Stackplot weighted_wiggle zero-area fix

[egpbos]

See #6313.

[tacaswell]

Can you force-push just the first commit (with out the merge from master)?

Can you also add an image test that exercises this condition? It might be best to modify the test data in test_axes.test_stackplot_baseline and update the expected images.

Other wise 👍

[egpbos]

Force push done.

[tacaswell]

Will you have time to add a test for this?

[egpbos]

I'm trying, but I'm running on OSX in a Conda virtual environment, so having issues with that. The frameworkpython method doesn't seem to work for me, and I don't have that much time, unfortunately. Any ideas on getting my venv working?

[egpbos]

Just to be sure this makes sense, what I'd do is simply add

    d[50,:] = 0  # test for fixed weighted wiggle (issue #6313)

after line 1394 in matplotlib/tests/test_axis.py. To generate the baseline image, I have this script, but that's the one that won't run in my venv, so that's where I'm stuck.

[efiring]

Others might have better methods, but here is what I do to develop mpl from within a conda environment on OSX:

1. Use conda to make a test environment with the desired python version, and use conda to install matplotlib in it.
2. Use conda to uninstall matplotlib.
3. In my github clone of matplotlib, make a feature branch, and make the desired changes to it.
4. In the root of that clone directory, 'pip install .'
5. In a directory somewhere away from the cloned directory, to run your particular modified test, run:
   [path_to_matplotlib]/tests.py matplotlib.tests.test_axis.test_stackplot_baseline
6. If everything worked as desired, copy the result images from the new result_images subdirectory of your testing directory to the appropriate baseline_images location in the repo, git commit everything, and push. If it didn't work, then modify the code, run 'pip uninstall matplotlib', and go back to 4.

[egpbos]

@efiring Awesome, that did the trick, thanks! Well, almost, I had to combine it with this pythonw/python.app tip. Perhaps an idea to put this method in the documentation? I assume steps 1 and 2 are to automatically install dependencies, right? Anyway, will try to generate the test images using this setup asap.

[egpbos]

Ok, forgot to add matplotlib.style.use('classic') to the image generation script, that seems to fix the tests. By the way, is there a developers guide with info like this?

[egpbos]

@tacaswell: matplotlib.tests.test_mlab.rec2txt_testcase is failing in AppVeyor. Does that have anything to do with stackplot / this PR?

[tacaswell]

No, that is a separate issue.

[tacaswell]

Thanks! I think this is your first contribution to the mpl source. Congratulations and hope we hear from you again.

Backported to v2.x as 102872a