Performance tips when using regular expressions (regexps) in font-locking

Performance tips when using regular expressions (regexps) in font-locking

[Nordlöw]

I have enhanced my cc-mode with extra font-locking for assignments,
function calls, numerical literals, operators and even format strings
to the [sf]?printf-functions. The regular expressions describing these
context are quite complicated and as a result my cc-mode is no longer
as snappy as without my enhancement. Does any have any tips on how to
write regular expressions when performance is crucial? For example
should I use [:space:] instead of [ \t]? Should I post the code
aswell?

Thanks in advance,
Nordlöw

Re: Performance tips when using regular expressions (regexps) in font-locking

[Stefan Monnier]

> I have enhanced my cc-mode with extra font-locking for assignments,
> function calls, numerical literals, operators and even format strings
> to the [sf]?printf-functions. The regular expressions describing these
> context are quite complicated and as a result my cc-mode is no longer
> as snappy as without my enhancement. Does any have any tips on how to
> write regular expressions when performance is crucial? For example
> should I use [:space:] instead of [ \t]? Should I post the coden
> aswell?

I believe inmost cases micro-optimization such as choosing between
[[:space:]] and [ \t] will make no noticeable difference.
But you should pay attention to * and + operators, especially when
nested or when adjacent and make sure that there is no redundancy: there
should ideally be only one way for the regexp to match a given piece
of text.

E.g. avoid [ab]*a*


        Stefan

Re: Performance tips when using regular expressions (regexps) in font-locking

[David Combs]

In article <jwveje71yc7.fsf-monnier+gnu.emacs.help@gnu.org>,
Stefan Monnier  <monnier@iro.umontreal.ca> wrote:
>> I have enhanced my cc-mode with extra font-locking for assignments,
>> function calls, numerical literals, operators and even format strings
>> to the [sf]?printf-functions. The regular expressions describing these
>> context are quite complicated and as a result my cc-mode is no longer
>> as snappy as without my enhancement. Does any have any tips on how to
>> write regular expressions when performance is crucial? For example
>> should I use [:space:] instead of [ \t]? Should I post the coden
>> aswell?
>
>I believe inmost cases micro-optimization such as choosing between
>[[:space:]] and [ \t] will make no noticeable difference.
>But you should pay attention to * and + operators, especially when
>nested or when adjacent and make sure that there is no redundancy: there
>should ideally be only one way for the regexp to match a given piece
>of text.
>
>E.g. avoid [ab]*a*
>
>
>        Stefan

Friedel's "Regular Expressions" book (O'Reilly) (cheaper from Bookpool.com),
 the *bible* on regexps (*hundreds* of pages long) has an entire chapter
 devoted to optimizing your regexps.  Essential reading.

 (You may *think* you know something about regexps -- read that book,
  and you'll know different! :-)

David