> For a language where backreferences are available, I thought it would be
> interesting to see how often people make use of this feature.  So I
> downloaded the CPAN archive and analyzed the perl scripts there.  Here
> is what I found:
> 
>     - I found 165228 perl scripts/modules in CPAN [1].
>     - Of those, 68501 use regular expressions [2].
>     - Of those, 32359 (or 47 percent) use backreferences [3].
> 
> So, nearly half of all perl scripts on CPAN that use regular expressions
> make use of the backreference feature.  IMO this argues strongly in
> favor of supporting backreferences in C++.  (Backreferences can only be
> handled by a backtracking NFA engine, IIRC.)

What he means by backreference is a way to refer to a submatch.  For
instance, with the regular expression "^([^ ]*) *([^ ]*)", the
backreference "$1" will refer to the substring matched by the first
parenthesed subexpression "([^ ]*)".  As long as the references do not
occur in the regular expression itself, they can be handled perfectly
well with a DFA engine.  So, the numbers above do not prove anything.

> There are other features besides backreferences that can only be
> provided by a backtracking NFA.  These features include non-greedy
> quantification, positive and negative look-ahead and look-behind
> assertions, independent sub-expressions, conditional sub-expressions,
> and backreferences within the pattern itself.

I believe all these features but backreferences within a pattern can
be provided by a DFA engine (though my RE library only support one of
these features, non-greedy quantification, at the moment).

So, the real questions are:
- how often are backreferences used within a pattern?
- when they are used, is it just for convenience, or would it be hard
  to rewrite the pattern without using backreferences?

[Feel free to forward this mail back to the C++ commitee.]

-- Jerome
-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners