Re: [Boost-Users] Problems getting subexpressions in regex_search

From: Darin Adler (darin_at_[hidden])
Date: 2002-06-30 00:26:14

On Saturday, June 29, 2002, at 11:29  AM, soleani wrote:

>     boost::cmatch what;
>     if ( boost::regex_search( input, what, regexRedirectedSection,
> boost::regbase::perl ) )
>     {
>         cerr << "Matched new style section redirection!!!! Input
> is: " << input << endl;
>         cerr << "0 gives: --" << what[0].first << "-- and second: "
> << what[0].second << endl;
>         cerr << "1 gives: --" << what[1].first << "-- and second: "
> << what[1].second << endl;
>         cerr << "2 gives: --" << what[2].first << "-- and second: "
> << what[2].second << endl;
>         cerr << "3 gives: --" << what[3].first << "-- and second: "
> << what[3].second << endl;

[...]

> Matched new style section redirection!!!! Input is: [MY_SECTION =
> <FILE:SECTION>]
> 0 gives: --[MY_SECTION = <FILE:SECTION>]-- and second:
> 1 gives: --MY_SECTION = <FILE:SECTION>]-- and second:  =
> <FILE:SECTION>]
> 2 gives: --FILE:SECTION>]-- and second: :SECTION>]
> 3 gives: --SECTION>]-- and second: >]
>
> I don't understand why $1 is not giving my MY_SECTION only.
> It looks like the non-greedy operator is not working as I expect it.
> This same expression works in Perl and jakarta's regex engine in Java.

Your misunderstanding here is the nature of "first" and "second". These 
are iterators that can be used to pinpoint where in the original string 
you matched. The fact that you can stream them out to cerr without a 
compiler error is a sort of "accident", and the output you are getting 
is from the point where the iterator is pointing to the end of the 
string. If you simply used the matches as strings, you'd see the 
results you'd expect.

         cerr << "0 gives: --" << what[0] << endl;
         cerr << "1 gives: --" << what[1] << endl;
         cerr << "2 gives: --" << what[2] << endl;
         cerr << "3 gives: --" << what[3] << endl;

You only want to use what[0].first and what[0].second when you have a 
reason to want to locate the part of the original string that was 
matched, and even then, you won't want to use them on their own, but 
rather together.

Aside to John Maddock: Perhaps these should be named something more 
like begin and end rather than first and second?

     -- Darin