Bug summary

Whenever creating a 3D plot with vertical_axis = 'y' option the set_box_aspect() method scales the axes incorrectly. The 3-tuple is expected to behave as x:y:z, instead it behaves like z:x:y

Code for reproduction

"""
Created on Mon Oct 31 18:30:31 2022

@author: Michel Gordillo
"""

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

fig = plt.figure(num=4, clear=True)
ax = fig.add_subplot(1, 1, 1, projection='3d')
ax.view_init(vertical_axis='y')

(theta, phi) = np.meshgrid(np.linspace(0, 2 * np.pi, 41),
                           np.linspace(0, 2 * np.pi, 41))

x = (3 + np.cos(phi)) * np.cos(theta)
z = (3 + np.cos(phi)) * np.sin(theta)
y = np.sin(phi)

k = 4

xlim = [-k, k]  #2k
ylim = [-k/2, k/2] #k
zlim = [-k, k] #2k

box_aspect = (4,2,4) #(x,y,z)

def fun(t, f): return (np.cos(f + 2 * t) + 1) / 2


dplot = ax.plot_surface(x, y, z, facecolors=cm.jet(fun(theta, phi)))
ax.set(xlabel='x', 
       ylabel='y', 
       zlabel='z',
       xlim = xlim,
       ylim = ylim,
       zlim = zlim,
       # xticks = [-4, -2, 2, 4],
       # yticks = [-4, -2, 2, 4],
       # zticks = [-1, 0, 1],
       title='Donut!')

ax.set_box_aspect(aspect = box_aspect)

fig.tight_layout()

Actual outcome

The box_aspect tuple:
box_aspect = (4,2,4) #(x,y,z)

Actually the axis scaled as
box_aspect = (4,2,4) -> #(z,x,y)
Like if shifting y to the right.

Expected outcome

To produce this result i just applied the same logic in reverse:
box_aspect = (4,4,2) #(z,x,y) <- Expecting the outcome to be applied like this

And it actually does.

Additional information

This only happens as explained in the title when both vertical = y option is enabled and we try to manually scale_box_aspect(). The problem is probably from how was vertical axis selection actually implemented.

From the documentation:

set_box_aspect(aspect, *, zoom=1)
[...]
aspect3-tuple of floats or None
Changes the physical dimensions of the Axes3D, such that the ratio of the axis lengths in display units is x:y:z.

This behaviour is not respected when vertical_axis = 'y' in enabled. The behaviour appears to be: z:x:y instead.

To actually fix this I just created a tuple with the new definition in mind, this could just be patched by shifting the tuple to make the vertical axis be the last element of the tuple

(x,y,z) -> input tuple

vertical axis == 'z' -> (x,y,z) no problem
vertical axis == 'y' -> (z,x,y) rotate right the tuple
vertical axis == 'x' -> (y,z,x) rotate right twice the tuple

Operating system

Windows

Matplotlib Version

'3.6.1'

Matplotlib Backend

Qt5Agg

Python version

Python 3.8.5

Jupyter version

Spyder 4.1.5

Installation

conda