Re^5: Suggestion for regular expression speed improvement.

[demerphq]

You need to think out of the box. Make your system handle the various cases. A regex, a split, an unpack, maybe even an xml parser....

Definitely doing it the way you are is wrong. If you are really insistent on ignoring our advice (unwise really), then at least use inversion and you might want to investigate the (?>...) construct "atomic matching". As well as adding anchors to your pattern.

[moritz]

Make your system handle the various cases. A regex, a split, an unpack, maybe even an xml parser....

I agree. The simplest solution is to make the caller supply a code reference, which you then call and which does the parsing/matching/unpacking/whatever.

[bala.linux]

Exactly. This was my approach but before taking that up I wanted to check with the community whether Im doing anything wrong with regular expression. May be I expected too much from the regular expression. Anyway, thanks for your inputs :)

[demerphq]

Well no, its just that if you use a big huge complex machine to do the work that you could do with a knife that you arent being efficient.

And split does use the regex engine, its just it uses it in special ways to be efficient.

[bala.linux]

Yes. My initial idea was to keep CSV, RE parsing separate but I thought may be If could get the same performance from RE, then need not to have a CSV parser. But now I understand that is not possible to think of. So, need to switch back to my old approach where I do two parsing separately. So, from the user interface user will have simple/advanced parsing modes. Thanks for your suggestion.