Re: [Maxima-discuss] defmvar, remvalue, remprop and kill

[Raymond T. <toy...@gm...>]

On Tue, Feb 11, 2025 at 9:59 AM Leo Butler <Leo...@um...> wrote:

> On Mon, Feb 10 2025, Raymond Toy <toy...@gm...> wrote:
>
> > On 2/10/25 2:05 PM, Leo Butler wrote:
> >> On Mon, Feb 10 2025, Raymond Toy<toy...@gm...> wrote:
> >>
> >>> On 2/10/25 9:36 AM, Raymond Toy wrote:
> >>>
> >>>> On 2/10/25 8:12 AM, Raymond Toy wrote:
> >>>>
> >>>>> On 2/5/25 10:51 AM, Leo Butler wrote:
> >>>>>
> >>>>>
> >>>>> An aside: Why is there the list |*builtin-symbols-with-values*| and
> >>>>> also the hash table |*builtin-symbol-values*|. Can't we just use
> >>>>> the hash table? One less thing to keep synchronized. I assume they
> >>>>> are, but I didn't check.
> >>>>>
> >>>> I looked. init-cl has this:
> >>>>
> >>>> |(dolist (s *builtin-symbols*) (when (boundp s) (push s
> >>>> *builtin-symbols-with-values*))) (dolist (s
> >>>> *builtin-symbols-with-values*) (setf (gethash s
> >>>> *builtin-symbol-values*) (symbol-value s))) |
> >>>>
> >>> While looking at this, I see that it seems that |*builtin-symbols*|
> >>> basically has the same info as |*builtin-symbol-props*|. We could
> >>> probably get rid of |*builtin-symbols*| as well, once suprv1.lisp is
> >>> updated to use the hashtable instead of the list. Again, suprv1.lisp
> >>> uses |member| with a test of |equal|. I think |eql| or even |eq| would
> >>> have worked. Of course, this won't matter if we use the hashtable,
> >>> which defaults to |eql| .
> >>> &#8203;
> >> I saw that there were 4 variables that look like they could all be
> >> consolidated into a single hashtable, maybe
> >> *BUILTIN-SYMBOLS-WITH-VALUES*? Looking with `magit-blame', it looks like
> >> the 4 variables are quite old...
> > Don't we need one for values, as you mention, and also one for the
> > properties?
>
> You are likely right.
>
> An alternative, though, would be to store both values and properties in
> an alist for each symbol. Perhaps that is too much work, though.
>

We could do that as well.  I think I like two hashtables, but could be
easily convinced that one is better.

I also found out that there's `*variable-initial-values*`.  Not exactly
sure how that differs from `*builtin-symbols-with-values*`, but the former
is used by `reset` to reset the value of a variable.  This also explains
why `file_search_maxima` isn't reset: it's not in the hashtable.  It
probably should be.

In any case, I think I'll file a bug that all these things should be
rationalized/consolidated if possible.

>
> Leo
>
> >>
> >> On a slightly different note, I cannot find a single variable that has
> >> the ASSIGN property set to SPECIAL. Indeed, this form
> >>
> >> (let (x) (do-symbols (s :maxima x)
> >>       (when (let ((w (get s 'assign)))
> >>             (or (and (listp w) (member 'special w)) (eq w 'special)))
> >>                 (push s x))))
> >>
> >> returns NIL. (By contrast, when SPECIAL is replaced by NEVERSET, I get a
> >> list of length 26.)'
> > That's what I saw too; no symbols with a property of special.
> >>
> >> I think we can safely replace the case in REMVALUE which checks if X has
> >> the SPECIAL property with a hash-table lookup in
> >> *BUILTIN-SYMBOLS-WITH-VALUES*.
> >>
> >> WDYT?
> >>
> >> Leo
>
> --
> ---
> Best regards,
> Dr Butler



-- 
Ray