bug#26540: 25.2; [PATCH] Add cl-set-equal to test for set equality

[Damien Cassou <damien@cassou.me>]

Hi Michael,

Michael Heerdegen <michael_heerdegen@web.de> writes:
> Damien Cassou <damien@cassou.me> writes: 
>> it makes sense and I will try this way. Nevertheless, it also 
>> means giving up on the :key feature. I guess it's ok. 
> 
> OTOH I see no reason not to support it.  There is no reason to 
> provide a function in a library specializing on sequences with 
> less features than in some other lib.


I agree with you that having :key would be nice. Nevertheless, my 
implementation currently relies on functions of seq.el (i.e., 
seq-contains) which would have to be adapted to support :key. I 
didn't want to do that.


> [...] with this implementation using hash-tables: 
>[] 
> #+begin_src emacs-lisp (defun seq-set-equal-2 (sequence1 
> sequence2) 
>   (let ((table1 (make-hash-table :size (length sequence1))) 
>         (table2 (make-hash-table :size (length sequence2)))) 
>     (seq-doseq (elt sequence1) (puthash elt t table1)) 
>     (seq-doseq (elt sequence2) (puthash elt t table2)) (and 
>     (seq-every-p (lambda (elt) (gethash elt table2)) sequence1) 
>          (seq-every-p (lambda (elt) (gethash elt table1)) 
>          sequence2)))) 
> #+end_src 

as far as I can tell, little effort has been put in optimizing 
seq.el the way you describe it so I guess such an implementation 
of seq-set-equal would feel a bit alien in the current code 
base. Moreover, is your implementation faster on very small sets? 
Finally, making your implementation of seq-set-equal accepting a 
TESTFN parameter would be a bit complex as you would have to pass 
that to `make-hash-table` which also requires a hash function. 

-- 
Damien Cassou
http://damiencassou.seasidehosting.st

"Success is the ability to go from one failure to another without
losing enthusiasm." --Winston Churchill