Front page | perl.perl5.porters |
Postings from November 2010
This is a question about perl regexp usage. I hope it's okay to post it here;
it seemed too advanced for the beginners list. It may lead on to a possible
feature request for the perl core.
I would like to perform match-and-substitute finding all possible matches of
a regular expression within a string. By 'all possible matches' I mean all
productions from that regular expression, not just the production that perl's
regexp engine happened to find first. So given the regexp /\Ax*/ and the
string 'xxx', I would expect to see matches '', 'x', 'xx', and 'xxx'.
At first sight it appears that the /g flag will provide this, but in fact it
just tells perl to keep matching once it has consumed some of the input string,
while I want it to keep backtracking and finding more possible matches even
after a success.
The perlre manual page suggests that using (*FAIL) together with ?{} can do it:
'xxx' =~ /\Ax*(?{say $&})(*FAIL)/;
This gives the expected result. Great! But now how to do a s/// substitution?
I would like to find all possible matches and get all resulting substitutions.
So far I have hand-rolled code for that in perl:
my $string = 'xxx';
my @m;
$string =~
/\A(x*)(?{push @m, [ ${^PREMATCH}, ${^POSTMATCH}, [ @- ], [ @+ ] ]})(*FAIL)/p;
my $replacement = '.$1.';
foreach (@m) {
my ($pre, $post, $starts, $ends) = @$_;
my @captured;
foreach my $n (1 .. $#$starts) {
my $start = $starts->[$n] // die;
my $end = $ends->[$n] // die;
push @captured, substr($string, $start, $end - $start);
}
(my $r = $replacement)
=~ s{\$(\d+)}{$captured[$1 - 1] // die}ge;
say $pre . $r . $post;
}
This appears to work but it is unsatisfactory. Can I coax perl's regexp
engine into finding all possible matches and gathering them into an array?
Would there be an appetite for adding a new regexp flag to find all possible
matches? It would be a useful teaching tool, and would also make it possible
to find the longest or shortest match.
--
Ed Avis <eda@waniasset.com>
Thread Next