[Chicken-janitors] #1422: for-each behaves inconsistently when compiled versus interpreted on lists of unequal length

[Chicken Trac]

#1422: for-each behaves inconsistently when compiled versus interpreted on lists
of unequal length
-------------------------------------+-------------------------------------
 Reporter:  sjamaan                  |                 Owner:
     Type:  defect                   |                Status:  new
 Priority:  major                    |             Milestone:  5.1
Component:  core libraries           |               Version:  5.0
 Keywords:  compiler syntax, for-    |  Estimated difficulty:  medium
  each, inconsistency                |
-------------------------------------+-------------------------------------
 There's a compiler syntax registered for {{{for-each}}} (and for
 {{{map}}}, too) which rewrite/expands most {{{for-each}}} calls to a loop
 for performance reasons.

 While working on the "scheme" module at some point I forgot to patch this
 to apply to {{{scheme#for-each}}} so it didn't get applied, and I got an
 error in both the scrutinizer and a test. In both cases, for-each was
 being used on two lists of unequal length, causing for-each to bail out
 with an error.

 Example:

 {{{
 (for-each (lambda (x y) (print (+ x y))) '(1 2 3) '(1 2))
 }}}

 When compiled, this prints {{{2\n4}}}, when interpreted this gives the
 following (quite useless) error:

 {{{
 2
 4

 Error: (for-each) lists are not of same length: (())
 }}}

 I think this error message really needs to be improved (which will be a
 bit more costly in terms of garbage collection because the head is held
 onto regardless of how far we got in the loop). I believe {{{map}}} has
 the same issue.

 Also, giving the error after executing the procedure is questionable;
 maybe it should first check the lengths of the lists and ''then'' start
 traversing them?

 Finally, we need to decide how we want {{{for-each}}} to behave: like the
 compiler syntax or like the procedure. Either is fine by me, and judging
 what the standard says, I think both are fine too. It doesn't explicitly
 say "it is an error" when the lists aren't the same length, but it just
 says "the lists must be the same length".  R7RS explicitly allows the
 lists to be of unequal length, so maybe we should just allow this as an
 allowed extension to R5RS and get rid of the check.

--
Ticket URL: <https://bugs.call-cc.org/ticket/1422>
CHICKEN Scheme <https://www.call-cc.org/>
CHICKEN Scheme is a compiler for the Scheme programming language.