GNU bug report logs -
#63225
Compiling regexp patterns (and REGEXP_CACHE_SIZE in search.c)
Previous Next
Full log
View this message in rfc822 format
2 maj 2023 kl. 18.14 skrev Ihor Radchenko <yantar92 <at> posteo.net>:
> | Cache size | Hit | Miss | % miss from total | ~org-element-parse-buffer~ time |
> |--------------+---------+---------+-------------------+---------------------------------|
> | 20 (default) | 3219470 | 1491165 | 31.66 | 21.035765s (1.091127s in 2 GCs) |
> | 40 | 4418377 | 293805 | 6.24 | 18.294018s (1.123854s in 2 GCs) |
> | 42 | 4550483 | 161820 | 3.43 | 17.946184s (1.073528s in 2 GCs) |
> | 45 | 4636222 | 76582 | 1.62 | 18.410150s (1.078844s in 2 GCs) |
Good, this quite solidly puts the working set size at 40-odd elements.
> The Org parser is basically a giant `cond' of a number of regexp
> matches. See `org-element--current-element'.
A common way to handle this is to build a big regexp to match many cases at the same time, essentially transforming
(cond ((looking-at RE1) ...)
((looking-at RE2) ...)
...)
to
(looking-at (rx (or (group RE1) (group RE2) ...)))
(cond ((match-beginning 1) ...)
((match-beginning 2) ...)
...)
This reduces the number of regexps used and is also typically faster.
(Essentially this is what `syntax-propertize-rules` does but in a more specialised context.)
Using tree-sitter for this could very well be even faster but it's not guaranteed to be available.
Otherwise it's very much a matter of optimisation of everything, including regexps. Minimise backtracking.
If you want to match five or more dashes, use "------*" instead of "-\\{5,\\}". And so on.
It's also obviously a good idea not to generate regexps dynamically each time if you can help it, and minimise consing in general.
>> Introducing regexp objects that could store compiled regexps and be
>> used instead of strings would be quite some work but probably
>> worthwhile.
>
> What exactly needs to be done? Assuming that regexp objects are not
> going to be readable, for simplicity.
A proper design, for starters. For example, we probably want them to be usable in customised variables which calls for them to be readable.
> If this is something to be used in practice, it will be more convenient
> to provide a macro like (with-regexp-cache-size N <body>).
Maybe, we'll see if it's something we need to add.
This bug report was last modified 2 years and 37 days ago.
Previous Next
GNU bug tracking system
Copyright (C) 1999 Darren O. Benham,
1997,2003 nCipher Corporation Ltd,
1994-97 Ian Jackson.