Python 3.4 fixes

[takluyver]

I ran the test suite under Python 3.4b1 and hit a couple of inconsequential test failures.

[takluyver]

This one is not really related to Python 3.4, it was just that the test was out of line with the implementation.

[minrk]

I guess the fixed test should be backported to 1.x, right?

[takluyver]

Yes, I think so.

[ivanov]

I don't really understand what the point of the original test is, but I read it as having insufficient arguments (ns not passed) or effectively a name error (a is not defined), and so its not clear why it should be changed for 3.4.

Can we shed more light on these tests via comments or docstrings?

[takluyver]

The docstrings of the formatter classes have some more description, but should possibly be updated. In the format strings, : separates the field name from the start of the format specification, so {a[:]} was previously parsed as a field name of a[ and a format specifier of ]. Trying to evaluate a[ gave the SyntaxError. That would be the same even if a was passed. However, in 3.4b1, it seems to do bracket matching, so it parses it all as the field name.

Hopefully this means that once we rely on Python 3.4 or above, we can drop FullEvalFormatter, because EvalFormatter will behave the same way.

[ivanov]

thanks for the explanation Thomas.

Hopefully this means that once we rely on Python 3.4 or above, we can drop FullEvalFormatter, because EvalFormatter will behave the same way.

Can we add that as a TODO: or XXX: note on FullEvalFormatter, so that years from now our grand-kids remember to do that ;)

[takluyver]

Good idea, done.

[ivanov]

Thanks for thinking of the posterity: let's call it "Seventh Generation Development"

[ivanov]

7GD or SGD to be TLA compliant

[takluyver]

Future release codename: "IPython: The Next Generation"?

[juliantaylor]

concerning backport to 1.x, I'm trying to run the testsuite with 3.4b2 and there are a lot more failures. E.g. the parametric tests are all broken, the removal is backportable but there are probably still more failures than I can handle in time.
Is 3.4 support for 1.x planed to be done on your side?

[takluyver]

I'm guessing the failures are just problems with the tests, not things that will actually break real usage? I doubt we'll bother about making all the tests pass with 1.x on Python 3.4 if doing so would take much effort, but we'd probably accept fixes for actually running 1.x on 3.4.

[takluyver]

I'll merge this later today or tomorrow unless anyone stops me.