求助：解决Cannot describe a DataFrame without columns错误

问题背景 #

两天前完成某数据集的简短数据分析，今日基于该工作开展新项目时复制部分代码，新项目运行正常，但旧项目出现ValueError: Cannot describe a DataFrame without columns错误。此前通过新建Notebook临时解决，现需无需重复新建Notebook的稳定方案。

工作环境：Jupyter Notebook、Python 3.6（虚拟环境）、Linux 22.04

相关代码及报错信息 #

categorical_features = dtype[dtype == 'object'].index

readable_df[numerical_features].describe()
# Split features into categorical and numerical, print numerical
dtype = readable_df.dtypes
numerical_features = dtype[dtype == 'int64'].index
categorical_features = dtype[dtype == 'object'].index

readable_df[numerical_features].describe()

报错堆栈：

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Input In [7], in <cell line: 6>()
      3 numerical_features = dtype[dtype == 'int64'].index
      4 categorical_features = dtype[dtype == 'object'].index
----> 6 readable_df[numerical_features].describe()

File ~/python-env/python-env/lib/python3.9/site-packages/pandas/core/generic.py:10227, in NDFrame.describe(self, percentiles, include, exclude, datetime_is_numeric)
   9978 @final
   9979 def describe(
   9980     self: NDFrameT,
   (...)
   9984     datetime_is_numeric=False,
   9985 ) -> NDFrameT:
   9986     """
   9987     Generate descriptive statistics.
   9988 
   (...)
  10225     max            NaN      3.0
  10226     """
> 10227     return describe_ndframe(
  10228         obj=self,
  10229         include=include,
  10230         exclude=exclude,
  10231         datetime_is_numeric=datetime_is_numeric,
  10232         percentiles=percentiles,
  10233     )

File ~/python-env/python-env/lib/python3.9/site-packages/pandas/core/describe.py:87, in describe_ndframe(obj, include, exclude, datetime_is_numeric, percentiles)
     82     describer = SeriesDescriber(
     83         obj=cast("Series", obj),
     84         datetime_is_numeric=datetime_is_numeric,
     85     )
     86 else:
--> 87     describer = DataFrameDescriber(
     88         obj=cast("DataFrame", obj),
     89         include=include,
     90         exclude=exclude,
     91         datetime_is_numeric=datetime_is_numeric,
     92     )
     94 result = describer.describe(percentiles=percentiles)
     95 return cast(NDFrameT, result)

File ~/python-env/python-env/lib/python3.9/site-packages/pandas/core/describe.py:164, in DataFrameDescriber.__init__(self, obj, include, exclude, datetime_is_numeric)
    161 self.exclude = exclude
    163 if obj.ndim == 2 and obj.columns.size == 0:
--> 164     raise ValueError("Cannot describe a DataFrame without columns")
    166 super().__init__(obj, datetime_is_numeric=datetime_is_numeric)

ValueError: Cannot describe a DataFrame without columns

报错原因 #

报错核心是readable_df[numerical_features]返回了无列名的空DataFrame，常见诱因：

1. Jupyter内核状态污染：旧Notebook的内核中，readable_df或numerical_features被后续单元格代码意外覆盖，或之前运行的单元格导致变量状态异常
2. 数据加载异常：旧Notebook中readable_df的加载路径（如相对路径）失效，或文件本身被修改，导致加载后的DataFrame无int64类型列，进而numerical_features为空数组
3. 内核环境不匹配：旧Notebook误使用了其他虚拟环境的内核，不同pandas版本对空索引的处理逻辑存在差异

稳定解决方案 #

1. 重置内核并重新运行全量代码 #

直接清空当前内核的所有变量状态，从头执行所有代码：

• 在Jupyter界面点击菜单栏 Kernel -> Restart & Run All

2. 增加变量校验逻辑 #

在调用describe()前添加校验，避免空索引导致的报错：

# Split features into categorical and numerical, print numerical
dtype = readable_df.dtypes
numerical_features = dtype[dtype == 'int64'].index
categorical_features = dtype[dtype == 'object'].index

# 校验数值特征是否存在
if len(numerical_features) == 0:
    print("警告：未找到int64类型的数值特征")
    # 可选：尝试匹配float64类型特征
    numerical_features = dtype[dtype == 'float64'].index
    if len(numerical_features) == 0:
        print("无可用数值特征，跳过describe操作")
    else:
        print(readable_df[numerical_features].describe())
else:
    print(readable_df[numerical_features].describe())

3. 确认Notebook使用正确的虚拟环境内核 #

• 检查界面右上角显示的内核名称，确认是你创建的Python3.6虚拟环境
• 若内核不对，点击 Kernel -> Change Kernel 选择对应环境
• 若虚拟环境未添加到Jupyter内核，激活虚拟环境后执行：

pip install ipykernel
python -m ipykernel install --user --name=python3.6-env --display-name="Python 3.6 (虚拟环境)"

4. 固化数据加载路径 #

将数据加载的相对路径改为绝对路径，避免路径失效问题：

import pandas as pd
# 替换为你的数据集绝对路径
readable_df = pd.read_csv("/home/your_username/your_dataset_path/data.csv")

内容的提问来源于stack exchange，提问作者Solaris