Skip to content

README.md

Latest commit

 

History

History
134 lines (94 loc) · 4.46 KB

README.md

File metadata and controls

134 lines (94 loc) · 4.46 KB

LinAlgKit

Docs CI Wheels PyPI

LinAlgKit Banner

LinAlgKit is a lightweight, NumPy-powered linear algebra toolkit for Python. It offers a minimal, clean API for matrices with scientific computing essentials: construction, arithmetic, transpose, trace, and determinant — all with first-class NumPy interoperability.

Features

  • Supports multiple numeric dtypes: int, float32, float64
  • Clean matrix API: +, -, * (matrix and scalar), transpose(), trace(), determinant()
  • Constructors: identity(n), zeros(r, c), ones(r, c)
  • NumPy interop: .from_numpy(ndarray), .to_numpy()
  • Pure Python packaging — quick pip install on any platform

Installation

pip install -U pip
pip install LinAlgKit

Editable install for development:

pip install -U pip
pip install -e .

Quickstart

import numpy as np
import LinAlgKit as lk

# Construct from NumPy
A = lk.Matrix.from_numpy(np.array([[1.0, 2.0], [3.0, 4.0]]))
B = lk.Matrix.identity(2)

# Core ops
C = A + B
AT = A.transpose()
detA = A.determinant()

print("C =\n", C.to_numpy())
print("AT =\n", AT.to_numpy())
print("det(A) =", detA)

Python API overview

  • Matrix, MatrixF, MatrixI classes with common operations
  • Functional helpers: array, zeros, ones, eye, matmul, transpose, trace, det
  • NumPy interop by design (copy in both directions for safety)

Design Philosophy

  • Vectorization-first: prefer NumPy operations and shapes that compose well.
  • Minimal surface area: focus on the 80% of linear algebra tasks used daily.
  • Explicit data flow: .from_numpy() and .to_numpy() are copy-based and clear.
  • Pythonic ergonomics: a small, predictable API that reads like the math.
  • Interop-ready: functions also accept NumPy arrays where sensible.

Why LinAlgKit? (vs. raw NumPy)

  • Matrix-first API with clear semantics (Matrix, transpose(), determinant()), helpful for pedagogy and readability.
  • Convenience constructors (identity, zeros, ones) aligned with matrix mental models.
  • Gentle learning curve for users coming from linear algebra courses before diving into broader NumPy idioms.
  • Clean separation between object API and functional helpers so you can mix OO and vectorized styles.

Testing

python -m pip install -U pytest
pytest -q

NumPy interop

  • Matrix/MatrixF/MatrixI.from_numpy(array) constructs from a 2D ndarray (copy)
  • .to_numpy() returns a 2D ndarray (copy)
  • For vectorized workflows, you can also use the functional API (matmul, trace, det) directly on NumPy arrays

Examples

from LinAlgKit import array, eye, matmul, det

A = array([[1., 2.], [3., 4.]])
I = eye(2)
print(det(A))         # -> -2.0
print(matmul(A, I))   # -> A

Scientific notes

  • Determinant, trace, and matmul are delegated to NumPy/numpy.linalg where applicable
  • API emphasizes clarity and composability; best used together with NumPy idioms

Benchmarks

For research-grade benchmarking, consider asv or simple scripts using timeit with NumPy arrays. A basic harness can be added in scripts/ if needed.

Roadmap

  • Convenience APIs (slicing helpers, broadcasting-aware ops)
  • Optional SciPy interop (sparse CSR constructors)
  • Expanded tests and property-based testing
  • Example notebooks and gallery in docs/

Citation

If you use LinAlgKit in academic work, please cite this repository:

@software{linalgkit2025,
  author  = {SciComputeOrg},
  title   = {LinAlgKit: A Lightweight Linear Algebra Toolkit for Python},
  year    = {2025},
  url     = {https://github.com/SciComputeOrg/LinAlgKit}
}

License

This project is licensed under the Apache License 2.0 — see the LICENSE file for details.

Contributing

Contributions are welcome! Please open issues and pull requests. For larger features (e.g., sparse matrices), open a design discussion first.