URL regex doesn't recognise important TLDs like .cat

[FauxFaux]

Hovering URLs like http://nyan.cat/ and clicking take you to http://nyan.ca/ , as the .cat TLD isn't recognised.

TLD specific code should probably be removed anyway, due to the introduction of arbitrary TLDs.

[ljani]

There are also other problems with the regex. Hosts such as http://localhostr.com get detected as http://localhost/

[joosera]

Also .pro is affected.
.[a-zA-Z][a-zA-Z] could possibly be changed to something like .[a-zA-Z]{3} (not sure what regex markup putty tray uses, this didn't seem to work when i manually edited the regex, though it didn't break the old behaviour of only hilighting the .pr part so not really sure what's going on there)

[wodim]

Bump.

[incognico]

Why not just simplify the regexp? The current one leads to nowhere. Why not use something like (pcre syntax that is) \b((?:[hH][tT][tT][pP]://|[wW][wW][wW]\.)[^\s)'"]+) or even \b((?:[a-zA-Z]+://|[wW][wW][wW]\.)[^\s'")]+) for any protocol? I don't know if the used libary can do word boundaries but in general this regexp needs to be minimal. Imagine the new gTLDs are starting :p

[FauxFaux]

The aim of the regex has previously been to match only plausible URLs, not anything that could conceivably be a URL; I'm guessing this is what people expect (it's what I expect).

It's customisable; if you want to use something more liberal feel free.

Also, no, the regex engine supports basically nothing. #4 is to fix that, but it's hard work.

[incognico]

Everything that conceivably is a URL is also a plausible URL in my opinion.

So, if anyone is interested:
I am now using (([a-zA-Z]+://|[wW][wW][wW]\.)[^ '")>]+).

In contrast to the very complex default regexp which needs a lot of maintence it allows me to click the following URIs:

www.example.newgtld
http://foo.intern
http://hostname
http://user:[email protected]:8081/foo
whatever://protocol

I also giggled at * @(#)regexp.c 1.3 of 18 April 87

[FauxFaux]

I've added a default option of @incognico's suggestion, and liberalised the "classic" default a bit.

I've additionally removed the nasty browser detection code, as this was just generally broken even launching urls like "www.google.com".