如何从ipywidgets输出中获取Pandas DataFrame？

嘿Tom，我懂你现在的困扰——你能在Jupyter的Output组件里看到过滤后的表格，但就是没办法把这些数据提取出来当成普通的Pandas DataFrame来用，对吧？其实问题很简单，咱们只要把过滤后的结果存到一个外部能访问的地方就行，给你两种靠谱的解决方案：

方案一：用全局变量快速实现（适合简单场景） #

原代码里的过滤结果是common_filtering函数里的局部变量，函数执行完就会被销毁，所以外部访问不到。咱们只需要在函数外定义一个全局变量，每次过滤后把结果赋值给它就行：

import pandas as pd
import numpy as np
import ipywidgets as widgets
from ipywidgets import Layout, AppLayout
from IPython.display import display
import functools

data = {'year': ['2000', '2000','2000','2000','2001','2001','2001','2001', '2002', '2002', '2002', '2002', '2003','2003','2003','2003','2004', '2004','2004','2004', '2005', '2005', '2005', '2005', '2006', '2006', '2006', '2006', '2006', '2007', '2007', '2007', '2007', '2008', '2008', '2008', '2008', '2009', '2009', '2009', '2009'], 'purpose':['Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday' ], 'market':['Belgium', 'Luxembourg', 'France', 'Spain', 'Norway', 'Sweden', 'Germany', 'Austria', 'Denmark', 'Portugal', 'Greece', 'Croatia', 'Belgium', 'Luxembourg', 'France', 'Spain', 'Norway', 'Sweden', 'Germany', 'Austria', 'Denmark', 'Portugal', 'Greece', 'Croatia', 'Belgium', 'Luxembourg', 'France', 'Spain', 'Norway', 'Sweden', 'Germany', 'Austria', 'Denmark', 'Portugal', 'Greece', 'Croatia', 'Belgium', 'Luxembourg', 'France', 'Spain', 'Norway' ]}
df_london = pd.DataFrame (data, columns = ['year','purpose', 'market'])

# Get our unique values
ALL = 'ALL'
def unique_sorted_values_plus_ALL(array):
    unique = array.unique().tolist()
    unique.sort()
    unique.insert(0, ALL)
    return unique

output = widgets.Output()
# 定义全局变量存储过滤后的结果
filtered_df = df_london.copy()

# Dropdown listbox
dropdown_year = widgets.Dropdown(description='Year', options = unique_sorted_values_plus_ALL(df_london.year))

# Function to filter our dropdown listboxe
def common_filtering(year):
    global filtered_df  # 声明要使用全局变量
    df = df_london.copy()
    filters = []
    # Evaluate our dropdown listbox and return booleans for our selections
    if year is not ALL:
        filters.append(df['year'] == year)
    output.clear_output()
    with output:
        if filters:
            df_filter = functools.reduce(lambda x,y: x&y, filters)
            filtered_df = df.loc[df_filter]  # 将过滤结果赋值给全局变量
            display(filtered_df)
        else:
            filtered_df = df  # 未过滤时返回原数据
            display(df)

def dropdown_year_eventhandler(change):
    common_filtering(change.new)

dropdown_year.observe(dropdown_year_eventhandler, names='value')
ui = widgets.HBox([dropdown_year])
display(ui, output)

现在你只要在Jupyter单元格里输入filtered_df，就能直接拿到最新的过滤结果，还能对它做任何Pandas操作，比如filtered_df.groupby('purpose').size()。

方案二：用类封装（适合复杂多过滤器场景） #

如果之后你要加更多过滤器（比如purpose、market的下拉框），用全局变量会显得混乱。这时候用类来封装状态会更优雅：

import pandas as pd
import numpy as np
import ipywidgets as widgets
from IPython.display import display
import functools

data = {'year': ['2000', '2000','2000','2000','2001','2001','2001','2001', '2002', '2002', '2002', '2002', '2003','2003','2003','2003','2004', '2004','2004','2004', '2005', '2005', '2005', '2005', '2006', '2006', '2006', '2006', '2006', '2007', '2007', '2007', '2007', '2008', '2008', '2008', '2008', '2009', '2009', '2009', '2009'], 'purpose':['Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday', 'Business', 'VFR', 'Study', 'Holiday' ], 'market':['Belgium', 'Luxembourg', 'France', 'Spain', 'Norway', 'Sweden', 'Germany', 'Austria', 'Denmark', 'Portugal', 'Greece', 'Croatia', 'Belgium', 'Luxembourg', 'France', 'Spain', 'Norway', 'Sweden', 'Germany', 'Austria', 'Denmark', 'Portugal', 'Greece', 'Croatia', 'Belgium', 'Luxembourg', 'France', 'Spain', 'Norway', 'Sweden', 'Germany', 'Austria', 'Denmark', 'Portugal', 'Greece', 'Croatia', 'Belgium', 'Luxembourg', 'France', 'Spain', 'Norway' ]}
df_london = pd.DataFrame (data, columns = ['year','purpose', 'market'])

ALL = 'ALL'
def unique_sorted_values_plus_ALL(array):
    unique = array.unique().tolist()
    unique.sort()
    unique.insert(0, ALL)
    return unique

class DataFilter:
    def __init__(self, df):
        self.raw_df = df
        self.filtered_df = df.copy()
        
        # 创建控件
        self.dropdown_year = widgets.Dropdown(description='Year', options=unique_sorted_values_plus_ALL(df.year))
        self.output = widgets.Output()
        
        # 绑定事件
        self.dropdown_year.observe(self._on_year_change, names='value')
        
        # 初始化显示
        self._common_filtering(self.dropdown_year.value)
    
    def _common_filtering(self, year):
        df = self.raw_df.copy()
        filters = []
        if year is not ALL:
            filters.append(df['year'] == year)
        
        self.output.clear_output()
        with self.output:
            if filters:
                df_filter = functools.reduce(lambda x,y: x&y, filters)
                self.filtered_df = df.loc[df_filter]
                display(self.filtered_df)
            else:
                self.filtered_df = df
                display(df)
    
    def _on_year_change(self, change):
        self._common_filtering(change.new)
    
    def show_ui(self):
        ui = widgets.HBox([self.dropdown_year])
        display(ui, self.output)

# 创建过滤器实例
filter_app = DataFilter(df_london)
filter_app.show_ui()

这个方案里，过滤后的结果存在filter_app.filtered_df里，你随时可以调用它。后续要加新的过滤器，只要在类里新增控件和过滤逻辑就行，代码结构会非常清晰。

内容的提问来源于stack exchange，提问作者twelsh37