🎰 Slotscheck

Adding __slots__ to a class in Python is a great way to improve performance. But to work properly, all base classes need to implement it — without overlap! It's easy to get wrong, and what's worse: there is nothing warning you that you messed up.

✨ Until now! ✨

slotscheck helps you validate your slots are working properly. You can even use it to enforce the use of slots across (parts of) your codebase.

See my talk at EuroPython for the details of how slots work and what can go wrong with them.

Quickstart

Usage is quick from the command line:

python -m slotscheck [FILES]...
# or
slotscheck -m [MODULES]...

For example:

$ slotscheck -m sanic
ERROR: 'sanic.app:Sanic' defines overlapping slots.
ERROR: 'sanic.response:HTTPResponse' has slots but superclass does not.
Oh no, found some problems!
Scanned 72 module(s), 111 class(es).

Now get to fixing — and add slotscheck to your CI pipeline or pre-commit to prevent mistakes from creeping in again! See here and here for examples.

Features

• Detect broken slots inheritance
• Detect overlapping slots
• Detect duplicate slots
• Pre-commit hook
• (Optionally) enforce the use of slots

See the documentation for more details and configuration options.

Why not a flake8 plugin?

Flake8 plugins need to work without running the code. Many libraries use conditional imports, star imports, re-exports, and define slots with decorators or metaclasses. This all but requires running the code to determine the slots and class tree.

There's an issue to discuss the matter.

Why not pylint or ruff?

Pylint and Ruff do have some checks related to __slots__, but they are still static analyzers. They cannot see slots that are computed at import time—or follow deep inheritance trees, which are common in real code. People often generate slots automatically when optimizing class hierarchies, and frameworks like dataclasses, attrs, and pydantic also generate slots for you. slotscheck runs the code and inspects the resulting class tree, and it can also detect unused slots. Ruff and Pylint can still catch some more basic slot mistakes, such as invalid __slots__ declarations.

Notes

• slotscheck does not work on alternative Python implementations such as PyPy, Jython, or IronPython. This is for two reasons:
  1. Other implementations such as PyPy handle slots differently, and running slotscheck wouldn't be useful.
  2. To perform its checks, slotscheck uses CPython-specific features
• slotscheck will try to import all submodules of the given package. If there are scripts without if __name__ == "__main__": blocks, they may be executed.
• Even in the case that slots are not inherited properly, there may still be an advantage to using them (i.e. attribute access speed and some memory savings). However, in most cases this is unintentional. slotscheck allows you to ignore specific cases.
• Because slotscheck imports your code in arbitrary order, it can—in rare cases—result in confusing and randomly-occurring import errors in third-party libraries. In such a case, it is recommended to omit modules such as your tests and mypy plugins from the slotscheck run. See here. Alternatively, you can use PYTHONHASHSEED=<any value> to make the import order deterministic. A solution to this problem is being worked on in this issue.

Installation

It's available on PyPI.

pip install slotscheck