[mew-int 2933] Re: Identify urls

On Fri, 5 Nov 2010 13:07:30 -0200, Diogo F.S.Ramos wrote:
> 
> > I am not sure allowing closing braces inside URLs is the way to go.
> > Sure some URLs contain braces but these are usually balanced.  I
> > personally use the following regex to allow "depth 1" matching braces.
> 
> You have a valid point, although I don't see the disadvantage of
> allowing closing braces inside URLs, but I guess it should correctly
> identify URLs like `(http://www.example.com/foo(bar))baz' as
> `http://www.example.com/foo(bar)'.

The problem with (http://www.example.com/foo(bar))baz is that it is
ambiguous.  Did the user want to say that http://www.example.com/foo(bar)
is the URL and forgot the space after the closing brace?  The only
case of URLs with braces I have seen are from Microsoft (MSDN) and
these have matching braces.  I have never seen URLs with two braces
like http://www.example.com/foo(bar)(zzz) or like
http://www.example.com/foo(b(a)r).

> Unfortunately I tried your solution with `re-builder' but it always
> stop recognizing the URL at a closing parentheses. Could you verify if
> it is working there for you?

You are correct; here is a better (although not yet perfect) version:

(setq mew-regex-url
  (let ((u "[^ 	\n>()\"`']*"))
    (concat "\\b\\(\\(\\(file\\|news\\|mailto\\):\\)"
	    "\\|\\(\\(s?https?\\|ftp\\|gopher\\|telnet\\|wais\\)://\\)\\)"
	    "\\((" u ")\\|" u "[^ 	\n>()\"`'.,:]\\)+")))

Best,
C.