get_tag_uri in /django/utils/feedgenerator.py breaks with port numbers

[nslater]

The following function:

def get_tag_uri(url, date):
    "Creates a TagURI. See http://diveintomark.org/archives/2004/05/28/howto-atom-id"
    tag = re.sub('^http://', '', url)
    if date is not None:
        tag = re.sub('/', ',%s:/' % date.strftime('%Y-%m-%d'), tag, 1)
    tag = re.sub('#', '/', tag)
    return u'tag:' + tag
}}

- http://code.djangoproject.com/browser/django/trunk/django/utils/feedgenerator.py#L48

... breaks for domain names with a port number, such as http://example.org:8080/ as this produces the following TAG value:

{{{
tag:example.org:8080,2007-09-21:/
}}}

From http://feedvalidator.org/docs/error/InvalidTAG.html and http://tools.ietf.org/html/rfc4151#section-2.1 you can see that the TAG URI should not contain the port number.

The following patch should be able to extract the domain name from the link:

{{{
-    tag = re.sub('^http://', '', url)
+    tag = str(urllib.splitport(urllib.splithost(urllib.splittype(url)[1])[0])[0])
}}}

[Daniel Pope <dan@…>]

Why not use urlparse.urlparse(url).hostname ?

[Arthur Koziel]

regression test

[Arthur Koziel]

I added a diff and a unit test for a new get_tag_uri method which relies on urlparse instead of going through some regular expressions. Please review.

[Russell Keith-Magee]

(In [12338]) Fixed #6188, #6304, #6618, #6969, #8758, #8989, #10334, #11069, #11973 and #12403 -- Modified the syndication framework to use class-based views. Thanks to Ben Firshman for his work on this patch.

[Russell Keith-Magee]

(In [12340]) [1.1.X] Fixed #8758 -- Corrected handling of tag creation in feeds when the URL contains a port number.

Partial backport of r12338 from trunk.