导入Pandas触发AttributeError：pyarrow模块无__version__属性

问题背景 #

在Jupyter Notebook和VS Code的Anaconda环境中，导入numpy正常，但导入pandas时触发上述报错。已多次重装numpy和pandas，问题仍反复出现。

复现代码 #

import numpy as np
import pandas as pd

报错信息 #

AttributeError                            Traceback (most recent call last)
c:\Users\apodo\Downloads\Untitled.py in line 2
      1 import numpy as np
----> 2 import pandas as pd

File c:\Users\apodo\anaconda3\Lib\site-packages\pandas\__init__.py:36
     32 del _hard_dependencies, _dependency, _missing_dependencies
     34 try:
     35     # numpy compat
---> 36     from pandas.compat import (
     37         is_numpy_dev as _is_numpy_dev,  # pyright: ignore[reportUnusedImport] # noqa: F401,E501
     38     )
     39 except ImportError as _err:  # pragma: no cover
     40     _module = _err.name

File c:\Users\apodo\anaconda3\Lib\site-packages\pandas\compat\__init__.py:27
     25 import pandas.compat.compressors
     26 from pandas.compat.numpy import is_numpy_dev
---> 27 from pandas.compat.pyarrow import (
     28     pa_version_under7p0,
     29     pa_version_under8p0,
     30     pa_version_under9p0,
     31     pa_version_under11p0,
     32     pa_version_under13p0,
...
---> 10     _palv = Version(Version(pa.__version__).base_version)
     11     pa_version_under7p0 = _palv < Version("7.0.0")
     12     pa_version_under8p0 = _palv < Version("8.0.0")

AttributeError: module 'pyarrow' has no attribute '__version__'

解决方案 #

报错根源是pyarrow模块损坏或版本与pandas不兼容，和numpy、pandas本身无关，按以下步骤修复：

• 彻底卸载pyarrow：
  打开Anaconda Prompt或终端，执行：

  pip uninstall -y pyarrow
  conda remove -y pyarrow
  

• 删除残留文件：
  手动删除Anaconda环境下的pyarrow文件夹，路径示例：c:\Users\apodo\anaconda3\Lib\site-packages\pyarrow
• 重新安装兼容版本：
  先查看当前pandas版本，再安装匹配的pyarrow版本（比如pandas 2.x适配pyarrow 14.x-16.x）：

  pip show pandas
  pip install pyarrow==15.0.0
  

• 验证：
  重新运行导入代码，确认报错消失。

内容的提问来源于stack exchange，提问作者Alex