Tachyon,
I have a similar problem, over six hundred complex regexen to match against a busy logfile and related messages to issue depending on which regexps were matched. much the same problem as the OP If I understand the regexp engine caching the compiled version of a regex if it is not going to change then I think this should be a reasonably efficient approach. Am I on the right tracks ? and is the /o unrequired as I have already interpolated the variable when the regex is first called ?
#!/usr/local/bin/perl -w
use strict;
my ($i, $compile_me, @names)=(1, "{my \@matches;", "no match");
while (<DATA>) {
next if /^\s*$/;
last if /END CONFIG/;
chomp;
my ($name, $reg)=split;
push @names, $name;
$compile_me.="push \@matches, $i if /$reg/o;";
$i++;
}
$compile_me.="\@matches}";
while (<DATA>) {
chomp;
print "\nmatches found for $_\n";
my @matches=eval $compile_me;
foreach (@matches) {
print $names[$_] , "\n"
}
}
__DATA__
Fred_and_Friends fr.d
Paul_and_co paul
some_numbers \d{2}
freud_likes_fred fr
END CONFIG
freud
fred
NaNa
pauline
12312sdfsdf
2
Update with speed test
I have now run a comparative test over 300^H^H^H, sorry 416 lines of log, with my 672 pattern matches. First using the eval of a string containing all the regexen and returning match index numbers as above. Second is my old naive code holding an array with the regexen and doing a foreach through it against each line. I did not use the /o for the reasons given above it works fine without it
>time ./Monitor.fast
real 0m1.49s
user 0m0.68s
sys 0m0.58s
>time ./Monitor
real 0m19.47s
user 0m14.69s
sys 0m0.50s
>
I think the numbers speak for themselves
Cheers, R. |