symbol equality --cl-rest-- help-function-arglist member vs member* and also equal, eql, and eq

symbol equality --cl-rest-- help-function-arglist member vs member* and also equal, eql, and eq

[MON KEY]

Why do symbol equality tests/comparisons with --cl-rest-- and
`help-function-arglist' be so freaky?

The only references I can find to --cl-rest-- in non-byte-compiled code are in:
`cl-do-arglist'       -> cl-macs.el'
`cl-macroexpand-all'  -> cl-extra.el

Is there a reason for the following behavior that I'm missing or is
this a bug (of some sort)?

;;; ==============================
(defun* test-cl-rest (arg1 arg2 &key arg-key)
  arg-key)

(test-cl-rest nil nil :arg-key "bubba")
;=> "bubba"

(help-function-arglist 'test-cl-rest)
;=> (arg1 arg2 &rest --cl-rest--)

(member '&rest
 (help-function-arglist 'test-cl-rest))
;=> (&rest --cl-rest--)

(member '--cl-rest--
 (help-function-arglist 'test-cl-rest))
;=> nil

(cadr (member '&rest (help-function-arglist 'test-cl-rest)))
;=> --cl-rest--

(cadr (member* '&rest '(bubba &rest --cl-rest--)))
;=> --cl-rest--

;; member & member
(equal
 (cadr (member '&rest (help-function-arglist 'test-cl-rest)))
 (cadr (member '&rest (help-function-arglist 'test-cl-rest))))
;=> t

;; member* & member*
(equal
 (cadr (member* '&rest '(bubba &rest --cl-rest--)))
 (cadr (member* '&rest '(bubba &rest --cl-rest--))))
;=> t

;; member & member*
(equal
 (cadr (member '&rest (help-function-arglist 'test-cl-rest)))
 (cadr (member* '&rest (help-function-arglist 'test-cl-rest))))
;=> t

;; member of arg list
(equal  '--cl-rest-- (cadr (member '&rest (help-function-arglist
'test-cl-rest))))
;=> nil

;; member* of arg list
(equal  '--cl-rest-- (cadr (member* '&rest (help-function-arglist
'test-cl-rest))))
;=> nil

;; member same symbol new list
(member '--cl-rest-- (member '&rest '(bubba &rest --cl-rest--)))
;=> (--cl-rest--)

;; member* same symbol new list
(member* '--cl-rest-- (member* '&rest '(bubba &rest --cl-rest--)))
;=> (--cl-rest--)

;; member same symbol new list
(equal '--cl-rest-- (car (member '--cl-rest-- (member '&rest '(bubba
&rest --cl-rest--)))))
;=> t

;; member* same symbol new list
(equal '--cl-rest-- (car (member* '--cl-rest-- (member* '&rest '(bubba
&rest --cl-rest--)))))
;=> t

;; member
(let (a)
  (setq a (member '&rest (help-function-arglist 'test-cl-rest)))
  (equal '--cl-rest-- (cadr a)))
;=> nil

;; member*
(let (a)
  (setq a (cadr (member* '&rest (help-function-arglist 'test-cl-rest))))
  (equal '--cl-rest-- a))
;=> nil

(let (c)
  (mapc #'(lambda (y) (when (or (eq y '--cl-rest)
                                (eql y '--cl-rest--)
                                (equal y '--cl-rest--))
                        (push y c )))
        (help-function-arglist 'test-cl-rest))
 c)
;=> nil

(equal (car '(--cl-rest--)) (car '(--cl-rest--)))
;=> t

(eql (car '(--cl-rest--)) (car '(--cl-rest--)))
;=> t

(eq (car '(--cl-rest-- 666)) (cadr '(666 --cl-rest--)))
;=> t

;; memq/member
(memq '--cl-rest-- (member '&rest (help-function-arglist 'test-cl-rest)))
;=> nil

;; memql/member
(memql '--cl-rest-- (member '&rest (help-function-arglist 'test-cl-rest)))
;=> nil

;; memq/member*
(memq '--cl-rest-- (member* '&rest (help-function-arglist 'test-cl-rest)))
=> nil

;; memq/member*
(memql '--cl-rest-- (member* '&rest (help-function-arglist 'test-cl-rest)))
=> nil

;; memql/member same-symbol new list
(memql '--cl-rest-- (member '&rest '(bubba &rest --cl-rest--)))
;=> (--cl-rest--)

;; memq same-symbol new list
(memq '--cl-rest-- '(arg1 arg2 &rest --cl-rest-- bubba))
;=> (--cl-rest-- bubba)

;; member*
(memql (cadr (member* '&rest '(bubba &rest --cl-rest--)))
 (help-function-arglist 'test-cl-rest))
;=> nil

;;; ==============================
s_P

Re: symbol equality --cl-rest-- help-function-arglist member vs member* and also equal, eql, and eq

[Miles Bader]

MON KEY <monkey@sandpframing.com> writes:
> Is there a reason for the following behavior that I'm missing

It looks like `--cl-rest--' is an uninterned symbol.

This is a good thing.

-Miles

-- 
Cannon, n. An instrument employed in the rectification of national boundaries.

Re: symbol equality --cl-rest-- help-function-arglist member vs member* and also equal, eql, and eq

[MON KEY]

Stephen
> Apparently --cl-rest-- is an uninterned symbol.  This is done so that
> the symbol cannot shadow an definition in code outside the cl-arglist
> processing module.

Miles
> It looks like `--cl-rest--' is an uninterned symbol.
> This is a good thing.

Ok, thanks to you both for explaining.

However, I'm not able to make that to jibe... still missing something
I'm sure.

I understand why there needs to be a reserved 'anonymous' symbol for the
--rest-of-cl-- e.g. --cl-rest-- and that this symbol should remain uninterned
but what I am missing on is why --cl-keys-- is converted to `cl-keys' whereas
--cl-rest-- can't/isn't in the `cl-do-arglist' function? For example,

In the following both `body' and `--cl-rest--' show as being interned whereas
the `*-wombats' aren't:

(intern-soft "body") ;=> t
(intern-soft "--cl-rest--") ;=> t
(intern-soft "scary-flying-wombats") ;=> nil

(mapatoms (lambda (x)
            (when (string= (symbol-name x)
                           ;; "body")       ;=> (found body)
                           ;; "--cl-rest--") ;=> (found --cl-rest--)
                           "scary-flying-wombats") ;=> nil
              (prin1 `(found ,x))))
          obarray)

(intern-soft "more-wombats") ;=> nil
(make-symbol "more-wombats") ;=> more-wombats

(mapatoms (lambda (x)
            (when (string= (symbol-name x) "more-wombats")
              (prin1 `(found ,x))))
          obarray)
 ;=> nil

(member 'cl-keys (help-function-arglist 'reduce)) ;=> (cl-keys)
(member 'body (help-function-arglist 'unless)) ;=> (body)

(BTW Everything above was evaluated on two systems and with `emacs -Q'.
With the same results.)

Is not the the now-maybe arg/keyword acting acting in a manner similar
to &key's --cl-keys--, e.g. it
is let bound not setq'd:

(defun* 3rd-degree (q z w &key now-maybe)
  (let ((3d (make-symbol "now-maybe")))
    (when now-maybe
    3d)))

(help-function-arglist '3rd-degree)
;=> (q z w &rest --cl-rest--)

(member 'now-maybe (help-function-arglist '3rd-degree))
;=> nil

(intern-soft "now-maybe")
;=> now-maybe

(mapatoms (lambda (x)
            (when (string= (symbol-name x) "now-maybe")
              (prin1 `(found ,x))))
          obarray)
;=> (found now-maybe)

IOW `now-maybe' is interned directly via the let form whereas `--cl-rest--'
appears to be indirected through let via restarg in `cl-do-arglist' e.g.

(setq restarg (make-symbol "--cl-rest--")

I'm _not_ advocating a change but it would be good to understand vis a vis
Miles' admonition:

"That is a good thing."

What would be the consequences if user level keys following &rest i.e.
:mon-key0 :mon-key1
:mon-key2 _were_ to be interned directly (maybe on a partitioned
obarray) instead of
passing the buck to --cl-rest--?

Again, asked out of curiosity :)

s_P

Re: symbol equality --cl-rest-- help-function-arglist member vs member* and also equal, eql, and eq

[David Kastrup]

MON KEY <monkey@sandpframing.com> writes:

> Stephen
>> Apparently --cl-rest-- is an uninterned symbol.  This is done so that
>> the symbol cannot shadow an definition in code outside the cl-arglist
>> processing module.
>
> Miles
>> It looks like `--cl-rest--' is an uninterned symbol.
>> This is a good thing.
>
> Ok, thanks to you both for explaining.
>
> However, I'm not able to make that to jibe... still missing something
> I'm sure.
>
> I understand why there needs to be a reserved 'anonymous' symbol for
> the --rest-of-cl-- e.g. --cl-rest-- and that this symbol should remain
> uninterned but what I am missing on is why --cl-keys-- is converted to
> `cl-keys' whereas --cl-rest-- can't/isn't in the `cl-do-arglist'
> function? For example,
>
> In the following both `body' and `--cl-rest--' show as being interned whereas
> the `*-wombats' aren't:
>
> (intern-soft "body") ;=> t
> (intern-soft "--cl-rest--") ;=> t
> (intern-soft "scary-flying-wombats") ;=> nil

For what it's worth, cl-macroexpand-all uses the _symbols_ body and
--cl-rest--.  Note that even if you let-bind or lambda-bind a literal
symbol in Elisp, it gets interned permanently by the Lisp reader.

That does not mean that you can't use make-symbol to generate
_different_ symbols with the _same_ symbol-name.

-- 
David Kastrup

symbol equality --cl-rest-- help-function-arglist member vs member* and also equal, eql, and eq

[Stephen J. Turnbull]

MON KEY writes:

 > Why do symbol equality tests/comparisons with --cl-rest-- and
 > `help-function-arglist' be so freaky?

Apparently --cl-rest-- is an uninterned symbol.  This is done so that
the symbol cannot shadow an definition in code outside the cl-arglist
processing module.