Change colorbar for contour to have the proper axes limits... #13506
Conversation
963bc27 to
9ad0d94
Compare
3bc20eb to
982616c
Compare
|
Thanks for doing this! |
|
@wsoofi. Thanks, It was your PR that inspired the need for it. It’s really not good to have the colorbar behave differently for different mappables but there were some subtleties that needed some work. |
| were made children of the normal axes tickers. | ||
|
|
||
| This version of Matplotlib extends that to mappables made by contours, and | ||
| allows the axes to run between the lesser the lowest boundary in the contour |
There was a problem hiding this comment.
| allows the axes to run between the lesser the lowest boundary in the contour | |
| allows the axes to run between the lowest boundary in the contour |
190e9cb to
4e0d077
Compare
|
@dstansby, thanks fixed. The except happens when the |
|
It doesn't work with contour, which means our test framework is missing something critical. With z = np.arange(24).reshape(4, 6)
cs = contourf(z, levels=[2, 4, 6, 10, 20])
plt.colorbar()
plt.savefig('contour_jk.png')yields the following incorrect colorbar: We also need to test the following: plt.clf(); cs = contourf(z, levels=[2, 4, 6, 10, 20])
plt.colorbar(spacing='proportional')
plt.savefig('contour_proportional.png')which yields this (correct) result with master: I suspect there are quite a few other cases that will need to be tested. |
|
Ok I guess the logic I had assumed proportional was true. Im not sure I get the other case It’s like we are treating the contours as categories, which seems strange to me. |
|
An example of where this is useful is in contouring bottom topography with contourf. One might want contour levels at 0, 50, 100, 250, 500, 1000, 2000, 5000. Yes, in a sense each interval is a category that gets a color. The interval boundaries need to be labeled, and each interval in the colorbar has to be large enough so that the labels don't collide. Hence the 'uniform' option, in which each interval gets the same space on the colorbar. |
|
Thanks @efiring, I think I fixed this. I'll add tests as well! |
6281da5 to
7d190d3
Compare
7d190d3 to
7b3ea96
Compare
d12c113 to
6d7c885
Compare
|
Test added. Not sure if there should be more... |
| @@ -578,13 +580,11 @@ def _reset_locator_formatter_scale(self): | |||
| """ | |||
| self.locator = None | |||
| self.formatter = None | |||
| if (isinstance(self.norm, colors.LogNorm) | |||
| and self._use_auto_colorbar_locator()): | |||
There was a problem hiding this comment.
I think we want the axes to be logarithmic in this case, but I agree its open to interpretation:
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.colors as mcolors
z = np.arange(24).reshape(4, 6) + 1
z = np.power(2, z)
fig, axs = plt.subplots(1, 2, constrained_layout=True)
ax = axs[0]
cs = ax.contourf(z, levels=np.power(10., [1, 2, 6.5, 10] ), norm=mcolors.LogNorm())
fig.colorbar(cs, ax=ax, spacing='proportional')
ax = axs[1]
cs = ax.contourf(z, levels=np.power(10, [1, 2, 6.5, 10]), norm=mcolors.LogNorm())
fig.colorbar(cs, ax=ax, spacing='uniform')
plt.show()With this change:
Without this change:
Before this PR:
| @@ -578,13 +580,11 @@ def _reset_locator_formatter_scale(self): | |||
| """ | |||
| self.locator = None | |||
| self.formatter = None | |||
| if (isinstance(self.norm, colors.LogNorm) | |||
| and self._use_auto_colorbar_locator()): | |||
|
ping @efiring, or anyone else interested in colorbars |
PR Summary
We've changed colorbars to mostly behave like other axes, so for instance,
print(cbar.ax.get_ylim())actually returns the axes limits of the colorbar.However, that did not work for colorbars made with
contour, which were still normalized with limits between 0 and 1. Now this axes are normalized between vmin and vmax for the colorbar (not necessarily the vmin and vmax of the colormap norm, a subtlety that confused me for a while).This closes #13494 because that example would now simply read:
cbar.ax.axhline(desired_cbar_yvalue, color='r', linewidth=5)rather than normalizing that between 0 and 1.
PR Checklist