迭代创建Seaborn子图后无法移除多余图例，需添加全局图例

问题背景 #

现有如下结构的数据集（实际数据量更大）：

Sample Name Treatment Replicate  Time Congener  Concentration (ng/L)
0     Blank-1     Blank         1     0   PbTx-1                     0
1      Ctrl-1      Ctrl         1     0   PbTx-1                  8800
2      Ctrl-2      Ctrl         2     0   PbTx-1                  7500
3      Ctrl-3      Ctrl         3     0   PbTx-1                  9100
4     10ppm-1     10ppm         1     0   PbTx-1                  9800

需求为：创建一组子图，每个子图对应一个时间周期，对比不同Treatment，展示各Congener，并在整个图的Treatment x轴下方添加一个含4列的全局图例。

当前已生成接近理想的图表，但所有子图的图例都可见；尝试移除子图图例时出现以下错误：

• 设置legend=False时报错：'Rectangle' object has no property 'legend'
• 调用ax.legend_.set_visible(False)或ax.get_legend().remove()时报错：'NoneType' object has no attribute 'set_visible'/'NoneType' object has no attribute 'remove'

原代码如下：

#Setting the variables outside of the actual arguments to improve readability
first_dimension = "Treatment"
second_dimension = "Concentration (ng/L)"
third_dimension = "Congener"

#Setting the y-axis limit for toxin
ylim_toxin = td['Concentration (ng/L)'].max() * 1.05
    
#This gets the total number of subplots
num_plots = len(treatments)
fig, axes = plt.subplots(ncols = num_plots, sharey=True, sharex=True, figsize = (5*num_plots,5))

#Removes the space between the plots, so they share an X and are lined up nicely by time the sample was taken
fig.subplots_adjust(hspace=0, wspace=0)

#Set overarching title based on previously input project name
fig.suptitle(f'{pdate}', fontweight='bold')
fig.supxlabel(f'Treatment')

    
#Create a bar plot of the data using seaborn, looping through the treatments list and each axis using count
count = 0
for ax in axes:
    axes[count].set_title(f"T = {treatments[count]}")
    treat = treatments[count]
    tdtreat = td.loc[td['Time'] == treat]
    sns.barplot(
                x = first_dimension , 
                y = second_dimension , 
                hue = third_dimension , 
                data = tdtreat ,
                capsize = 0.05 ,
                errwidth = '1' ,
                errcolor = '0' ,
                edgecolor="0" ,
                errorbar = 'sd',
                ax = ax)
    count = count + 1
    
ax.set(xlabel=None)
ax.set(ylabel=None)
    
#Setting a y-label with supylabel created a y label that was far to the left, simply labelling the first y-axis
axes[0].set_ylabel("Concentration (ng/L)", fontsize = 12)

错误原因 #

1. 调用ax.legend_.set_visible(False)时机错误：在绘制barplot之前执行该语句，此时轴对象还未生成图例，ax.legend_为None，因此报错。
2. legend=False参数问题：部分Seaborn版本中，barplot的legend参数存在兼容问题，被错误解析为绘图元素的属性而非绘图控制参数。

解决方法 #

步骤1：修正循环逻辑，移除所有子图图例 #

先绘制barplot，再移除当前子图的图例，确保ax.legend_已存在。

步骤2：创建全局图例 #

从任意一个子图中获取图例的handles和labels，通过fig.legend()创建全局图例，并设置4列布局及底部位置。

完整修正代码 #

#Setting the variables outside of the actual arguments to improve readability
first_dimension = "Treatment"
second_dimension = "Concentration (ng/L)"
third_dimension = "Congener"

#Setting the y-axis limit for toxin
ylim_toxin = td['Concentration (ng/L)'].max() * 1.05
    
#This gets the total number of subplots
num_plots = len(treatments)
# 增加图高度，预留图例空间
fig, axes = plt.subplots(ncols = num_plots, sharey=True, sharex=True, figsize = (5*num_plots,6))  

# 调整子图间距，底部预留空间放全局图例
fig.subplots_adjust(hspace=0, wspace=0, bottom=0.2)  

#Set overarching title based on previously input project name
fig.suptitle(f'{pdate}', fontweight='bold')
fig.supxlabel(f'Treatment')

# 存储图例的handles和labels
handles, labels = None, None

#Create a bar plot of the data using seaborn
for count, ax in enumerate(axes):
    ax.set_title(f"T = {treatments[count]}")
    treat = treatments[count]
    tdtreat = td.loc[td['Time'] == treat]
    
    # 绘制barplot
    sns.barplot(
                x = first_dimension , 
                y = second_dimension , 
                hue = third_dimension , 
                data = tdtreat ,
                capsize = 0.05 ,
                errwidth = 1 ,
                errcolor = '0' ,
                edgecolor="0" ,
                errorbar = 'sd',
                ax = ax)
    
    # 仅从第一个子图获取图例数据
    if count == 0:
        handles, labels = ax.get_legend_handles_labels()
    # 移除当前子图的图例
    ax.legend_.remove()

# 设置轴标签
ax.set(xlabel=None)
ax.set(ylabel=None)
axes[0].set_ylabel("Concentration (ng/L)", fontsize = 12)

# 添加全局图例：4列布局，底部中央位置
fig.legend(handles=handles, labels=labels, ncol=4, loc='lower center', bbox_to_anchor=(0.5, 0.02), fontsize=10)

plt.show()

关键说明 #

1. 使用enumerate简化循环计数，避免手动维护count变量。
2. 调整fig.subplots_adjust(bottom=0.2)预留底部空间，防止全局图例遮挡子图内容。
3. 仅从第一个子图获取图例的handles和labels，确保图例内容完整且统一。
4. 绘制每个子图后立即调用ax.legend_.remove()，此时ax.legend_已由barplot生成，不会出现NoneType错误。

内容的提问来源于stack exchange，提问作者RedTideSide