bug#71503: 30.0.50; pcase-lambda (with "and" and "guard") does not work as expected

[Philip Kaludercic <philipk@posteo.net>]

[-- Attachment #1: Type: text/plain, Size: 1805 bytes --]

Eli Zaretskii <eliz@gnu.org> writes:

> Ping! Ping!
>
>> Cc: mekeor@posteo.de, monnier@iro.umontreal.ca, 71503@debbugs.gnu.org
>> Date: Sat, 06 Jul 2024 10:38:00 +0300
>> From: Eli Zaretskii <eliz@gnu.org>
>> 
>> Ping!
>> 
>> > Cc: mekeor@posteo.de, 71503@debbugs.gnu.org
>> > Date: Sat, 22 Jun 2024 11:37:02 +0300
>> > From: Eli Zaretskii <eliz@gnu.org>
>> > 
>> > > Cc: 71503@debbugs.gnu.org
>> > > From: Philip Kaludercic <philipk@posteo.net>
>> > > Date: Fri, 14 Jun 2024 08:42:25 +0000
>> > > 
>> > > Mekeor Melire <mekeor@posteo.de> writes:
>> > > 
>> > > > I was expecting these two expressions to evaluate to the same value.
>> > > > First, we call `pcase' on a value with a pattern involving `and' and
>> > > > `guard':
>> > > >
>> > > >     (pcase "value"
>> > > >       ((and v (guard (string= "not-value" v))) v))
>> > > >     ;; => nil
>> > > >
>> > > > Second, let's use the same value and pattern, but this time using
>> > > > `pcase-lambda':
>> > > >
>> > > >     (funcall
>> > > >       (pcase-lambda
>> > > >         ((and v (guard (string= "not-value" v)))) v)
>> > > >       "value")
>> > > >     ;; => "value"
>> > > >
>> > > > Am I missing something or is this a bug?
>> > > 
>> > > The difference is that pcase-lambda doesn't do case-distinction, but
>> > > just pattern matching/destruncting.  So if the pattern-matching fails,
>> > > then the variable is just not bound, instead of the entire expression
>> > > falling back to returning no value/nil.  I am guessing you wanted to
>> > > have something like Scheme's `case-lambda'[0]?  Or we could clarify this
>> > > point in the docstring.
>> > 
>> > Would you mind suggesting a clarification for the doc string (and the
>> > ELisp manual as well)?

Sorry, I didn't realise you were pining me.  I was thinking of something
like


[-- Attachment #2: Type: text/plain, Size: 920 bytes --]

diff --git a/lisp/emacs-lisp/pcase.el b/lisp/emacs-lisp/pcase.el
index 5a7f3995311..f546ba34c7c 100644
--- a/lisp/emacs-lisp/pcase.el
+++ b/lisp/emacs-lisp/pcase.el
@@ -241,9 +241,10 @@ pcase-exhaustive
 ;;;###autoload
 (defmacro pcase-lambda (lambda-list &rest body)
   "Like `lambda' but allow each argument to be a pattern.
-I.e. accepts the usual &optional and &rest keywords, but every
-formal argument can be any pattern accepted by `pcase' (a mere
-variable name being but a special case of it)."
+I.e. accepts the usual &optional and &rest keywords, but every formal
+argument can be any pattern accepted by `pcase' (a mere variable name
+being but a special case of it).  Keep in mind that BODY is always
+evaluated, regardless of whether the argument-patterns match or not."
   (declare (doc-string 2) (indent defun)
            (debug (&define (&rest pcase-PAT) lambda-doc def-body)))
   (let* ((bindings ())

[-- Attachment #3: Type: text/plain, Size: 493 bytes --]


but now I notice that `pcase-let*' documents that all expressions should
match,

  Each EXP should match its respective PATTERN (i.e. be of structure
  compatible to PATTERN); a mismatch may signal an error or may go
  undetected, binding variables to arbitrary values, such as nil.

Since `pcase-lambda' inherits these semantics we would have to propagate
this promise -- or be more specific about what happens, e.g. binding
values to nil.

>> > Thanks.

-- 
	Philip Kaludercic on peregrine