Regex profiler

[phizel]

Does a profiler exist for regular expressions? The only thing I could find was re's debug mode, but the output isn't very intuitive. I am imagining something like Devel::NYTProf with flame graphs indicating which sections of a regex need optimizing. Take the common whitespace trimming example s/^\s+|\s+$//g; it turns out to sub-optimal because the alternation negates the anchor optimization. Since I can't make heads or tails of re's debug mode output, the only tool in my toolbox for diagnosing regexen is to compare runtimes with Benchmark. Since regular expressions are a feature of many other languages, a better solution might even exist outside of perl.

[Corion]

Regexp::Debugger can create a heatmap of a match.

[choroba]

> a better solution might even exist outside of perl.

Note that other languages might have similar regex syntax to Perl, but the implementation is definitely different, so profiling/benchmarking it doesn't really help if you're going to use the regex in Perl.

[LanX]

Depending on what you mean by "profiling" you can inject Perl (?{code}) snippets into your regex to record the current time. (IMHO that's the most likely approach a third party profiler would take)

But please be aware that you might get a Heisenbug situation then, because your snippet was hindering some of the many internal heuristics or optimizations.

Looking into re debug mode before and after those injections is always a good approach.