Very nice and much needed pull request.

I wonder, however, whether we need to test all three of png, pdf and svg for these. The pdf and svg files in particular are quite large. (Yes, I know our test suite already has some files of that size, but all the more reason to avoid expanding further... 😉). Maybe we should test PNG only (as those are the smallest files), unless there is some known bug that shows up only with the other formats. Note, we currently do this for the delaunay triangulation tests for the same reason.

Yes. Also, there still appears to be some text on some of the output files.

I've restricted it to only pngs and removed the text from the images

Fixed for python 2.6

👍

Looks good -- and thanks for amending the commit so the PDFs and SVGs don't make it into the history.

Merging.