Support unicode objects

[emosenkis]

natsort silently fails to do anything useful with unicode strings:

>>> natsort.natsort_key('AAAS 5A')
('AAAS ', 5.0, 'A')
>>> natsort.natsort_key(u'AAAS 5A')
(u'AAAS 5A',)
# should be (u'AAAS ', 5.0, u'A')

[emosenkis]

This seems to be caused by explicitly checking for str:
https://github.com/SethMMorton/natsort/blob/master/natsort/natsort.py#L212

The re module supports unicode, so it might just work if str is replaced with basestring. If it were me, I would probably raise an exception if natsort_key was called on a non-string instead of just returning a tuple containing it.

[SethMMorton]

I will take a look at this. I don't want to raise an exception for any non-string since it is designed to handle ints and floats as well. I'll make sure it supports unicode as well.

[SethMMorton]

OK. I added unicode support. I made it more explicit in the docs that natsort will accept ints and floats as well.

I will close this issue if you find that this is fixed. If you used pypi, this version is uploaded there.

[emosenkis]

Works for me. Thanks! I've never gotten such a quick response before.