Re: [Emacs-diffs] feature/byte-switch 44c95c5: bytecomp.el: Don't store non-keyword symbols in jump-tables.

Re: [Emacs-diffs] feature/byte-switch 44c95c5: bytecomp.el: Don't store non-keyword symbols in jump-tables.

[Stefan Monnier]

>  (defun byte-compile-cond-valid-obj2-p (obj)
> -  (if (consp obj)
> -      (and (eq (car obj) 'quote)
> -           (= (length obj) 2)
> -           (symbolp (cadr obj)))
> -    t))
> +  (cond
> +   ((consp obj)
> +    (and (eq (car obj) 'quote)
> +         (= (length obj) 2)
> +         (symbolp (cadr obj))))
> +   ((symbolp obj) (keywordp obj))
> +   (t t)))

I suggest you document the property that you need this to satisfy.
E.g. something like "OBJ is a constant expression whose value can be
compared with `eq`" or something like that.  And you might like to see if
maybe you can use macroexp-const-p in here (there are many different
definitions of "constant", tho).


        Stefan

Re: [Emacs-diffs] feature/byte-switch 44c95c5: bytecomp.el: Don't store non-keyword symbols in jump-tables.

[Vibhav Pant]

On Sun, Feb 5, 2017 at 8:44 PM, Stefan Monnier <monnier@iro.umontreal.ca> wrote:
> I suggest you document the property that you need this to satisfy.
> E.g. something like "OBJ is a constant expression whose value can be
> compared with `eq`" or something like that.  And you might like to see if
> maybe you can use macroexp-const-p in here (there are many different
> definitions of "constant", tho).

macroexp-const-p seems to be a much better alternative to this. I've replaced
byte-compile-cond-valid-obj2-p altogether with this. It seems to return non-nil
for constant expressions that would work in a jump table implementation.

At some point, I would like to extend this to forms that call functions, but
that would need additional code to make sure that the callee function doesn't
modify the value of the variable we're comparing against (is there some way
to do that already?).

Thanks,
Vibhav

-- 
Vibhav Pant
vibhavp@gmail.com

Re: [Emacs-diffs] feature/byte-switch 44c95c5: bytecomp.el: Don't store non-keyword symbols in jump-tables.

[Clément Pit-Claudel]

On 2017-02-05 12:24, Vibhav Pant wrote:
> that would need additional code to make sure that the callee function doesn't
> modify the value of the variable we're comparing against (is there some way
> to do that already?).

I've seen (declare (side-effect-free t)) and (declare (pure t)) in the sources, and there's a file called unsafep.el; maybe one of these is a good starting point?
(This branch is pretty exciting, keep up the good work!)

Cheers,
Clément.