make plotting independent of tulip

[johnyf]

polytope.polytope and polytope.plotboth depend ontulip.graphicsfor plotting (remnants from a time whenpolytopewas a subpackage oftulip). For example:

polytope/polytope/polytope.py

Line 397 in 3a721bc

from tulip.graphics import newax

polytope/polytope/plot.py

Line 81 in 4b3d1c3

from tulip.graphics import newax

polytope is a dependency of tulip. Therefore, it cannot itself depend on tulip, because this results in a cyclic dependency (hence the labeling as "bug").
The usage is very limited: tulip.graphics.newax and tulip.graphics.dom2vec, both of which are small or eliminable.

The module tulip.graphics originates from my porting to Python part of plotting utilities that I had developed in Matlab. In Python, either some things are not needed at all, or there is a much better and simpler way to do them, or someone - like matplotlib and friends - has already written what is needed. Also, tulip.graphics derives from code that was handling seamless plotting in 2 or 3 dimensions, which in practice is not used in tulip or polytope (and I don't think it would be very practical to plot 3D partitions).

[johnyf]

A second flaw in the design (by me) is that polytope.polytope.Polytope.plot returns silently (!) if tulip.graphics.newax fails to import. It should raise ImportError, because requesting a plot and not getting one is an error, and "Errors should never pass silently. Unless explicitly silenced." (PEP 20).

[slivingston]

Thanks for creating this issue. I noticed the unnecessary dependence on tulip previously but had not taken action.

[johnyf]

Addressed as of 9d96bc0. Plotting needs to be tested more, but that is a separate issue.