bug#19023: 24.4; Long pause when evaluating numeric expression

bug#19023: 24.4; Long pause when evaluating numeric expression

[Gareth Rees]

In GNU Emacs 24.4.1 (x86_64-apple-darwin13.4.0, NS apple-appkit-1265.21)
of 2014-10-20 on builder10-9.porkrind.org

In a fresh emacs -Q, if I type:

    M-: 1000000 RET

there is a pause of about 10 seconds before Emacs displays:

    1000000 (#o3641100, #xf4240)

I suggest that the length of this pause be reduced.

bug#19023: 24.4; Long pause when evaluating numeric expression

[Eli Zaretskii]

> From: Gareth Rees <gdr@garethrees.org>
> Date: Tue, 11 Nov 2014 20:01:08 +0000
> 
> In GNU Emacs 24.4.1 (x86_64-apple-darwin13.4.0, NS apple-appkit-1265.21)
> of 2014-10-20 on builder10-9.porkrind.org
> 
> In a fresh emacs -Q, if I type:
> 
>     M-: 1000000 RET
> 
> there is a pause of about 10 seconds before Emacs displays:
> 
>     1000000 (#o3641100, #xf4240)
> 
> I suggest that the length of this pause be reduced.

My crystal ball says that this pause is due to Emacs trying to locate
a font that can display a character whose codepoint is U+F4240.  In
which case, the pause is not n Emacs, it is in system libraries that
search the available fonts.

bug#19023: 24.4; Long pause when evaluating numeric expression

[Gareth Rees]

> My crystal ball says that this pause is due to Emacs trying to locate
> a font that can display a character whose codepoint is U+F4240.  In
> which case, the pause is not n Emacs, it is in system libraries that
> search the available fonts.

That was my guess too, and if I was hoping to see a character at that code point, I would have to be prepared to wait for the system to search all the fonts to find it.

But normally when I evaluate a numeric expression, I do not want to see a character—I’m just using M-: as a calculator—and so the wait is unnecessary (and a regression from 24.3).

bug#19023: 24.4; Long pause when evaluating numeric expression

[Eli Zaretskii]

> From: Gareth Rees <gdr@garethrees.org>
> Date: Tue, 11 Nov 2014 20:51:29 +0000
> 
> But normally when I evaluate a numeric expression, I do not want to see a character—I’m just using M-: as a calculator

But Emacs doesn't (and cannot) know what you want.

bug#19023: 24.4; Long pause when evaluating numeric expression

[Gareth Rees]

> But Emacs doesn't (and cannot) know what you want.

Surely Emacs should know that I don’t want to wait 10 seconds to see the result of evaluating a numeric expression? Or if this is really in doubt, Emacs could provide a user option by which I could tell it how long I’m prepared to wait. Or Emacs could show me the numeric result *while* it is searching the fonts, allowing me to hit C-g to cancel the search if I am happy with just the number.

For reference, the commit that introduced this problem appears to have been 3223d013c2f94e6a0dd37879cf5faec8c0b2261d

For the moment, I am working around the problem by advising eval-expression-print-format, like this:

(defun old-format (orig-fun value)
  (and (integerp value) (eq standard-output t)
       (format " (#o%o, #x%x)" value value)))

(advice-add 'eval-expression-print-format :around #'old-format)

bug#19023: 24.4; Long pause when evaluating numeric expression

[Lars Magne Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

>> But normally when I evaluate a numeric expression, I do not want to
>> see a character—I’m just using M-: as a calculator
>
> But Emacs doesn't (and cannot) know what you want.

I've always thought it rather strange that Emacs tries to display the
character corresponding to the integer value when you M-: stuff.  How
useful is that, really?

In my experience, I've never wanted to know the octal value or the
corresponding Unicode codepoint of the integer.  I have found the hex
value useful, I think, once.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#19023: 24.4; Long pause when evaluating numeric expression

[Stefan Monnier]

[ Side comment.  ]

> (defun old-format (orig-fun value)
>   (and (integerp value) (eq standard-output t)
>        (format " (#o%o, #x%x)" value value)))

> (advice-add 'eval-expression-print-format :around #'old-format)

If you don't use `orig-fun', you might as well use :override rather
than :around.


        Stefan

bug#19023: 24.4; Long pause when evaluating numeric expression

[Stefan Monnier]

> I've always thought it rather strange that Emacs tries to display the
> character corresponding to the integer value when you M-: stuff.  How
> useful is that, really?

If you ask it this way, then it's probably never useful.
But when you use M-: on an expression which returns a character, then
I find it handy to see the character rather than only its integer value.

But maybe this feature is more trouble than it's worth.  After all, you
can always use M-: (string <exp>) RET to force Emacs to show you the
character.


        Stefan

bug#19023: 24.4; Long pause when evaluating numeric expression

[Lars Magne Ingebrigtsen]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> But maybe this feature is more trouble than it's worth.  After all, you
> can always use M-: (string <exp>) RET to force Emacs to show you the
> character.

Yeah.

I tried reproducing the problem reported here, and I got nothing.

But I evaluated a number of digits, and one of them took a couple of
seconds to display.  (This is with X over ssh over wifi.)  So I'm able
to reproduce the problem, and the slower the X connection is, the more
severe it probably is.

Unless somebody else thinks this is really useful functionality, I think
we should probably just ditch it.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#19023: 24.4; Long pause when evaluating numeric expression

[Lars Magne Ingebrigtsen]

On the other hand, M-: (following-char) currently returns

110 (#o156, #x6e, ?n)

which is useful, come to think of it.  Er.  So I don't know...

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no