homepage

When using a javascript URL with only digits as paths, the urlsplit() functions behaves different in Python 2.7 than in 2.6:
$ python2.6 -c "import urlparse; print urlparse.urlsplit('javascript:123')"
SplitResult(scheme='javascript', netloc='', path='123', query='', fragment='')
$ python2.7 -c "import urlparse; print urlparse.urlsplit('javascript:123')"
SplitResult(scheme='', netloc='', path='javascript:123', query='', fragment='')
Python 3.2 has the same regression:
$ python3.2 -c "import urllib.parse; print(urllib.parse.urlsplit('javascript:123'))"
SplitResult(scheme='', netloc='', path='javascript:123', query='', fragment='')
I consider the Python 2.6 behaviour to be correct, ie. the current behaviour is buggy.

The behaviour change is caused by the fix for issue #754016.

What kind of url is 'javascript:123' and how do you (/ we) say that python2.6 behavior was correct?

Regarding the correctness of the Python 2.6 implementation: http://www.faqs.org/rfcs/rfc1738.html specifies URLs of the form <scheme>:<scheme-specific-part> where the scheme specific part is allowed to consist only of digits.
I agree that the example URL is not a good one and it is artificially constructed.
Some better examples demonstrating the same issue might be
clsid:85bbd92o-42a0-1o69-a2e4-08002b30309d
or
mailto:1337@example.org 

To make the previous comment more precise: URLs where
the scheme specific part begins with a digit are affected.

I'm attaching a patch with a fix and a unittest using the email example. I put this in a new test_RFC2368 (the mailto URL scheme) method. Seems like there is no unittest for parsing mailto scheme to begin with.

Oops, wrong revision base.

Santoso, Quick review comments:
1. The patch looks good.
2. I would use a temporary 'throw-away' variable instead of _, but don't bother to change it, before committing I shall take care.
3. Important - Did you find any regression with the earlier builds, also could you run the full test suite to ensure that all tests pass?

Senthil,
Thanks for the review! I was initially thinking of `port = ...` but opted for _, arbitrarily, instead.
regrtest on Darwin-10.6.0-i386-64bit ran fine.

New changeset 7a693e283c68 by Senthil Kumaran in branch '2.7':
Issue #11467: Fix urlparse behavior when handling urls which contains scheme
http://hg.python.org/cpython/rev/7a693e283c68 

New changeset 495d12196487 by Senthil Kumaran in branch '3.1':
Issue #11467: Fix urlparse behavior when handling urls which contains scheme specific part only digits.
http://hg.python.org/cpython/rev/495d12196487 

Fixed this in all codelines. Thanks Santoso.