FixedInt

Custom implementation of fixed-size integer classes in Python.

A nice learning exercise on bit manipulation as well as a refresher on more advanced Python topics such as operator overloading, function closures, and the class factory and singleton design patterns.

Setup

This package was written for Python versions 3.8+.

Create and activate a virtual environment for the project directory. Then, you can install the package in editable mode:

python -m venv .venv
source .venv/bin/activate
pip install -e .

NOTE: This project uses pyproject.toml for metadata and setup instead of the legacy setup.py approach. If the installation fails, you probably have an outdated version of pip:
python -m pip install --upgrade pip

Usage

Import the FixedInt class factory and call it to create classes representing integers with specific bit widths.

from fixedint import FixedInt, FixedSignedInt, FixedUnsignedInt

# Create FixedIntTypes.
UInt12 = FixedInt(12, signed=False)
Int8 = FixedSignedInt(8)
UInt8 = FixedUnsignedInt(8)

# Create instances of the FixedIntTypes.
foo = UInt8(5)
bar = Int8(200)  # Handles overflow
print(foo, bar)  # 5 -56

Features

Easily get the zero-padded, Two's complement binary representation:

>>> FixedSignedInt(8)(101).as_binary()
'01100101'
>>> FixedSignedInt(8)(-45).as_binary()
'11010011'

Supports operations between instances:

>>> num1 = UInt8(25)
>>> num2 = UInt8(7)
>>> print(num1 + num2)
32

Supports operations between other FixedIntTypes. The returned type is the type of the first operand:

>>> num3 = UInt8(100)
>>> num4 = FixedSignedInt(12)(2040)
>>> num3 - num4  # Overflows
FixedInt(size=8, signed=False)(108)
>>> num4 - num3
FixedInt(size=12, signed=True)(1_940)

Supports operations with the built-in int. The returned type is the type of the first operand:

>>> Int8(5) * 8
FixedInt(size=8, signed=True)(40)
>>> 8 * Int8(5)
40
>>> type(_)
<class 'int'>

Polymorphic inheritance checking:

>>> from fixedint import FixedIntType, FixedSignedInt
>>> num1 = FixedSignedInt(10)(25)
>>> num2 = FixedSignedInt(10)(40)
>>> type(num1) is type(num2)
True
>>> isinstance(num1, FixedSignedInt(10))
True
>>> isinstance(num1, FixedIntType)
True
>>> isinstance(num1, int)
True

All types created by the FixedInt() and its factory wrapper functions are interned within the FixedIntType parent class. This class can be imported for use as a type hint but also doubles as a singleton manager. All numbers with the same (size, signed) properties are guaranteed to share the same class object in memory.

Testing

I've included a unit test script to sanity check the behavior of FixedIntType instances and how they interact with each other and with the built-in int:

make test

FixedIntType itself subclasses the built-in int, allowing it to be treated as one almost anywhere a normal int would. Obviously, this practice is always a double-edged sword because my custom type might also break some internals of the normal int, causing strange bugs.

Disclaimer

This was an afternoon hobby project by an undergraduate student with too much free time. I do not claim it to be safe or reliable for use in production-level projects.