Automatic "__slots__".
from autoslot import Slots
class Compact(Slots):
def __init__(self, a, b):
self.x = a
self.y = bThis produces exactly the same class as if you had done:
class Compact:
__slots__ = {'x', 'y'}
def __init__(self, a, b):
self.x = a
self.y = bSimply: the code inside __init__() is scanned to find all assignments
to attributes on self, and these are added as __slots__.
The benefit of using autoslot.Slots over a manual slots declaration is
that you can modify the
code inside the __init__() method to add more attributes, and those
changes will automatically be reflected in the __slots__ definition.
You can also have the best of both worlds: slots for fields you expect,
as well as a __dict__ for those you don't:
from autoslot import SlotsPlusDict
class SemiCompact(SlotsPlusDict):
def __init__(self, a, b):
self.x = a
self.y = b
inst = SemiCompact(1, 2)
inst.z = 123 # <-- This won't fail!Attributes x and y will be stored in slots, while all other
dynamically-assigned attributes will go into the usual __dict__ instance
inside the class. If most of your class's attributes appear in the __init__()
method (these will become slots), then the space bloat caused by dictionary
hash-table expansion will be contained to only the dynamically-assigned
attributes.
See for yourself! The code is tiny.
In words: the metaclass finds the __init__() method, if present, and
accesses its bytecode. It looks for all assignments to attributes of
self, and considers those to be desired __slots__ entries. Then the
metaclass injects __slots__ into the namespace of the class definition
and thereafter allows class creation to proceed as normal.
When __slots__ are used, weak references (e.g. using the weakref
standard library module) won't work. If you need weak references, just
set it up on a new __slots__ class variable as you would normally
do without using autoslot:
from autoslot import Slots
class Compact(Slots):
__slots__ = ['__weakref__']
def __init__(self, a, b):
self.x = a
self.y = bEverything else will still work, and instances of Compact will now
also play nicely with the weakref module.
This project uses nox with the uv backend for running tests across multiple Python versions.
# Create a virtual environment
uv venv --python=3.13 .venv
source .venv/bin/activate
# Install nox with uv support
uv tool install 'nox[uv]'
# Run tests on all Python versions (auto-downloads missing interpreters)
nox --download-python=always
# Run tests on a specific version
nox -s tests-3.14
# List available test sessions
nox -l