Re: Surprisingly poor regex performance

There is no need to do all this anchoring. ".*" does not match newlines (unless you use the /s option), so the following will also work (and should be just as fast):

  my $re = qr/(.*$pat.*)/;
  while ($mmap =~ m/$re/g)
  {
    print $1,"\n";
  }
[download]

Comment on Re: Surprisingly poor regex performance Download Code

Replies are listed 'Best First'.
Re^2: Surprisingly poor regex performance by sgifford (Prior) on Dec 14, 2004 at 18:34 UTC
You're right, that is roughly as fast as `/(?:\A\|\n)(.$pat.\n)/omg`. But it's still slower than `index`/`rindex`.	[reply] [d/l] [select]
Re^3: Surprisingly poor regex performance by tall_man (Parson) on Dec 14, 2004 at 21:49 UTC
You were asking why the index/rindex is so good. It's because you're matching a pure literal string, which can use the highly-efficient Boyer-Moore search (the longer the literal string, the faster it tends to be). For some reason, the "^" is not being optimized properly as an anchor for "/mg" searches. Leaving it out entirely, or substituting the alternate `"(?:\A\|\n)"` pattern, results in searches that are much faster. That seems like a bug in the optimizer.	[reply] [d/l]