Mailman 3 January 2017 - Matplotlib-devel

Confusion about draw_artist
by Matthew Brett April 1, 2017

April 1, 2017

Hi, We're using 'draw_artist' in an image viewer, and I just found that our viewer is crashing using the macosx backend. Here's a minimal example: import matplotlib matplotlib.use('macosx') import matplotlib.pyplot as plt fig, ax = plt.subplots(1, 1) im = plt.imshow([[1, 2], [3, 4]]) fig.canvas.draw() ax.draw_artist(im) This generates: Traceback (most recent call last): File "draw_artist.py", line 8, in <module> ax.draw_artist(im) File "/Users/mb312/dev_trees/matplotlib/lib/matplotlib/axes/_base.py", line 2423, in draw_artist raise AttributeError(msg) AttributeError: draw_artist can only be used after an initial draw which caches the render for the macosx backend, and works correctly for the tkagg backend. I can't see any way to cache the renderer here - could someone give me a pointer? Cheers, Matthew

2 5

2016-01-30 phone call notes
by Thomas Caswell Jan. 30, 2017

Jan. 30, 2017

In attendance were my self, Nelle Varoquaux, and Ryan May. - went over 2.0.1 release critical bugs - most have patches - imshow needs c level debugging :( - #7991 needs attention - the rest need minor fixes updates - discussed pytest convsion - issues with conftests.py - issues with pgf not failing when we expect - sorted it out, bugs in test framework, have been 'default' style sense #6757 - Ryan will update codecov to check that all tests are run - Ryan's talk went well - Discussed automating wheel building - plan for 2.0.1 next weekend (2017-02-04) Tom

1 0

Reorganising the test files?
by Nelle Varoquaux Jan. 29, 2017

Jan. 29, 2017

Hello everyone, Now that the pytest migration is almost over (thanks Elliott!), I was wondering if this could be a good opportunity to reorganize the test files to follow a "normal" python package structure. The recommended python package structure is to place a tests folder in each submodule, and the relevant tests files in this folder. To take a concrete example, it would for example require moving the tests corresponding of matplotlib.sphinx_ext.mathmpl.py to matplotlib/sphinxext/tests/test_mathmpl.py This would be a more or less incompatible change in the sense that matplotlib.tests seems to be an importable module (it probably shouldn't), with about a dozen functions. We could easily make this backward compatible in case anyone does use this module anywhere. ['__builtins__', '__cached__', '__doc__', '__file__', '__loader__', '__name__', '__package__', '__path__', '__spec__', '_multiprocess_can_split_', 'absolute_import', 'assert_str_equal', 'difflib', 'division', 'os', 'print_function', 'setup', 'six', 'unicode_literals'] If we go down that route, we need to completely finish the pytest migration first. Cheers, N

7 8

user mailing list policy
by Thomas Caswell Jan. 29, 2017

Jan. 29, 2017

Going through the moderation backlog on matplolib-user we have a way high false-negative rate of non-subscribing posting with reasonable emails (you just got a pile of old emails because I cleared the queue). I think we should either open the user list to any posters or get some sort of rotation going so that the moderation queue gets handled (ideally) once a day. Tom

3 3

Fwd: our tests
by Elliott Sales de Andrade Jan. 28, 2017

Jan. 28, 2017

Hi Eric, Sorry if you receive this twice. On 24 November 2016 at 18:49, Eric Firing <efiring.ocean(a)gmail.com> wrote: > Does anyone know why: > > 1) In our Travis suite, the USE_PYTEST run takes much longer than the > others. > > Nelle has answered this one already. > 2) On master (approximately), when I try to run a single function with > py.test, it isn't found. Example: > > py.test lib/matplotlib/tests/test_axes.py::test_hist_steplog > =============================================== test session starts > ================================================ > platform darwin -- Python 3.5.1, pytest-2.8.5, py-1.4.31, pluggy-0.3.1 > rootdir: /Users/efiring/work/programs/py/mpl/matplotlib, inifile: > pytest.ini > collecting 530 items > =========================================== no tests ran in 0.39 seconds > =========================================== > ERROR: not found: /Users/efiring/work/programs/p > y/mpl/matplotlib/lib/matplotlib/tests/test_axes.py::test_hist_steplog > (no name '/Users/efiring/work/programs/py/mpl/matplotlib/lib/matplotl > ib/tests/test_axes.py::test_hist_steplog' in any of [<Module > 'lib/matplotlib/tests/test_axes.py'>]) > > This works for me; I think you need to update your version of pytest: py.test lib/matplotlib/tests/test_axes.py::test_hist_steplog ============================= test session starts ============================= platform linux -- Python 3.5.2, pytest-3.0.3, py-1.4.31, pluggy-0.4.0 rootdir: /home/elliott/code/matplotlib, inifile: pytest.ini collected 505 items lib/matplotlib/tests/test_axes.py ... ========================== 3 passed in 3.78 seconds =========================== Note, there are 3 tests because of the 3 extensions that are checked by the image_comparison decorator. > 3) Also on master, I get errors with tests.py. Example: > > python tests.py lib/matplotlib/tests/test_axes.py > > [snip] > > > This will not work at the moment. tests.py continues to use nose, so it will not run pytest-based files. The list of files that only work with nose is very short right now. As soon as my other PR is merged, I will open a PR that should make this work again (there are conflicts). -- Elliott -- Elliott

1 0

our tests
by Eric Firing Jan. 28, 2017

Jan. 28, 2017

Does anyone know why: 1) In our Travis suite, the USE_PYTEST run takes much longer than the others. 2) On master (approximately), when I try to run a single function with py.test, it isn't found. Example: py.test lib/matplotlib/tests/test_axes.py::test_hist_steplog =============================================== test session starts ================================================ platform darwin -- Python 3.5.1, pytest-2.8.5, py-1.4.31, pluggy-0.3.1 rootdir: /Users/efiring/work/programs/py/mpl/matplotlib, inifile: pytest.ini collecting 530 items =========================================== no tests ran in 0.39 seconds =========================================== ERROR: not found: /Users/efiring/work/programs/py/mpl/matplotlib/lib/matplotlib/tests/test_axes.py::test_hist_steplog (no name '/Users/efiring/work/programs/py/mpl/matplotlib/lib/matplotlib/tests/test_axes.py::test_hist_steplog' in any of [<Module 'lib/matplotlib/tests/test_axes.py'>]) 3) Also on master, I get errors with tests.py. Example: python tests.py lib/matplotlib/tests/test_axes.py [snip] ====================================================================== ERROR: matplotlib.tests.test_axes.test_errorbar_inputs_shotgun ---------------------------------------------------------------------- Traceback (most recent call last): File "/Users/efiring/anaconda/envs/test/lib/python3.5/site-packages/nose/case.py", line 198, in runTest self.test(*self.arg) File "/Users/efiring/work/programs/py/mpl/matplotlib/lib/matplotlib/testing/decorators.py", line 150, in wrapped_callable func(*args, **kwargs) TypeError: test_errorbar_inputs_shotgun() missing 1 required positional argument: 'kwargs' ====================================================================== ERROR: matplotlib.tests.test_axes.test_remove_shared_axes ---------------------------------------------------------------------- Traceback (most recent call last): File "/Users/efiring/anaconda/envs/test/lib/python3.5/site-packages/nose/case.py", line 198, in runTest self.test(*self.arg) File "/Users/efiring/work/programs/py/mpl/matplotlib/lib/matplotlib/testing/decorators.py", line 150, in wrapped_callable func(*args, **kwargs) TypeError: test_remove_shared_axes() missing 2 required positional arguments: 'shared_axes_generator' and 'shared_axis_remover' ---------------------------------------------------------------------- Ran 437 tests in 142.437s FAILED (SKIP=1, errors=2) Exception ignored in: <generator object TestLoader.loadTestsFromGenerator.<locals>.generate at 0x108f40ca8> RuntimeError: generator ignored GeneratorExit Eric

4 3

typo in matplotlib docs
by Howard, Tom Jan. 28, 2017

Jan. 28, 2017

Hi I noticed a (fairly trivial) typo on page http://matplotlib.org/users/recipes.html I couldn't work out how to report it - hope this is a suitable forum. Details: ax.plot(t, mu1, lw=2, label='mean population 1', color='blue') ax.plot(t, mu1, lw=2, label='mean population 2', color='yellow') should be ax.plot(t, mu1, lw=2, label='mean population 1', color='blue') ax.plot(t, mu2, lw=2, label='mean population 2', color='yellow') Then the image makes more sense, too! ================================================ Best wishes Tom https://drtomhoward.wordpress.com/about/ Tel: +44(0)1392 886678 Dr Tom Howard, Met Office Hadley Centre FitzRoy Road, Exeter, EX1 3PB, UK

3 2

Internation Matplotlib sprint
by Nelle Varoquaux Jan. 25, 2017

Jan. 25, 2017

Hello everyone, As Thomas mentioned in the meeting summary report, we are considering organizing a sprint. Because our team is so international, we need to find a convenient location for all/most of us, and estimate how much funding we need to organize such an event. A possibility would be to organize it just after Scipy, in Austin, as some of us will be funded to attend. Please note that the bathroom bill is currently under discussion by the Texas Congress, and thus it is possible that publicly funded organisation won't be able to reimburse travel to Scipy 2016 this year. In order to estimate the budget, please fill in this very short survey: https://docs.google.com/forms/d/e/1FAIpQLSfUB86RsrYGkJoJEEPxmlS-VwYrAi8Jtmu… This is nothing definite: we're just trying to see what's possible :) Thanks, Nelle

1 0

project email addresses
by Thomas Caswell Jan. 25, 2017

Jan. 25, 2017

Folks, We just got matplotlib(a)numfocus.org as a usable email address. The initial impetus to get this address was for use with appveyor, but I suspect it may be useful for other reasons. If you would like / need access send me an email off-list and we will sort how to get you the password. Tom

1 0

2017-01-23 phone call notes
by Thomas Caswell Jan. 24, 2017

Jan. 24, 2017

Folks, Today we had my self, nellev and efiring. - Discussed that we have had surprisingly few 2.0 related bug reports, unclear if we have not had uptake yet or if we really didn't break all that much - Discussed existing branches on upstream, tried to see if we could restrict permissions to create new branches on upstream, seems like not (coupled to write permissions). Please be careful to not push new branches to upstream! - CI is taking too long, how to speed up: - drop builds on PRs (and only do on master), candidates are dropping 3.4 or 3.5 - reduce appveyor matrix on PRs - (not discussed on call) raise ~500$USD / yr to get 2 appveyor builds. - make the doc build faster (can we cache anything?) - auto-canceling super-ceded PR builds is in beta on travis. Just being patient might work here. - BIDS is working on a distributed 'doc-a-thon', (Nelle will follow up with details if mpl wants to participate). - discussed difficulty of reviewing doc changes (are they correct? is english correct? is it formatted properly?) - discussed possibility of a in-person sprint (again Nelle will follow up) Please correct me if I missed anything, wrote this from memory after the fact. Tom PS Sorry for stealing some of your thunder Nelle :(

3 2