Hello dominic01,

The Camel Book (4^th Edition, 2012) has a section on “Time Efficiency” which contains the following (p. 692):

Short-circuit alternation is often faster than the corresponding regex. So:

print if /one-hump/ || /two/;
[download]

is likely to be faster than:

print if /one-hump|two/;
[download]

at least for certain values of one-hump and two.

Also, if your regex is more likely to find a space than a punctuation character, test for the space first. Only profiling will show whether these kinds of tweaks make significant improvements, but they’re worth a try.

Hope that helps,

> at least for certain values of one-hump and two.

Indeed, as trie optimization since 5.10 can make a huge difference.

Cheers Rolf
_{(addicted to the Perl Programming Language and ☆☆☆☆ :)}

PS: Je suis Charlie!

Don't guess at which things in your code might be slow, measure it, e.g. Devel::NYTProf

Also, using the same regex 250 times sounds like your code might be in need of some refactoring, like pulling any repeated code blocks into subroutines.

If I define it in qr/(?:[.,:;] ?| )/ and use the variable in the regular expression ...

This does not directly address your performance concerns, but when qr//-defined regex objects are interpolated into other regexes, either qr//, m// or s///, the process is more or less like string interpolation; the new regex does not somehow "call back" to the interpolated regex as a subroutine call within another subroutine would do.

To get a feel for this process, write and run (with full warnings and strictures) some code like the following, which just prints compiled regexes. (This is untested because I can't provide a working example at the moment.) Note that qr// automatically encapsulates its pattern in a  (?:...) non-capturing group that preserves the regex modifier flags, so the non-capturing grouping within  qr/(?:[.,:;] ?| )/ is redundant (but does no harm); try it both ways.

my $foo = qr/[.,:;] ?| /;
print $foo, "\n";

my $bar = qr{ $foo+ (?: hic | hac | hoc) }xms;
print $bar, "\n";

my $baz = qr{ $bar{42} f[eio]e }xms;
print $baz, "\n";
[download]

It also depends how you use the results. You do not seem to capture the result from your match. So in some circumstances I could imagine that

might do the job as well (like matching the bit between two words in a text, assuming no two punctuation signs or blanks in sequence). More context would be helpful to provide more support!