urlize template filter raises exception in some cases

[rivolaks@…]

In some cases, urlize template filter fails because it's trying to parse ipv6 url and exception is raised:

>>> from django.utils.html import urlize
>>> urlize('www.ee]')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "django/utils/functional.py", line 193, in wrapper
    return func(*args, **kwargs)
  File "django/utils/html.py", line 213, in urlize
    url = smart_urlquote('http://%s' % middle)
  File "django/utils/html.py", line 152, in smart_urlquote
    scheme, netloc, path, query, fragment = urlsplit(url)
  File "/usr/lib/python2.7/urlparse.py", line 183, in urlsplit
    raise ValueError("Invalid IPv6 URL")
ValueError: Invalid IPv6 URL

'www.google.com]' also breaks, but 'google.com]' doesn't.

IMHO urlize filter should never raise exceptions. In my case it is used on plaintext user input which we have no control over, and I'm assuming that errors in Django's template rendering are always handled internally and silently. In this case I'd expect urlize() to catch the ValueError from urlsplit and handle it somehow, e.g. by skipping the url.

Happens in both 1.4 as well as master, 1.3 is not affected.

[Claude Paroz]

Fixing the square bracket issue

[Claude Paroz]

Attached a patch which fixes the specific square bracket issue (note that the issue is only accurate from Python 2.7).

The question remains whether we should swallow exceptions around smart_urlquote. Aymeric?

[Jan Bednařík]

Patch reviewed and it's RFC.

Perhaps the question about swallowing exceptions shoud be discussed in django-developers or in separate ticket.

[Andrew Godwin]

I'm going to commit the patch since it's the precise fix for this bug and we're getting affected by this - if we want to discuss the general exceptions issue let's do it elsewhere.

[Andrew Godwin <andrew@…>]

In 7f75460fd6befbef805fee3c91608efb0e9f444d:

Fixed #19070 -- urlize filter no longer raises exceptions on 2.7

Thanks to claudep for the patch.

[tom@…]

With the above patch applied / git HEAD:

>>> from django.utils.html import urlize
>>> urlize('www.ee]')
u'<a href="http://www.ee">www.ee</a>]'
>>> urlize('www[foo.com')
Traceback (most recent call last):
  File "<console>", line 1, in <module>
  File "/Users/tomi/Projects/feedify/venv/lib/python2.7/site-packages/django/utils/functional.py", line 176, in wrapper
    return func(*args, **kwargs)
  File "/Users/tomi/Projects/feedify/venv/lib/python2.7/site-packages/django/utils/html.py", line 168, in urlize
    url = smart_urlquote('http://%s' % middle)
  File "/Users/tomi/Projects/feedify/venv/lib/python2.7/site-packages/django/utils/html.py", line 107, in smart_urlquote
    scheme, netloc, path, query, fragment = urlparse.urlsplit(url)
  File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/urlparse.py", line 182, in urlsplit
    raise ValueError("Invalid IPv6 URL")
ValueError: Invalid IPv6 URL
>>> 

pull request at ​https://github.com/django/django/pull/573 / patch attached.

[Andrew Godwin <andrew@…>]

In 501c7a221cfd022b6d0021fd9a54005c5ffc7a23:

Merge pull request #573 from tominsam/master

Fixed #19070: urlize template filter raises exception in some cases

[EmilStenstrom]

It still seems possible to break urlize in Django 1.5.1:

Testcase:

>>> import django
>>> django.VERSION
(1, 5, 1, 'final', 0)
>>> from django.template.defaultfilters import urlize
>>> urlize('[http://168.192.0.1](http://168.192.0.1)')
Traceback (most recent call last):
  File "<console>", line 1, in <module>
  File "/Users/EmilStenstrom/Envs/kundo/lib/python2.7/site-packages/django/template/defaultfilters.py", line 45, in _dec
    return func(*args, **kwargs)
  File "/Users/EmilStenstrom/Envs/kundo/lib/python2.7/site-packages/django/template/defaultfilters.py", line 336, in urlize
    return mark_safe(urlize_impl(value, nofollow=True, autoescape=autoescape))
  File "/Users/EmilStenstrom/Envs/kundo/lib/python2.7/site-packages/django/utils/functional.py", line 194, in wrapper
    return func(*args, **kwargs)
  File "/Users/EmilStenstrom/Envs/kundo/lib/python2.7/site-packages/django/utils/html.py", line 212, in urlize
    url = smart_urlquote(middle)
  File "/Users/EmilStenstrom/Envs/kundo/lib/python2.7/site-packages/django/utils/html.py", line 153, in smart_urlquote
    scheme, netloc, path, query, fragment = urlsplit(url)
  File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/urlparse.py", line 182, in urlsplit
    raise ValueError("Invalid IPv6 URL")
ValueError: Invalid IPv6 URL

[Tim Graham]

@EmilStenstrom, I think opening a new ticket would have been the best course of action here since the commits for this ticket were already released, but moving this back to "Unreviewed" for the new issue.

[andrea_crotti]

On django (1, 6, 0, 'alpha', 0)
I get a different output
u'[<a href="​http://168.192.0.1](http://168.192.0.1)" rel="nofollow">​http://168.192.0.1](http://168.192.0.1)</a>'
I'm not entirely sure that this result is correct either though..

[andrea_crotti]

This adds a test that shows that the behaviour of urlize described here does not fail on 1.6.0

[Sasha Romijn]

As this problem was reported, I think it would be good to add the test to the testsuite. It's a realistic use, and means we can be sure this does not break at any time. I'll clean the patch a little bit and make a PR.

[Sasha Romijn]

Slight patch clean up, PR files: ​https://github.com/django/django/pull/1161

[Erik Romijn <eromijn@…>]

In 1927bf7c91761040a7e46197fc2d19b66f2ae332:

Fix #19070 -- Additional test for urlize and brackets

[Aymeric Augustin <aymeric.augustin@…>]

In d34b1c29e294b19e51a47918125314a1540c01d4:

Merge pull request #1161 from erikr/urlize-additional-test

Fixed #19070 -- Additional test for urlize and brackets