Help Hacking Gruber's Liberal URL Regex

I've taken the Liberal URL Regex from Daring Fireball, merged it with some of Alan Storm improvements and hacked my way into fixing some bugs like support for IDN chars inside parentheses. This is what I've:

/(?:[\w-]+:\/\/?|www[.])[^\s()<>]+(?:(?:\([^\s()<>]*\)[^\s()<>]*)+|[^[:punct:]\s]|\/)/

However I've encountered a bug that I'm not being able to solve:

'www.dsd(sd)sdsd.com' // can also be the valid 'www.dsd.com/whatever(whatever)'

The above URL is being recognized as www.dsd(sd)sdsd.com' (or www.dsd.com/whatever(whatever)') instead of www.dsd(sd)sdsd.com (or www.dsd.com/whatever(whatever)). This only seems to happen when the URL has parentheses, since the following URL:

'www.sampleurl.com'

Is correctly being recognized as www.sampleurl.com开发者_StackOverflow.

I think the [^[:punct:]\s]|\/ part of the regex is not being executed when the URL has parentheses, I've been trying for some time but I can't seem to find a solution. Can anyone help me?

For commodity, I've set up a Rubular permalink with the regex and some test data (the last URL fails).

---

I think that Gruber's regex was a little rushed, for instance it doesn't match URL's like:

http://en.wikipedia.org/wiki/Something_(Special)_For_You

I'm even more impressed by seeing that both Gruber and Alan missed this really simple typo:

\([\w\d]+\)

Wouldn't \(\w+\) be enough? :S

---

Seems like Gruber has revised his regular expression:

\b((?:[a-z][\w-]+:(?:\/{1,3}|[a-z0-9%])|www\d{0,3}[.])(?:[^\s()<>]+|\([^\s()<>]+\))+(?:\([^\s()<>]+\)|[^`!()\[\]{};:'".,<>?«»“”‘’\s]))

Works just fine now.

---

www.dsd(sd)sdsd.com is not a valid domain name.

If you had 'www.dsd.com/whatever(whatever)', it would be recognized correctly. (Or at least is in my tests)

---

 /(?:[\w-]+:\/\/?|www[.])[^\s()<>]+(?:(?:\([^\s()<>]*\)[^\s()<>]*)+|[^[:punct:]\s]|\/)/
  www.                   |               |            |
                          dsd            |            |
                                          (sd)        |
                                                       sdsd.com'

That's how I think this breaks down... the bit of the regex above (sd) starts with an escaped open paren, then a stared char class matching sd, then an escaped closing paren, and the next thing is [^\s()<>]* which matches sdsd.com'.