It seems a bit odd for IPython to test for an unreleased version of SymPy.

A related question: why isn't sympy shipping this extension itself, so that users could do:

%load_ext sympy.ipythonextensions

?

Obviously, there are advantages to it being on either side, but I think it belongs in sympy in the long run.

I agree. See http://code.google.com/p/sympy/issues/detail?id=2945. The short answer is that we haven't gotten around to it yet.

I included the version check because this only works with the latest development version. Before the SymPy fix, latex(set) would just return the string form of the set, which doesn't play well with MathJax (obviously), so if you release before we do, printing of sets will be broken for anyone using the latest released versions of SymPy and IPython (and even if you don't it will be broken for anyone using an older version of SymPy).

I appreciate that, but I would want to just test for 0.7.2, which will always get the right answer for released software.

That's fine too. I just used this because I don't really know how soon there will be a 0.7.2, and I thought some users might want this now. But I will gladly change it if that's what you want.

So which would you prefer? It doesn't matter to me. Unlike with the previous PR, there aren't really any major SymPy functions that return sets or frozensets, so this probably isn't a huge deal to anyone actually using the SymPy extension with the notebook.

Is there a test equivalent to version_info > (0,7,1) that would properly identify any version after 0.7.1 including dev? I would do that.

Ah yes, of course. That just requires reordering the logic.

Thanks, merging.