bug#59328: 29.0.50; `seq-keep' implementation only valid for lists

[Eli Zaretskii]

> From: Michael Heerdegen <michael_heerdegen@web.de>
> Cc: 59328@debbugs.gnu.org,  larsi@gnus.org,  jonas@bernoul.li
> Date: Fri, 25 Nov 2022 10:47:31 +0100
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > Can tests for this be written in a way that they are only run if the
> > relevant packages are available on the user's system?  If so, I'd
> > prefer to have that than no tests at all.
> 
> I don't know.

AFAIK, 'require' can return nil if asked not to error out.

> Alternatively we could implement `seq-map' for an ad-hoc defined
> sequence type and test using that type, e.g. this expression:
> 
> #+begin_src emacs-lisp
> (progn
>   (defvar gensym)
>   (let ((gensym (make-symbol "foo")))
>     (eval `(cl-defmethod seq-map (function (thing (head ,gensym)))
>              (append (list (car thing) (cadr thing)) (seq-map function (cddr 
> thing))))
>           t)
>     (equal (list gensym nil 4 46)
>            (seq-keep (lambda (x) (and (integerp x) (* 2 x)))
>                      (list gensym nil 2 'x gensym 23)))))
> #+end_src
> 
> returns t with my patch installed and nil else and works without relying
> on something external.  I'm not sure if defining methods (for seq-map in
> this case) that are globally visible is allowed in tests, so I
> implemented the example above in a way that the change of the generic
> function is not visible from the outside (thus the "secret" gensym).
> 
> Would something like that be acceptable?
> 
> Sorry for my ignorance, I didn't write much tests before.

Sounds like over-engineering to me.

Like I said: it's your call.  If you see too many complications to adding a
test, and my suggestions don't convince you, I won't object to installing
your original proposal without a test.

Thanks.