Regular expression challenge for stripping single comments.

[DXRW4E]

That's very difficult to read to me and requires a bit of thinking to figure out its logic.  I'm not saying its bad or wrong, just a bit complicated looking.

Is all OK, however is only this

'(?:[^;"''\r\n]|''[^''\r\n]*''|"[^"\r\n]*")*'

is the ABC of regex the first thing that you see all over the web

'(?:[^"]|"[^"]*")*'

I was referring to this

the example above refers to read the string into quotes, for this is a bit more complicated

Ciao.

Edited December 31, 2013 by DXRW4E

[jchd]

Ascend4nt,

 

Although I still hope future versions of PCRE add the ability to search for anything not-of-a-capture-group back..

I was about posting a question about this point on the PCRE list but before I'd like to come up with a real-world example where such a feature is useful.

If you have a captured group, then want to skip until that group, that means that you expect to see that again, so tell the engine you want to match that.

(abc|def) .*? \g-1
matches
abcXXXXXX YYYYYYabc

If you're not so sure that the group reappears, make it optional:

(abc|def) .*? \g-1?
matches
abcXXXXXX YYYYYYabc
and
abcXXXXXX YYYYYY

From this point of view I don't have an example of a situation where a back-ref to a group would be needed in a class. Can anyone surprise me?

Of course if you want to match a different final boundary you can still use a reference to subroutine:

(abc|def) .*? (?-1)
matches
abcXXXXXX YYYYYYabc
and
abcXXXXXX YYYYYYdef
(and other combinations)

[Ascend4nt]

jchd, hmm.. good question.  I was thinking about the possiblity of using that to detect duplicate things like 'patient_id = 124ab' but there are other means to do that. And from a few trials with the old engine it seems like the feature was actually broken, as it skipped over the captured groups in most cases.  Oh, and oddly it seems to require curly braces for backreferences g{-1} and I swore in one case I had to use it with an absolute number even if followed by a bracket. 

Overall I suppose its not something to pester them about (unless its deviating from a Posix standard?).  The 'detect duplicates' thought I had can be solved with lookahead and lookbehind assertions.  And as you point out, if we're looking for a specific group to reappear we have other tools at our disposal.

[jchd]

I asked the question on the pcre list and I'll keep you informed of the outcome.