Commit schedule for matplotlib-inline

[kishkash555]

The last release of matplotlib-inline is 0.1.3 from 9 months ago (Sept. 2021).
There have been several code changes since.
Specifically I am interested in commit a433a89 from May 12/13th which includes dropping some Matplotlib rcParams overrides that were specific to ipython notebooks.
I would like to know if any schedule is currently in place for this commit to be published.
Thanks.

[fperez]

Yup, thanks for raising this issue @kishkash555!

@agoose77, @minrk & team - any objection to making a release soon? I'm going to be teaching with this code in a few weeks and would rather not deal with that annoying rcParams issue any more (I have a bunch of HW assignments with the hacky fix in there).

I haven't made a release of any of our projects in ages, but I'm happy to dust off my keys for this one :) Any objections?

[fperez]

Ping @martinRenou @Carreau @SylvainCorlay - any of you have bandwidth to cut a release? If not LMK and I can dust things off to do it , though I don't think I have the auth on PyPI for this particular package.

I also see the conda-forge feedstock is @martinRenou's; I'm going to guess that doesn't need any updates, but I'm also willing to help there if needed.

[martinRenou]

Thanks for pinging me @fperez, I added you as an owner on Pypi :)

[fperez]

Thx @martinRenou!

I've accepted the invite and updated my PyPI account with 2FA, now that I'm using it again...

Anything else I should keep in mind before cutting a release? I haven't released in PyPI in ages (probably an old IPython 0.11 series or so)...

It seems this project simply uploads a wheel and a source package...

I don't see a pyproject.toml file, but for now I'm happy leaving things as-is with the setup.cfg we have...

I ran a quick test and I can build the wheel and source dist just fine.

Anything else people can think of before a release? If not LMK and I'll update the version #...

On that topic: I plan on just calling it 0.1.4, as the changes are pretty minimal. But if folks prefer we can call it 0.2.0 too.

[fperez]

Also @martinRenou one question: once I make the PyPI release, does the conda-forge build get triggered automatically, or do I need to do anything else on that end?

[martinRenou]

I usually just do:

python -m build

Then I check that the wheel and source dist are correct under dist/. Then I do:

twine upload dist/*

does the conda-forge build get triggered automatically, or do I need to do anything else on that end?

I am not sure how conda-forge gets triggered honestly. But pushing a tag on the repo and pushing the Pypi release seems to trigger it.

You can also push a PR manually to the feedstock if you want, adding yourself as a maintainer https://github.com/conda-forge/matplotlib-inline-feedstock

[fperez]

Ok, thx @martinRenou!

Given there's no objections so far, folks like @juanitorduz need this, and I have very limited time, I'm going to do it now. Tagging 0.1.4 and pushing out, 🤞 :)

[juanitorduz]

Thank you @fperez for pushing this. Much appreciated! It is definitively a big plus for users like me but it is definitively not urgent so no rush / stress 😅 !

[fperez]

Wohoo!!.

Thx @martinRenou & @juanitorduz! It was fun to dust off my "release manager" hat, even if for a tiny bit :) I had a badly outdated pypirc file with the old .io urls so it took me a bit of head-scratching.

Glad to have this out - I hope the conda build happens automaticall, as I pushed a 0.1.4 tag too into this repo. If it hasn't happened in a day or two we can check again.

[juanitorduz]

Thank you !

[fperez]

Mmh - hold off. I might have messed up!

[fperez]

🤦

[fperez]

I checked the build, etc, but I mistakenly was working off an outdated fork, so I think I missed a few commits. Let me check before you install.

[fperez]

Ok, now for real!, also on PyPI.

I had to yank 0.1.4 from PyPI and just made a 0.1.5.

Oh well, classic mix-up from not doing this all the time: I was working off my own fork which I hadn't updated from upstream in ages, and ended up making 0.1.4 out of a weird mix of old upstream + local fixes.

I tell my students to always watch their remotes when doing this kind of thing, yet here we are :)

In any case, now it's actually done, and I can "get back to work" :)

[juanitorduz]

Thanks @fperez ! I can confirm that

matplotlib==3.5.3
matplotlib-inline==0.1.5

are working as expected (at least in a couple of tests).

Before "going back to work" get yourself a nice coffee / drink ;)

[martinRenou]

Thank you!

[sryza]

After this new release, over at Dagster, we're seeing failures in our tests that do plotting inside notebooks. The error we're seeing is:

[2022-08-17T00:37:46Z] /workdir/examples/docs_snippets/.tox/py37/lib/python3.7/site-packages/matplotlib/pyplot.py in subplots(nrows, ncols, sharex, sharey, squeeze, subplot_kw, gridspec_kw, **fig_kw)
--
  | [2022-08-17T00:37:46Z]    1451
  | [2022-08-17T00:37:46Z]    1452     """
  | [2022-08-17T00:37:46Z] -> 1453     fig = figure(**fig_kw)
  | [2022-08-17T00:37:46Z]    1454     axs = fig.subplots(nrows=nrows, ncols=ncols, sharex=sharex, sharey=sharey,
  | [2022-08-17T00:37:46Z]    1455                        squeeze=squeeze, subplot_kw=subplot_kw,
  | [2022-08-17T00:37:46Z]
  | [2022-08-17T00:37:46Z] /workdir/examples/docs_snippets/.tox/py37/lib/python3.7/site-packages/matplotlib/pyplot.py in figure(num, figsize, dpi, facecolor, edgecolor, frameon, FigureClass, clear, **kwargs)
  | [2022-08-17T00:37:46Z]     808             facecolor=facecolor, edgecolor=edgecolor, frameon=frameon,
  | [2022-08-17T00:37:46Z]     809             FigureClass=FigureClass, **kwargs)
  | [2022-08-17T00:37:46Z] --> 810         fig = manager.canvas.figure
  | [2022-08-17T00:37:46Z]     811         if fig_label:
  | [2022-08-17T00:37:46Z]     812             fig.set_label(fig_label)
  | [2022-08-17T00:37:46Z]
  | [2022-08-17T00:37:46Z] AttributeError: 'NoneType' object has no attribute 'canvas'

[fperez]

Mmh, sorry to hear that @sryza! Any chance you could provide us with

a) a self-contained small example that produces the problem
b) version info on matplotlib that you're using?

I'm happy to investigate, but so far I can't seem to reproduce your failure with some simple attempts.

[sryza]

Here's the notebook that's triggering it: https://github.com/dagster-io/dagster/blob/master/examples/docs_snippets/docs_snippets/legacy/data_science/iris-kmeans_2.ipynb.

This is with matplotlib 3.5.3.

We're running it through papermill.

[fperez]

Great, thanks for the reproducible case @sryza! I've managed to boil it down to this. A notebook with just this in one cell will show the problem:

import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns

# if the next line is uncommented, the bug disappears!
#import matplotlib_inline

g = sns.PairGrid(pd.DataFrame(dict(a=[1])))
plt.figure()

Oh, and as I was looking for the PR culprit, I see that @1kastner in #19 kindly offered what I think is the right fix :)

Let me double check and we'll provide an update quickly.