web-mode.el and speed

web-mode.el and speed

[Bois Francois-Xavier]

[-- Attachment #1: Type: text/plain, Size: 608 bytes --]

Dear Emacsians,

I am the author of web-mode.el (http://web-mode.org) a major mode for
editing web templates (html files with js/css parts, and server blocks).

The performance of this mode is good, but I would like, as a challenge, to
speed up the loading of really huge HTML documents (my test file is
http://www.w3.org/html/wg/drafts/html/master/single-page.html)

I think that using a search-chars-forward (native) function could improve
the performance
(search-chars-foward '(?\< ?\>))

Do you think that such a function could be useful for other modes ? Does a
core developer could implement it ?


Fx

[-- Attachment #2: Type: text/html, Size: 936 bytes --]

Re: web-mode.el and speed

[Eli Zaretskii]

> From: Bois Francois-Xavier <fxbois@gmail.com>
> Date: Mon, 7 Oct 2013 21:29:42 +0200
> 
> I think that using a search-chars-forward (native) function could improve
> the performance
> (search-chars-foward '(?\< ?\>))

What's wrong with skip-chars-forward?

Re: web-mode.el and speed

[Bois Francois-Xavier]

[-- Attachment #1: Type: text/plain, Size: 779 bytes --]

I think that a function that would not rely on regexp could improve
drastically the performance.
For seconds are needed to load the html5-spec page. And most of the time is
spent identifying tags. So any micro optimization can have a huge impact.

(I have previously won 20% in perfomance when I modified the tag attributes
parser function to use chars instead of strings. I would have never thought
that the gain would be so large)


On Mon, Oct 7, 2013 at 10:05 PM, Eli Zaretskii <eliz@gnu.org> wrote:

> > From: Bois Francois-Xavier <fxbois@gmail.com>
> > Date: Mon, 7 Oct 2013 21:29:42 +0200
> >
> > I think that using a search-chars-forward (native) function could improve
> > the performance
> > (search-chars-foward '(?\< ?\>))
>
> What's wrong with skip-chars-forward?
>

[-- Attachment #2: Type: text/html, Size: 1292 bytes --]

Re: web-mode.el and speed

[Andreas Schwab]

Bois Francois-Xavier <fxbois@gmail.com> writes:

> I think that a function that would not rely on regexp could improve
> drastically the performance.

skip-chars-forward doesn't use a regexp.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: web-mode.el and speed

[Stefan Monnier]

> I think that a function that would not rely on regexp could improve
> drastically the performance.

Did you profile the code?


        Stefan "it doesn't use a regexp, but (re-search-forward "[<>]"),
                which does, should be about as fast anyway."

Re: web-mode.el and speed

[Eli Zaretskii]

> From: Bois Francois-Xavier <fxbois@gmail.com>
> Date: Mon, 7 Oct 2013 22:21:30 +0200
> Cc: emacs-devel@gnu.org
> 
> I think that a function that would not rely on regexp could improve
> drastically the performance.
> For seconds are needed to load the html5-spec page. And most of the time is
> spent identifying tags. So any micro optimization can have a huge impact.
> 
> (I have previously won 20% in perfomance when I modified the tag attributes
> parser function to use chars instead of strings. I would have never thought
> that the gain would be so large)

As others have pointed out, you are trying to optimize without making
any measurements.  This methodology is known to lead to wrong
conclusions.  So you are well advised to try skip-chars-forward and
see if it is fast enough for your use case.

Re: web-mode.el and speed

[Andreas Röhler]

Am 07.10.2013 22:21, schrieb Bois Francois-Xavier:
> I think that a function that would not rely on regexp could improve
> drastically the performance.
> For seconds are needed to load the html5-spec page. And most of the time is
> spent identifying tags. So any micro optimization can have a huge impact.
>
> (I have previously won 20% in perfomance when I modified the tag attributes
> parser function to use chars instead of strings. I would have never thought
> that the gain would be so large)
>
>
> On Mon, Oct 7, 2013 at 10:05 PM, Eli Zaretskii <eliz@gnu.org> wrote:
>
>>> From: Bois Francois-Xavier <fxbois@gmail.com>
>>> Date: Mon, 7 Oct 2013 21:29:42 +0200
>>>
>>> I think that using a search-chars-forward (native) function could improve
>>> the performance
>>> (search-chars-foward '(?\< ?\>))
>>
>> What's wrong with skip-chars-forward?
>>
>

As it was said: skip-chars-forward

there is still the string/comment issue, where it probably shouldn't match

(while (and (< 0 (abs (skip-chars-forward CHARS)))(nth 8 (syntax-ppss))))