TST: add future unicode_literals test (#786)

[olivierverdier]

add simple test to check that the bug #786 is indeed fixed in master

[takluyver]

I'm not quite sure if this is reliably cross-platform. On Windows, \xe9 may be equivalent to the single byte \xe9 (although due to limitations in Python, it may not be). Can we use an isinstance test instead of len?

Also, we're aiming for Python 3 compatibility for 0.12, so we need to be careful with this test. We probably need to add a skipif_python3 decorator for this. Have a look at IPython.testing.decorators (or I can add that, if you prefer).

[olivierverdier]

Ok, I used isinstance instead of length, which is indeed more elegant. I'm not sure how to skip the test in Python 3, so I let you do that.

[takluyver]

Great, thanks, that's definitely neater. In fact, I think it would now pass after conversion to Python 3, but I'll probably add a decorator to skip it anyway.

One minor niggle: the second argument to assert is the message that shows up if the assertion fails, and you've worded it as a description of what happens if it passes. I think it makes more sense for the message to say what went wrong, e.g.

assert isinstance(ip.user_ns['unicode_str'], unicode), 'strings literals are not being interpreted as unicode'

[olivierverdier]

ok, you're right about the assert message. Since I find it a bit awkward, I remove them altogether.

[takluyver]

Excellent. Thanks, Olivier - it's merged.