How can I match a URL but exclude terminators from the match?

I want to match urls in text and replace them with anchor tags, but I want to exclude some terminators just like how Twitter matches urls in tweets.

So far I've got this, but it's obviously not working too well.

(http[s]?\:\/\/[a-zA-Z0-9\-\.]+\.[a-zA-Z]{2,3}(\/\S*)?)

EDIT: Some example urls. In all cases below I only want to match "http://www.example.com"

http://www.example.com.

http://www.example.com:

"http://www.example.com"

http://www.example.com;

http://www.example.c开发者_运维问答om!

[http://www.example.com]

{http://www.example.com}

http://www.example.com*

I looked into this very issue last year and developed a solution that you may want to look at - See: URL Linkification (HTTP/FTP) This link is a test page for the Javascript solution with many examples of difficult-to-linkify URLs.

My regex solution, written for both PHP and Javascript - (but could easily be translated to Ruby) is not simple (but neither is the problem as it turns out.) For more information I would recommend also reading:

The Problem With URLs by Jeff Atwood, and
An Improved Liberal, Accurate Regex Pattern for Matching URLs by John Gruber

The comments following Jeff's blog post are a must read if you want to do this right...

Ruby's URI module has a extract method that is used to parse out URLs from text. Parsing the returned values lets you piggyback on the heuristics in the module to extract the scheme and host information from a URL, avoiding reinventing the wheel.

text = '
http://www.example.com.
http://www.example.com:
"http://www.example.com"
http://www.example.com;
http://www.example.com!
[http://www.example.com]
{http://www.example.com}
http://www.example.com*
http://www.example.com/foo/bar?q=foobar
http://www.example.com:81
'

require 'uri'

puts URI::extract(text).map{ |u| uri = URI.parse(u); "#{ uri.scheme }://#{ uri.host[/(^.+?)\.?$/, 1] }" }

# >> http://www.example.com
# >> http://www.example.com
# >> http://www.example.com
# >> http://www.example.com
# >> http://www.example.com
# >> http://www.example.com
# >> http://www.example.com
# >> http://www.example.com
# >> http://www.example.com
# >> http://www.example.com

The only gotcha, is that a period '.' is a legitimate character in a host name, so URI#host won't strip it. Those get caught in the map statement where the URL is rebuilt. Note that URI is stripping off the path and query information.

A pragmatic and easy understandable solution is:

regex = %r!"(https?://[-.\w]+\.\w{2,6})"!

Some notes:

• With %r we can choose the start and end delimiter. In this case I used exclamation mark, since I want to use slash unescaped in the regex.
• The optional quantifier (i.e. '?') binds only to the preceding expression, in this case 's'. There's no need to put the 's' in a character class [s]?. It's the same as s?.
• Inside the character class [-.\w] we don't need to escape dash and dot in order to make them match dot and dash literally. Dash should be first, however, to not mean range.
• \w matches [A-Za-z0-9_] in Ruby. It's not exactly the full definition of URL characters, but combined with dash and dot it may be enough for our needs.
• Top domains are between 2 and 6 characters long, e.g. '.se' and '.travel'
• I'm not sure what you mean by I want to exclude some terminators but this regex matches only the wanted one in your example.

• We want to use the first capture group, e.g. like this:

  if input =~ %r!"(https?://[-.\w]+.\w{2,6})"!

  match = $~[1]

  else

  match = ""

  end

What about this?

%r|https?://[-\w.]*\w|