Patch for issue #6009

[afvincent]

Here is a patch for issue #6009 about the EngFormatter in matplotlib.ticker.

It simply enforces a space when there is no SI prefix but yet a unit symbol. For example “10 seconds” should be formatted as “10 s” by the EngFormatter (previously “10s”).

[WeatherGod]

Should add a unit test to exercise this. Preferably not an image test (we have plenty of those), but one that tests the strings coming out of EngFormatter.

[tacaswell]

I think this should be == not is. It is just a little less brittle.

[tacaswell]

Left minor comment, but other wise 👍

[afvincent]

I don't understand why my last commit broke the Travis CI build test: it seems to me that the changes are really minor, aren't they? Furthermore, when looking at the failure logs in “Details” section of Travis, one can find errors like FAIL: matplotlib.tests.test_colorbar.test_colorbar_closed_patch.test and all the 6 (unexpected) failures relate to test_colorbar. How can it be impacted by a change in ticker.EngFormatter?

[jenshnielsen]

Travis is broken because the merge of another pr broke master is is not caused by your pr

[afvincent]

Ok, thank you for the explanation :).

[tacaswell]

Can you rebase so we can merge it?

[afvincent]

I think I've done a mistake on this branch with my git local repo (what a change…). I used the same branch to implement the enhancement I presented in #6533 (which also fix this issue btw). Should I go back in time with git and then rebase, or do something else? I guess opening a new PR (on a branch from an up-to-date master) which fixes this small issue and also implements the space_sep new option isn't really a clean option, is it?

[afvincent]

I rebased after deleting my last commit (the one with the enhancement #6533). Hopefully it will be OK.

[tacaswell]

backported to v2.x as 59e0ab9