the value of the variable .codn *gensym-counter* , after incrementing it. Note: the variation in name is not the basis of the uniqueness assurance offered by .code make-sym and .codn gensym ; the basis is that the returned symbol is a freshly instantiated object. .code make-sym still returns unique symbols even if repeatedly called with the same string. .coNP Special variable @ *gensym-counter* .desc This variable is initialized to 0. Each time the .code gensym function is called, it is incremented. The incremented value forms the basis of the numeric suffix which gensym uses to form the name of the new symbol. .coNP Function @ make-package .synb .mets (make-package << name ) .syne .desc The .code make-package function creates and returns a package named .metn name , where .meta name is a string. It is an error if a package by that name exists already. .coNP Function @ packagep .synb .mets (packagep << obj ) .syne .desc The .code packagep function returns .codet if .meta obj is a package, otherwise it returns .codn nil . .coNP Function @ find-package .synb .mets (find-package << name ) .syne .desc The argument .meta name should be a string. If a package called .meta name exists, then it is returned. Otherwise .code nil is returned. .coNP Function @ intern .synb .mets (intern < name <> [ package ]) .syne .desc The argument .meta name should be a symbol. The optional argument .meta package should be a package. If .meta package is not supplied, then the value taken is that of .codn *user-package* . The .code intern function searches .meta package for a symbol called .metn name . If that symbol is found, it is returned. If that symbol is not found, then a new symbol called .meta name is created and inserted into .metn package , and that symbol is returned. In this case, the package becomes the symbol's home package. .coNP Function @ rehome-sym .synb .mets (rehome-sym < symbol <> [ package ]) .syne .desc The arguments .meta symbol and .meta package must be a symbol and package object, respectively. If .meta package is not given, then it defaults to the value of .codn *user-package* . The .code rehome-sym function moves .meta symbol into .metn package . If .meta symbol is already in a package, it is first removed from that package. If a symbol of the same name exists in .meta package that symbol is first removed from .metn package . .coNP Function @ symbolp .synb .mets (symbolp << obj ) .syne .desc The .code symbolp function returns .code t if .meta obj is a symbol, otherwise it returns .codn nil . .coNP Function @ symbol-name .synb .mets (symbol-name << symbol ) .syne .desc The .code symbol-name function returns the name of .metn symbol . .coNP Function @ symbol-package .synb .mets (symbol-package << symbol ) .syne .desc The .code symbol-package function returns the home package of .metn symbol . .coNP Function @ packagep .synb .mets (packagep << obj ) .syne .desc The .code packagep function returns .code t if .meta obj is a package, otherwise it returns .codn nil . .coNP Function @ keywordp .synb .mets (keywordp << obj ) .syne .desc The .code keywordp function returns .code t if .meta obj is a keyword symbol, otherwise it returns .codn nil . .coNP Function @ bindable .synb .mets (bindable << obj ) .syne .desc The .code bindable function returns .code t if .meta obj is a bindable symbol, otherwise it returns .codn nil . All symbols are bindable, except for keyword symbols, and the special symbols .code t and .codn nil. .SS* Pseudo-random Numbers .coNP Special variable @ *random-state* .desc The .code *random-state* variable holds an object which encapsulates the state of a pseudo-random number generator. This variable is the default argument value for the .code random-fixnum and .codn random functions , for the convenience of writing programs which are not concerned about the management of random state. On the other hand, programs can create and manage random states, making it possible to obtain repeatable sequences of pseudo-random numbers which do not interfere with each other. For instance objects or modules in a program can have their own independent streams of random numbers which are repeatable, independently of other modules making calls to the random number functions. When \*(TX starts up, the .code *random-state* variable is initialized with a newly created random state object, which is produced as if by the call .codn (make-random-state 42) . .coNP Function @ make-random-state .synb .mets (make-random-state <> [ seed ]) .syne .desc The .code make-random-state function creates and returns a new random state, an object of the same kind as what is stored in the .code *random-state* variable. The seed, if specified, must be either an integer value, or an existing random state object. Note that the sign of the seed is ignored, so that negative seed values are equivalent to their additive inverses. If the seed is not specified, then .code make-random-state produces a seed based on some information in the process environment, such as current time of day. It is not guaranteed that two calls to .code (make-random-state) that are separated by less than some minimum increment of real time produce different seeds. The minimum time increment depends on the platform. On a platform with a millisecond-resolution real-time clock, the minimum time increment is a millisecond. Calls to make-random-state less than a millisecond apart may predictably produce the same seed. If an integer seed is specified, then the integer value is mapped to a pseudo-random sequence, in a platform-independent way. If a random state is specified as a seed, then it is duplicated. The returned random state object is a distinct object which is in the same state as the input object. It will produce the same remaining pseudo-random number sequence, as will the input object. .coNP Function @ random-state-p .synb .mets (random-state-p << obj ) .syne .desc The .code random-state-p function returns .code t if .meta obj is a random state, otherwise it returns .codn nil . .coNP Functions @, random-fixnum @ random and @ rand .synb .mets (random-fixnum <> [ random-state ]) .mets (random < random-state << modulus ) .mets (rand < modulus <> [ random-state ]) .syne .desc All three functions produce pseudo-random numbers, which are positive integers. The numbers are obtained from a WELLS 512 pseudo-random number generator, whose state is stored in the random state object. The .code random-fixnum function produces a random fixnum integer: a reduced range integer which fits into a value that does not have to be heap-allocated. The .code random and .code rand functions produce a value in the range [0, .metn modulus ). They differ only in the order of arguments. In the .code rand function, the random state object is the second argument and is optional. If it is omitted, the global .code *random-state* object is used. .SS* Time .coNP Functions @ time and @ time-usec .synb (time) (time-usec) .syne .desc The .code time function returns the number of seconds that have elapsed since midnight, January 1, 1970, in the UTC timezone. The .code time-usec function returns a cons cell whose .code car field holds the seconds measured in the same way, and whose .code cdr field extends the precision by giving number of microseconds as an integer value between 0 and 999999. .coNP Functions @ time-string-local and @ time-string-utc .synb .mets (time-string-local < time << format ) .mets (time-string-utc < time << format ) .syne .desc These functions take the numeric time returned by the .code time function, and convert it to a textual representation in a flexible way, according to the contents of the .meta format string. The .code time-string-local function converts the time to the local timezone of the host system. The .code time-string-utc function produces time in UTC. The .meta format argument is a string, and follows exactly the same conventions as the format string of the C library function .codn strftime . The .meta time argument is an integer representing seconds obtained from the time function or from the .code car field of the cons returned by the .code time-usec function. .coNP Functions @ time-fields-local and @ time-fields-utc .synb .mets (time-fields-local << time ) .mets (time-fields-utc << time ) .syne .desc These functions take the numeric time returned by the time function, and convert it to a list of seven fields. The .code time-string-local function converts the time to the local timezone of the host system. The .code time-string-utc function produces time in UTC. The fields returned as a list consist of six integers, and a boolean value. The six integers represent the year, month, day, hour, minute and second. The boolean value indicates whether daylight savings time is in effect (always .code nil in the case of .codn time-fields-utc ). The .meta time argument is an integer representing seconds obtained from the .code time function or from the .code time-usec function. .coNP Functions @ make-time and @ make-time-utc .synb .mets (make-time < year < month < day < hour < minute < second << dst-advice ) .mets (make-time-utc < year < month < day < hour < minute < second << dst-advice ) .syne .desc The .code make-time function returns a time value, similar to the one returned by the .code time function. The .code time value is constructed not from the system clock, but from a date and time specified as arguments. The .meta year argument is a calendar year, like 2014. The .meta month argument ranges from 1 to 12. The .meta hour argument is a 24-hour time, ranging from 0 to 23. These arguments represent a local time, in the current time zone. The .meta dst-advice argument specifies whether the time is expressed in daylight savings time (DST). It takes on three possible values: .codn nil , the keyword .codn :auto , or else the symbol .codn t . Any other value has the same interpretation as .codn t . If .meta dst-advice is .codn t , then the time is assumed to be expressed in DST. If the argument is .codn nil , then the time is assumed not to be in DST. If .meta dst-advice is .codn :auto , then the function tries to determine whether DST is in effect in the current time zone for the specified date and time. The .code make-time-utc function is similar to .codn make-time , except that it treats the time as UTC rather than in the local time zone. The .meta dst-advice argument is supported by .code make-time-utc for function call compatibility with .codn make-time . It may or may not have any effect on the output (since the UTC zone by definition doesn't have daylight savings time). .SS* Environment Variables and Command Line Note that environment variable names, their values, and command line arguments are all regarded as being externally encoded in UTF-8. \*(TX performs the encoding and decoding automatically. .coNP Special variables @ *args* and @ *args-full* .desc The .code *args* variable holds a list of strings representing the remaining arguments which follow any options processed by the \*(TX executable, and the script name. The .code *args-full* variable holds the original, complete list of arguments passed from the operating system. Note: the .code *args* variable is .code nil during the processing of the command line, so \*(TL expressions invoked using the .code -p or .code -e option cannot use it. .coNP Function @ env .synb (env) .syne .desc The .code env function retrieves the list of environment variables. Each variable is represented by a single entry in the list: a string which contains an .code = (equal) character somewhere, separating the variable name from its value. See also: the .code env-hash function. .coNP Function @ env-hash .synb (env-hash) .syne .desc The .code env-hash function constructs and returns an .code :equal-based hash. The hash is populated with the environment variables, represented as key-value pairs. .coNP Functions @, getenv @, setenv and @ unsetenv .synb .mets (getenv << name ) .mets (setenv < name < value <> [ overwrite-p ]) .mets (unsetenv << name ) .syne .desc These functions provide access to, as well as manipulation of, environment variables. Of these three, .code setenv and .code unsetenv might not be available on some platforms, or .code unsetenv might be be present in a simulated form which sets the variable .meta name to the empty string rather than deleting it. The .code getenv function searches the environment for the environment variable whose name is .metn name . If the variable is found, its value is returned. Otherwise .code nil is returned. The .code setenv function creates or modifies the environment variable indicated by .metn name . The .meta value string argument specifies the new value for the variable. If the .meta overwrite-p argument is specified, and is true, then the variable is overwritten if it already exists. If the argument is false, then the variable is not modified if it already exists. If the argument is not specified, it defaults to the value .metn t, effectively giving rise to a two-argument form of .code setenv which creates or overwrites environment variables. The .code setenv function unconditionally returns .meta value regardless of whether or not it overwrites an existing variable. The .code unsetenv function removes the environment variable specified by .metn name , if it exists. On some platforms, it instead sets the environment variable to the empty string. .SS* System Programming .coNP Accessor @ errno .synb .mets (errno <> [ new-errno ]) .mets (set (errno) << new-value ) .syne .desc The .code errno function retrieves the current value of the C library error variable .codn errno . If the argument .meta new-errno is present and is not .codn nil , then it specifies a value which is stored into .codn errno . The value returned is the prior value. The place form of .code errno does not take an argument. .coNP Function @ exit .synb .mets (exit << status ) .syne .desc The .code exit function terminates the entire process (running \*(TX image), specifying the termination status to the operating system. Values of .meta status may be .codn nil , .codn t , or an integer value. The value .code nil corresponds to the C constant .codn EXIT_FAILURE , and .code t corresponds to .codn EXIT_SUCCESS . These are platform-independent indicators of failed or successful termination. The numeric value 0 also indicates success. .coNP Function @ abort .synb .mets (abort) .syne .desc The .code abort function terminates the entire process (running \*(TX image), specifying an abnormal termination status to the process. Note: .code abort calls the C library function .code abort which works by raising the .code SIG_ABRT signal, known in \*(TX as the .code sig-abrt variable. Abnormal termination of the process is this signal's default action. .coNP Function @ usleep .synb .mets (usleep << usec ) .syne .desc The .code usleep function suspends the execution of the program for at least .meta usec microseconds. The return value is .code t if the sleep was successfully executed. A .code nil value indicates premature wakeup or complete failure. Note: the actual sleep resolution is not guaranteed, and depends on granularity of the system timer. Actual sleep times may be rounded up to the nearest 10 millisecond multiple on a system where timed suspensions are triggered by a 100 Hz tick. .coNP Functions @ mkdir and @ ensure-dir .synb .mets (mkdir < path <> [ mode ]) .mets (ensure-dir < path <> [ mode ]) .syne .desc .code mkdir tries to create the directory named .meta path using the POSIX .code mkdir function. An exception of type .code file-error is thrown if the function fails. Returns .code t on success. The .meta mode argument specifies the request numeric permissions for the newly created directory. If omitted, the requested permissions are .code #o777 (511): readable and writable to everyone. The requested permissions are subject to the system .codn umask . The function .code ensure-dir is similar to .code mkdir except that it attempts to create all the missing parent directories as necessary, and does not throw an error if the directory exists. .coNP Function @ chdir .synb .mets (chdir << path ) .syne .desc .code chdir changes the current working directory to .metn path , and returns .metn t , or else throws an exception of type .codn file-error . .coNP Function @ pwd .synb (pwd) .syne .desc The .code pwd function retrieves the current working directory. If the underlying .code getcwd C library function fails with an .code errno other than .codn ERANGE , an exception will be thrown. .coNP Functions @ sh and @ run .synb .mets (sh << system-command ) .mets (run < program <> [ argument-list ]) .syne .desc The .code sh function executes .meta system-command using the system command interpreter. The run function spawns a .metn program , searching for it using the system PATH. Using either method, the executed process receives environment variables from the parent. \*(TX blocks until the process finishes executing. If the program terminates normally, then its integer exit status is returned. The value zero indicates successful termination. The return value .code nil indicates an abnormal termination, or the inability to run the process at all. In the case of the .code run function, if the child process is created successfully but the program cannot be executed, then the exit status will be an .code errno value from the failed .code exec attempt. .SS* Unix Filesystem Manipulation .coNP Function @ stat .synb .mets (stat << path ) .syne .desc The .code stat function inquires the filesystem about the existence of an object denoted by the string .metn path . If the object is not found or cannot be accessed, an exception is thrown. Otherwise, information is retrieved about the object. The information takes the form of a property list in which keyword symbols denote numerous properties. An example such property list is: .cblk (:dev 2049 :ino 669944 :mode 16832 :nlink 23 :uid 500 :gid 500 :rdev 0 :size 12288 :blksize 4096 :blocks 24 :atime 1347933533 :mtime 1347933534 :ctime 1347933534) .cble These properties correspond to the similarly-named entries of the .code struct stat structure in POSIX. For instance, the .code :dev property has the same value as the .code st_dev field. .coNP Special variables @, s-ifmt @, s-iflnk @, s-ifreg @, s-ifblk ... , @ s-ixoth The following variables exist, having integer values. These are bitmasks which can be applied against the value given by the .code :mode property in the property list returned by the function stat: .codn s-ifmt , .codn s-ifsock , .codn s-iflnk , .codn s-ifreg , .codn s-ifblk , .codn s-ifdir , .codn s-ifchr , .codn s-ififo , .codn s-isuid , .codn s-isgid , .codn s-isvtx , .codn s-irwxu , .codn s-irusr , .codn s-iwusr , .codn s-ixusr , .codn s-irwxg , .codn s-irgrp , .codn s-iwgrp , .codn s-ixgrp , .codn s-irwxo , .codn s-iroth , .code s-iwoth and .codn s-ixoth . These variables correspond to the C language constants from POSIX: .codn S_IFMT , .codn S_IFLNK , .code S_IFREG and so forth. The .code logtest function can be used to test these against values of mode. For example .code (logtest mode s-irgrp) tests for the group read permission. .coNP Functions @, makedev @ minor and @ major .synb .mets (makedev < minor << major ) .mets (minor << dev ) .mets (major << dev ) .syne .desc The parameters .metn minor , .meta major and .meta dev are all integers. The .code makedev function constructs a combined device number from a minor and major pair (by calling the Unix .code makedev function). This device number is suitable as an argument to the .codee mknod function (see below). Device numbers also appear the .code :dev property returned by the .code stat function. The .code minor and .code major functions extract the minor and major device number from a combined device number. .coNP Function @ mknod .synb .mets (chmod < path << mode ) .syne .desc The .code chmod function changes the permissions of the filesystem objects specified by .metn path . It is a direct wrapper for the POSIX C library function of the same name. The permissions are specified by .metn mode , an integer argument. The existing permissions may be obtained using the .code stat function. The function throws a .code file-error exception if an error occurs, otherwise it returns .codn t. .TP* Example: .cblk ;; Set permissions of foo.txt to "rw-r--r--" ;; (owner can read and write; group owner ;; and other users can only read). ;; numerically: (chmod "foo.txt" #o644) ;; symbolically: (chmod "foo.txt" (logior s-irusr s-iwusr s-irgrp s-iroth)) .cble .coNP Function @ mknod .synb .mets (mknod < path < mode <> [ dev ]) .syne .desc The .code mknod function tries to create an entry in the filesystem: a file, FIFO, or a device special file, under the name .metn path . If it is successful, it returns .codn t , otherwise it throws an exception of type .codn file-error . The .meta mode argument is a bitwise or combination of the requested permissions, and the type of object to create: one of the constants .codn s-ifreg , .codn s-ififo , .codn s-ifchr , .code s-ifblk or .codn s-ifsock . The permissions are subject to the system .codn umask . If a block or character special device .cod2 ( s-ifchr or .codn s-ifblk ) is being created, then the .meta dev argument specifies the major and minor numbers of the device. A suitable value can be constructed from a major and minor pair using the .code makedev function. .TP* Example: .cblk ;; make a character device (8, 3) called /dev/foo ;; requesting rwx------ permissions (mknod "dev/foo" (logior #o700 s-ifchr) (makedev 8 3)) .cble .coNP Functions @ symlink and @ link .synb .mets (symlink < target << path ) .mets (link < target << path ) .syne .desc The .code symlink function creates a symbolic link called .meta path whose contents are the absolute or relative path .metn target . .meta target does not actually have to exist. The link function creates a hard link. The object at .meta target is installed into the filesystem at .meta path also. If these functions succeed, they return .codn t . Otherwise they throw an exception of type .codn file-error . .coNP Function @ readlink .synb .mets (readlink << path ) .syne .desc If .meta path names a filesystem object which is a symbolic link, the .code readlink function reads the contents of that symbolic link and returns it as a string. Otherwise, it fails by throwing an exception of type .codn file-error . .SS* Unix Signal Handling On platforms where certain advanced features of POSIX signal handling are available at the C API level, \*(TX exposes signal-handling functionality. A \*(TX program can install a \*(TL function (such as an anonymous. .codn lambda , or the function object associated with a named function) as the handler for a signal. When that signal is delivered, \*(TX will intercept it with its own safe, internal handler, mark the signal as deferred (in a \*(TX sense) and then dispatch the registered function at a convenient time. Handlers currently are not permitted to interrupt the execution of most \*(TX internal code. Immediate, asynchronous execution of handlers is currently enabled only while \*(TX is blocked on I/O operations or sleeping. Additionally, the .code sig-check function can be used to dispatch and clear deferred signals. These handlers are then safely called if they were subroutines of .codn sig-check , and not asynchronous interrupts. .coNP Special variables @, sig-hup @, sig-int @, sig-quit @, sig-ill @, sig-trap @, sig-abrt @, sig-bus @, sig-fpe @, sig-kill @, sig-usr1 @, sig-segv @, sig-usr2 @, sig-pipe @, sig-alrm @, sig-term @, sig-chld @, sig-cont @, sig-stop @, sig-tstp @, sig-ttin @, sig-ttou @, sig-urg @, sig-xcpu @, sig-xfsz @, sig-vtalrm @, sig-prof @, sig-poll @, sig-sys @, sig-winch @, sig-iot @, sig-stkflt @, sig-io @ sig-lost and @ sig-pwr .desc These variables correspond to the C signal constants .codn SIGHUP , .code SIGINT and so forth. The variables .codn sig-winch , .codn sig-iot , .codn sig-stk flt, .codn sig-io , .code sig-lost and .code sig-pwr may not be available since a system may lack the corresponding signal constants. See notes for the function .codn log-authpriv . The highest signal number is 31. .coNP Functions @ set-sig-handler and @ get-sig-handler .synb .mets (set-sig-handler < signal-number << handling-spec ) .mets (get-sig-handler << signal-number ) .syne .desc The .code set-sig-handler function is used to specify the handling for a signal, such as the installation of a handler function. It updates the signal handling for a signal whose number is .meta signal-number (usually one of the constants like .codn sig-hup , .code sig-int and so forth), and returns the previous value. The .code get-sig-handler function returns the current value. The .meta signal-number must be an integer the range 1 to 31. Initially, all 31 signal handling specifications are set to the value .codn t . The .meta handling-spec parameter may be a function. If a function is specified, then the signal is enabled and connected to that function until another call to .code set-sig-handler changes the handling for that signal. If .meta handling-spec is the symbol .codn nil , then the function previously associated with the signal, if any, is removed, and the signal is disabled. For a signal to be disabled means that the signal is set to the .code SIG_IGN disposition (refer to the C API). If .meta handling-spec is the symbol .codn t , then the function previously associated with the signal, if any, is removed, and the signal is set to its default disposition. This means that it is set to .code SIG_DFL (refer to the C API). Some signals terminate the process if they are generated while the handling is configured to the default disposition. Note that the certain signals like .code sig-quit and .code sig-kill cannot be ignored or handled. Please observe the signal documentation in the IEEE POSIX standard, and your platform. A signal handling function must take two arguments. It is of the form: .cblk .mets (lambda >> ( signal << async-p ) ...) .cble The .meta signal argument is an integer indicating the signal number for which the handler is being invoked. The .meta asyncp-p argument is a boolean value. If it is .codn t , it indicates that the handler is being invoked asynchronously\(emdirectly in a signal handling context. If it is .codn nil , then it is a deferred call. Handlers may do more things in a deferred call, such as terminate by throwing exceptions, and perform I/O. The return value of a handler is normally ignored. However if it invoked asynchronously (the .meta async-p argument is true), then if the handler returns a .cod2 non- nil value, it is understood that the handler requesting that it be deferred. This means that the signal will be marked as deferred, and the handler will be called again at some later time in a deferred context, whereby .meta async-p is .codn nil . This is not guaranteed, however; it's possible that another signal will arrive before that happens, possibly resulting in another async call, so the handler must be prepared to deal with an async call at any time. If a handler is invoked synchronously, then its return value is ignored. In the current implementation, signals do not queue. If a signal is delivered to the process again, while it is marked as deferred, it simply stays deferred; there is no counter associated with a signal, only a boolean flag. .coNP Function @ sig-check .synb (sig-check) .syne .desc The .code sig-check function tests whether any signals are deferred, and for each deferred signal in turn, it executes the corresponding handler. For a signal to be deferred means that the signal was caught by an internal handler in \*(TX and the event was recorded by a flag. If a handler function is removed while a signal is deferred, the deferred flag is cleared for that signal. Calls to the .code sig-check function may be inserted into CPU-intensive code that has no opportunity to be interrupted by signals, because it doesn't invoke any I/O functions. .coNP Function @ kill .synb .mets (kill < process-id <> [ signal ]) .syne .desc The .code kill function is used for sending a signal to a process group or process. It is a wrapper for the POSIX .code kill function. If the .meta signal argument is omitted, it defaults to the same value as .codn sig-term . .SS* Unix Processes .coNP Functions @ fork and @ wait .synb .mets (fork) .mets (wait >> [ pid <> [ flags ]]) .syne .desc The .code fork and .code wait functions are interfaces to the Unix functions .code fork and .codn waitpid . The .code fork function creates a child process which is a replica of the parent. Both processes return from the function. In the child process, the return value is zero. In the parent, it is an integer representing the process ID of the child. If the function fails to create a child, it returns .code nil rather than an integer. In this case, the .code errno function can be used to inquire about the cause. The .code wait function, if successful, returns a cons cell consisting of a pair of integers. The .code car of the cons is the process ID of the process or group which was successfully waited on, and the .code cdr is the status. If .code wait fails, it returns .codn nil . The .code errno function can be used to inquire about the cause. The .meta process-id argument, if not supplied, defaults to -1, which means that .code wait waits for any process, rather than a specific process. Certain other values have special meaning, as documented in the POSIX standard for the .code waitpid function. The .meta flags argument defaults to zero. If it is specified as nonzero, it should be a bitwise combination (via the .code logior function) of the constants .codn w-nohang , .codn w-untraced and .codn w-continued . If .code w-nohang is used, then .code wait returns a cons cell whose .code car specifies a process ID value of zero in the situation that at least one of the processes designated by .code process-id exist and are children of the calling process, but have not changed state. In this case, the status value in the .code cdr is unspecified. Status values may be inspected with the functions .codn w-ifexited , .codn w-exitstatus , .codn w-ifsignaled , .codn w-termsig , .codn w-coredump , .codn w-ifstopped , .code w-stopsig and .codn w-ifcontinued . .coNP Functions @, w-ifexited @, w-exitstatus @, w-ifsignaled @, w-termsig @, w-coredump @ w-ifstopped and @ w-stopsig .synb .mets (w-ifexited << status ) .mets (w-exitstatus << status ) .mets (w-ifsignaled << status ) .mets (w-termsig << status ) .mets (w-coredump << status ) .mets (w-ifstopped << status ) .mets (w-stopsig << status ) .mets (w-ifcontinued << status ) .syne .desc These functions analyze process exit values produced by the .code wait function. They are closely based on the POSIX macros .codn WIFEXITED , .code WEXITSTATUS , and so on. The .meta status value is either an integer, or a cons cell. In this case, the cons cell is expected to have an integer in its .code cdr which is used as the status. The .codn w-ifexited , .codn w-ifsignaled , .codn w-coredump , .code w-ifstopped and .code w-ifcontinued functions have Lisp boolean return semantics, unlike their C language counterparts: they return .code t or .codn nil , rather than zero or nonzero. The others return integer values. .coNP Function @ exec .synb .mets (exec < file <> [ args ]) .syne .desc The exec function replaces the process image with the executable specified by string argument .metn file . The executable is found by searching the system path. The .meta file argument becomes the first argument of the executable, argument zero. If .meta args is specified, it is a list of strings. These are passed as the additional arguments of the executable. If .code exec fails, an exception of type .code file-error is thrown. .coNP Function @ exit* .synb .mets (exit* << status ) .syne .desc The .code exit* function terminates the entire process (running \*(TX image), specifying the termination status to the operating system. The .meta status argument is treated exactly like that of the .code exit function. Unlike that function, this one exits the process immediately, cleaning up only low-level operating system resources such as closing file descriptors and releasing memory mappings, without performing user-space cleanup. .code exit* is implemented using a call to the POSIX function .codn _exit . .coNP Functions @ getpid and @ getppid .synb (getpid) (getppid) .syne .desc These functions retrieve the current process ID and the parent process ID respectively. They are wrappers for the POSIX functions .code getpid and .codn getppid . .coNP Function @ daemon .synb .mets (daemon < nochdir-p << noclose-p ) .syne .desc This is a wrapper for the function .code daemon which originated in BSD Unix. It returns .code t if successful, .code nil otherwise, and the .code errno variable is set in that case. .SS* Unix File Descriptors .coNP Function @ open-fileno .synb .mets (open-fileno < file-descriptor <> [ mode-string ]) .syne The .code open-fileno function creates a \*(TX stream over a file descriptor. The .meta file-descriptor argument must be an integer denoting a valid file descriptor. For a description of .metn mode-string , see the .code open-file function. .coNP Function @ fileno .synb .mets (fileno << stream ) .syne .desc The .code fileno function returns the underlying file descriptor of .metn stream , if it has one. Otherwise, it returns .codn nil. This is equivalent to querying the stream using .code stream-get-prop for the .code :fd property. .coNP Function @ dupfd .synb .mets (dupfd < old-fileno <> [ new-fileno ]) .syne .desc The .code dupfd function provides an interface to the POSIX functions .code dup or .codn dup2 , when called with one or two arguments, respectively. .coNP Function @ pipe .synb (pipe) .syne .desc The .code pipe function, if successful, returns a pair of integer file descriptors as a cons cell pair. The descriptor in the .code car field of the pair is the read end of the pipe. The .code cdr holds the write end. If the function fails, it throws an exception of type .codn file-error . .coNP Function @ poll .synb .mets (poll < poll-list <> [ timeout ]) .syne .desc The .code poll function suspends execution while monitoring one or more file descriptors for specified events. It is a wrapper for the same-named POSIX function. The .meta poll-list argument is a list of .code cons pairs. The .code car of each pair is either an integer file descriptor, or else a stream object which has a file descriptor (the .code fileno function can be applied to that stream to retrieve a descriptor). The .code cdr of each pair is an integer bit mask specifying the events, whose occurrence the file descriptor is to be monitored for. The variables .codn poll-in , .codn poll-out , .code poll-err and several others are available which hold bitmask values corresponding to the constants .codn POLLIN , .codn POLLOUT , .code POLLERR used with the C language .code poll function. The .meta timeout argument, if absent, defaults to the value -1, which specifies an indefinite wait. A nonnegative value specifies a wait with a timeout, measured in milliseconds. The function returns a list of pairs representing the descriptors or streams which were successfully polled. If the function times out, it returns an empty list. If an error occurs, an exception is thrown. The returned list is similar in structure to the input list. However, it holds only entries which polled positive. The .code cdr of every pair now holds a bitmask of the events which were to have occurred. .SS* Unix Itimers Itimers ("interval timers") can be used in combination with signal handling to execute asynchronous actions. Itimers deliver delayed, one-time signals, and also periodically recurring signals. For more information, consult the POSIX specification. .coNP Variables @, itimer-real @, itimer-virtual and @ itimer-prof .desc These variables correspond to the POSIX constants .codn ITIMER_REAL , .code ITIMER_VIRTUAL and .codn ITIMER_PROF . Their values are suitable as the .meta timer argument of the .code getitimer and .code setitimer functions. .coNP Functions @ getitimer and @ setitimer .synb .mets (getitimer << timer ) .mets (setitimer < timer < interval << value ) .syne .desc The .code getitimer function returns the current value of the specified timer, which must be .codn itimer-real , .code itimer-virtual or .codn itimer-prof . The current value consists of a list of two integer values, which represents microseconds. The first value is the timer interval, and the second value is the timer's current value. Like .codn getitimer , the .code setitimer function also retrieves the specified timer. In addition, it stores a new value in the timer, which is given by the two arguments, expressed in microseconds. .SS* Unix Syslog On platforms where a Unix-like syslog API is available, \*(TX exports this interface. \*(TX programs can configure logging via the .code openlog function, control the logging mask via .code setlogmask and generate logs via .codn syslog , or using special syslog streams. .coNP Special variables @, log-pid @, log-cons @, log-ndelay @, log-odelay @ log-nowait and @ log-perror .desc These variables take on the values of the corresponding C preprocessor constants from the .code header: .codn LOG_PID , .codn LOG_CON S, etc. These integer values represent logging options used in the option argument to the .code openlog function. Note: .code LOG_PERROR is not in POSIX, and so .code log-perror might not be available. See notes about .code LOG_AUTHPRIV in the documentation for .codn log-authpriv . .coNP Special variables @, log-user @, log-daemon @ log-auth and @ log-authpriv .desc These variables take on the values of the corresponding C preprocessor constants from the .code header: .codn LOG_USER , .codn LOG_DAEMON , .code LOG_AUTH and .codn LOG_AUTHPRIV . These are the integer facility codes specified in the .code openlog function. Note: .code LOG_AUTHPRIV is not in POSIX, and so .code log-authpriv might not be available. For portability use code like .code (or (symbol-value 'log-authpriv) 0) to evaluate to 0 if .code log-authpriv doesn't exist, or else check for its existence using .codn (boundp 'log-authpriv) . .coNP Special variables @, log-emerg @, log-alert @, log-crit @, log-err @, log-warning @, log-notice @ log-info and @ log-debug These variables take on the values of the corresponding C preprocessor constants from the .code header: .codn LOG_EMERG , .codn LOG_ALERT , etc. These are the integer priority codes specified in the .code syslog function. .coNP The @ *stdlog* special variable .desc The .code *stdlog* variable holds a special kind of stream: a syslog stream. Each newline-terminated line of text sent to this stream becomes a log message. The stream internally maintains a priority value that is applied when it generates messages. By default, this value is that of .codn log-info . The stream holds the priority as the value of the .code :prio stream property, which may be changed with the .code stream-set-prop function. The latest priority value which has been configured on the stream is used at the time the newline character is processed and the log message is generated, not necessarily the value which was in effect at the time the accumulation of a line began to take place. Messages sent to .code *stdlog* are delimited by newline characters. That is to say, each line of text written to the stream is a new log. .coNP Function @ openlog .synb .mets (openlog < id-string >> [ options <> [ facility ]]) .syne .desc The .code openlog function is a wrapper for the .code openlog C function, and the arguments have the same semantics. It is not necessary to use .code openlog in order to call the .code syslog function or to write data to .codn *stdlog* . The call is necessary in order to override the default identifying string, to set options, such as having the PID (process ID) recorded in log messages, and to specify the facility. The .meta id-string argument is mandatory. The .meta option argument is a bitwise mask (see the logior function) of option values such as .code log-pid and .codn log-cons . If it is missing, then a value of 0 is used, specifying the absence of any options. The .meta facility argument is one of the values .codn log-user , .code log-daemon or .codn log-auth . If it is missing, then .code log-user is assumed. .coNP Function @ closelog .synb (closelog) .syne .desc The .code closelog function is a wrapper for the C function .codn closelog . .coNP Function @ setlogmask .synb .mets (setlogmask << bitmask-integer ) .syne .desc The .code setlogmask function interfaces to the corresponding C function, and has the same argument and return value semantics. The .meta bitmask-integer argument is a mask of priority values to enable. The return value is the prior value. Note that if the argument is zero, then the function doesn't set the mask to zero; it only returns the current value of the mask. Note that the priority values like .code log-emerg and .code log-debug are integer enumerations, not bitmasks. These values cannot be combined directly to create a bitmask. Rather, the .code mask function should be used on these values. .TP* Example: .cblk ;; Enable LOG_EMERG and LOG_ALERT messages, ;; suppressing all others (setlogmask (mask log-emerg log-alert)) .cble .coNP Function @ syslog .synb .mets (syslog < priority < format << format-arg *) .syne .desc This function is the interface to the .code syslog C function. The .code printf formatting capabilities of the function are not used; the .meta format argument follows the conventions of the \*(TL .code format function instead. Note in particular that the .code %m convention for interpolating the value of strerror(errno) which is available in some versions of the .code syslog C function is currently not supported. Note that syslog messages are not newline-terminated. .SS* Unix Path Globbing On platforms where the POSIX .code glob function is available \*(TX provides this functionality in the form of a like-named function, and some numeric constants. .coNP Special variables @, glob-err @, glob-mark @, glob-nosort @, glob-nocheck @, glob-noescape @, glob-period @, glob-altdirfunc @, glob-brace @, glob-nomagic @, glob-tilde @ glob-tilde-check and @ glob-onlydir These variables take on the values of the corresponding C preprocessor constants from the .code header: .codn GLOB_ERR , .codn GLOB_MARK , .codn GLOB_NOSORT , etc. These values are passed as the optional second argument of the .code glob function. They are bitmasks and so multiple values can be combined using the .code logior function. Note that the .codn glob-period , .codn glob-altdirfunc , .codn glob-brace , .codn glob-nomagic , .codn glob-tilde , .code glob-tilde-check and .code glob-onlydir variables may not be available. They are extensions in the GNU C library implementation of .codn glob . .coNP Function @ glob .synb .mets (glob < pattern >> [ flags <> [ error-func ]]) .syne .desc The .code glob function is a interface to the Unix function of the same name. The .meta pattern argument must be a string, which holds a glob pattern: a pattern which matches zero or more path names, similar to a regular expression. The function tries to expand the pattern and return a list of strings representing the matching path names in the file system. If there are no matches, then an empty list is returned. The optional .meta flags argument defaults to zero. If given, it may be a bitwise combination of the values of the variables .codn glob-err , .codn glob-mark , .code glob-nosort and others. If the .meta error-func argument is specified, it gives a callback function which is invoked when .code glob encounters errors accessing paths. The function takes two arguments: the pathname and the .code errno value which occurred for that pathname. The function's return value is boolean. If the function returns true, then .code glob will terminate. Details of the semantics of the .code glob function, and the meaning of all the .meta flags arguments are given in the documentation for the C function. .SS* Web Programming Support .coNP Functions @ url-encode and @ url-decode .synb .mets (url-encode < string <> [ space-plus-p ]) .mets (url-decode < string <> [ space-plus-p ]) .syne .desc These functions convert character strings to and from a form which is suitable for embedding into the request portions of URL syntax. Encoding a string for URL use means identifying in it certain characters that might have a special meaning in the URL syntax and representing it using "percent encoding": the percent character, followed by the ASCII value of the character. Spaces and control characters are also encoded, as are all byte values greater than or equal to 127 (7F hex). The printable ASCII characters which are percent-encoded consist of this set: .cblk :/?#[]@!$&'()*+,;=% .cble More generally, strings can consists of Unicode characters, but the URL encoding consists only of printable ASCII characters. Unicode characters in the original string are encoded by expanding into UTF-8, and applying percent-encoding the UTF-8 bytes, which are all in the range .codn \exx80-\exxFF . Decoding is the reverse process: reconstituting the UTF-8 byte sequence specified by the URL-encoding, and then decoding the UTF-8 sequence into the string of Unicode characters. There is an additional complication: whether or not to encode spaces as plus, and to decode plus characters to spaces. In encoding, if spaces are not encoded to the plus character, then they are encoded as .codn %20 , since spaces are reserved characters that must be encoded. In decoding, if plus characters are not decoded to spaces, then they are left alone: they become plus characters in the decoded string. The .code url-encode function performs the encoding process. If the .code space-plus-p argument is omitted or specified as .codn nil , then spaces are encoded as .codn %20 . If the argument is a value other than .codn nil , then spaces are encoded as the character .code + .codn (plus) . The .code url-decode function performs the decoding process. If the .code space-plus-p argument is omitted or specified as .codn nil , then .code + .code (plus) characters in the encoded data are retained as .code + characters in the decoded strings. Otherwise, plus characters are converted to spaces. .coNP Functions @ html-encode and @ html-decode .synb .mets (html-encode << text-string ) .mets (html-decode << html-string ) .syne .desc The .code html-encode and .code html-decode functions convert between an HTML and raw representation of of text. The .code html-encode function returns a string which is based on the content of .metn text-string , but in which all characters which have special meaning in HTML have been replaced by HTML codes for representing those characters literally. The returned string is the HTML-encoded verbatim representation of .metn text-string . The .code html-decode function converts .metn html-string , which may contain HTML character encodings, into a string which contains the actual characters represented by those encodings. The function composition .code (html-decode (html-encode text)) returns a string which is equal to .codn text . The reverse composition .code (html-encode (html-decode html)) does not necessarily return a string equal to .codn html . For instance if html is the string .strn "

Hello, world!

" , then .code html-decode produces .strn "

Hello, world!

" . From this, .code html-encode produces .strn "<p>Hello, world!</p>" . .SS* Filter Module The filter module provides a trie (pronounced "try") data structure, which is suitable for representing dictionaries for efficient filtering. Dictionaries are unordered collections of keys, which are strings, which have associated values, which are also strings. A trie can be used to filter text, such that keys appearing in the text are replaced by the corresponding values. A trie supports this filtering operation by providing an efficient prefix-based lookup method which only looks at each input character ones, and which does not require knowledge of the length of the key in advance. .coNP Function @ make-trie .synb (make-trie) .syne .desc The .code make-trie function creates an empty trie. There is no special data type for a trie; a trie is some existing type such as a hash table. .coNP Function @ trie-add .synb .mets (trie-add < trie < key << value ) .syne .desc The .code trie-add function adds the string .meta key to the trie, associating it with .metn value . If .meta key already exists in .metn trie , then the value is updated with .metn value . The .meta trie must not have been compressed with .metn trie-compress . A trie can contain keys which are prefixes of other keys. For instance it can contain .str dog and .strn dogma . When a trie is used for matching and substitution, the longest match is used. If the input presents the text .strn doggy , then the match is .strn dog . If the input is .strn dogmatic , then .str dogma matches. .coNP Function @ trie-compress .synb .mets (trie-compress << trie ) .syne .desc The .code trie-compress function changes the representation of .meta trie to a representation which occupies less space and supports faster lookups. The new representation is returned. The compressed representation of a trie does not support the .code trie-add function. This function destructively manipulates .metn trie , and may return an object that is the same object as .codn trie , or it may return a different object, while at the same time still modifying the internals of .metn trie . Consequently, the program should not retain the input object .codn trie , but use the returned object in its place. .coNP Function @ trie-lookup-begin .synb .mets (trie-lookup-begin << trie ) .syne .desc The .code trie-lookup-begin function returns a context object for performing an open-coded lookup traversal of a trie. The .meta tri argument is expected to be a trie that was created by the .code make-trie function. .coNP Function @ trie-lookup-feed-char .synb .mets (trie-lookup-feed-char < trie-context << char ) .syne .desc The .code trie-lookup-feed-char function performs a one character step in a trie lookup. The .meta trie-context argument must be a trie context returned by .metn trie-lookup-begin , or by some previous call to .codn trie-lookup-feed-char . The .meta char argument is the next character to match. If the lookup is successful (the match through the trie can continue with the given character) then a new trie context object is returned. The old trie context remains valid. If the lookup is unsuccessful, .code nil is returned. Note: determining whether a given string is stored in a trie can be performed looking up every character of the string successively with .codn trie-lookup-feed-char , using the newly returned context for each successive operation. If every character is found, it means that either that exact string is found in the trie, or a prefix. The ambiguity can be resolved by testing whether the trie has a value at the last node using .codn tree-value-at . For instance, if .str catalog is inserted into an empty trie with value .strn foo , then .str cat will look up successfully, being a prefix of .strn catalog ; however, the value at .str cat is .codn nil , indicating that .str cat is only a prefix of one or more entries in the trie. .coNP Function tree-value-at .synb .mets (trie-value-at << trie-context ) .syne .desc The .code trie-value-at function returns the value stored at the node in in the trie given by .metn trie-context . Nodes which have not been given a value hold the value .codn nil . .coNP Function @ filter-string-tree .synb .mets (filter-string-tree < filter << obj ) .syne .desc The .code filter-string-tree a tree structure similar to .metn obj , in which all of the string atoms have been filtered through .metn filter . The .meta obj argument is a string tree structure: either the symbol .codn nil , denoting an empty structure; a string; or a list of tree structures. If .meta obj is .codn nil , then .code filter-string-tree returns .codn nil . The .meta filter argument is a filter: it is either a trie, a function, or nil. If .meta filter is .codn nil , then .code filter-string-trie just returns .metn obj . If .meta filter is a function, it must be a function that can be called with one argument. The strings of the string tree are filtered by passing each one into the function and substituting the return value into the corresponding place in the returned structure. Otherwise if .meta filter is a trie, then this trie is used for filtering, the string elements similarly to a function. For each string, a new string is returned in which occurrences of the keys in the trie are replaced by the values in the trie. .coNP Function @ filter-equal .synb .mets (filter-equal < filter-1 < filter-2 < obj-1 << obj-2 ) .syne .desc The .code filter-equal function tests whether two string trees are equal under the given filters. The precise semantics can be given by this expression: .cblk .mets (equal (filter-string-tree < filter-1 << obj-1 ) .mets \ \ \ \ \ \ (filter-string-tree < filter-2 << obj-2 )) .cble The string tree .meta obj-1 is filtered through .metn filter-1 , as if by the .code filter-string-tree function, and similarly, .meta obj-2 is filtered through .metn filter-2 . The resulting structures are compared using .codn equal , and the result of that is returned. .SS* Access To TXR Pattern Language From Lisp It is useful to be able to invoke the abilities of the \*(TX pattern Language from \*(TL. An interface for doing this provided in the form of the .code match-fun function, which is used for invoking a \*(TX pattern function. The .code match-fun function has a cumbersome interface which requires the \*(TL program to explicitly deal with the variable bindings emerging from the pattern match in the form of an association list. To make it the interface easier to use, \*(TX provides the macros .codn txr-if , .codn txr-when and .codn txr-case . .coNP Function match-fun .synb .mets (match-fun < name < args < input << files ) .syne .desc The .code match-fun function invokes a \*(TX pattern function whose name is given by .metn name , which must be a symbol. The .meta args argument is a list of expressions. The expressions may be symbols which will be interpreted as pattern variables, and may be bound or unbound. If they are not symbols, then they are treated as expressions (of the pattern language, not \*(TL) and evaluated accordingly. The .meta input argument is a list of strings, which may be lazy. It represents the lines of the text stream to be processed. The .meta file argument is a list of filename specifications, which follow the same conventions as files given on the \*(TX command line. If the pattern function uses the .code @(next) directive, it can process these additional files. The .code match-fun function's return value falls into three cases. If there is a match failure, it returns .codn nil . Otherwise it returns a cons cell. The .code car field of the cons cell holds the list of captured bindings. The .code cdr of the cons cell is one of two values. If the entire input was processed, the cdr field holds the symbol .codn t . Otherwise it holds another cons cell whose .code car is the remainder of the list of lines which were not matched, and whose .code cdr is the line number. .TP* Example: .cblk @(define foo (x y)) @x:@y @line @(end) @(do (format t "~s\en" (match-fun 'foo '(a b) '("alpha:beta" "gamma" "omega") nil))) Output: (((a . "alpha") (b . "beta")) ("omega") . 3) .cble In the above example, the pattern function .code foo is called with arguments .codn (a b) . These are unbound variables, so they correspond to parameters .code x and .code y of the function. If .code x and .code y get bound, those values propagate to .code a and .codn b . The data being matched consists of the lines .strn alpha:beta , .str gamma and .strn omega . Inside .codn foo, .code x and .code y bind to .str alpha and .strn beta , and then the line variable binds to .strn gamma . The input stream is left with .strn omega . Hence, the return value consists of the bindings of .code x and .code y transferred to .code a and .codn b , and the second cons cell which gives information about the rest of the stream: it is the part starting at .strn omega , which is line 3. Note that the binding for the .code line variable does not propagate out of the pattern function .codn foo ; it is local inside it. .coNP Macro txr-if .synb .mets (txr-if < name <> ( argument *) < input < then-expr <> [ else-expr ]) .syne .desc The .code txr-if macro invokes the \*(TX pattern matching function .metn name on some input given by the .meta input parameter, which is a list of strings, or a single string. If .meta name succeeds, then .meta then-expr is evaluated, and if it fails, .meta else-expr is evaluated instead. In the successful case, .meta then-expr is evaluated in a scope in which the bindings emerging from the .meta name function are turned into \*(TL variables. The result of .code txr-if is that of .metn then-expr . In the failed case, .meta else-expr is evaluated in a scope which does not have any new bindings. The result of .code txr-if is that of .metn else-expr . If .meta else-expr is missing, the result is .codn nil . The .meta argument forms supply arguments to the pattern function .metn name . There must be as many of these arguments as the function has parameters. Any argument which is a symbol is treated, for the purposes of calling the pattern function, as an unbound pattern variable. The function may or may not produce a binding for that variable. Also, every argument which is a symbol also denotes a local variable that is established around .meta then-expr if the function succeeds. For any such pattern variable for which the function produces a binding, the corresponding local variable will be initialized with the value of that pattern variable. For any such pattern variable which is left unbound by the function, the corresponding local variable will be set to .codn nil . Any .meta argument can be a form other than a symbol. In this situation, the argument is evaluated, and will be passed to the pattern function as the value of the binding for the corresponding argument. .TP* Example: .cblk @(define date (year month day)) @{year /\ed\ed\ed\ed/}-@{month /\ed\ed/}-@{day /\ed\ed/} @(end) @(do (each ((date '("09-10-20" "2009-10-20" "July-15-2014" "foo"))) (txr-if date (y m d) date (put-line `match: year @y, month @m, day @d`) (put-line `no match for @date`)))) Output: no match for 09-10-20 match: year 2009, month 10, day 20 no match for July-15-2014 no match for foo .cble .coNP Macro @ txr-when .synb .mets (txr-when < name <> ( argument *) < input << form *) .syne .desc The .code txr-when macro is based on .codn txr-if . It is equivalent to .code .cblk .meti \ \ (txr-if < name <> ( argument *) < input (progn << form *)) .cble If the pattern function .meta name produces a match, then each .meta form is evaluated in the scope of the variables established by the .meta argument expressions. The result of the .code txr-when form is that of the last .metn form . If the pattern function fails then the forms are not evaluated, and the result value is .codn nil . .coNP Macro @ txr-case .synb .mets (txr-case < input-form .mets \ \ >> {( name <> ( argument *) << form *)}* .mets \ \ >> [( t << form *)]) .syne .desc The .code txr-case macro evaluates .meta input-form and then uses the value as an input to zero or more test clauses. Each test clause invokes the pattern function named by that clause's .meta name argument. If the function succeeds, then each .meta form is evaluated, and the value of the last .meta form is taken to be the result value of .codn txr-case , which terminates. If there are no forms, then .code txr-case terminates with a .code nil result. The forms are evaluated in an environment in which variables are bound based on the .meta argument forms, with values depending on the result of the invocation of the .meta name pattern function, in the same manner as documented in detail for the .code txr-if macro. If the function fails, then the forms are not evaluated, and control passes to the next clause. A clause which begins with the symbol .code t executes unconditionally and causes .code txr-case to terminate. If it has no forms, then .code txr-case yields .codn nil , otherwise the forms are evaluated in order and the value of the last one specifies the result of .codn txr-case . .SS* Quote/Quasiquote Operator Syntax .coNP Operator @ quote .synb .mets (quote << form ) .syne .desc The .code quote operator, when evaluated, suppresses the evaluation of .metn form , and instead returns .meta form itself as an object. For example, if .meta form is a symbol, then .meta form is not evaluated to the symbol's value; rather the symbol itself is returned. Note: the quote syntax .cblk .meti >> '
.cble is translated to .cblk .meti (quote << form ). .cble .TP* Example: .cblk (quote a) ;; yields a (quote (+ 2 2)) ;; yields (+ 2 2), not 4. .cble .coNP Macro @ qquote .synb .mets (qquote << form ) .syne .desc The .code qquote (quasi-quote) macro operator implements a notation for convenient list construction. If .meta form is an atom, or a list structure which does not contain any .code unquote or .code splice operators, then .cblk .meti (qquote << form ) .cble is equivalent to .cblk .meti (qquote << form ). .cble If .metn form , however, is a list structure which contains .code unquote or .code splice operators, then the substitutions implied by those operators are performed on .metn form , and the .code qquote operator returns the resulting structure. Note: how the qquote operator actually works is that it is compiled into code. It becomes a Lisp expression which, when evaluated, computes the resulting structure. A .code qquote can contain another .codn qquote . If an .code unquote or .code splice operator occurs within a nested .codn qquote , it belongs to that .codn qquote , and not to the outer one. However, an unquote operator which occurs inside another one belongs one level higher. For instance in .cblk (qquote (qquote (unquote (unquote x)))) .cble the leftmost .code qquote belongs with the rightmost unquote, and the inner .code qquote and .code unquote belong together. When the outer .code qquote is evaluated, it will insert the value of .codn x , resulting in the object .codn (qquote (unquote [value-of-x])) . If this resulting qquote value is evaluated again as Lisp syntax, then it will yield .codn [value-of-value-of-x] , the value of .code [value-of-x] when treated as a Lisp expression and evaluated. .TP* Examples: .cblk (qquote a) -> a (qquote (a b c)) -> (a b c) (qquote (1 2 3 (unquote (+ 2 2)) (+ 2 3))) -> (1 2 3 4 (+ 2 3)) (qquote (unquote (+ 2 2))) -> 4 .cble In the second-to-last example, the .code 1 2 3 and the .code (+ 2 3) are quoted verbatim. Whereas the .code (unquote (+ 2 2)) operator caused the evaluation of .code (+ 2 2) and the substitution of the resulting value. The last example shows that .meta form can itself (the entire argument of .codn qquote ) can be an unquote operator. However, note: .code (quote (splice form)) is not valid. Note: a way to understand the nesting behavior is a via a possible model of quasi-quote expansion which recursively compiles any nested quasi quotes first, and then treats the result of their expansion. For instance, in the processing of .cblk (qquote (qquote (unquote (unquote x)))) .cble the .code qquote operator first encounters the embedded .code (qquote ...) and compiles it to code. During that recursive compilation, the syntax .code (unquote (unquote x)) is encountered. The inner quote processes the outer unquote which belongs to it, and the inner .code (unquote x) becomes material that is embedded verbatim in the compilation, which will then be found when the recursion pops back to the outer quasiquote, which will then traverse the result of the inner compilation and find the .codn (unquote x) . .TP* "Dialect note:" In Lisp dialects which have a published quasiquoting operator syntax, there is the expectation that the quasiquote read syntax corresponds to it. That is to say, that for instance the read syntax .code ^(a b ,c) is expected translated to .codn (qquote b (unquote c)) . In \*(TL, this is not true! Although .code ^(b b ,c) is translated to a quasiquoting macro, it is an internal one, not based on the public .codn qquote , .code unquote and .code splice symbols being documented here. This idea exists for hygiene. The quasiquote read syntax is not confused by the presence of the symbols .codn qquote , .code unquote or .code splice in the template, since it doesn't treat them specially. This also allows programmers to use the quasiquote read syntax to construct quasiquote macros. For instance .cblk ^(qquote (unquote ,x)) ;; does not mean ^^,x .cble To the quasiquote reader, the .code qquote and .code unquote symbols mean nothing special, and so this syntax simply means that if the value of .code x is .codn foo , the result will be .codn (qquote (unquote foo)) . The form's expansion is actually this: .cblk (sys:qquote (qquote (unquote (sys:unquote x)))) .cble the .code sys:qquote macro recognizes .code sys:unquote embedded in the form, and the other symbols not in the .code sys: package are just static template material. The .code sys:quote macro and its associated .code sys:unquote and .code sys:splice operators work exactly like their ordinary counterparts. So in effect, \*(TX has two nearly identical, independent quasi-quote implementations, one of which is tied to the read syntax, and one of which isn't. This is useful for writing quasiquotes which write quasiquotes. .coNP Operator @ unquote .synb .mets (qquote (... (unquote << form ) ...)) .mets (qquote (unquote << form )) .syne .desc The .code unquote operator is not an operator .I per .IR se . The .code unquote symbol has no binding in the global environment. It is a special syntax that is recognized within a .code qquote form, to indicate forms within the quasiquote which are to be evaluated and inserted into the resulting structure. The syntax .cblk .meti (qquote (unquote << form )) .cblk is equivalent to .metn form : the .code qquote and .code unquote "cancel out". .coNP Operator splice .synb .mets (qquote (... (splice << form ) ...)) .syne .desc The .code splice operator is not an operator .I per .IR se . The .code splice symbol has no binding in the global environment. It is a special syntax that is recognized within a .code qquote form, to indicate forms within the quasiquote which are to be evaluated and inserted into the resulting structure. The syntax .cblk .meti (qquote (splice << form )) .cble is not permitted and raises an exception if evaluated. The .code splice syntax must occur within a list, and not in the dotted position. The .code splice form differs from unquote in that .cblk .meti (splice << form ) .cble requires that .meta form must evaluate to a list. That list is integrated into the surrounding list. .SS* Macros \*(TL supports structural macros. \*(TX's model of macroexpansion is that \*(TL code is processed in two phases: the expansion phase and the evaluation phase. The expansion phase is invoked on Lisp code early during the processing of source code. For instance when a \*(TX file containing a .code @(do ...) directive is loaded, expansion of the Lisp forms are its arguments takes place during the parsing of the entire source file, and is complete before any of the code in that file is executed. If the .code @(do ...) form is later executed, the expanded forms are then evaluated. \*(TL also supports symbol macros, which are symbolic forms that stand for forms, with which they are replaced at macro expansion time. When Lisp data is processed as code by the .code eval function, it is first expanded, and so processed in its entirety in the expansion phase. Then it is processed in the evaluation phase. .coNP Macro parameter lists \*(TX macros support destructuring, similarly to Common Lisp macros. This means that macro parameter lists are like function argument lists, but support nesting. A macro parameter list can specify a nested parameter list in every place where a function argument list allows only a parameter name. For instance, consider this macro parameter list: .cblk ((a (b c)) : (c frm) ((d e) frm2 de-p) . g) .cble The top level of this list has four elements: the mandatory parameter .codn (a (b c)) , the optional parameter .code c (with default init form .codn frm ), the optional parameter .code (d e) (with default init form .code frm2 and presence-indicating variable .codn de-p ), and the dot-position parameter .code g which captures trailing arguments. Note that some of the parameters are compounds: .code (a (b c)) and .codn (d e) . These compounds express nested macro parameter lists. Macro parameter lists match a similar tree structure to their own. For instance a mandatory parameter .code (a (b c)) matches a structure like .codn (1 (2 3)) , such that the parameters .codn a , .code b and .code c will end up bound to .codn 1 , .code 2 and .codn 3 , respectively. The binding strictness is relaxed for optional parameters. If .code (a (b c)) is optional, and the argument is, say, .codn (1) , then .code a gets .codn 1 , and .code b and .code c receive .codn nil . Macro parameter lists also support two special keywords, namely .code :env and .codn :whole . The parameter list .code (:whole x :env y) will bind parameter .code x to the entire macro form, and bind parameter .code y to the macro environment. The .code :whole and .code :env notation can occur anywhere in a macro parameter list. .coNP Operator @ macro-time .synb .mets (macro-time << form *) .syne .desc The .code macro-time operator has a syntax similar to the .code progn operator. Each .meta form is evaluated from left to right, and the resulting value is that of the last form. The special behavior of .code macro-time is that the evaluation takes place during the expansion phase, rather than during the evaluation phase. During the expansion phase, all .code macro-time expressions which occur in a context that calls for evaluation are evaluated, and replaced by their quoted values. For instance .code (macro-time (list 1 2 3)) evaluates .code (list 1 2 3) to the object .code (1 2 3) and the entire .code macro-time form is replaced by that value, quoted: .codn '(1 2 3) . If the form is evaluated again at evaluation-time, the resulting value will be that of the quote, in this case .codn (1 2 3) . .code macro-time forms do not see the surrounding lexical environment; the see only global function and variable bindings and macros. Note 1: .code macro-time is intended for defining helper functions and variables that are used by macros. A macro cannot "see" a .code defun function or .code defvar variable because .code defun and .code defvar forms are evaluated at evaluation time, which occurs after expansion time. The macro-time operator hoists the evaluation of these forms to macro-expansion time. Note 2: .code defmacro forms are implicitly in macro-time; they do not have to be wrapped with the .code macro-time operator. The .code macro-time operator doesn't have to be used in programs whose macros do not make references to variables or functions. .coNP Operator @ defmacro .synb .mets (defmacro < name <> ( param * [: << opt-param * ] [. < rest-param ]) .mets \ \ << body-form *) .syne .desc The .code defmacro operator is evaluated at expansion time. It defines a macro-expander function under the name .metn name , effectively creating a new operator. Note that the parameter list is a macro parameter list, and not a function parameter list. This means that each .meta param and .meta opt-param can be not only a symbol, but it can itself be a parameter list. The corresponding argument is then treated as a structure which matches that parameter list. This nesting of parameter lists can be carried to an arbitrary depth. A macro is called like any other operator, and resembles a function. Unlike in a function call, the macro receives the argument expressions themselves, rather than their values. Therefore it operates on syntax rather than on values. Also, unlike a function call, a macro call occurs in the expansion phase, rather than the evaluation phase. The return value of the macro is the macro expansion. It is substituted in place of the entire macro call form. That form is then expanded again; it may itself be another macro call, or contain more macro calls. .TP* Example: .cblk ;; dolist macro similar to Common Lisp's: ;; ;; The following will print 1, 2 and 3 ;; on separate lines: ;; and return 42. ;; ;; (dolist (x '(1 2 3) 42) ;; (format t "~s\en")) (defmacro dolist ((var list : result) . body) (let ((i (my-gensym))) ^(for ((i ,list)) (i ,result) ((set i (cdr i))) (let ((,var (car i))) ,*body)))) .cble .coNP Operator @ macrolet .synb .mets (macrolet >> ({( name < macro-style-params << macro-body-form *)}*) .mets \ \ << body-form *) .syne .desc The .code macrolet binding operator extends the macro-time lexical environment by making zero or more new local macros visible. The .code macrolet symbol is followed by a list of macro definitions. Each definition is a form which begins with a .metn name , followed by .meta macro-style-params which is a macro parameter list, and zero or more .metn macro-body-form s. These macro definitions are similar to those globally defined by the .code defmacro operator, except that they are in a local environment. The macro definitions are followed by optional .metn body-forms . The macros specified in the definitions are visible to these forms. Forms inside the macro definitions such as the .metn macro-body-form s, and initializer forms appearing in the .meta macro-style-params are subject to macro-expansion in a scope in which none of the new macros being defined are yet visible. Once the macro definitions are themselves macro-expanded, they are placed into a new macro environment, which is then used for macro expanding the .metn body-form s. A .code macrolet form is fully processed in the expansion phase of a form, and is effectively replaced by .code progn form which contains expanded versions of .metn body-form s. This expanded structure shows no evidence that any macrolet forms ever existed in it. Therefore, it is impossible for the code evaluated in the bodies and parameter lists of .code macrolet macros to have any visibility to any surrounding lexical variable bindings, which are only instantiated in the evaluation phase, after expansion is done and macros no longer exist. .coNP Function @ macro-form-p .synb .mets (macro-form-p < obj <> [ env ]) .syne .desc The .code macro-form-p function returns .code t if .meta obj represents the syntax of a form which is a macro form: either a compound macro or a symbol macro. Otherwise it returns .codn nil . A macro form is one that will transform under .code macroexpand-1 or .codn macroexpand ; an object which isn't a macro form will not undergo expansion. The optional .meta env parameter is a macroexpansion environment. A macroexpansion environment is passed down to macros and can be received via their special .code :env parameter. .meta env is used by .code macro-form-p to determine whether .meta obj is a macro in a lexical macro environment. If .meta env is not specified or is .codn nil , then .code macro-form-p only recognizes global macros. .TP* Example: .cblk ;; macro which translates to 'yes if its ;; argument is a macro from, or otherwise ;; transforms to the form 'no. (defmacro ismacro (:env menv form) (if (macro-form-p form menv) ''yes ''no)) (macrolet ((local ())) (ismacro (local))) ;; yields yes (ismacro (local)) ;; yields no (ismacro (ismacro foo)) ;; yields yes .cble During macro expansion, the global macro .code ismacro is handed the macro-expansion environment via .codn :env menv . When the macro is invoked within the macrolet, this environment includes the macro-time lexical scope in which the .code local macro is defined. So when global checks whether the argument form .code (local) is a macro, the conclusion is yes: the (local) form is a macro call in that environment: .code macro-form-p yields .codn t . When .code (global (local)) is invoked outside of the macrolet, no local macro is visible is there, and so .code macro-form-p yields .codn nil . .coNP Functions @ macroexpand-1 and @ macroexpand .synb .mets (macroexpand-1 < obj <> [ env ]) .mets (macroexpand < obj <> [ env ]) .syne .desc If .meta obj is a macro form (an object for which .code macro-form-p returns .codn t ), these functions expand the macro form and return the expanded form. Otherwise, they return .metn obj . .code macroexpand-1 performs a single expansion, expanding just the macro that is referenced by the symbol in the first position of .metn obj , and returns the expansion. That expansion may itself be a macro form. .code macroexpand performs an expansion similar to .codn macroexpand-1 . If the result is a macro form, then it expands that form, and keeps repeating this process until the expansion yields a non-macro-form. That non-macro-form is then returned. The optional .meta env parameter is a macroexpansion environment. A macroexpansion environment is passed down to macros and can be received via their special .code :env parameter. The environment they receive is their lexically apparent macro-time environment in which local macros may be visible. A macro can use this environment to "manually" expand some form in the context of that environment. .TP* Example: .cblk ;; (foo x) expands x, and if x begins with a number, ;; it removes the number and returns the resulting ;; form. Otherwise, it returns the entire form. (defmacro rem-num (:env menv some-form) (let ((expanded (macroexpand some-form menv))) (if (numberp (car expanded)) (cdr expanded) some-form))) ;; The following yields (42 a). (macrolet ((foo () '(1 list 42)) (bar () '(list 'a))) (list (rem-num (foo)) (rem-num (bar))))) .cble The .code rem-num macro is able to expand the .code (foo) and .code (bar) forms it receives as the .code some-form argument, even though these forms use local macro that are only visible in their local scope. This is thanks to the macro environment passed to .codn rem-num . It is correctly able to work with the expansions .code (1 list 42) and .code (list 'a) to produce .code (list 42) and .code (list 'a) which evaluate to .code 42 and .code a respectively. .coNP Functions @ lexical-var-p and @ lexical-fun-p .synb .mets (lexical-var-p < env << form ) .mets (lexical-fun-p < env << form ) .syne .desc These two functions are useful to macro writers. They are intended to be called from the bodies of macro expanders, such as the bodies of .code defmacro or .code macrolet forms. The .meta env argument is a macro-time environment, which is available to macros via the special .code :env parameter. Using these functions, a macro can enquire whether a given .meta form is a symbol which has a variable binding or a function binding in the lexical environment. This information is known during macro expansion. The macro expander recognizes lexical function and variable bindings, because these bindings can shadow macros. .TP* Example: .cblk ;; ;; this macro replaces itself with :lexical-var if its ;; argument is a lexical variable, :lexical-fun if ;; its argument is a lexical function, or with ;; :not-lex-fun-var if neither is the case. ;; (defmacro classify (sym :env e) (cond ((lexical-var-p e expr) :lexical-var) ((lexical-fun-p e expr) :lexical-fun) (t :not-lex-fun-var))) ;; ;; This returns: ;; ;; (:lexical-var :not-lex-fun-var :lexical-fun) ;; (let ((x 1) (y 2)) (symacrolet ((y x)) (flet ((f () (+ 2 2))) (list (classify x) (classify y) (classify z))))) .cble .TP* Note: These functions do not call .code macroexpand on the form. In most cases, it is necessary for the macro writers to do so. Not that in the above example, symbol .code y is classified as neither a lexical function nor variable. However, it can be macro-expanded to .code x which is a lexical variable. .coNP Function @ lexical-lisp1-binding .synb .mets (lexical-lisp1-binding < env << symbol ) .syne .desc The .code lexical-lisp1-binding function inspects the macro-time environment .meta env to determine what kind of binding, if any, does .meta symbol have in that environment, from a Lisp-1 perspective. That is to say, it considers function bindings, variable bindings and symbol macro bindings to be in a single name space and finds the innermost binding of one of these types for .metn symbol . If such a binding is found, then the function returns one of the three keyword symbols .codn :var , .codn :fun , or .codn :symacro . If no such lexical binding is found, then the function returns .codn nil . Note that a .code nil return doesn't mean that the symbol doesn't have a lexical binding. It could have an operator macro lexical binding (a macro binding in the function namespace established by .codn macrolet ). .coNP Operator @ defsymacro .synb .mets (defsymacro < sym << form ) .syne .desc A .code defsymacro form introduces a symbol macro: a symbol macro is a parameterless substitution keyed to a symbol. In contexts where a symbol macro definition of .meta sym is visible, if the form .meta sym appears such that its evaluation is called for, it is subject to replacement by .metn form . After replacement takes place, .meta form itself is then processed for further replacement of macros and symbol macros. Symbol macros are also recognized in contexts where .meta sym denotes a place which is the target of an assignment operation like .code set and similar. A .code defsymacro form is implicitly executed at expansion time, and thus need not be wrapped in a .code macro-time form, just like .codn defmacro . Note: if a symbol macro expands to itself directly, expansion stops. However, if a symbol macro expands to itself through a chain of expansions, an infinite expansion time loop results. .coNP Operator @ symacrolet .synb .mets (symacrolet >> ({( sym << form )}*) << body-form *) .syne .desc The .code symacrolet operator binds local, lexically scoped macros that are similar to the global symbol macros introduced by .codn defsymacro . Each .meta sym in the bindings list is bound to its corresponding form, creating a new extension of the expansion-time lexical macro environment. Each .meta body-form is subsequently macro-expanded in this new environment in which the new symbol macros are visible. Note: ordinary lexical bindings such as those introduced by let or by function parameters lists shadow symbol macros. If a symbol .code x is bound by nested instances of .code macrolet and a .codn let , then the scope enclosed by both constructs will see whichever of the two bindings is more inner, even though the bindings are active in completely separate phases of processing. From the perspective of the arguments of a .code dwim form, lexical function bindings also shadow symbol macros. This is consistent with the Lisp-1-style name resolution which applies inside a .code dwim form. Of course, lexical operator macros do not shadow symbol macros under any circumstances. .coNP Macros @ placelet and @ placelet* .synb .mets (placelet >> ({( sym << place )}*) << body-form *) .mets (placelet* >> ({( sym << place )}*) << body-form *) .syne .desc The .code placelet macro binds lexically scoped symbol macros in such a way that they behave as aliases for places denoted by place forms. Each .meta place must be an expression denoting a syntactic place. The corresponding .meta sym is established as an alias for the storage location which that place denotes, over the scope of the .metn body-form -s. This binding takes place in such a way that each .meta place is evaluated exactly once, only in order to determine its storage location. The corresponding .meta sym then serves as an alias for that location, over the scope of the .metn body-form -s. This means that whenever .meta sym is evaluated, it stands for the value of the storage location, and whenever a value is apparently stored into .metn sym , it is actually the storage location which receives it. The .code placelet* variant implements an alternative scoping rule, which allows a later .meta place form to refer to a .meta sym bound to an earlier .meta place form. In other words, a given .meta sym binding is visible not only to the .metn body-form -s but also to .meta place forms which occur later. Note: certain kinds of places, notably .cblk .meti (force << promise ) .cble expressions, must be accessed before they can be stored, and this restriction continues to hold when those places are accessed through .code placelet aliases. Note: .code placelet differs from .code symacrolet in that the forms themselves are not aliased, but the storage locations which they denote. .code (symacrolet ((x y) z) performs the syntactic substitution of symbol .code x by form .codn y , wherever .code x appears inside .code z as an evaluated form, and is not shadowed by any inner binding. Whereas .code (placelet ((x y)) z) generates code which arranges for .code y to be evaluated to a storage location, and syntactically replaces occurrences of .code x with a form which directly denotes that storage location, wherever .code x appears inside .code z as an evaluated form, and is not shadowed by any inner binding. Also, .code x is not necessarily substituted by a single, fixed form, as in the case of .codn symacrolet . Rather it may be substituted by one kind of form when it is treated as a pure value, and another kind of form when it is treated as a place. .coNP Operators @ tree-bind and @ mac-param-bind .synb .mets (tree-bind < macro-style-params < expr << form *) .mets (mac-param-bind < context-expr < macro-style-params < expr << form *) .syne .desc The .code tree-bind operator evaluates .codn expr , and then uses the resulting value as a counterpart to a macro-style parameter list. If the value has a tree structure which matches the parameters, then those parameters are established as bindings, and the .metn form s, if any, are evaluated in the scope of those bindings. The value of the last .meta form is returned. If there are no forms, .code nil is returned. Note: this operator throws an exception if there is a structural mismatch between the parameters and the value of .codn expr . One way to avoid this exception is to use .codn tree-case . The .code mac-param-bind operator is similar to .code tree-bind except that it takes an extra argument, .metn context-expr. This argument is an expression which is evaluated. It is expected to evaluate to a compound form. If an error occurs during binding, the error diagnostic message is based on information obtained from this form. By contrast, the .code tree-bind operator's error diagnostic refers to the .code tree-bind form, which is cryptic if the binding is used for the implementation of some other construct, hidden from the user of that construct. .coNP Operator @ tree-case .synb .mets (tree-case < expr >> {( macro-style-params << form *)}*) .syne .desc The .code tree-case operator evaluates .meta expr and matches it against a succession of zero or more cases. Each case defines a pattern match, expressed as a macro style parameter list .metn macro-style-params . If the object produced by .meta expr matches .metn macro-style-params , then the parameters are bound, becoming local variables, and the .metn form s, if any, are evaluated in order in the environment in which those variables are visible. If there are forms, the value of the last .meta form becomes the result value of the case, otherwise the result value of the case is nil. If the result value of a case is the object .code : (the colon symbol), then processing continues with the next case. Otherwise the evaluation of .code tree-case terminates, returning the result value. If the value of .meta expr does not match the .meta macro-style-params parameter list of a case, processing continues with the next case. If no cases match, then .code tree-case terminates, returning .codn nil . .TP* Example: .cblk ;; reverse function implemented using tree-case (defun tb-reverse (obj) (tree-case obj (() ()) ;; the empty list is just returned ((a) obj) ;; one-element list returned (unnecessary case) ((a . b) ^(,*(tb-reverse b) ,a)) ;; car/cdr recursion (a a))) ;; atom is just returned .cble Note that in this example, the atom case is placed last, because an argument list which consists of a symbol is a "catch all" match that matches any object. We know that it matches an atom, because the previous .code (a . b) case matches conses. In general, the order of the cases in .code tree-case is important: even more so than the order of cases in a .code cond or .codn caseql . .coNP Macro @ tb .synb .mets (tb < macro-style-params << form *) .syne .desc The .code tb macro is similar to the .code lambda operator but its argument binding is based on a macro-style parameter list. The name is an abbreviation of .codn tree-bind . A .code tb form evaluates to a function which takes a variable number of arguments. When that function is called, those arguments are taken as a list object which is matched against .meta macro-style-params as if by .metn tree-bind . If the match is successful, then the parameters are bound to the corresponding elements from the argument structure and each successive .meta form is evaluated an environment in which those bindings are visible. The value of the last .meta form is the return value of the function. If there are no forms, the function's return value is .codn nil . The following equivalence holds, where .code args should be understood to be a globally unique symbol: .cblk (tb pattern body ...) <--> (lambda (. args) (tree-bind pattern args body ...)) .cble .coNP Macro @ tc .synb .mets (tc >> {( macro-style-params << form *)}*) .syne .desc The .code tc macro produces an anonymous function whose behavior is closely based on the .code tree-case operator. Its name is an abbreviation of .codn tree-case . The anonymous function takes a variable number of arguments. Its argument list is taken to be the value macro is tested against the multiple pattern clauses of an implicit .codn tree-bind . The return value of the function is that of the implied .codn tree-bind . The following equivalence holds, where .code args should be understood to be a globally unique symbol: .cblk (tc clause1 clause2 ...) <--> (lambda (. args) (tree-bind args clause1 clause2 ...)) .cble .SS* User-Defined Places and Place Operators \*(TL provides a number of place-modifying operators such as .codn set , .codn push , and .codn inc . It also provides a variety of kinds of syntactic places which may be used with these operators. Both of these categories are open-ended: \*(TL programs may extend the set of place-modifying operators, as well as the vocabulary of forms which are recognized as syntactic places. Regarding place operators, it might seem obvious that new place operators can be developed, since they are macros, and macros can expand to uses of existing place operators. As an example, it may seem that .code inc operator could be written as a macro which uses .codn set : .cblk (defmacro new-inc (place : (delta 1)) ^(set ,place (+ ,place ,delta))) .cble However, the above .code new-inc macro has a problem: the .code place argument form is inserted into two places in the expansion, which leads to two evaluations. This is visibly incorrect if the place form contains any side effects. It is also potentially inefficient. \*(TL provides a framework for writing place update macros which evaluate their argument forms once, even if they have to access and update the same places. The framework also supports the development of new kinds of place forms as capsules of code which introduce the right kind of material into the lexical environment of the body of an update macro, to enable this special evaluation. .NP* Place-Expander Functions The central design concept in \*(TL syntactic places are .IR "place-expander functions" . Each compound place is defined by up to three place-expander functions, which are associated with the place via the leftmost operator symbol of the place form. One place-expander, the .IR "update expander" , is mandatory. Optionally, a place may also provide a .I "clobber expander" as well as a .IR "delete expander" . An update expander provides the expertise for evaluating a place form once in its proper run-time context to determine its actual run-time storage location, and to access and modify the storage location. A clobber expander provides an optimized mechanism for uses that perform a one-time store to a place without requiring its prior value. If a place definition does not supply a clobber expander, then the syntactic places framework uses the update expander to achieve the functionality. A delete expander provides the expertise for determining the actual run-time storage location corresponding to a place, and obliterating it, returning its prior value. If a place does not supply a delete expander, then the place does not support deletion. Operators which require deletion, such as .code del will raise an error when applied to that place. The expanders operate independently, and it is expected that place-modifying operators choose one of the three, and use only that expander. For example, accessing a place with an update expander and then overwriting its value with a clobber expander may result in incorrect code which contains multiple evaluations of the place form. The programmer who implements a new place does not write expanders directly, but rather defines them via the .code defplace macro. The programmer who implements a new place update macro likewise does not call the expanders directly. Usually, they are invoked via the macros .codn with-update-expander , .codn with-clobber-expander and .codn with-delete-expander . These are sufficient for most kind of macros. In certain complicated cases, expanders may be invoked using the wrapper functions .codn call-update-expander , .codn call-clobber-expander and .codn call-delete-expander . These convenience macros and functions perform certain common chores, like macro-expanding the place in the correct environment, and choosing the appropriate function. The expanders are described in the following sections. .NP* The Update Expander .synb .mets (lambda >> ( getter-sym < setter-sym < place-form .mets \ \ \ \ \ \ \ \ << body-form ) ...) .syne .desc The update expander is a code-writer. It takes a .meta body-form argument, representing code, and returns a larger form which surrounds this code with additional code. This larger form returned by the update expander can be regarded as having two abstract actions, when it is substituted and evaluated in the context where .meta place-form occurs. The first abstract action is to evaluate .meta place-form exactly one time, in order to determine the actual run-time location to which that form refers. The second abstract action is to evaluate the caller's .metn body-form -s, in a lexical environment in which bindings exist for some lexical functions or (more usually) lexical macros. These lexical macros are explicitly referenced by the .metn body-form ; the update expander just provides their definition, under the names it is given via the .meta getter-sym and .meta setter-sym arguments. The update expander writes local functions or macros under these names: a getter function and a setter function. Usually, update expanders write macros rather than functions, possibly in combination with some lexical anonymous variables which hold temporary objects. Therefore the getter and setter are henceforth referred to as macros. The code being generated is with regard to some concrete instance of .metn place-form . This argument is the actual form which occurs in a program. For instance, the update expander for the .code car place might be called with an arbitrary variant of the .meta place-form might look like .codn (car (inc (third some-list))) . In the abstract semantics, upfront code wrapped around the .meta body-form by the update expander provides the logic to evaluate this place to a location, which is retained in some hidden local context. The getter local macro named by .meta getter-sym must provide the logic for retrieving the value of this place. The getter macro takes no arguments. The .meta body-form makes free use of the getter function; they may call it multiple times, which must not trigger multiple evaluations of the original place form. The setter local macro named by .meta setter-sym must generate the logic for storing a new value into the once-evaluated version of .metn place-form . The setter function takes exactly one argument, whose value specifies the value to be stored into the place. It is the caller's responsibility to ensure that the argument form which produces the value to be stored via the setter is evaluated only once, and in the correct order. The setter does not concern itself with this form. Multiple calls to the setter can be expected to result in multiple evaluations of its argument. Thus, if necessary, the caller must supply the code to evaluate the new value form to a temporary variable, and then pass the temporary variable to the setter. This code can be embedded in the .meta body-form or can be added to the code returned by a call to the update expander. The setter local macro or function must return the new value which is stored. That is to say, when .meta body-form invokes this local macro or function, it may rely on it yielding the new value which was stored, as part of achieving its own semantics. The update expander does not macro-expand .codn place-form . It is assumed that the expander is invoked in such a way that the place has been expanded in the correct environment. In other words, the form matches the type of place which the expander handles. If the expander had to macro-expand the place form, it would sometimes have to come to the conclusion that the place form must be handled by a different expander. No such consideration is the case: when an expander is called on a form, that is final; it is certain that it is the correct expander, which matches the symbol in the .code car position of the form, which is not a macro in the context where it occurs. An update expander is free to assume that any place which is stored (the setter local macro is invoked on it) is accessed at least once by an invocation of the getter. A place update macro which relies on an update expander, but uses only the store macro, might not work properly. An example of an update expander which relies on this assumption is the expander for the .cblk .meti (force << promise ) .cble place type. If .meta promise has not yet been forced, and only the setter is used, then .meta promise might remain unforced as its internal value location is updated. A subsequent access to the place will incorrectly trigger a force, which will overwrite the value. The expected behavior is that storing a value in an unforced .code force place changes the place to forced state, preempting the evaluation of the delayed form. Afterward, the promise exhibits the value which was thus assigned. The update expander is not responsible for all issues of evaluation order. A place update macro may consist of numerous places, as well as numerous value-producing forms which are not places. Each of the places can provide its registered update expander which provides code for evaluating just that place, and a means of accessing and storing the values. The place update macro must call the place expanders in the correct order, and generate any additional code in the correct order, so that the macro achieves its required documented evaluation order. .TP* "Example Update Expander Call:" .cblk ;; First, capture the update expander ;; function for (car ...) places ;; in a variable, for clarity. (defvar car-update-expander [*place-update-expander* 'car]) ;; Next, call it for the place (car [a 0]). ;; The body form specifies logic for ;; incrementing the place by one and ;; returning the new value. (call car-update-expander 'getit 'setit '(car [a 0]) '(setit (+ (getit) 1))) --> ;; Resulting code: (rlet ((#:g0032 [a 0])) (macrolet ((getit nil (append (list 'car) (list '#:g0032))) (setit (val) (append (list 'sys:rplaca) (list '#:g0032) (list val)))) (setit (+ (getit) 1)))) ;; Same expander call as above, with a sys:expand to show the ;; fully expanded version of the returned code, in which the ;; setit and getit calls have disappeared, replaced by their ;; macro-expansions. (sys:expand (call car-update-expander 'getit 'setit '(car [a 0]) '(setit (+ (getit) 1)))) --> (let ((#:g0032 [a 0])) (sys:rplaca #:g0032 (+ (car #:g0032) 1))) .cble The main noteworthy points about the generated code are: .RS .IP - the .code (car [a 0]) place is evaluated by evaluating the embedded form .code [a 0] and storing storing the resulting object into a hidden local variable. That's as close a reference as we can make to the .code car field. .IP - the getter macro expands to code which simply calls the .code car function on the cell. .IP - the setter uses a system function called .codn sys:rplaca , which differs from .code rplaca in that it returns the stored value, rather than the cell. .RE .NP* The Clobber Expander .synb .mets (lambda >> ( simple-setter-sym < place-form .mets \ \ \ \ \ \ \ \ << body-form ) ...) .syne .desc The clobber expander is a code-writer similar to the update expander. It takes a .meta body-form argument, and returns a larger form which surrounds this form with additional program code. The returned block of code has one main abstract action. It must arrange for the evaluation of .meta body-form in a lexical environment in which a lexical macro or lexical function exists which has the name requested by the .meta simple-setter-sym argument. The simple setter local macro written by the clobber expander is similar to the local setter written by the update expander. It has exactly the same interface, performs the same action of storing a value into the place, and returns the new value. The difference is that its logic may be considerably simplified by the assumption that the place is being subject to exactly one store, and no access. A place update macro which uses a clobber expander, and calls it more than once, break the assumption; doing so may result in multiple evaluations of the .metn place-form . .NP* The Delete Expander .synb .mets (lambda >> ( deleter-sym < place-form .mets \ \ \ \ \ \ \ \ << body-form ) ...) .syne .desc The delete expander is a code-writer similar to clobber expander. It takes a .meta body-form arguments, and returns a larger form which surrounds this form with additional program code. The returned block of code has one main abstract action. It must arrange for the evaluation of .meta body-form in a lexical environment in which a lexical macro or lexical function exists which has the name requested by the .meta deleter-sym argument. The deleter macro written by the clobber expander takes no arguments. It may be called at most once. It returns the previous value of the place, and arranges for its obliteration, whatever that means for that particular kind of place. .coNP Macro @ with-update-expander .synb .mets (with-update-expander >> ( getter << setter ) < place < env .mets \ << body-form ) .syne .desc The .code with-update-expander macro evaluates the .meta body-form argument, whose result is expected to be a Lisp form. The macro adds additional code around this code, and the result is returned. This additional code is called the .IR "place-access code" . The .meta getter and .meta setter arguments must be symbols. Over the evaluation of the .metn body-form , these symbols are bound to the names of local functions which are provided in the place-access code. The .meta place argument is a form which evaluates to a syntactic place. The generated place-access code is based on this place. The .meta env argument is a form which evaluates to a macro-expansion-time environment. The .code with-update-expander macro uses this environment to perform macro-expansion on the value of the .meta place form, to obtain the correct update expander function for the fully macro-expanded place. The place-access code is generated by calling the update expander for the expanded version of .codn place . .TP* "Example:" The following is an implementation of the .code swap macro, which exchanges the contents of two places. Two places are involved, and, correspondingly, the .code with-update-expander macro is used twice, to add two instances of place-update code to the macro's body. .cblk (defmacro swap (place-0 place-1 :env env) (with-gensyms (tmp) (with-update-expander (getter-0 setter-0) place-0 env (with-update-expander (getter-1 setter-1) place-1 env ^(let ((,tmp (,getter-0))) (,setter-0 (,getter-1)) (,setter-1 ,tmp)))))) .cble The basic logic for swapping two places is contained in the code template: .cblk ^(let ((,tmp (,getter-0))) (,setter-0 (,getter-1)) (,setter-1 ,tmp)) .cble The temporary variable named by the .code gensym symbol .code tmp is initialized by calling the getter function for .metn place-0 . Then the setter function of .meta place-0 is called in order to store the value of .meta place-1 into .metn place-0 . Finally, the setter for .meta place-1 is invoked to store the previously saved temporary value into that place. The name for the temporary variable is provided by the .code with-gensyms macro, but establishing the variable is the caller's responsibility; this is seen as an explicit .code let binding in the code template. The names of the getter and setter functions are similarly provided by the .code with-update-expander macros. However, binding those functions is the responsibility of that macro. To achieve this, it adds the place-access code to the code generated by the .code ^(let ...) backquote template. In the following example macro-expansion, the additional code added around the template is seen. It takes the form of two .code macrolet binding blocks, each added by an invocation of .codn with-update-expander : .cblk (macroexpand '(swap a b)) --> (macrolet ((#:g0036 () 'a) ;; getter macro for a (#:g0037 (val-expr) ;; setter macro for a (append (list 'sys:setq) (list 'a) (list val-expr)))) (macrolet ((#:g0038 () 'b) ;; getter macro for b (#:g0039 (val-expr) ;; setter macro for b (append (list 'sys:setq) (list 'b) (list val-expr)))) (let ((#:g0035 (#:g0036))) ;; temp <- a (#:g0037 (#:g0038)) ;; a <- b (#:g0039 #:g0035)))) ;; b <- temp .cble In this expansion, for example .code #:g0036 is the generated symbol which forms the value of the .code getter-0 variable in the .code swap macro. The getter is a macro which simply expands to a .codn a : straightforward access to the variable a. Of course, .code #:g0035 is nothing but the value of the .code tmp variable. Thus the swap macro's .cblk ^(let ((,tmp (,getter-0))) ...) .cble has turned into .cblk ^(let ((#:g0035 (#:g0036))) ...) .cble A full expansion, with the .code macrolet local macros expanded out: .cblk (sys:expand '(swap a b)) --> (let ((#:g0035 a)) (sys:setq a b) (sys:setq b #:g0035)) .cble In other words, the original syntax .cblk (,getter-0) .cble became .cblk (#:g0036) .cble and finally just .codn a . Similarly, .cblk (,setter-0 (,getter-1)) .cble became the .code macrolet invocations .cblk (#:g0037 (#:g0038)) .cble which finally turned into: .codn "(sys:setq a b)" . .coNP Macro @ with-clobber-expander .synb .mets (with-clobber-expander <> ( simple-setter ) < place < env .mets \ << body-form ) .syne .desc The .code with-clobber-expander macro evaluates .metn body-form , whose result is expected to be a Lisp form. The macro adds additional code around this form, and the result is returned. This additional code is called the .IR "place-access code" . The .meta simple-setter argument must be a symbol. Over the evaluation of the .metn body-form , this symbol is bound to the name of a functions which are provided in the place-access code. The .meta place argument is a form which evaluates to a syntactic place. The generated place-access code is based on this place. The .meta env argument is a form which evaluates to a macro-expansion-time environment. The .code with-clobber-expander macro uses this environment to perform macro-expansion on the value of the .meta place form, to obtain the correct update expander function for the fully macro-expanded place. The place-access code is generated by calling the update expander for the expanded version of .codn place . .TP* "Example:" The following implements a simple assignment statement, similar to .code set except that it only handles exactly two arguments: .cblk (defmacro assign (place new-value :env env) (with-clobber-expander (setter) place env ^(,setter ,new-value))) .cble Note that the correct evaluation order of .code place and .code new-value is taken care of, because .code with-clobber-expander generates the code which performs all the necessary evaluations of .codn place . This evaluation occurs before the code which is generated by .cblk ^(,setter ,new-value) .cble part is evaluated, and that code is what evaluates .codn new-value . Suppose that a macro were desired which allows assignment to be notated in a right to left style, as in: .cblk (assign 42 a) ;; store 42 in variable a .cble Now, the new value must be evaluated prior to the place, if left to right evaluation order is to be maintained. The standard .code push macro has this property: the push value is on the left, and the place is on the right. Now, the code has to explicitly take care of the order, like this: .cblk ;; WRONG! We can't just swap the parameters; ;; place is still evaluated first, then new-value: (defmacro assign (new-value place :env env) (with-clobber-expander (setter) place env ^(,setter ,new-value))) ;; Correct: arrange for evaluation of new-value first, ;; then place: (defmacro assign (new-value place :env env) (with-gensym (tmp) ^(let ((,tmp ,new-value)) ,(with-clobber-expander (setter) place env ^(,setter ,tmp))))) .cble .coNP Macro @ with-delete-expander .synb .mets (with-delete-expander <> ( deleter ) < place < env .mets \ << body-form ) .syne .desc The .code with-delete-expander macro evaluates .metn body-form , whose result is expected to be a Lisp form. The macro adds additional code around this code, and the resulting code is returned. This additional code is called the .IR "place-access code" . The .meta deleter argument must be a symbol. Over the evaluation of the .metn body-form , this symbol is bound to the name of a functions which are provided in the place-access code. The .meta place argument is a form which evaluates to a syntactic place. The generated place-access code is based on this place. The .meta env argument is a form which evaluates to a macro-expansion-time environment. The .code with-delete-expander macro uses this environment to perform macro-expansion on the value of the .meta place form, to obtain the correct update expander function for the fully macro-expanded place. The place-access code is generated by calling the update expander for the expanded version of .codn place . .TP* "Example:" The following implements the .code del macro: .cblk (defmacro del (place :env env) (with-delete-expander (deleter) place env ^(,deleter))) .cble .coNP Function @ call-update-expander .synb .mets (call-update-expander < getter < setter < place < env << body-form ) .syne .desc The .code call-update-expander function provides an alternative interface for making use of an update expander, complementary to .codn with-update-expander . Arguments .meta getter and .meta setter are symbols, provided by the caller. These are passed to the update expander function, and are used for naming local functions in the generated code which the update expander adds to .metn body-form . The .meta place argument is a place which has not been subject to macro-expansion. The .code call-update-expander function takes on the responsibility for macro-expanding the place. The .meta env parameter is the macro-expansion environment object required to correctly expand .code place in its original environment. The .meta body-form argument represents the source code of a place update operation. This code makes references to the local functions whose names are given by .meta getter and .metn setter . Those arguments allow the update expander to write these functions with the matching names expected by .metn body-form . The return value is an object representing source code which incorporates the .metn body-form , augmenting it with additional code which evaluates .code place to determine its location, and provides place accessor local functions expected by the .metn body-form . .TP* "Example:" The following shows how to implement a .code with-update-expander macro using .codn call-update-expander : .cblk (defmacro with-update-expander ((getter setter) unex-place env body) ^(with-gensyms (,getter ,setter) (call-update-expander ,getter ,setter ,unex-place ,env ,body))) .cble Essentially, all that .code with-update-expander does is to choose the names for the local functions, and bind them to the local variable names it is given as arguments. Then it calls .codn call-update-expander . .TP* "Example:" Implement the swap macro using .codn call-update-expander : .cblk (defmacro swap (place-0 place-1 :env env) (with-gensyms (tmp getter-0 setter-0 getter-1 setter-1) (call-update-expander getter-0 setter-0 place-0 env (call-update-expander getter-1 setter-1 place-1 env ^(let ((,tmp (,getter-0))) (,setter-0 (,getter-1)) (,setter-1 ,tmp)))))) .cble .coNP Function @ call-clobber-expander .synb .mets (call-clobber-expander < simple-setter < place < env << body-form ) .syne .desc The .code call-clobber-expander function provides an alternative interface for making use of a clobber expander, complementary to .codn with-clobber-expander . Argument .meta simple-setter is a symbol, provided by the caller. It is passed to the clobber expander function, and is used for naming a local function in the generated code which the update expander adds to .metn body-form . The .meta place argument is a place which has not been subject to macro-expansion. The .code call-clobber-expander function takes on the responsibility for macro-expanding the place. The .meta env parameter is the macro-expansion environment object required to correctly expand .code place in its original environment. The .metn body-form argument represents the source code of a place update operation. This code makes references to the local function whose name is given by .metn simple-setter . That argument allows the update expander to write this function with the matching name expected by .metn body-form . The return value is an object representing source code which incorporates the .metn body-form , augmenting it with additional code which evaluates .code place to determine its location, and provides the clobber local function to the .metn body-form . .coNP Function @ call-delete-expander .synb .mets (call-delete-expander < deleter < place < env << body-form ) .syne .desc The .code call-delete-expander function provides an alternative interface for making use of a delete expander, complementary to .codn with-delete-expander . Argument .meta deleter is a symbol, provided by the caller. It is passed to the delete expander function, and is used for naming a local function in the generated code which the update expander adds to .metn body-form . The .meta place argument is a place which has not been subject to macro-expansion. The .code call-delete-expander function takes on the responsibility for macro-expanding the place. The .meta env parameter is the macro-expansion environment object required to correctly expand .code place in its original environment. The .meta body-form argument represents the source code of a place delete operation. This code makes references to the local function whose name is given by .metn deleter . That argument allows the update expander to write this function with the matching name expected by .metn body-form . The return value is an object representing source code which incorporates the .metn body-form , augmenting it with additional code which evaluates .code place to determine its location, and provides the delete local function to the .metn body-form . .coNP Macro @ define-modify-macro .synb .mets (define-modify-macro < name < parameter-list << function-name ) .syne .desc The .code define-modify-macro macro provides a simplified way to write certain kinds of place update macros. Specifically, it provides a way to write place update macros which modify a place by retrieving the previous value, pass it through a function (perhaps together with some additional arguments), and then store the resulting value back into the place and return it. The .meta name parameter specifies the name for the place update macro to be written. The .meta function-name parameter must specify a symbol: the name of the update function. The update macro and update function both take at least one parameter: the place to be updated, and its value, respectively. The .meta parameter-list specifies the additional parameters for update function, which will also become additional parameters of the macro. Because it is a function parameter list, it cannot use the special destructuring features of macro parameter lists, or the .code :env or .code :whole special parameters. It can use optional parameters. Of course, it may be empty. The .code define-modify-macro macro writes a macro called .metn name . The leftmost parameter of this macro is a place, followed by the additional arguments specified by .metn parameter-list . The macro will arrange for the evaluation of the place argument to determine the place location. It will then retrieve and save the prior value of the place, and evaluate the remaining arguments. The prior value of the place, and the values of the additional arguments, are all passed to .meta function and the resulting value is then stored back into the location previously determined for .metn place . .TP* "Example:" Some standard place update macros are implementable using .codn define-modify-macro , such as .codn inc . The .code inc macro reads the old value of the place, then passes it through the .code + (plus) function, along with an extra argument: the delta value, which defaults to one. The .code inc macro could be written using .code define-modify-macro as follows: .cblk (define-modify-macro inc (: (delta 1)) +) .cble Note that the argument list .code (: (delta 1)) doesn't specify the place, because the place is the implicit leftmost argument of the macro which isn't given a name. With the above definition in place, when .code (inc (car a)) is invoked, then .code (car a) is first reduced to a location, and that location's value is retrieved and saved. Then the .code delta parameter s evaluated to its value, which has defaulted to 1, since the argument was omitted. Then these two values are passed to the .code + function, and so 1 is added to the value previously retrieved from .codn (car a) . The resulting sum is then stored back .code (car a) without, of course, evaluating .code (car a) again. .coNP Macro @ defplace .synb .mets (defplace < place-destructuring-args < body-sym .mets \ \ \ \ \ \ \ \ \ >> ( getter-sym < setter-sym << update-body ) .mets \ \ \ \ \ \ \ \ \ >> [( ssetter-sym << clobber-body ) .mets \ \ \ \ \ \ \ \ \ \ >> [( deleter-sym << delete-body )]]) .syne .desc The .code defplace macro is used to introduce a new kind of syntactic place. It writes the update expander, and optionally clobber and delete expander functions, from a simpler, more compact specification, and automatically registers the resulting functions. The compact specification of a .code defplace call contains only code fragments for the expander functions. The name and syntax of the place is determined by the .meta place-destructuring-args argument, which is macro-style parameter list whose structure mimics that of the the place. In particular, its leftmost symbol gives the name under which the place is registered. The .code defplace macro provides automatic destructuring of the syntactic place, so that the expander code fragments can refer to the components of a place by name. The .meta body-sym parameter must be be a symbol. This symbol will capture the .meta body-forms parameter which is passed to the update expander, clobber expander or delete expander. The code fragments then have access to the the body forms via this name. The .metn getter-sym , .metn setter-sym , and .metn update-body parenthesized triplet specify the update expander fragment. The .code defplace macro will bind .meta getter-sym and .meta setter-sym to symbols. The .meta update-body must then specify a template of code which evaluates the syntactic place to determine its storage location, and provides a pair of local functions, using these two symbols as their name. The template must also insert the .meta body-sym forms into the scope of these local functions, and the place determining code. The .meta setter-sym and .meta clobber-body arguments similarly specify an optional clobber expander fragment, as a single optional argument. If specified, the .meta clobber-body must generate a local function named using .meta setter-sym wrapped around .meta body-sym forms. The .meta deleter-sym and .meta deleter-body likewise specify a delete expander fragment. If this is omitted, then the place shall not support deletion. .TP* "Example:" Implementation of the place denoting the .code car field of .code cons cells: .cblk (defplace (car cell) body ;; the update expander fragment (getter setter (with-gensyms (cell-sym) ;; temporary symbol for cell ^(let ((,cell-sym ,cell)) ;; evaluate place to cell ;; getter and setter access cell via temp var (macrolet ((,getter () ^(car ,',cell-sym)) (,setter (val) ^(sys:rplaca ,',cell-sym ,val))) ;; insert body form from place update macro ,body)))) ;; clobber expander fragment: simpler: no need ;; to evaluate cell to temporary variable. (ssetter ^(macrolet ((,ssetter (val) ^(sys:rplaca ,',cell ,val))) ,body)) ;; deleter: delegate to pop semantics: ;; (del (car a)) == (pop a). (deleter ^(macrolet ((,deleter () ^(pop ,',cell))) ,body))) .cble .coNP Macro @ define-place-macro .synb .mets (define-place-macro < name < macro-style-params .mets \ \ << body-form *) .syne .desc In some situations, an equivalence exists between two forms, only one of which is recognized as a place. The .code define-place-macro macro can be used to establish a form as a place in terms of a translation to an equivalent form which is already a place. The .code define-place-macro has the same syntax as .codn defmacro . It specifies a macro transformation for a compound form which has the .meta name symbol in its leftmost position. This macro expansion is applied when such a form is used as a place. It is applied after all other expansions, and no other macro-expansions are applied afterward. .TP* "Example:" Implementation of .code first in terms of .codn car : .cblk (define-place-macro first (obj) ^(car ,obj)) .cble .coNP Operator @ defmacro .coNP Macro @ rlet .synb .mets (rlet >> ({( sym << init-form )}*) << body-form *) .syne .desc The macro .code rlet is similar to the .code let operator. It establishes bindings for one or more .metn sym -s, which are initialized using the values of .metn init-form -s. Note that the simplified syntax for a variable which initializes to .code nil by default is not supported by .codn rlet ; that is to say, the syntax .meta sym cannot be used in place of the .meti >> ( sym << init-form ) syntax when .meta sym is to be initialized to .codn nil . The .code rlet macro differs from .code let in that .code rlet assumes that those .metn sym -s which have constant .metn init-form -s (according to the .code constantp function) may be safely implemented as a symbol macro rather than a lexical variable. Therefore .code rlet is suitable in situations in which simpler code is desired from the output of certain kinds of machine-generated code, which binds local symbols: code with fewer temporary variables. On the other hand, .code rlet is not suitable in situations when true variables are required, which are assignable, and provide temporary storage. .TP* "Example:" .cblk ;; WRONG! Exchange two variables, a and b: (rlet ((temp a)) (set a b) (set b temp)) ;; Demonstration of constant-propagation (let ((a 42)) (rlet ((x 1) (y a)) (+ x y))) --> 43 (sys:expand '(let ((a 42)) (rlet ((x 1) (y a)) (+ x y)))) --> (let ((a 42)) (let ((y a)) (+ 1 y))) .cble The last example shows that the .code x variable has disappeared in the expansion. The .code rlet macro turned it into into a .code symacrolet denoting the constant 1, which then propagated to the use site, turning the expression .code (+ x y) into .codn (+ 1 y) . .coNP Macro @ with-gensyms .synb .mets (with-gensyms <> ( sym *) << body-form *) .syne .desc The .code with-gensyms evaluates the .metn body-form -s in an environment in which each variable name symbol .meta sym is bound to a new uninterned symbol ("gensym"). .TP* "Example:" The code: .cblk (let ((x (gensym)) (y (gensym)) (z (gensym))) ^(,x ,y ,z)) .cble may be expressed more conveniently using the .code with-gensyms shorthand: .cblk (with-gensyms (x y z) ^(,x ,y ,z)) .cble .SS* Debugging Functions .coNP Functions source-loc and source-loc-str .synb .mets (source-loc << form ) .mets (source-loc-str << form ) .syne .desc These functions map an expression in a \*(TX program to the file name and line number of the source code where that form came from. The .code source-loc function returns the raw information as a cons cell whose .cod3 car / cdr consist of the line number, and file name. The .code source-loc-str function formats the information as a string. If .meta form is not a piece of the program source code that was constructed by the \*(TX parser, then .code source-loc returns .codn nil , and .code source-loc-str returns a string whose text says that source location is not available. .coNP Function @ rlcp .synb .mets (rlcp < dest-form << source-form ) .syne .desc The .code rlcp function copies the source code location info ("rl" means "read location") from the .meta source-form object to the .meta dest-form object. These objects are pieces of list-based syntax. Note: the function is intended to be used in macros. If a macro transforms .meta source-form to .metn dest-form , this function can be used to propagate the source code location info also, so that when the \*(TL evaluator encounters errors in transformed code, it can give diagnostics which refer to the original untransformed source code. .SS* Profiling .coNP Operator @ prof .synb .mets (prof << form *) .syne .desc The .code prof operator evaluates the enclosed forms from left to right similarly to .codn progn , while determining the memory allocation requests and time consumed by the evaluation of the forms. If there are no forms, the prof operator measures the smallest measurable operation of evaluating nothing and producing .codn nil . If the evaluation terminates normally (not abruptly by a non-local control transfer), then .code prof yields a list consisting of: .cblk .mets >> ( value < malloc-bytes < gc-bytes << milliseconds ) .cble where .meta value is the value returned by the rightmost .metn form , or .code nil if there are no forms, .meta malloc-bytes is the total number of bytes of all memory allocation requests (or at least those known to the \*(TX runtime, such as those of all internal objects), .meta gc-bytes is the total number of bytes drawn from the garbage-collected heaps, and .meta milliseconds is the total processor time consumed over the execution of those forms. Notes: The bytes allocated by the garbage collector from the C function .code malloc to create heap areas are not counted as .metn malloc-bytes . .meta malloc-bytes includes storage such as the space used for dynamic strings, vectors and bignums (in addition to their gc-heap allocated nodes), and the various structures used by the .code cobj type objects such as streams and hashes. Objects in external libraries that use uninstrumented allocators are not counted: for instance the C .code FILE * streams. .coNP Macro @ pprof .synb .mets (pprof << form *) .syne .desc The .code pprof (pretty-printing .codn prof ) macro is similar to .codn progn . It evaluates .metn form s, and returns the rightmost one, or .code nil if there are no forms. Over the evaluation of .metn form s, it counts memory allocations, and measures CPU time. If .metn form s terminate normally, then just prior to returning, .code pprof prints these statistics in a concise report on the .codn *std-output* . The .code pprof macro relies on the .code prof operator. .SS* Garbage Collection .coNP Function @ sys:gc .synb (sys:gc) .syne .desc The .code gc function triggers garbage collection. Garbage collection means that unreachable objects are identified and reclaimed, so that their storage can be re-used. The function returns .code nil if garbage collection is disabled (and consequently nothing is done), otherwise .codn t . .coNP Function @ sys:gc-set-delta .synb .mets (sys:gc-set-delta << bytes ) .syne .desc The .code gc-set-delta function sets the GC delta parameter. Note: This function may disappear in a future release of \*(TX or suffer a backward-incompatible change in its syntax or behavior. When the amount of new dynamic memory allocated since the last garbage collection equals or exceeds the GC delta, a garbage collection pass is triggered. From that point, a new delta begins to be accumulated. Dynamic memory is used for allocating heaps of small garbage-collected objects such as cons cells, as well as the satellite data attached to some objects: like the storage arrays of vectors, strings or bignum integers. Most garbage collector behaviors are based on counting objects in the heaps. Sometimes a program works with a small number of objects which are very large, frequently allocating new, large objects and turning old ones into garbage. For instance a single large integer could be many megabytes long. In such a situation, a small number of heap objects therefore control a large amount of memory. This requires garbage collection to be triggered much more often than when working with small objects, such as conses, to prevent runaway allocation of memory. It is for this reason that the garbage collector uses the GC delta. There is a default GC delta of 64 megabytes. This may be overridden in special builds of \*(TX for small systems. .coNP Function @ finalize .synb .mets (finalize < object << function ) .syne .desc The .code finalize function registers .meta function to be invoked in the situation when .meta object is identified by the garbage collector as unreachable. This function is called a finalizer. If and when this situation occurs, the finalizer .meta function will be called with .meta object as its only argument. Finalizers are called in the same order in which they are registered: newer registrations are called after older registrations. If .meta object is registered multiple times by multiple calls to .codn finalize , then if those finalizers are called, they are all called, in the order of registration. After a finalization call takes place, its registration is removed; .meta object and .meta function are no longer associated. However, neither .meta object nor .meta function are reclaimed immediately; they are treated as if they were reachable objects until at least the next garbage collection pass. It is therefore safe for .meta function to store somewhere a persistent reference to .meta object or to itself, thereby reinstating these objects as reachable. .meta function is itself permitted to call .code finalize to register the original .code object or any other object for finalization. Such registrations made during finalization execution are not eligible for the current phase of finalization processing; they will be processed in a later garbage collection pass. .SS* Modularization .coNP Special variable @ *self-path* .desc This variable holds the invocation path name of the \*(TX program. .coNP Special variable @ stdlib The .code stdlib variable expands to the directory where the \*(TX standard library is installed. Note: there is no need to use the value of this variable to load library modules. Library modules are keyed to specific symbols, and lazily loaded. When a \*(TL library function, macro or variable is referenced for the first time, the library module which defines it is loaded. This automatic loading happens during the code expansion phase, or "macro time", so it works for macros. In the middle of a syntax tree of code, code expander can encounters a symbol which triggers a library load which defines a macro. When that load completes, the code expander can expand that macro using the newly created definition. .coNP Macro @ load .synb .mets (load << target ) .syne .desc The .code load macro causes a file of \*(TL code to be read and evaluated. The .meta target argument is a string. If .meta target specifies a relative pathname, then it is assumed to be a reference relative to the directory of the file in which the .code load macro form occurs. If .meta target has no suffix, then .code load first tries to load the un-suffixed name. If that cannot be opened, the .str .tl suffix is added to the path and another attempt is made. If that fails, an exception is thrown. If .meta target has a .str .txr suffix, it is assumed to be a \*(TX query language file, and an exception of type .code eval-error is thrown, since this is not supported. If .meta target is successfully resolved and opened, \*(TL forms are read from the file in succession. Each form is evaluated as if by the .code eval function, before the next form is read. If a syntax error is encountered, an exception of type .code eval-error is thrown. Parser error messages are directed to the .code *stderr* stream. .SS* Scoped Resource Management .coNP Macro @ with-resources .synb .mets (with-resources >> ({ sym >> [ init-form <> [ cleanup-form ])}*) .mets \ \ << body-form *) .syne .desc The .code with-resources macro provides a sequential binding construct similar to .codn let* . Every .meta sym is established as a variable which is visible to the .metn init-form -s of subsequent variables, to all subsequent .metn cleanup-form -s including that of the same variable, and to the .metn body-form -s. If no .meta init-form is supplied, then .meta sym is bound to the value .codn nil . If an .meta init-form is supplied, but no .metn cleanup-form , then .meta sym is bound to the value of the .metn init-form . If a .meta cleanup-form is supplied in addition to .metn init-form , it specifies code to be executed upon the termination of the entire .code with-resources construct. When an instance of .code with-resources terminates, all of the .metn cleanup-form -s specified in its binding clauses are evaluated, in reverse (right-to-left) order. The value of the last .meta body-form is returned, or else .code nil if no .metn body-form -s are present. .TP* "Example:" The following opens a text file and reads a line from it, returning that line, while ensuring that the stream is closed immediately: .cblk (with-resources ((f (open-file "/etc/motd") (close-stream f))) (whilet ((l (get-line f))) (put-line l))) .cble .SS* Debugger \*(TX has a simple, crude, built-in debugger. The debugger is invoked by adding the .code -d command line option to an invocation of \*(TX. In this debugger it is possible to step through code, set breakpoints, and examine the variable binding environment. Prior to executing any code, the debugger waits at the .code txr> prompt, allowing for the opportunity to set breakpoints. Help can be obtained with the .code h or .code ? command. Whenever the program stops at the debugger, it prints the Lisp-ified piece of syntax tree that is about to be interpreted. It also shows the context of the input being matched. The s command can be used to step into a form; n to step over. Sometimes the behavior seems counter-intuitive. For instance stepping over a .code @(next) directive actually means skipping everything which follows it. This is because the query material after a .code @(next) is actually child nodes in the abstract syntax tree node of the .code next directive, whereas the surface syntax appears flat. .coNP Sample Session Here is an example of the debugger being applied to a web scraping program which connects to a US NAVY clock server to retrieve a dynamically-generated web page, from which the current time is extracted, in various time zones. The handling of the web request is done by the wget command; the \*(TX query opens a wget command as and scans the body of the HTTP response containing HTML. This is the code, saved in a file called navytime.txr: .cblk @(next `!wget -c http://tycho.usno.navy.mil/cgi-bin/timer.pl -O - 2> /dev/null`) What time is it?

US Naval Observatory Master Clock Time

  @(collect :vars (MO DD HH MM SS (PM "  ") TZ TZNAME))
  
@MO. @DD, @HH:@MM:@SS @(maybe)@{PM /PM/} @(end)@TZ@/\et+/@TZNAME
  @  (until)
@/.*/ @(end)

US Naval Observatory @(output) @ (repeat) @MO-@DD @HH:@MM:@SS @PM @TZ @ (end) @(end) .cble This is the debug session: .cblk $ txr -d navytime.txr stopped at line 1 of navytime.txr form: (next (sys:quasi "!wget -c http://tycho.usno.navy.mil/cgi-bin/timer.pl -O - 2> /dev/null")) depth: 1 data (nil): nil .cble The user types .code s to step into the .code (next ...) form. .cblk txr> s stopped at line 2 of navytime.txr form: (sys:text " 1+ #\espace) "HTML" (# 1+ #\espace) "PUBLIC" (# 1+ #\espace) "\e"-//W3C//DTD" (# 1+ #\espace) "HTML" (# 1+ #\espace) "3.2" (# 1+ #\espace) "Final\e"//EN>") depth: 2 data (1): "" txr> s .cble The current form now is a .code sys:text form which is an internal representation of a block of horizontal material. The pattern matching is in vertical mode at this point, and so the line of data is printed without an indication of character position. .cblk stopped at line 2 of navytime.txr form: (sys:text " 1+ #\espace) "HTML" (# 1+ #\espace) "PUBLIC" (# 1+ #\espace) "\e"-//W3C//DTD" (# 1+ #\espace) "HTML" (# 1+ #\espace) "3.2" (# 1+ #\espace) "Final\e"//EN>") depth: 3 data (1:0): "" . "" .cble The user types .code s to step in. .cblk txr> s stopped at line 2 of navytime.txr form: "" .cble Now, the form about to be processed is the first item of the .codn (sys:text ...) , the string .strn s stopped at line 2 of navytime.txr form: (# 1+ #\espace) depth: 4 data (1:9): "" .cble Control has now passed to the second element of the .codn (sys:text ...) , a regular expression which matches one or more spaces, generated by a single space in the source code according to the language rules. The input context shows that .str s stopped at line 2 of navytime.txr form: "HTML" depth: 4 data (1:10): "" .cble Now, the regular expression has matched and moved the position past the space; the facing input is now .strn "HTML ..." . The programmer then repeats the .code s command by hitting Enter. .cblk txr> stopped at line 2 of navytime.txr form: (# 1+ #\espace) depth: 4 data (1:14): "" txr> stopped at line 2 of navytime.txr form: "PUBLIC" depth: 4 data (1:15): "" txr> stopped at line 2 of navytime.txr form: (# 1+ #\espace) depth: 4 data (1:21): "" txr> stopped at line 2 of navytime.txr form: "\e"-//W3C//DTD" depth: 4 data (1:22): "" txr> stopped at line 2 of navytime.txr form: (# 1+ #\espace) depth: 4 data (1:34): "" .cble It is not evident from the session transcript, but during interactive use, the input context appears to be animated. Whenever the programmer hits Enter, the new context is printed and the dot appears to advance. Eventually the programmer becomes bored and place a breakpoint on line 15, where the .code @(output) block begins, and invokes the .code c command to continue the execution: .cblk txr> b 15 txr> c stopped at line 15 of navytime.txr form: (output (((repeat nil (((sys:var MO nil nil) "-" (sys:var DD nil nil) " " (sys:var HH nil nil) ":" (sys:var MM nil nil) ":" (sys:var SS nil nil) " " (sys:var PM nil nil) " " (sys:var TZ nil nil))) nil nil nil nil nil nil)))) depth: 2 data (16): "" .cble The programmer issues a .code v command to take a look at the variable bindings, which indicate that the .code @(collect) has produced some lists: .cblk txr> v bindings: 0: ((PM " " "PM" "PM" "PM" "PM" "PM" "PM") (TZNAME "Universal Time" "Eastern Time" "Central Time" "Mountain Time" "Pacific Time" "Alaska Time" "Hawaii-Aleutian Time") (TZ "UTC" "EDT" "CDT" "MDT" "PDT" "AKDT" "HAST") (SS "35" "35" "35" "35" "35" "35" "35") (MM "32" "32" "32" "32" "32" "32" "32") (HH "23" "07" "06" "05" "04" "03" "01") (DD "30" "30" "30" "30" "30" "30" "30") (MO "Mar" "Mar" "Mar" "Mar" "Mar" "Mar" "Mar")) .cble Then a continue command, which finishes the program, whose output appears: .cblk txr> c Mar-30 23:22:52 UTC Mar-30 07:22:52 PM EDT Mar-30 06:22:52 PM CDT Mar-30 05:22:52 PM MDT Mar-30 04:22:52 PM PDT Mar-30 03:22:52 PM AKDT Mar-30 01:22:52 PM HAST .cble .SS* Compatibility New \*(TX versions are usually intended to be backward-compatible with prior releases in the sense that documented features will continue to work in the same way. Due to new features, new versions of \*(TX will supply new behaviors where old versions of \*(TX would have produced an error, such as a syntax error. Though, strictly speaking, this means that something is working differently in a new version, replacing an error situation with functionality is usually not considered a deviation from backward-compatibility. When a change is introduced which is not backward compatible, \*(TX's .code -C option can be used to request emulation of old behavior. The option was introduced in \*(TX 98, and so the oldest \*(TX version which can be emulated is \*(TX 97. Here are version values which have a special meaning as arguments to the .code -C option, along with a description of what behaviors are affected. For each of these version values, the described behaviors are provided if .code -C is given an argument which is equal or lower. For instance .code -C 103 selects the behaviors described below for version 105, but not those for 102. .IP 109 The optional trailing semicolon on hex and octal codes in the \*(TX pattern language was introduced in 110. The feature is disabled with 109 or lower compatibility, so that .code @\ex21;a encodes .code !;a rather than the current behavior of encoding .codn !a . Also, in 109 and earlier, newlines were allowed in word list literals and word list quasiliterals. They were treated as a word-separating space. A backslash-escaped newline, and all whitespace around it, was deleted just like in ordinary literals, and did not separate words. The old behavior is emulated. .IP 107 Up through \*(TX 107, by accident, there was a function called .code flip as well as an operator by the same name. The function was renamed to .codn flipargs . Version 107 compatibility or earlier provides the function under the original name also. .IP 105 Provides the behavior that the .code open-file function automatically marks a stream open on a TTY devices as a real-time stream (subject to the availability of the POSIX .code isatty function). Also allows unrecognized backslash escape sequences in regular expression syntax to simply denote the escaped character literally, as was historically the case prior to \*(TX 106, so that .code \ez for instance denotes .codn z . As of \*(TX 106, these are diagnosed as errors. .IP 102 Up to \*(TX 102, the .code get-string function did not close the stream. This old behavior is emulated. .IP 101 Up to \*(TX 101, the .code make-like function incorrectly returned .code nil when converting the empty list .code nil to string type. This affects numerous generic sequence functions, causing their result to be .code nil instead of an empty string. .IP 100 Up to \*(TX 100, the .code split-str function had an undocumented behavior. When the .code sep argument was an empty string, it split the string into individual characters as if by calling .codn list-str . This behavior changed to the currently documented behavior starting in \*(TX 101. Also, the arguments of the .code where function, which introduced in \*(TX 91, were reversed starting in \*(TX 101. .IP 99 Up to \*(TX 99, the substitution of TXR Lisp expressions in .code @(output) directives and in the quasistrings of the pattern language exhibited the buggy behavior that if the TXR Lisp expression produced a list, the list was rendered as a parenthesized representation, or the text .code nil in the empty list case. Moreover, in the .code @(output) case, the value of TXR Lisp expressions was not subject to filtering. Starting with \*(TX 100, these issues are fixed, making the the behavior is consistent with the behavior of TXR Lisp quasiliterals. .IP 97 Up to \*(TX 97, the error exception symbols such as .code file-error were named with underscores, as in .codn file_error . These error symbols existed: .codn type_error , .codn internal_error , .codn numeric_error , .codn range_error , .codn query_error , .code file_error and .codn process_error . .coNP Variables @ *txr-version* and @ *lib-version* .desc The .code *txr-version* variable gives the version of the \*(TX executable. Programs can express conditional variable based on detecting the version. The .code *lib-version* variable gives the version of the installed library of \*(TX code accompanying the executable. It is expected that these two variables have an identical value. Any discrepancy in their value indicates an installation whose library or \*(TX executable were upgraded independently. Should such a situation arise in any system and cause a problem, \*(TX programs can be defensively coded against it with the help of these variables. Some features of the \*(TX library are built into the executable, whereas others are in the library directory. This aspect of library symbols isn't specified in this manual; knowing which of these two variables is relevant to a library feature requires familiarity with the implementation. .SH* Appendix .SS* A. NOTES ON EXOTIC REGULAR EXPRESSIONS Users familiar with regular expressions may not be familiar with the complement and intersection operators, which are often absent from text processing tools that support regular expressions. The following remarks are offered in hope that they are of some use. .TP* "Equivalence to Sets" Regexp intersection is not essential; it may be obtained from complement and union as follows, since De Morgan's law applies to regular expression algebra: .code (R1)&(R2) = .codn ~(~(R1)|~(R2)) . (The complement of the union of the complements of R1 and R2 constitutes the intersection.) This law works because the regular expression operators denote set operations in a straightforward way. A regular expression denotes a set of strings (a potentially infinite one) in a condensed way. The union of two regular expressions .code R1|R2 denotes the union of the set of texts denoted by .code R1 and that denoted by .codn R2 . Similarly .code R1&R2 denotes a set intersection, and .code ~R denotes a set complement. Thus algebraic laws that apply to set operations apply to regular expressions. It's useful to keep in mind this relationship between regular expressions and sets in understanding intersection and complement. Given a finite set of strings, like the set .cblk { "abc", "def" } .cble which corresponds to the regular expression .codn (abc|def) , the complement is the set which contains an infinite number of strings: it consists of all possible strings except .str abc and .strn def . It includes the empty string, all strings of length 1, all strings of length 2, all strings of length 3 other than .str abc and .strn def , all strings of length 4, etc. This means that a "harmless looking" expression like .code ~(abc|def) can actually match arbitrarily long inputs. .TP* "Set Difference" How about matching only three-character-long strings other than .str abc or .strn def ? To express this, regex intersection can be used: these strings are the intersection of the set of all three-character strings, and the set of all strings which are not .str abc or .strn def . The straightforward set-based reasoning leads us to this: .codn ...&~(abc|def) . This .code A&~B idiom is also called set difference, sometimes notated with a minus sign: .code A-B (which is not supported in \*(TX regular expression syntax). Elements which are in the set .codn A , but not .codn B , are those elements which are in the intersection of .code A with the complement of .codn B . This is similar to the arithmetic rule .codn A - B = A + -B : subtraction is equivalent to addition of the additive inverse. Set difference is a useful tool: it enables us to write a positive match which captures a more general set than what is intended (but one whose regular expression is far simpler than a positive match for the exact set we want), then we can intersect this over-generalized set with the complemented set of another regular expression which matches the particulars that we wish excluded. .TP* "Expressivity versus Power" It turns out that regular expressions which do not make use of the complement or intersection operators are just as powerful as expressions that do. That is to say, with or without these operators, regular expressions can match the same sets of strings (all regular languages). This means that for a given regular expression which uses intersection and complement, it is possible to find a regular expression which doesn't use these operators, yet matches the same set of strings. But, though they exist, such equivalent regular expressions are often much more complicated, which makes them difficult to design. Such expressions do not necessarily . B express what it is they match; they merely capture the equivalent set. They perform a job, without making it obvious what it is they do. The use of complement and intersection leads to natural ways of expressing many kinds of matching sets, which not only demonstrate the power to carry out an operation, but also easily express the concept. .TP* "Example: Matching C Language Comments" For instance, using complement, we can write a straightforward regular expression which matches C language comments. A C language comment is the digraph .codn /* , followed by any string which does not contain the closing sequence .codn */ , followed by that closing sequence. Examples of valid comments are .codn /**/ , .code /* abc */ or .codn /***/ . But C comments do not nest (cannot contain comments), so that .code /* /* nested */ */ actually consists of the comment .codn /* /* nested */ , which is followed by the trailing junk .codn */ . Our simple characterization of interior part of a C comment as a string which does not contain the terminating digraph makes use of the complement, and can be expressed using the complemented regular expression like this: .codn (~.*[*][/].*) . That is to say, strings which contain .code */ are matched by the expression .codn .*[*][/].* : zero or more arbitrary characters, followed by .codn */ , followed by zero or more arbitrary characters. Therefore, the complement of this expression matches all other strings: those which do not contain .codn */ . These strings make up the inside of a C comment between the .code /* and .codn */ . The equivalent simple regex is quite a bit more complicated. Without complement, we must somehow write a positive match for all strings such that we avoid matching .codn */ . Obviously, sequences of characters other than .code * are included: .codn [^*]* . Occurrences of .code * are also allowed, but only if followed by something other than a slash, so let's include this via union: .cblk ([^*]|[*][^/])*. .cble Alas, already, we have a bug in this expression. The subexpression .code [*][^/] can match .codn ** , since a .code * is not a .codn / . If the next character in the input is .codn / , we missed a comment close. To fix the problem we revise to this: .cblk ([^*]|[*][^*/])* .cble (The interior of a C language comment is a any mixture of zero or more non-asterisks, or digraphs consisting of an asterisk followed by something other than a slash or another asterisk). Oops, now we have a problem again. What if two asterisks occur in a comment? They are not matched by .codn [^*] , and they are not matched by .codn [*][^*/] . Actually, our regex must not simply match asterisk-non-asterisk digraphs, but rather sequences of one or more asterisks followed by a non-asterisk: .cblk ([^*]|[*]*[^*/])* .cble This is still not right, because, for instance, it fails to match the interior of a comment which is terminated by asterisks, including the simple test cases where the comment interior is nothing but asterisks. We have no provision in our expression for this case; the expression requires all runs of asterisks to be followed by something which is not a slash or asterisk. The way to fix this is to add on a subexpression which optionally matches a run of zero or more interior asterisks before the comment close: .cblk ([^*]|[*]*[^*/])*[*]* .cble Thus our the semi-final regular expression is .cblk [/][*]([^*]|[*]*[^*/])*[*]*[*][/] .cble (Interpretation: a C comment is an interior string enclosed in .codn /* */ , where this interior part consists of a mixture of non-asterisk characters, as well as runs of asterisk characters which are terminated by a character other than a slash, except for possibly one rightmost run of asterisks which extends to the end of the interior, touching the comment close. Phew!) One final simplification is possible: the tail part .code [*]*[*][/] can be reduced to .code [*]+[/] such that the final run of asterisks is regarded as part of an extended comment terminator which consists of one or more asterisks followed by a slash. The regular expression works, but it's cryptic; to someone who has not developed it, it isn't obvious what it is intended to match. Working out complemented matching without complement support from the language is not impossible, but it may be difficult and error-prone, possibly requiring multiple iterations of trial-and-error development involving numerous test cases, resulting in an expression that doesn't have a straightforward relationship to the original idea. .TP* "The Non-Greedy Operator" The non-greedy operator .code % is actually defined in terms of a set difference, which is in turn based on intersection and complement. The uninteresting case .code (R%) where the right operand is empty reduces to .codn (R*) : if there is no trailing context, the non-greedy operator matches .code R as far as possible, possibly to the end of the input, exactly like the greedy operator. The interesting case .code (R%T) is defined as a "syntactic sugar" which expands to the expression .code ((R*)&(~.*(T&.+).*))T which means: match the longest string which is matched by .codn R* , but which does not contain a non-empty match for .codn T ; then, match .codn T . This is a useful and expressive notation. With it, we can write the regular expression for matching C language comments simply like this: .code [/][*].%[*][/] (match the opening sequence .codn /* , then match a sequence of zero or more characters non-greedily, and then the closing sequence .codn */ . With the non-greedy operator, we don't have to think about the interior of the comment as set of strings which excludes .codn */ . Though the non-greedy operator appears expressive, its apparent simplicity may be deceptive. It looks as if it works "magically" by itself; "somehow" this .code .% part "knows" only to consume enough characters so that it doesn't swallow an occurrence of the trailing context. Care must be taken that the trailing context passed to the operator really is the correct text that should be excluded by the non-greedy match. For instance, take the expression .codn .%abc . If you intend the trailing context to be merely .codn a , you must be careful to write .codn (.%a)bc . Otherwise, the trailing context is .codn abc , and this means that the .code .% match will consume the longest string that does not contain .codn abc , when in fact what was intended was to consume the longest string that does not contain .codn a . The change in behavior of the .code % operator upon modifying the trailing context is not as intuitive as that of the * operator, because the trailing context is deeply involved in its logic. On a related note, for single-character trailing contexts, it may be a good idea to use a complemented character class instead. That is to say, rather than .codn (.%a)bc , consider .codn [^a]*abc . The set of strings which don't contain the character a is adequately expressed by .codn [^a]* .