axis limits come out reversed a lot

[beckermr]

Description

Right now in many cases, the plot limits come out reverse (e.g., the x-axis goes from 1 to -4 instead of -4 to 1).

Steps to reproduce

A "Minimal, Complete and Verifiable Example" will make it much easier for maintainers to help you.

import proplot as pplt
import numpy as np

rng = np.random.RandomState(seed=10)
x = rng.uniform(size=1000)*4 - 6
y = rng.uniform(size=1000)*2 + 6

fig, axs = pplt.subplots()
axs.hexbin(x, y)

Expected behavior: axis limits numerically ordered (should be -6 to -2)

Actual behavior: came out reversed (-2 to -6)

Equivalent steps in matplotlib

Please try to make sure this bug is related to a proplot-specific feature. If you're not sure, try to replicate it with the native matplotlib API. Matplotlib bugs belong on the matplotlib github page.

import matplotlib.pyplot as plt
import numpy as np

rng = np.random.RandomState(seed=10)
x = rng.uniform(size=1000)*4 - 6
y = rng.uniform(size=1000)*2 + 6

fig, axs = plt.subplots(nrows=1, ncols=1)
axs.hexbin(x, y)

Proplot version

Paste the results of import matplotlib; print(matplotlib.__version__); import proplot; print(proplot.version)here.

3.5.0
0.9.5

[lukelbd]

This is related to #300 (also documented on this page). Proplot automatically reverses the axis when it detect "dependent" variable coordinates and the axes were not previously fixed -- however this behavior should definitely be disabled for hexbin and hist2d, and it looks like this used to be disabled for hist and histh but the #191 fix is now obsolete. Sorry for the annoying behavior! I'll fix this later today.

Note also that the #300 fix make this feature more conservative; for example when passing stuff to ax.plot(x, y) an axis will only be reversed if 1) the x limits were not previously fixed with e.g. set_xlim and 2) the x coordinates are monotonically decreasing rather than simply having a negative initial step size.

[beckermr]

Thank you!

[lukelbd]

Fixed by 275bf30. Here's an example:

fig, axs = pplt.subplots(ncols=2, share=False)
x = np.sort(np.random.rand(100) * 100)
x = x[::-1]
axs[0].hist(x)
axs[1].hexbin(x, x)
print(pplt.internals.inputs._is_descending(x))

Result is True but axes are not reversed:

[beckermr]

Thank you! Looking forward to a release with the patch!

[esheldon]

I'm still seeing this for the quiver command, even in new releases, does that need to be addressed separately?