Kindly requesting some help here: I honestly can't see what the test failure is due to or how it is related to this PR.

The test failure is likely intermittent and not related to your work. I have restarted the test

Looks like @jenshnielsen was right here. The tests are now passing.

Just played with this and one subtle thing that struck me is that the special values of "None" and "nothing" are now sorted somewhere in the middle (by n obviously) and it might be nice to have those at the top.

Just played with this and one subtle thing that struck me is that the special values of "None" and "nothing" are now sorted somewhere in the middle (by n obviously) and it might be nice to have those at the top.

The current behavior is that the multiple entries of None (or none or nothing, depending on whoever PYTHONHASHSEED decided to pick) are randomly scattered across the combobox, so sorting is already an improvemvent.

I could special-case them to the top, but if we go that route we may as well come up with a better order of the markers (such as one that keeps triangles, squares, pentagons and hexagons next to each other and in that order).

The code for uniqueifying the symbols seems too complex/sutble for what it is doing. I think what I proposed in the now folded comment is easier to read/understand.

Perhaps this new version is clearer?

Yes, that is much clearer

@mdboom I think sorting alphabetically makes enough sense here.

@anntzer Sorry, I just sorted out what which way you are de-duplicating the mapping and I was worried over nothing.