def getIcoUrl(url):
    """Return the URL of the favourite icon associated with input.
    Returns None if input isn't a valid address."""
    try:
        fp = urllib.urlopen(url)
    except IOError, e:
        if e[1] == "unknown url type":
            # The given URL isn't a web address so adding
            # /favicon.ico won't work. Might as well
            # jump straight to giving up
            return None
    else:
        if fp.headers['content-type'].startswith('text/html') or fp.headers['content-type'].startswith('application/xhtml+xml'):
            data = fp.read()
            fp.close()
            # Try to find any links to shortcut icons
            ICONREGEX = "<link rel="shortcut icon" (?:.*)href=(?:['|"])(.[^'|"|\s]{0,})(?:['|"])(?:.*)>"
            results = re.search(ICONREGEX, data, re.I)
            if results != None:
                # Found the requested URL of the favourite icon
                icourl = results.groups()[0]
                # the base url is already known to be valid, so we
                # don't need to check that the join works.
                # If the icon's url is invalid, we'll let
                # loadResizeImage handle the inability to retrieve it
                return urlparse.urljoin(url, icourl)
    else:
        fp.close()
    # Return the default of url/favicon.ico
    return urlparse.urljoin(url, '/favicon.ico')

I put that in code and pre blocks in the hope that it would not try to parse it, but the preview doesn’t look like it is going to work. It isn’t very hard to write yourself, but hopefully you can guess what it looks like untextiled (the numbered lists used to be comments) or get at the source somehow.
If you want to use this function you’ll have to rework the file naming since you probably don’t want to call this function for every request just to see what the cached image is/should be called, but that shouldn’t be too hard; saving the images as the user inputted URL instead of the image’s URL would eliminate the space saving feature of http://website/foobar always pointing to the file http—-website-favicon.ico for any value of foobar, but if you add in some code to normalise an URL (like always remove trailing slashes) it shouldn’t prove too much of a problem since most people would enter the root of their namespace. Or you could go all out and cache the results of this function in some sort of mapping file. – Testing:
>>>print getIcoUrl(‘http://www.python.org/doc/2.3/modindex.html‘)
‘http://www.python.org/doc/2.3/icons/pyfav.gif’
>>> print getIcoUrl(‘http://www.google.com‘)
‘http://www.google.com/favicon.ico’ – Transparency: If the icon is a gif (all I tested with), you seem to be able to get transparency working fine by doing i.save(thumb,‘PNG‘,transparency=i.info[‘transparency‘]) in loadResizeImage() but ICO files don’t get a ‘transparency’ key in their info dictionary (so you’d want to add a has_key() check before doing that form of save).
I have no idea really how transparency works in PIL or file formats in general, so I may be missing something obvious, but getcolors() only shows 4 colors for one of the icons with transparancy problems on your comments (simon.incutio.com/favicon.ico) and it looks like it uses 4 colors so one isn’t set aside just for transparency and converting to an RGBA image lists each of the 4 colors with a 4th value of 255 (fully opaque, I assume, but all the same value at any rate). Doing the same with the python.org gif gives you an extra color and lists one of the colors as having a 4th channel of value 0 (vs. 255 for the others) if converted to RGBA. So ICO files don’t support the same way of getting at the transparency info of other files, and I suspect it doesn’t support transparency on them at all. ]]> http://www.kryogenix.org/days/2005/01/31/favatarsUrlparse/comment-page-1#comment-742 N/A Thu, 01 Jan 1970 01:00:00 +0000 http://www.kryogenix.org/adpb/2005/01/31/favatarsUrlparse/#comment-742 <p>Opps, I messed up the formatting a bit. The else: fp.close() should be indented one block.</p> Opps, I messed up the formatting a bit. The else: fp.close() should be indented one block.

RewriteRule ^favicon.ico$ http://introvert.net/favicon.png [R=301]

I’m pretty sure Python’s urllib will handle this redirect transparently.