nan support

[maartenbreddels]

traittypes/traittypes/traittypes.py

Line 112 in 609b402

if not np.array_equal(old_value, new_value):

will think arrays are not equal always when they contain a nan, the proper code might be sth like:

            try:
                return np.allclose(a, b, 0, 0, equal_nan=True)
            except (TypeError, ValueError):
                return False

There are more places around, like in ipywidgets

• https://github.com/jupyter-widgets/ipywidgets/blob/463cb0444d1a2f31c3d5cacfe6b38d067f231406/ipywidgets/widgets/widget.py#L530
  I'm not sure we should have a common strategy, or it should be a per-situation solution, such as the referenced bqplot issue (coming soon).