[Bug]: Some styles trigger pcolormesh grid deprecation

Bug summary

Calling plot_surface followed by fig.colorbar results in the following deprecation warning:

Auto-removal of grids by pcolor() and pcolormesh() is deprecated since 3.5
and will be removed two minor releases later; please call grid(False) first.

However, setting ax.grid(False) first does not help, the Warning is still shown.

Code for reproduction

Edit: As noted further down in the discussion: The warning does not appear for the default style. But if you load, e.g., plt.style.use('bmh'), you will see the warning.

# https://matplotlib.org/stable/gallery/mplot3d/surface3d.html

import matplotlib.pyplot as plt
from matplotlib import cm
from matplotlib.ticker import LinearLocator
import numpy as np

plt.style.use('bmh')  # Edit: One of the styles with which you will see the warning

fig, ax = plt.subplots(subplot_kw={"projection": "3d"})

# Make data.
X = np.arange(-5, 5, 0.25)
Y = np.arange(-5, 5, 0.25)
X, Y = np.meshgrid(X, Y)
R = np.sqrt(X**2 + Y**2)
Z = np.sin(R)

# Plot the surface.
surf = ax.plot_surface(X, Y, Z, cmap=cm.coolwarm,
                       linewidth=0, antialiased=False)

# Customize the z axis.
ax.set_zlim(-1.01, 1.01)
ax.zaxis.set_major_locator(LinearLocator(10))
# A StrMethodFormatter is used automatically
ax.zaxis.set_major_formatter('{x:.02f}')

# Add a color bar which maps values to colors.
fig.colorbar(surf, shrink=0.5, aspect=5)

plt.show()

Actual outcome

 MatplotlibDeprecationWarning: Auto-removal of grids by pcolor() and pcolormesh() is deprecated
 since 3.5 and will be removed two minor releases later; please call grid(False) first.
 fig.colorbar(surf, shrink=0.5, aspect=5)

Expected outcome

No Warning or a warning with a fix that works.

Additional information

Works fine for matplotlib<3.5.

Operating system

Linux

Matplotlib Version

3.5.0

Matplotlib Backend

module://ipykernel.pylab.backend_inline

Python version

3.88

Jupyter version

6.4.0

Installation

conda

I cannot reproduce this in a script or in Jupyter Notebook. Maybe you have a custom matplotlibrc or style setting?

Good point, and I tried. By default it indeed does not happen. However, it happens if you use, e.g., plt.style.use('bmh') or plt.style.use('ggplot').

@anntzer

OK, so this has nothing to do with colorbar explicitly, and the same warning pops up for pcolromesh by itself:

import matplotlib.pyplot as plt
import numpy as np
plt.style.use('ggplot')

fig, ax = plt.subplots()
# Make data.
X = np.arange(-5, 5, 0.25)
Y = np.arange(-5, 5, 0.25)
X, Y = np.meshgrid(X, Y)
R = np.sqrt(X**2 + Y**2)
Z = np.sin(R)

surf = ax.pcolormesh(X, Y, Z)
plt.show()

MatplotlibDeprecationWarning: Auto-removal of grids by pcolor() and pcolormesh() is deprecated since 3.5 and will be removed two minor releases later; please call grid(False) first.
  surf = ax.pcolormesh(X, Y, Z)

This came in in #15604 (@anntzer) In that PR we stopped pcolormesh from removing the grid lines automatically, which was the old behaviour.

OTOH, I guess I am not clear why we are deprecating this, or if it is even possible to do so cleanly. Its impossible for pcolormesh to know if the user expects the grids to stay or go.

[tl/dr] I needed this write-down to find out what do to. Go to Solution if you don't want to follow the details.

I'll try to summarize the situation.

History

Basic conflict:

depending on style we may or may not have grids on; plus maybe explicitly set by the user.
pcolormesh with a grid on top is not wanted most of the time.

Before #15604 the canonical resolution is that the mesh trumps the grid and pcolormesh deactivates the grid.
This is what you want in many cases, but not always, e.g. The mesh doesn't fill the whole Axes (#15600 which brought up the topic) or a user could have explicitly activated a grid.

Conceptually, auto-remove-grid through a plotting function is questionable. That's beyond their responsibility.

So #15604 deprecated the behavior.

We can get away with this because it doesn't change appearance for most standard cases:

All shipped styles with grid use axes.axisbelow: True and thus the grid vanishes behind the mesh
The default style has axes.axisbelow: line but no grid

Problem here

The problem is that these cases still see the (irrelevant) deprecation message.

Solution

I think we can remove most of the deprecation annoyance by not warning in case of axes.axisbelow: True.

Alternative

For completeness only, after writing this I decided that actually removing grid manipulation is the correct thing to do.
imshow() doesn't remove the grid either.

One could also reconsider that grid-removal is somewhat meaningful for pcolormesh, but make it configurable via a remove_grid=True parameter. That way, users can deactivate the behavior if needed. This has still a small discoverability problem, but bearable:

I run my code and the grid vanishes
I debug and find that pcolormesh removes it
I check the docs and find remove_grid

Note: I don't think that defaulting to False would make sense. With either case one would change the default only in rare cases. But adding a parameter that optionally executes a command for rare cases, does not make sense if I can do the same with a simple ax.grid(False) in the next line.

Hello,
I'm not sure if I'm adding to this discussion, but I'm seeing the same warning. I'm importing seaborn and calling sns.set() and sns.set_palette('bright')

I feel it was a misfeature that we ever removed the grid in the first place just for this artists type.
I don't see a way to sensibly deprecate this since it is not something the user is actively doing. So we can
1. stop the behaviour without deprecation, or
2. live with it.

I personally prefer 1 as I consider this a bug, but can understand why some would think this was a feature and would prefer 2.

... added to the weekly dev call to discuss, unless we can come to consensus here.

Was there anything decided in this regard? I see the same problem appearing in v3.5.1 as well.

@prisae

@prisae No I don't think there was a discussion. I'll re-add to this weeks call. Thanks for the ping...

The consensus on the call is to go with option 3 (leave the warning, but only when the grid is above the colormesh).

re-open because the above is still not implemented to my knowledge.

plt.rcParams['axes.grid'] = False
this stop the warning

Hello. I think this might be affecting astropy too. Please see astropy/astropy#12753 . Thanks!

Don't warn on grid removal deprecation is grid is hidden

subfigure _axstack property

Navigation Menu

Repository navigation

[Bug]: Some styles trigger pcolormesh grid deprecation #21723

Bug summary

Code for reproduction

Actual outcome

Expected outcome

Additional information

Operating system

Matplotlib Version

Matplotlib Backend

Python version

Jupyter version

Installation

History

Problem here

Solution

Alternative

4 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[Bug]: Some styles trigger pcolormesh grid deprecation #21723

Description

Bug summary

Code for reproduction

Actual outcome

Expected outcome

Additional information

Operating system

Matplotlib Version

Matplotlib Backend

Python version

Jupyter version

Installation

Activity

QuLogic commented on Nov 23, 2021

prisae commented on Nov 23, 2021

jklymak commented on Nov 23, 2021

timhoffm commented on Nov 23, 2021

History

Problem here

Solution

Alternative

phollox-arch commented on Nov 24, 2021

jklymak commented on Nov 24, 2021

jklymak commented on Nov 24, 2021

prisae commented on Jan 4, 2022

jklymak commented on Jan 4, 2022

4 remaining items

tacaswell commented on Jan 13, 2022

jklymak commented on Jan 14, 2022

AzharUddinSheikh commented on Jan 19, 2022

pllim commented on Jan 20, 2022

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions