为何向subgridspec添加轴会干扰wspace参数，导致子图间距异常？

我创建了一个包含4个子图和2个覆盖轴的布局，用于为前两个与后两个子图设置共用x轴标签（代码如下）。尽管相关参数设置为相同值，但左侧两个子图的间距与右侧两个子图的间距仍存在细微差异，请问这是为什么？若不添加覆盖轴，间距则符合预期。

import matplotlib.pyplot as plt
from matplotlib.gridspec import GridSpec
# make the grid
fig = plt.figure(constrained_layout=True, figsize=(8,3))
grsp = GridSpec(1,4, figure=fig, wspace=0.6, width_ratios=[1,0.05,1,0.05])

grsp_one = grsp[0,0].subgridspec(1,2, wspace=0.01)
grsp_two = grsp[0,2].subgridspec(1,2, wspace=0.01)

axs_one = grsp_one.subplots(sharey='row')
axs_two = grsp_two.subplots(sharey='row')
# create the ovelapping ax on selection of axes
axmaster_1 = fig.add_subplot(grsp_one[0,0:2])
axmaster_2 = fig.add_subplot(grsp_two[0,0:2])

for axmaster in [axmaster_1, axmaster_2]:
    # hide the ticks of the master axis
    axmaster.xaxis.set_major_locator(matplotlib.ticker.NullLocator())
    axmaster.yaxis.set_major_locator(matplotlib.ticker.NullLocator())
    axmaster.spines[['bottom', 'right', 'top', 'left']].set_visible(False)
    axmaster.set_zorder(-10)


axmaster_1.set_xlabel('xlabel_1', labelpad=20)
axmaster_2.set_xlabel('xlabel_2', labelpad=20)

axmaster_1.set_ylabel('ylabel_1', labelpad=30)
axmaster_2.set_ylabel('ylabel_2', labelpad=30)

plt.show()

原因分析 #

问题出在constrained_layout=True的布局机制上：

• 你添加的覆盖轴（axmaster_1和axmaster_2）会被constrained_layout纳入整体布局计算，即使你隐藏了轴的线条和刻度，它们的x/y轴标签仍会占用空间。
• 左右两组覆盖轴的标签会触发constrained_layout对各自区域的空间分配进行微调，由于标签位置、padding等参数的细微影响，最终导致两组子图的内部间距出现差异。
• 当没有覆盖轴时，constrained_layout仅基于4个子图的参数计算，所以间距完全一致。

解决方案 #

方法1：让覆盖轴不参与布局计算 #

通过set_in_layout(False)将覆盖轴排除在constrained_layout的计算逻辑之外，既保留共用标签，又不干扰子图间距：

import matplotlib.pyplot as plt
from matplotlib.gridspec import GridSpec
import matplotlib.ticker

fig = plt.figure(constrained_layout=True, figsize=(8,3))
grsp = GridSpec(1,4, figure=fig, wspace=0.6, width_ratios=[1,0.05,1,0.05])

grsp_one = grsp[0,0].subgridspec(1,2, wspace=0.01)
grsp_two = grsp[0,2].subgridspec(1,2, wspace=0.01)

axs_one = grsp_one.subplots(sharey='row')
axs_two = grsp_two.subplots(sharey='row')

axmaster_1 = fig.add_subplot(grsp_one[0,0:2])
axmaster_2 = fig.add_subplot(grsp_two[0,0:2])

for axmaster in [axmaster_1, axmaster_2]:
    axmaster.xaxis.set_major_locator(matplotlib.ticker.NullLocator())
    axmaster.yaxis.set_major_locator(matplotlib.ticker.NullLocator())
    axmaster.spines[['bottom', 'right', 'top', 'left']].set_visible(False)
    axmaster.set_zorder(-10)
    axmaster.set_in_layout(False)  # 关键：排除该轴的布局计算

axmaster_1.set_xlabel('xlabel_1', labelpad=20)
axmaster_2.set_xlabel('xlabel_2', labelpad=20)

axmaster_1.set_ylabel('ylabel_1', labelpad=30)
axmaster_2.set_ylabel('ylabel_2', labelpad=30)

plt.show()

方法2：替换constrained_layout为tight_layout #

关闭动态布局，改用手动调整的tight_layout，避免覆盖轴的干扰：

import matplotlib.pyplot as plt
from matplotlib.gridspec import GridSpec
import matplotlib.ticker

fig = plt.figure(figsize=(8,3))  # 移除constrained_layout=True
grsp = GridSpec(1,4, figure=fig, wspace=0.6, width_ratios=[1,0.05,1,0.05])

grsp_one = grsp[0,0].subgridspec(1,2, wspace=0.01)
grsp_two = grsp[0,2].subgridspec(1,2, wspace=0.01)

axs_one = grsp_one.subplots(sharey='row')
axs_two = grsp_two.subplots(sharey='row')

axmaster_1 = fig.add_subplot(grsp_one[0,0:2])
axmaster_2 = fig.add_subplot(grsp_two[0,0:2])

for axmaster in [axmaster_1, axmaster_2]:
    axmaster.xaxis.set_major_locator(matplotlib.ticker.NullLocator())
    axmaster.yaxis.set_major_locator(matplotlib.ticker.NullLocator())
    axmaster.spines[['bottom', 'right', 'top', 'left']].set_visible(False)
    axmaster.set_zorder(-10)

axmaster_1.set_xlabel('xlabel_1', labelpad=20)
axmaster_2.set_xlabel('xlabel_2', labelpad=20)

axmaster_1.set_ylabel('ylabel_1', labelpad=30)
axmaster_2.set_ylabel('ylabel_2', labelpad=30)

plt.tight_layout()  # 用tight_layout手动调整布局
plt.show()

方法3：直接使用全局文本标签（推荐） #

如果只是需要共用标签，完全不需要覆盖轴，直接用fig.text()在指定位置添加标签更简洁：

import matplotlib.pyplot as plt
from matplotlib.gridspec import GridSpec

fig = plt.figure(constrained_layout=True, figsize=(8,3))
grsp = GridSpec(1,2, figure=fig, wspace=0.6)  # 简化GridSpec结构

grsp_one = grsp[0,0].subgridspec(1,2, wspace=0.01)
grsp_two = grsp[0,1].subgridspec(1,2, wspace=0.01)

axs_one = grsp_one.subplots(sharey='row')
axs_two = grsp_two.subplots(sharey='row')

# 手动添加共用标签，通过坐标调整位置
fig.text(0.22, -0.1, 'xlabel_1', ha='center', fontsize=12)
fig.text(0.78, -0.1, 'xlabel_2', ha='center', fontsize=12)
fig.text(-0.05, 0.5, 'ylabel_1', va='center', rotation='vertical', fontsize=12)
fig.text(0.45, 0.5, 'ylabel_2', va='center', rotation='vertical', fontsize=12)

plt.show()

内容的提问来源于stack exchange，提问作者Stef M