如何解决resample后Series使用apply()多聚合操作的报错？

问题背景 #

对DataFrame执行重采样后使用apply()方法时，因操作对象变为Series而非DataFrame，触发如下错误：

'Series' object has no attribute 'columns'

报错代码 #

df = pd.DataFrame(data)
x = df[df.columns[0]].fillna(0)
x_r = x.resample("s")
x = x_r.apply(['mean', np.max, np.min])

完整报错栈 #

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
/tmp/ipykernel_296/635968.py in <module>
      2 print(x_mm)
      3 
----> 4 x_mm.apply(['mean',np.max,np.min])

/local/opt/anaconda/anaconda3/envs/lib/python3.7/site-packages/pandas/core/resample.py in aggregate(self, func, *args, **kwargs)
    333     def aggregate(self, func, *args, **kwargs):
    334 
---> 335         result = ResamplerWindowApply(self, func, args=args, kwargs=kwargs).agg()
    336         if result is None:
    337             how = func

/local/opt/anaconda/anaconda3/envs/lib/python3.7/site-packages/pandas/core/apply.py in agg(self)
    162         elif is_list_like(arg):
    163             # we require a list, but not a 'str'
---> 164             return self.agg_list_like()
    165 
    166         if callable(arg):

/local/opt/anaconda/anaconda3/envs/lib/python3.7/site-packages/pandas/core/apply.py in agg_list_like(self)
    334         if selected_obj.ndim == 1:
    335             for a in arg:
---> 336                 colg = obj._gotitem(selected_obj.name, ndim=1, subset=selected_obj)
    337                 try:
    338                     new_res = colg.aggregate(a)

/local/opt/anaconda/anaconda3/envs/lib/python3.7/site-packages/pandas/core/resample.py in _gotitem(self, key, ndim, subset)
    390         # try the key selection
    391         try:
---> 392             return grouped[key]
    393         except KeyError:
    394             return grouped

/local/opt/anaconda/anaconda3/envs/lib/python3.7/site-packages/pandas/core/base.py in __getitem__(self, key)
    218 
    219         if isinstance(key, (list, tuple, ABCSeries, ABCIndex, np.ndarray)):
---> 220             if len(self.obj.columns.intersection(key)) != len(key):
    221                 bad_keys = list(set(key).difference(self.obj.columns))
    222                 raise KeyError(f"Columns not found: {str(bad_keys)[1:-1]}")

/local/opt/anaconda/anaconda3/envs/lib/python3.7/site-packages/pandas/core/generic.py in __getattr__(self, name)
   5485         ):
   5486             return self[name]
-> 5487         return object.__getattribute__(self, name)
   5488 
   5489     def __setattr__(self, name: str, value) -> None:

AttributeError: 'Series' object has no attribute 'columns'

原因分析 #

代码中df[df.columns[0]]取DataFrame单列时返回的是Series对象，旧版本pandas中，对Series的Resampler传入列表形式的聚合函数（如['mean', np.max, np.min]）时，内部逻辑会尝试访问columns属性，而Series没有该属性，因此报错。

解决方案 #

方案1：保留DataFrame结构 #

使用双层方括号选取单列，确保操作对象始终是DataFrame：

import pandas as pd
import numpy as np

df = pd.DataFrame(data)
# 双层方括号保留DataFrame结构，而非转为Series
x = df[[df.columns[0]]].fillna(0)
x_r = x.resample("s")
x = x_r.apply(['mean', np.max, np.min])

方案2：使用agg方法替代apply（推荐） #

对于聚合操作，agg()比apply()更适配列表形式的聚合函数，且对Series同样友好：

import pandas as pd
import numpy as np

df = pd.DataFrame(data)
x = df[df.columns[0]].fillna(0)
x_r = x.resample("s")
# 用agg直接传入聚合函数列表
x = x_r.agg(['mean', np.max, np.min])

方案3：自定义函数处理Series #

如果必须使用apply()，可以自定义函数返回包含多个统计量的Series：

import pandas as pd
import numpy as np

def calculate_stats(s):
    return pd.Series({
        'mean': s.mean(),
        'max': np.max(s),
        'min': np.min(s)
    })

df = pd.DataFrame(data)
x = df[df.columns[0]].fillna(0)
x_r = x.resample("s")
x = x_r.apply(calculate_stats)

内容的提问来源于stack exchange，提问作者Art