Added tripcolor triangle-centred colour values.

[ianthomas23]

Implementation of issue #811.

triplot determines whether color values are specified at triangulation points or triangles based on the length of the color values array, as this is the most intuitive way to do it. The only complication is if there are the same number of points as triangles; in this case you may need to use the colorpoints kwarg as well. This is a little clumsy, but seems better than forcing the user to always use a kwarg to get the new behaviour.

Existing code is unaffected by the change.

[bfroehle]

I find the kwarg colorpoints confusing. Reading through code and seeing tripcolor(x, y, t, C, colorpoints=False) wouldn't indicate to me a solid shading of each face. I think a clearer interface would be tripcolor(x, y, t, facecolors=C).

If you want to allow it to be short circuited to tripcolor(x, y, t, C) depending, of course, on the size of C... well that seems reasonable too.

[mdboom]

Note that elsewhere in pcolormesh, the faceted option has been deprecated for some time and recently removed in favor of respecting the edgecolors kwarg. Does that make sense to do here as well?

[mdboom]

This looks like a nice improvement. In addition to my commented about faceted, this should have an example and a unit test.

[ianthomas23]

Apologies for taking so long, but I have now addressed the comments. The changes are, in order of my commits:

1. Deprecated shading='faceted' and supported the edgecolors kwarg instead, as in pcolormesh.
2. Removed the confusing colorpoints kwarg and used facecolors instead. I've kept the short-circuited option of not using facecolors if it is clear from the length of C whether it applies to faces or points.
3. Modified the existing tripcolor example to include a facecolors plot.
4. Added an image comparison unit test.

Note that the edgecolors kwarg isn't supported for shading='gouraud'. It would be nice to support this, but I cannot see how without changing many files in the backend code. The behaviour is at least consistent with pcolormesh. I've noticed that pcolor doesn't support gouraud shading at all - I presume the conclusion here is that support for gouraud shading is patchy and needs to be reconsidered sometime.

[efiring]

Thanks very much for the changes.
The merge button is grey, however, so it looks like you need to rebase against master.

[ianthomas23]

I've rebased against master as requested.

[efiring]

I think you mean "at triangulation points", not "at triangles"; the latter would be the same as using facecolors.

[ianthomas23]

You are correct; well spotted!