Re^2: Delimited Backtracking with Regex

[ikegami]

I've been asked why I used a package variable instead of a lexical variable. It's because regexps close around the lexicals that exist when they are first run.

                                       # pass  1   2   3
                                       #      --- --- ---
sub test {
   my @matches;
   '' =~ /
      (?{ push @matches, 'a' })
      (?{ print(scalar(@matches), "\n") })   # 1   2   3
   /xg;  
   print(scalar(@matches), "\n");            # 1   0   0
}

test() for 1..3;

A variable called @matches is created everytime test is called. The regexp always uses the variable from the first call.

[blokhead]

You can still do the same thing using lexicals (if you are allergic to using symbol-table variables). The only thing to be careful is to reuse the same push statement and same target array every time. This is adapted from the code in Re^3: Regexes: finding ALL matches (including overlap):

{
  my @matches;
  my $push = qr/(?{ push @matches, $1 })/;

  sub match_all_ways {
      my ($string, $regex) = @_;
      @matches = ();
      $string =~ m/($regex)$push(?!)/;
      return @matches;
  }
}

print match_all_ways( "TXXXABCDGXXXCCCDTGYYYCCCYYYCC", qr/XXX.*YYY/ );

Technically, $push is not needed -- you could just include (?{push @matches, $1}) in the m// statement inline. However, I like this way as it makes it a much more obvious that this part of the regex is only compiled once.