From: "Benjamin J. Tilly" <ben_tilly@...>
Date: 2001-05-30T00:47:56+09:00
Subject: [ruby-talk:15939] Re: Regexp (a\1)

matz@zetabits.com (Yukihiro Matsumoto) wrote:
>Hi,
>
>In message "[ruby-talk:15924] Re: Regexp (a\1)"
>    on 01/05/29, ts <decoux@moulon.inra.fr> writes:
>
>|Y>   * backreferences corresponding to unclosed/unmatched parentheses
>|Y>     should fail always.
>|
>| This is not what do perl, and this will be a difference from perl/ruby.
>|
>| For perl a reference exist only *after* it has found the close ')'
>|     ruby    ....                *when*    ......         open  '('
>
>Hmm, I don't get the point.

The following 2-line program is valid in both Ruby and Perl.  But
it gives different results:

  print "OK\n" if "ababba" =~ /^(a|(b\1))*$/;
  print "Not OK\n" if "ababbba" =~ /^(a|(b\1))*$/;

In Perl \1 matches the previous iteration through the group.

>| This mean that perl can re-use a previous backreference, for example this
>| is valid in perl
>|
>|  "aaaaaa" =~  /^(a\1?)(a\1?)(a\2?)(a\3?)$/
>|
>| and it must give the result $1 = "a", $2 = "aa", $3 = "a", $4 = "aa" (this
>| is extract from the tests of perl).
>
>Ruby gives same result (before and after the modification).
>Probably I'm missing something.

The example that I gave should give you different results before
and after your modification.  Neither will be what Perl gives.

Cheers,
Ben