如何在Matplotlib 3D绘图中移动坐标轴脊柱（Spines）？

Great question! The confusion here stems from a critical difference between 2D and 3D axes in Matplotlib: the spines property on an Axes3D object doesn't control the 3D coordinate axes (x/y/z) you see—it's a holdover from the 2D Axes class, referring to the 2D border of the plotting area. That's why modifying spines has no effect on your 3D axes, and why keys like "left"/"right" don't map to 3D axis directions.

To work with the actual x/y/z axes in a 3D plot, you need to interact with the Axis3D objects accessible via ax.w_xaxis, ax.w_yaxis, and ax.w_zaxis. These handle the 3D axis lines, ticks, and labels. Below are two practical solutions to achieve your goal:

Option 1: Manually Draw Custom Axes (Full Control) #

If you want precise control over where axes intersect (like at the origin (0,0,0)), the most reliable approach is to hide the default 3D axes and draw custom axis lines. This lets you place them exactly where you want:

import matplotlib.pyplot as plt
import numpy as np

fig = plt.figure()
ax = fig.add_subplot(111, projection="3d", aspect="equal")

# Plot your data
r = 1
theta = np.linspace(0, 2*np.pi, 100)
x, y = r*np.cos(theta), r*np.sin(theta)
z = np.zeros(x.shape)
ax.plot(x, y, z)

# Define consistent axis limits
axis_lim = [-r, r]
ax.set_xlim(axis_lim)
ax.set_ylim(axis_lim)
ax.set_zlim(axis_lim)

# Hide default axes and panes
ax.set_axis_off()

# Draw custom axes through the origin
# X-axis: from (-r, 0, 0) to (r, 0, 0)
ax.plot(axis_lim, [0, 0], [0, 0], color="black", linewidth=2)
# Y-axis: from (0, -r, 0) to (0, r, 0)
ax.plot([0, 0], axis_lim, [0, 0], color="black", linewidth=2)
# Z-axis: from (0, 0, -r) to (0, 0, r)
ax.plot([0, 0], [0, 0], axis_lim, color="black", linewidth=2)

# Optional: Add custom ticks and labels
for tick_val in np.linspace(-r, r, 5):
    # X-axis ticks
    ax.plot([tick_val, tick_val], [0, 0.05*r], [0, 0], color="black")
    ax.text(tick_val, 0.05*r, 0, f"{tick_val:.1f}", fontsize=8)
    # Y-axis ticks
    ax.plot([0, 0.05*r], [tick_val, tick_val], [0, 0], color="black")
    ax.text(0.05*r, tick_val, 0, f"{tick_val:.1f}", fontsize=8)

plt.show()

Option 2: Modify Existing 3D Axes (Partial Control) #

If you want to keep the default ticks and labels but reposition axis lines to pass through the origin, you can modify the line property of each Axis3D object. You'll also need to adjust tick positions to align with the new axes:

import matplotlib.pyplot as plt
import numpy as np

fig = plt.figure()
for i in range(2):
    ax = plt.subplot(1,2,i+1, projection="3d", aspect="equal")
    
    # Plot your data
    r = 1
    theta = np.linspace(0, 2*np.pi, 100)
    x, y = r*np.cos(theta), r*np.sin(theta)
    z = np.zeros(x.shape)
    ax.plot(x, y, z)
    
    # Get current axis limits
    xlim = ax.get_xlim()
    ylim = ax.get_ylim()
    zlim = ax.get_zlim()
    
    # Reposition X-axis to lie along y=0, z=0
    ax.w_xaxis.line.set_data([xlim[0], xlim[1]], [0, 0])
    ax.w_xaxis.line.set_3d_properties([0, 0])
    
    # Reposition Y-axis to lie along x=0, z=0
    ax.w_yaxis.line.set_data([0, 0], [ylim[0], ylim[1]])
    ax.w_yaxis.line.set_3d_properties([0, 0])
    
    # Reposition Z-axis to lie along x=0, y=0
    ax.w_zaxis.line.set_data([0, 0], [0, 0])
    ax.w_zaxis.line.set_3d_properties([zlim[0], zlim[1]])
    
    # Adjust tick lines to align with new axis positions
    # X-axis ticks: set y and z to 0
    for tick in ax.w_xaxis.get_ticklines():
        x_tick = tick.get_data()[0]
        tick.set_data(x_tick, np.zeros_like(x_tick))
        tick.set_3d_properties(np.zeros_like(x_tick))
    
    # Y-axis ticks: set x and z to 0
    for tick in ax.w_yaxis.get_ticklines():
        y_tick = tick.get_data()[1]
        tick.set_data(np.zeros_like(y_tick), y_tick)
        tick.set_3d_properties(np.zeros_like(y_tick))
    
    # Z-axis ticks: set x and y to 0
    for tick in ax.w_zaxis.get_ticklines():
        z_tick = tick.get_3d_properties()
        tick.set_data(np.zeros_like(z_tick), np.zeros_like(z_tick))
        tick.set_3d_properties(z_tick)

plt.show()

Why Your Original Code Failed #

To clarify: The ax.spines dictionary in Axes3D refers to the 2D border of the entire plotting area (like the left/right edges of the figure canvas), not the 3D coordinate axes. Modifying these spines only affects that outer 2D frame, which is usually hidden in 3D plots anyway—hence why you saw no change.

内容的提问来源于stack exchange，提问作者binnev