I'm trying to modify the url-matching regex at http://daringfireball.net/2010/07/improved_regex_for_matching_urls to not match anything that's already part of a valid URL tag or used as the link text.
For example, in the following string, I want to match http://www.foo.com, but NOT http://www.bar.com or http://www.baz.com
www.foo.com <a href="http://www.bar.com">http://www.baz.com</a>
I was trying to add a negative lookahead to exclude matches followed by " or <, but for some reason, it's only applying to the "m" in .com. So, this regex still returns http://www.bar.co and http://www.baz.co as matches.
I can't see what I'm doing wrong... any ideas?
\b((?:[a-z][\w-]+:(?:/{1,3}|[a-z0-9%])|www\d{0,3}[.]|[a-z0-9.\-]+[.][a-z]{2,4}/)(?:[^\s()<>]+|\(([^\s()<>]+|(\([^\s()<>]+\)))*\))+(?:\(([^\s()<>]+|(\([^\s()<>]+\)))*\)|[^\s`!()\[\]{};:'".,<>?«»“”‘’]))(?!["<])
Here is a simpler example too:
((((ht|f)tps?:\/\/)|(www.))[a-zA-Z0-9_\-.:#/~}?]+)(?!["<])
Yeah, its actually trivial to make it work if you just want to exclude trailing characters, just make your expression 'independent', then no backtracking will occurr in that segment.
(?>\b ...)(?!["<])
A perl test:
use strict;
use warnings;
my $str = 'www.foo.com <a href="http://www.bar.com">http://www.baz.com</a>http://www.some.com';
while ($str =~ m~
(?>
\b((?:[a-z][\w-]+:(?:/{1,3}|[a-z0-9%])|www\d{0,3}[.]|[a-z0-9.\-]+[.][a-z]{2,4}/)(?:[^\s()<>]+|\(([^\s()<>]+|(\([^\s()<>]+\)))*\))+(?:\(([^\s()<>]+|(\([^\s()<>]+\)))*\)|[^\s`!()\[\]{};:'".,<>?«»“”‘’]))
)
(?!["<])
~xg)
{
print "$1\n";
}
Output:
www.foo.com
http://www.some.com