Discovering which capture groups are required to match, from a compiled regexp ?

[chetmurthy]

I'm writing helper wrappers based on the OCaml Pcre library, and one of the things that would be useful is the ability to learn for each capture group whether, if the regexp matches, the capture group will match. I don't see any entrypoint for doing this, and truthfully I don't know enough about PCRE's regexp language to know whether it's even computable, but for simpler regexp languages it seems pretty straightforward.

I thought I'd ask here if this were something that's possible with Pcre.

[zherczeg]

I think it is computable. But generic pattern analysis in not a pcre thing, there are other projects for that:
https://github.com/zherczeg/repan

[PhilipHazel]

I would have thought, for any sensible regex, that any capture group might match - otherwise what is the point of having the capture group in the regex? Of course, one can write patterns with groups that can never match - a trivial example is /(*FAIL)(group)/ - but clearly such patterns are not useful. Or are you wanting to analyze patterns to check whether there are any unreachable groups? As Zoltan says, pattern analysis is not something PCRE goes in for. In any case, this is a general discussion rather than an issue (IMHO). There's a Google group for PCRE2 discussions.

[chetmurthy]

an example of such a regexp would be: (a)(b)?(c)- groups 1,3 MUST match, for the entire regexp to match, but group 2 MAY match.

I didn't realize there was a Google group for PCRE discussions; I'll go there, thanks!