awk macro: support regexes better in ranges.

* share/txr/stdlib/awk.tl (sys:awk-compile-time): New slot,
rng-rec-temp. Specifies the name of a temporary variable
which is scoped over the evaluation of ranges, and which
caches a reference to the current record string.
(sys:range-test): New function.
(awk:let): Rename the original rng macrolet to sys:rng,
and introduce rng as a wrapper around it which inserts
the logic for calling the regex or function that might
emerge from the evaluation of from-expr or to-expr.
(awk): If ranges are present in the syntax, then bind
the temporary variable which caches the record around
the evaluations of the ranges.

* txr.1: Document new special behavior of rng w.r.t
functions and regexes. Add admonition against modifying
rec and some other awk variables from range expressions.

2 files changed, 43 insertions, 5 deletions

diff --git a/share/txr/stdlib/awk.tl b/share/txr/stdlib/awk.tl
index 7c6ae504..07a5f3bc 100644
--- a/share/txr/stdlib/awk.tl
+++ b/share/txr/stdlib/awk.tl
@@ -48,6 +48,7 @@
   begin-actions end-actions
   cond-actions
   (nranges 0)
+  (rng-rec-temp (gensym))
   rng-expr-temps
   rng-exprs)
 
@@ -109,6 +110,11 @@
     (t (put-string self.rec)
        (put-string self.ors))))
 
+(defun sys:range-test (val rec)
+  (caseq (typeof val)
+    ((regex fun) (call val rec))
+    (t val)))
+
 (defun sys:awk-expander (clauses)
   (let ((awc (new sys:awk-compile-time)))
     (each ((cl clauses))
@@ -176,7 +182,7 @@
      (macrolet ((next () '(return-from :awk-rec))
                 (next-file () '(return-from :awk-file))
                 (prn (. args) ^(qref ,',aws-sym (prn ,*args)))
-                (rng (from-expr to-expr :env e)
+                (sys:rng (from-expr to-expr :env e)
                   (let ((ix (pinc (qref ,awc nranges)))
                         (rng-temp (gensym))
                         (from-expr-ex (sys:expand from-expr e))
@@ -193,6 +199,9 @@
                                (or (set flag ,flag-new) ,flag-old)))
                           (qref ,awc rng-exprs))
                     rng-temp))
+                (rng (from-expr to-expr)
+                  ^(sys:rng (sys:range-test ,from-expr ,(qref ,awc rng-rec-temp))
+                            (sys:range-test ,to-expr ,(qref ,awc rng-rec-temp))))
                 (ff (. opip-args)
                   ^(symacrolet ((f (rslot ,',aws-sym 'fields 'f-to-rec)))
                      (set f [(opip ,*opip-args) f])))
@@ -229,9 +238,10 @@
                               ,*awc.end-file-actions))))
                       (,awk-fun (lambda (,aws-sym)
                                   ,(if awc.rng-exprs
-                                     ^(let* ,(nreverse
-                                               (zip awc.rng-expr-temps
-                                                    awc.rng-exprs))
+                                     ^(let* ((,awc.rng-rec-temp rec)
+                                             ,*(nreverse
+                                                 (zip awc.rng-expr-temps
+                                                      awc.rng-exprs)))
                                         ,p-actions-xform)
                                      p-actions-xform))))
                  ,*awc.begin-actions
diff --git a/txr.1 b/txr.1
index 58506358..a47e776f 100644
--- a/txr.1
+++ b/txr.1
@@ -38771,7 +38771,25 @@ The expressions
 .meta from-condition
 and
 .meta to-condition
-are ordinary expressions which are evaluated; however, the are evaluated
+are ordinary expressions which are evaluated. However, their
+evaluation is unusual in two ways.
+
+Firstly, if either expression
+produces, as its result, a function or regular expression object,
+then that function or regular expression object is applied to
+the current record (value of the
+.code rec
+variable), and the result of that application is then taken
+as the result of the condition. This allows for expressions like
+.code "(rng (f^ #/start/) #/end/)"
+which denotes a range which begins with a record which
+begins with the prefix
+.str start
+and ends with a record which contains
+.str end
+as a substring.
+
+Secondly, the conditions are evaluated
 out of order with respect to the surrounding expression
 in which they occur.  Ranges and their constituent
 .meta from-condition
@@ -38813,6 +38831,16 @@ itself is evaluated, the
 expression merely fetches a previously computed Boolean value which indicates
 whether the range is active for this record.
 
+Also, the behavior is unspecified if range expressions attempt to modify
+the awk-special variables.
+.codn rec ,
+.codn f ,
+.codn fs ,
+.code ft
+or
+.codn kfs .
+It is not recommended to place any side effects into range expressions.
+
 Evaluation of ranges obeys the following logic, which is applied to
 each range, prior to the processing of condition-action clauses.
 If a range is not currently active, its