Use numbers abstract base class

[jdemeyer]

Be compliant with PEP 3141: replace code like isinstance(foo, (int, float)) by isinstance(foo, numbers.Real). This is in particular needed for better SageMath support.

[jdfreder]

Looks like the tests themselves are failing at a lower level, in the widgets themselves. We use traitlets to strictly type min and max to int and float, so I think the SageMath numbers are conflicting with that. You'll either need to implement custom trait types that does the type checking how you propose here and then sprinkle those throughout the widgets, or probably better, open a PR against ipython/traitlets fixing the validation logic of Int and Float accordingly.

[jdemeyer]

Looks like the tests themselves are failing at a lower level, in the widgets themselves

Can you please tell me how to run the testsuite? I have no idea...

[jdemeyer]

I spoke too quickly. How to run the tests is in README.md, not in CONTRIBUTING.md (which I would find more logical).

[jdemeyer]

I think the SageMath numbers are conflicting with that.

Why do you think that? I tested the @interact in SageMath and it works.

[jdfreder]

Can you please tell me how to run the testsuite? I have no idea...

I spoke too quickly. How to run the tests is in README.md, not in CONTRIBUTING.md (which I would find more logical).

Another way is to peak at how our CI (Travis) does it, defined in .travis.yml in the repo root.
nosetests --with-coverage --cover-package=ipywidgets ipywidgets runs the Python tests, which were failing.

Why do you think that? I tested the @interact in SageMath and it works.

Awesome, glad it's working. It was an erroneous conclusion on my part after a quick look at the test output.