plt.pcolormesh shape mismatch when shading='gouraud'

[egranstedt]

Bug report

Bug summary

When running plt.pcolormesh(X, Y, C, shading='flat'), if C.shape == (nrows, ncols), then if X and Y are 2-D arrays, their shapes must be either (nrows, ncols) or (nrows+1, ncols+1). If shading='gouraud', however, the latter results in an error. The code snippet below uses 1-D arrays for X and Y, but the same error occurs when they are the equivalent 2-D arrays.

Code for reproduction

import numpy as np
import matplotlib.pyplot as plt
x = np.arange(-4, 5)
y = np.arange(-6, 7)
xe = np.linspace(-4.5, 4.5, 10)
ye = np.linspace(-6.5, 6.5, 14)
c = np.exp(-0.5*0.125*(x.reshape(1, -1)**2 + y.reshape(-1, 1)**2))
plt.pcolormesh(xe, ye, c, shading='flat')  # works
plt.pcolormesh(x, y, c, shading='gouraud')  # works
plt.pcolormesh(xe, ye, c, shading='gouraud')  # doesn't work

Actual outcome

Traceback (most recent call last):

  File "<ipython-input-67-154f761638cd>", line 1, in <module>
    plt.pcolormesh(xe, ye, c, shading='gouraud')

  File "/home/egranstedt/anaconda3/lib/python3.6/site-packages/matplotlib/pyplot.py", line 3245, in pcolormesh
    ret = ax.pcolormesh(*args, **kwargs)

  File "/home/egranstedt/anaconda3/lib/python3.6/site-packages/matplotlib/__init__.py", line 1892, in inner
    return func(ax, *args, **kwargs)

  File "/home/egranstedt/anaconda3/lib/python3.6/site-packages/matplotlib/axes/_axes.py", line 5569, in pcolormesh
    X, Y, C = self._pcolorargs('pcolormesh', *args, allmatch=allmatch)

  File "/home/egranstedt/anaconda3/lib/python3.6/site-packages/matplotlib/axes/_axes.py", line 5187, in _pcolorargs
    C.shape, Nx, Ny, funcname))

TypeError: Dimensions of C (13, 9) are incompatible with X (10) and/or Y (14); see help(pcolormesh)

Expected outcome

I expected pcolormesh would accept the same arrays for X and Y irrespective of the setting of theshading` keyword argument.

Matplotlib version

• The error occurs with Matplotlib version 2.0.0 in the python 3.6 interpreter on Linux (installed via anaconda), but also seems to exist in older versions of Matplotlib (e.g., 1.5.1)

[tacaswell]

Exact work:

• [ ] sort out if we can handle the N+1 case for gourand shading
• [ ] if so, fix the bug, if not document the limitation

[AtharvaKhare]

Dimensions of c are wrong. Try this:

import numpy as np
import matplotlib.pyplot as plt
xe = np.linspace(-4.5, 4.5, 10)
ye = np.linspace(-6.5, 6.5, 14)
c = np.exp(-0.5*0.125*(xe.reshape(1, -1)**2 + ye.reshape(-1, 1)**2))
plt.pcolormesh(xe, ye, c, shading='gouraud') 
plt.show()

[dopplershift]

The actual problem here is that if you don't use the Gouraud shading, pcolormesh is capable of automatically adjusting dimensions so that the original example works. The question is what it will take for Gouraud shading to work when given the slightly mis-shaped data.

[Jwink3101]

Just checking if there has been an on this? I followed pull request 9594 but it isn't clear it got done.

There are two distinct modes of operation here. When flat shading, pcolormesh is like a histogram where you define edges and then center fill color. When then gouraud, it is defined at the vertices. This should be consistent!

[QuLogic]

@jklymak since you just finished working on the 'flat' case, I wonder if you have any thoughts here (and on the linked PR)?

[jklymak]

I think interpolating the grid midpoints is "fine" to make gouraud work, maybe with a warning. The linked PR died because the author insisted on a kwarg controlling whether that would happen or not, which I didn't agree with.

[github-actions[bot]]

This issue has been marked "inactive" because it has been 365 days since the last comment. If this issue is still present in recent Matplotlib releases, or the feature request is still wanted, please leave a comment and this label will be removed. If there are no updates in another 30 days, this issue will be automatically closed, but you are free to re-open or create a new issue if needed. We value issue reports, and this procedure is meant to help us resurface and prioritize issues that have not been addressed yet, not make them disappear. Thanks for your help!

[jklymak]

This is still an issue.