如何用Matplotlib的PdfPages实现每页4个图表的PDF输出？

已有可生成目标图表的Seaborn代码，使用matplotlib.backends.backend_pdf.PdfPages导出PDF时，希望每页展示4个图表。尝试用grouper函数分组处理后，所有图表重复出现在每一页，求正确实现方法或代码问题分析。

原有可正常生成所有图表的代码：

import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
from matplotlib.backends.backend_pdf import PdfPages
plt.style.use('ggplot')
%matplotlib inline

data = dict({'Variable_Grouping':['Type_A', 'Type_A', 'Type_A', 'Type_C', 'Type_C', 'Type_C', 'Type_C', 'Type_D', 'Type_D', 'Type_E', 'Type_E', 'Type_E', 'Type_H', 'Type_H'], 'Variable':['a1', 'a2', 'a3', 'c1', 'c2', 'c3', 'c4', 'd1', 'd2', 'e1', 'e2', 'e3', 'h1', 'h2'], 'Count':[5, 3, 8, 4, 3, 9, 5, 3, 8, 5, 3, 8, 5, 3],'Percent':[0.0625, 0.125, 0.4375, 0.0, 0.125, 0.5, 0.02, 0.125, 0.03, 0.0625, 0.05, 0.44, 0.07, 0.023]})
to_plot = pd.DataFrame(data)

g = sns.FacetGrid(to_plot, col='Variable_Grouping', col_wrap = 2, sharex=False, sharey = False, height = 5, aspect = 1, margin_titles=True)
g=g.map(plt.bar, "Variable","Count").add_legend()
for ax, (_, subdata) in zip(g.axes, to_plot.groupby('Variable_Grouping')):
    ax2=ax.twinx()
    subdata.plot(x='Variable',y='Percent', ax = ax2, legend=True, color='g', label = 'Percent')
    ax2.set_ylabel('Percent')
    ax2.grid(False)
for ax in g.axes.flatten():
    ax.tick_params(labelbottom=True, labelrotation = 90)
g.fig.suptitle('Analysis', fontsize=16, fontweight = 'demibold', y = 1.02)
g.fig.subplots_adjust(hspace=0.3, wspace=0.7, right = 0.9)
plt.show();

尝试分组导出的问题代码：

def grouper(iterable, n, fillvalue=None):
    from itertools import zip_longest
    args = [iter(iterable)] * n
    return zip_longest(*args, fillvalue=fillvalue)

if len(to_plot['Variable_Grouping'].unique()) < 4:
    N_plots_per_page =len(to_plot['Variable_Grouping'].unique())
elif len(to_plot['Variable_Grouping'].unique()) >= 4:
    N_plots_per_page = 4

with PdfPages('Analysis.pdf') as pdf:
    for cols in grouper(to_plot['Variable_Grouping'].unique(), N_plots_per_page):
        g = sns.FacetGrid(to_plot, col='Variable_Grouping', col_wrap = 2, sharex=False, sharey = False, height = 5, aspect = 1, margin_titles=True)
        g=g.map(plt.bar, "Variable","Count").add_legend()
        for ax, (_, subdata) in zip(g.axes, to_plot.groupby('Variable_Grouping')):
            ax2=ax.twinx()
            subdata.plot(x='Variable',y='Percent', ax = ax2, legend=True, color='g', label = 'Percent')
            ax2.set_ylabel('Percent')
            ax2.grid(False)
        for ax in g.axes.flatten():
            ax.tick_params(labelbottom=True, labelrotation = 90)
        g.fig.suptitle('Analysis', fontsize=16, fontweight = 'demibold', y = 1.02)
        g.fig.subplots_adjust(hspace=0.3, wspace=0.7, right = 0.9)
        pdf.savefig(bbox_inches = 'tight')
        plt.show()
        plt.close();

核心问题是分组循环完全没用到grouper生成的cols变量，每次创建FacetGrid都传入了完整的to_plot数据，导致每页都绘制所有Variable_Grouping的图表，自然出现重复。

另外还有两个小问题：

• grouper生成的cols可能包含fillvalue（比如最后一页不足4个分组时），需要过滤掉空值
• 循环中遍历to_plot.groupby('Variable_Grouping')时，没有和当前页的分组对应，会导致双轴数据匹配错误

import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
from matplotlib.backends.backend_pdf import PdfPages
from itertools import zip_longest
plt.style.use('ggplot')

data = dict({'Variable_Grouping':['Type_A', 'Type_A', 'Type_A', 'Type_C', 'Type_C', 'Type_C', 'Type_C', 'Type_D', 'Type_D', 'Type_E', 'Type_E', 'Type_E', 'Type_H', 'Type_H'], 'Variable':['a1', 'a2', 'a3', 'c1', 'c2', 'c3', 'c4', 'd1', 'd2', 'e1', 'e2', 'e3', 'h1', 'h2'], 'Count':[5, 3, 8, 4, 3, 9, 5, 3, 8, 5, 3, 8, 5, 3],'Percent':[0.0625, 0.125, 0.4375, 0.0, 0.125, 0.5, 0.02, 0.125, 0.03, 0.0625, 0.05, 0.44, 0.07, 0.023]})
to_plot = pd.DataFrame(data)

def grouper(iterable, n, fillvalue=None):
    args = [iter(iterable)] * n
    return zip_longest(*args, fillvalue=fillvalue)

unique_groups = to_plot['Variable_Grouping'].unique()
N_plots_per_page = min(4, len(unique_groups))

with PdfPages('Analysis.pdf') as pdf:
    for group_batch in grouper(unique_groups, N_plots_per_page):
        # 过滤掉grouper填充的空值
        current_groups = [g for g in group_batch if g is not None]
        # 筛选当前页要绘制的数据
        page_data = to_plot[to_plot['Variable_Grouping'].isin(current_groups)]
        
        # 创建仅包含当前分组的FacetGrid
        g = sns.FacetGrid(page_data, col='Variable_Grouping', col_wrap=2, 
                          sharex=False, sharey=False, height=5, aspect=1, margin_titles=True)
        g = g.map(plt.bar, "Variable", "Count").add_legend()
        
        # 遍历当前页的分组和对应的轴，匹配双轴数据
        grouped_data = page_data.groupby('Variable_Grouping')
        for ax, (group_name, subdata) in zip(g.axes, grouped_data):
            ax2 = ax.twinx()
            subdata.plot(x='Variable', y='Percent', ax=ax2, legend=True, color='g', label='Percent')
            ax2.set_ylabel('Percent')
            ax2.grid(False)
        
        # 调整轴标签
        for ax in g.axes.flatten():
            ax.tick_params(labelbottom=True, labelrotation=90)
        
        # 调整图表布局和标题
        g.fig.suptitle('Analysis', fontsize=16, fontweight='demibold', y=1.02)
        g.fig.subplots_adjust(hspace=0.3, wspace=0.7, right=0.9)
        
        pdf.savefig(bbox_inches='tight')
        plt.close()

• 每次循环筛选当前页对应的current_groups，生成仅包含这些分组的page_data
• 创建FacetGrid时传入page_data，而非完整数据集
• 遍历page_data.groupby('Variable_Grouping')确保双轴数据和当前页的图表一一对应
• 过滤grouper生成的空值，避免无效图表
• 简化N_plots_per_page的判断逻辑，用min(4, len(unique_groups))更简洁

内容的提问来源于stack exchange，提问作者Sid