为什么 python 包使用 __version__ 和 pip （同一目录）显示不同的版本

Question

我仔细检查了环境，两种方法显示同一个包目录：

>>> !pip show astropy
Name: astropy
Version: 5.1
Summary: Astronomy and astrophysics core library
Home-page: http://astropy.org/
Author: The Astropy Developers
Author-email: astropy.team@gmail.com
License: BSD 3-Clause License
Location: /home/duhc/anaconda3/lib/python3.9/site-packages
Requires: packaging, numpy, PyYAML, pyerfa
Required-by: sdss-marvin, photutils, mgefit, gwcs

>>> import astropy
>>> print(astropy.__version__)
>>> print(astropy.__path__)
4.3.1
['/home/duhc/anaconda3/lib/python3.9/site-packages/astropy']

我们看到，目录是一样的，都是/home/duhc/anaconda3/lib/python3.9/site-packages ，而版本不同。

Answer 1

这不会发生在新鲜的环境中：

$>conda create -n astropy python=3.9.7 astropy=5.1
$>conda activate astropy
$>pip show astropy
Name: astropy
Version: 5.1
Summary: Astronomy and astrophysics core library
Home-page: http://astropy.org
Author: The Astropy Developers
Author-email: astropy.team@gmail.com
License: BSD 3-Clause License
Location: ...\miniconda3\envs\astropy\lib\site-packages
Requires: numpy, packaging, pyerfa, PyYAML
Required-by:

$>python -c "import astropy; print(astropy.__version__)"
5.1

从 pypi 安装软件包时也不会发生

$>conda uninstall astropy
$>pip install astropy
$>python -c "import astropy; print(astropy.__version__)"
5.1

$>pip show astropy
Name: astropy
Version: 5.1
Summary: Astronomy and astrophysics core library
Home-page: http://astropy.org
Author: The Astropy Developers
Author-email: astropy.team@gmail.com
License: BSD 3-Clause License
Location: ...\miniconda3\envs\astropy\lib\site-packages
Requires: numpy, packaging, pyerfa, PyYAML
Required-by:

可能发生的事情是你用 conda 覆盖了一个 pip 安装的包。 完整的重现步骤：

$> conda create -n astropy python=3.9.7
$> conda activate astropy
$> pip install astropy
$> pip show astropy
Name: astropy
Version: 5.1
Summary: Astronomy and astrophysics core library
Home-page: http://astropy.org
Author: The Astropy Developers
Author-email: astropy.team@gmail.com
License: BSD 3-Clause License
Location: ...\miniconda3\envs\astropy\lib\site-packages
Requires: numpy, packaging, pyerfa, PyYAML
Required-by:
$> conda install astropy==4.3.1
$> pip show astropy
Name: astropy
Version: 5.1
Summary: Astronomy and astrophysics core library
Home-page: http://astropy.org
Author: The Astropy Developers
Author-email: astropy.team@gmail.com
License: BSD 3-Clause License
Location: ...\miniconda3\envs\astropy\lib\site-packages
Requires: numpy, packaging, pyerfa, PyYAML
Required-by:

$> python -c "import astropy; print(astropy.__version__)"
4.3.1

因此，看起来这不是包的错误，而是一个很好的例子，说明为什么你应该小心不要在同一个环境中随意混合conda和pip命令。 我特别建议不要在您的基础环境中这样做（就个人而言，我根本不会更改基础环境，除了更新）。