Matplotlib水平标签文本换行失效：如何通过内置功能设置文本框最大宽度

Great question! You're spot-on about the bounding box calculation difference being the root cause—Matplotlib's default wrap=True only kicks in for vertically rotated labels because it doesn't enforce width constraints for horizontal ones out of the box.

Luckily, you don't need external libraries like textwrap or workarounds like rotating labels. Matplotlib has built-in methods to manually set a maximum width for your horizontal tick labels and enable wrapping. Here's how to modify your example to make it work:

import matplotlib.pyplot as plt
x = "Very long label"
y = "Even much longer label"
z = "Now this is what I call a really long label"
l = [x, y, z]
d = [3, 1, 4]
fig, (axA, axB) = plt.subplots(1, 2, figsize=(10, 5))

# Case A: Vertical labels (default wrap works)
axA.bar(l, d)
font_dictA={"va": "center", "ha": "right", "rotation": 90, "wrap": True}
axA.set_xticklabels(l, fontdict= font_dictA)
axA.set_title("Case A: vertical labels")

# Case B: Horizontal labels with manual wrap width
axB.bar(l, d)
font_dictB={"va": "top", "ha": "center", "rotation": 0}
axB.set_xticklabels(l, fontdict= font_dictB)
axB.set_title("Case B: horizontal labels with wrap")

# Key step: Enable wrap and set max width for each tick label
for label in axB.get_xticklabels():
    label.set_wrap(True)
    # Set maximum width to 80 points (adjust this value as needed)
    label.set_wrap_width(80)

plt.subplots_adjust(bottom=0.2)
plt.show()

How this works: #

• Instead of setting wrap=True in the font dictionary (which does nothing for horizontal labels), we iterate over each tick label text object directly.
• The set_wrap(True) method explicitly enables text wrapping for the label.
• set_wrap_width() defines the maximum width (in points, where 1 point = 1/72 inch) before the text wraps to a new line. Tweak this number to control how early or late the text breaks—smaller values mean more wrapping, larger values mean longer lines.

For dynamic width adjustment: #

If you want the label width to automatically adapt to your axis size instead of using a fixed value, calculate the axis width in points and split it evenly among your labels:

# Calculate dynamic wrap width based on axis size
ax_bbox = axB.get_window_extent().transformed(fig.dpi_scale_trans.inverted())
ax_width_points = ax_bbox.width * 72  # Convert axis width to points
label_width = ax_width_points / len(l)  # Divide width by number of labels

for label in axB.get_xticklabels():
    label.set_wrap(True)
    label.set_wrap_width(label_width)

This ensures each label gets an equal share of the axis width, preventing overlap and making the layout responsive to figure resizing.

All of this uses only Matplotlib's built-in functionality, no external tools required—exactly what you asked for!

内容的提问来源于stack exchange，提问作者Mr. T