如何交叉导入子目录中的模块，以便它们既可以作为子模块又可以独立运行？

Question

I have a Python module that normally works as a stand-alone.

file1.py
file2.py
file3.py

However, I also want it to be part of a different project, in which the module is placed in a separate subdirectory.

__init.py__
build.py
└── compiler
    └── __init__.py
    └── file1.py
    └── file2.py
    └── file3.py

Since the module scripts use plenty of cross-imports, this is not possible. Once placed in a subdirectory, the imports no longer find the respective files because it looks in the top directory only.

To remedy the problem, I tried various things. I appended the subdirectory as an additional path in the top-most build.py script.

sys.path.append('compiler')

It did not solve the problem. Cross-imports are still not working.

I also tried relative imports but that breaks the stand-alone version of the module. So, I tried exception handling to catch them

try:
    from file1 import TestClass
except ImportError:
    from .file1 import TestClass

That did not work either and resulted, despite my best efforts in ImportError: attempted relative import with no known parent package errors.

I also tried all sorts of variations of these things, but none of it worked.

I know it has to be possible to do something like this and I am surprised that this is so hard to do. My Internet searches all came back with the same suggestions—the ones I outlined above, none of which worked in my case, particularly because they do not take into account the ability to run code as a stand-alone and as a sub-module.

I can't be the first person trying to write a module that can be used as a stand-alone package or as a sub-module in other projects. What am I missing here?

Answer 1

Relative imports, as the error tells you, require a parent package. Think of .file1 as shorthand for <this_module>.file1 . If there is no <this_module> , you can't ask it for file1 . In order to properly use relative imports you'll have to make a wrapper project to contain the shared module so it can be properly namespaced.

So your standalone module will instead look like this, matching the consumer:

__init.py__
standalone.py
└── compiler
    └── __init__.py
    └── file1.py
    └── file2.py
    └── file3.py

The other option is to make your shared module truly installable with a setup.py or pyproject.toml or whatever your favorite method is. Then you install it in the consuming project instead of directly including it.