Source root not correctly used

[Julfried]

Bug description

I have a package that sits in a child folder of the current working directory (./src). Now if I want to to run pyreverse for this module I add ./src as a source-root. However pyreverse does not find the module.

Here is an example:

ex/
└── src/
    ├── __init__.py
    └── mymodule/
        ├── __init__.py
        └── testmodule.py

Configuration

No response

Command used

pyreverse -o html --source-root ./src mymodule

Pylint output

raise ImportError(f"No module named {'.'.join(module_parts)}")
ImportError: No module named mymodule

Expected behavior

When specifiy source-roots, pyreverse should also search the modules in this path

Pylint version

pylint 3.3.1
astroid 3.3.5
Python 3.12.7 | packaged by conda-forge | (main, Oct 4 2024, 15:47:54) [MSC v.1941 64 bit (AMD64)]

OS / Environment

win64

Additional dependencies

No response

[webknjaz]

FTR the linked PR hasn't fixed a no-name-in-module I'm seeing in an src layout with an implicit PEP 420 namespace. No idea why, though.

[Julfried]

Thanks for pointing this out and extending the tests!

I believe the issue lies in the fallback logic here:

# Fall back to legacy discovery by looking for __init__.py upwards
while True:
    if not os.path.exists(os.path.join(dirname, "__init__.py")):
        return dirname

This approach works only for explicit namespaces but fails for implicit ones (PEP 420), as they don't require __init__.py. The "legacy" name might be misleading here. Instead, I think there should be two distinct logic paths:

1. Explicit namespaces: Identify the module by looking for __init__.py.
2. Implicit namespaces: If no __init__.py, check if .py files exist in the directory (though this might not be sufficient in all cases).
3. Error handling: Throw an exception if neither case is satisfied.

My fix mainly addressed scenarios where the source root is a sub-folder of the working directory. This issue seems related but also requires handling implicit namespaces properly. In the next few weeks ,I’ll be short on time, but happy to assist further once there's more clarity!

---

Reference: I think this could also be linked to the issues I observed in #10077