Enh/extend mosaic kwargs #24604

tacaswell · 2022-12-03T20:20:12Z

PR Summary

Closes #24571

Open questions:

do we want this?
what should the name be?
is it exciting enough to get a whats new?

PR Checklist

Documentation and Tests

Has pytest style unit tests (and pytest passes)
Documentation is sphinx and numpydoc compliant (the docs should build without error).
New plotting related features are documented with examples.

Release Notes

New features are marked with a .. versionadded:: directive in the docstring and documented in doc/users/next_whats_new/
API changes are marked with a .. versionchanged:: directive in the docstring and documented in doc/api/next_api_changes/
Release notes conform with instructions in next_whats_new/README.rst or next_api_changes/README.rst

rcomer · 2022-12-05T19:19:04Z

I would use it, and I’d be excited to read about it in the what’s new. For a name, maybe something like per_subplot_kw or mapped_subplot_kw?

story645 · 2022-12-05T20:01:11Z

For a name, maybe something like per_subplot_kw or mapped_subplot_kw?

I'm concerned about having two keyword arguments that are circular references to each other. What I mean is the docs are something like:

subplot_kw: dictionary of keyword arguments passed through for all subplots, per_subplot_kw takes precedence when there's a conflict
per_subplot_kw: dictionary of keyword arguments passed through for individual subplots, defaults to subplot_kw

My concern is that coupled parameters leads to confusion about which parameter to use and what is prioritized; for example, confusion around colors. If consensus is that keeping these parameters separate is a good idea for the default fall through, then I think they should at least follow each other in the documentation and not be separated by the gridspec_kw.

anntzer · 2022-12-05T20:51:42Z

per_subplot_kw doesn't default to subplot_kw, though. It would be "additional keyword arguments passed to the subplot constructors; take precedence over conflicting entries in subplot_kw".

tacaswell · 2022-12-05T20:57:38Z

I like mapped_subplot_kw. It is a bit long, but it makes the point that it is mapped (and "to the keys in the mosaic" is a logical thing to have shared keys with, and it is the keywords passed to add_subplot)

story645 · 2022-12-05T21:22:21Z

"take precedence over conflicting entries in subplot_kw"

As mentioned in #24571, when there's no conflict any arguments passed to 'subplot_kw' are applied (line 2063-2064), which is why I'm calling that the default/(edit is probably more appropriate to say that subplot_kw is the fallback). But more to my point, the dev calling this function has to keep track of what they passed to both parameters when either one seems to be behaving strangely.

tacaswell · 2022-12-07T21:26:50Z

Other proposals

by_subplot_kw
by_label_subplot_kw

and post fixing a modifier (from call with @QuLogic and @ksunden ).

tacaswell · 2022-12-08T22:07:59Z

per the call:

named per_subplot_kw
put next to subplot_kw docstring

doc/users/next_whats_new/per_subplot_mosaic.rst

lib/matplotlib/figure.py

timhoffm · 2022-12-13T17:10:37Z

lib/matplotlib/figure.py

@@ -1821,6 +1841,9 @@ def subplot_mosaic(self, mosaic, *, sharex=False, sharey=False,
            The string notation allows only single character Axes labels and
            does not support nesting but is very terse.

+            Tuples may be used instead of lists and tuples may not be used


Tuples may be used instead of lists

Do we want to allow this explicitly? I'm in favor of having one documented and recommended way, so that usages are consistent and not half of the people are using lists and the other half tuples.
It's a separate topic that tuples are accepted due to duck-typing and can be used in special situations.

Agree to drop the first half of this sentence as many things will "work":

import matplotlib.pyplot as plt from collections import deque a = deque([deque(['A', 'B']), deque(['C', 'D'])]) plt.subplot_mosaic(a) plt.show()

but we should only document one spelling. In the tutorial I think we also pass a 2D numpy array.

Maybe the input should be "something coercible to a numpy object array of hashables"?

This also "works" 🤦🏻

plt.subplot_mosaic(frozenset([frozenset('ab'), frozenset('cd')]))

I updated this wording.

On one hand, no one has complained that they can not use tuples as keys yet, on the other hand if we take "tuple is broadcast" in this PR then relaxing this in the future would cost of some tricky logic around sorting out what to do about:

plt.subplot_mosaic([[('a', 'b'), 'a'], ['b', 'c']], per_subplot_kw={('a', 'b'): {}})

I think you have to always treat the tuple as "can I broadcast this" first and then we decide if we fallback to "maybe this is a key?" (as if you do it the other way you can never broadcast to 'a' and 'b' in this case). We might also be able to break this ambiguity by requiring tuple keys to always be nested in another tuple for per_subplot_kw as {(('a', 'b'),): {}} which is annoying, but not the worst.

I think the wording as you have written it now is the correct solution:

The Axes identifiers may be str or a non-iterable hashable object (e.g. tuple s may not be used).

lib/matplotlib/figure.py

rcomer

I have some tiny doc suggestions. Also should the new keyword be included in the pyplot.subplot_mosaic docstring?

Otherwise, this looks good to me.

tutorials/provisional/mosaic.py

lib/matplotlib/figure.py

tacaswell · 2022-12-18T18:23:34Z

@rcomer pyplot should be updated, thank you for catching that!

tacaswell · 2022-12-19T14:56:40Z

Drat, I have been seeing those docs build failures locally for a while now and had not chased them down the source of the problem. Given that numpy 1.24 just came out I am now confident that it was a change to numpy, looking into this....

tacaswell · 2022-12-19T16:15:30Z

ok, I understand the doc failure and will open a separate PR to fix it.

This is already done at creation time.

Co-authored-by: Tim Hoffmann <[email protected]>

Co-authored-by: Ruth Comer <[email protected]>

doc/users/next_whats_new/per_subplot_mosaic.rst

lib/matplotlib/figure.py

tutorials/provisional/mosaic.py

Co-authored-by: Elliott Sales de Andrade <[email protected]>

tacaswell added this to the v3.7.0 milestone Dec 3, 2022

story645 marked this pull request as ready for review December 3, 2022 23:15

story645 marked this pull request as draft December 3, 2022 23:17

tacaswell force-pushed the enh/extend_mosaic_kwargs branch from 6ca4fbb to c058015 Compare December 8, 2022 22:06

tacaswell added the Release critical For bugs that make the library unusable (segfaults, incorrect plots, etc) and major regressions. label Dec 8, 2022

tacaswell marked this pull request as ready for review December 8, 2022 22:07

timhoffm reviewed Dec 9, 2022

View reviewed changes

doc/users/next_whats_new/per_subplot_mosaic.rst Outdated Show resolved Hide resolved

lib/matplotlib/figure.py Show resolved Hide resolved

lib/matplotlib/figure.py Outdated Show resolved Hide resolved

tacaswell force-pushed the enh/extend_mosaic_kwargs branch from a724652 to 562ba8d Compare December 9, 2022 18:08

timhoffm reviewed Dec 13, 2022

View reviewed changes

timhoffm approved these changes Dec 14, 2022

View reviewed changes

rcomer reviewed Dec 18, 2022

View reviewed changes

tutorials/provisional/mosaic.py Outdated Show resolved Hide resolved

lib/matplotlib/figure.py Outdated Show resolved Hide resolved

tacaswell and others added 7 commits December 21, 2022 15:26

MNT: remove explicit label setting

7760e5e

This is already done at creation time.

ENH: add ability to pass per-Axes subplot_kw through subplot_mosaic

f74fe35

DOC: clarify special case

e651574

Co-authored-by: Tim Hoffmann <[email protected]>

DOC: tweak wording on what are allowed keys in the mosaic

54189fc

DOC: fix markup

4e3143e

DOC: fix word choice and markup

f714df7

Co-authored-by: Ruth Comer <[email protected]>

DOC: add missing docs to pyplot

5b0fb4a

tacaswell force-pushed the enh/extend_mosaic_kwargs branch from 68d53c3 to 5b0fb4a Compare December 21, 2022 20:26

rcomer approved these changes Dec 21, 2022

View reviewed changes

QuLogic reviewed Dec 21, 2022

View reviewed changes

DOC/STY: improve wording and line wrapping

e56f876

Co-authored-by: Elliott Sales de Andrade <[email protected]>

QuLogic force-pushed the enh/extend_mosaic_kwargs branch from aa3cda5 to e56f876 Compare December 29, 2022 20:24

greglucas merged commit 65e54af into matplotlib:main Dec 29, 2022

tacaswell deleted the enh/extend_mosaic_kwargs branch January 3, 2023 22:17

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Enh/extend mosaic kwargs #24604

Enh/extend mosaic kwargs #24604

tacaswell commented Dec 3, 2022

rcomer commented Dec 5, 2022

story645 commented Dec 5, 2022

anntzer commented Dec 5, 2022

tacaswell commented Dec 5, 2022

story645 commented Dec 5, 2022 •

edited

Loading

tacaswell commented Dec 7, 2022

tacaswell commented Dec 8, 2022

timhoffm Dec 13, 2022

tacaswell Dec 13, 2022

tacaswell Dec 13, 2022

tacaswell Dec 13, 2022

timhoffm Dec 14, 2022

rcomer left a comment

tacaswell commented Dec 18, 2022

tacaswell commented Dec 19, 2022

tacaswell commented Dec 19, 2022

Enh/extend mosaic kwargs #24604

Enh/extend mosaic kwargs #24604

Conversation

tacaswell commented Dec 3, 2022

PR Summary

PR Checklist

rcomer commented Dec 5, 2022

story645 commented Dec 5, 2022

anntzer commented Dec 5, 2022

tacaswell commented Dec 5, 2022

story645 commented Dec 5, 2022 • edited Loading

tacaswell commented Dec 7, 2022

tacaswell commented Dec 8, 2022

timhoffm Dec 13, 2022

Choose a reason for hiding this comment

tacaswell Dec 13, 2022

Choose a reason for hiding this comment

tacaswell Dec 13, 2022

Choose a reason for hiding this comment

tacaswell Dec 13, 2022

Choose a reason for hiding this comment

timhoffm Dec 14, 2022

Choose a reason for hiding this comment

rcomer left a comment

Choose a reason for hiding this comment

tacaswell commented Dec 18, 2022

tacaswell commented Dec 19, 2022

tacaswell commented Dec 19, 2022

story645 commented Dec 5, 2022 •

edited

Loading