Skip to content

README.md

Latest commit

 

History

History
139 lines (107 loc) · 4.19 KB

README.md

File metadata and controls

139 lines (107 loc) · 4.19 KB

TinyMatrix

TinyMatrix is a minimal, dependency-free Python library that provides a small Matrix type with basic linear-algebra-like operations: construction, element-wise arithmetic, matrix multiplication, transpose, indexing/slicing, and a few static constructors (identity, zeroes, ones).

Designed for learning and small scripts where a tiny and readable matrix helper is preferred over a large dependency.

Install

Install editable from the project root for development:

python -m pip install -e .

Or install normally (if packaged):

python -m pip install .

Quick usage

from tinymatrix import Matrix

# Create a matrix
A = Matrix(matrix=[[1, 2], [3, 4]])

# Arithmetic
B = Matrix(matrix=[[5, 6], [7, 8]])
C = A + B  # element-wise addition

# Scalar multiply
D = A * 2

# Matrix multiplication
E = A @ Matrix(matrix=[[1], [2]])

# Transpose
T = A.T

# Indexing / slicing
val = A[0, 1]
sub = A[0:1, :]

# Static constructors
I = Matrix.identity(3)
Z = Matrix.zeroes(2, 3)
O = Matrix.ones(2, 2)

Refer to tests/test_matrix.py for additional examples and expected behavior.

Running tests

From the project root run:

pytest -q

License

Licensed under the MIT License. See the LICENSE file for full text.

🗺️ Roadmap / TODO

Phase 1: Foundation & Quality

  • Add full type hints for all public APIs
  • Add optional dtype support (float/int/Decimal/complex)
  • Add stricter input and type validation
  • Improve slice/index edge-case tests
  • Add GitHub Actions for lint, test, and build
  • Add code coverage tooling + badge

Phase 2: Core Matrix Operations

  • Add .row(i), .col(j), .flatten() helpers
  • Add .apply(func) for element-wise operations
  • Add broadcasting support (1×N and M×1 vectors)
  • Add random matrix constructors (uniform, normal)
  • Improve printing for large matrices (ellipsis, alignment)
  • Add .reshape(), .squeeze(), .expand_dims()
  • Add element-wise functions (abs, sqrt, exp, log, sin, cos, etc.)
  • Add reduction operations (sum, mean, std, min, max) with axis support
  • Add .diagonal(), .trace() operations

Phase 3: Linear Algebra Core

  • Implement determinant calculation
  • Implement matrix inverse
  • Implement matrix rank
  • Implement matrix norm (Frobenius, spectral, etc.)
  • Add matrix power (A ** n)
  • Add Kronecker product
  • Add Hadamard (element-wise) product operator

Phase 4: Decompositions & Solvers

  • Add LU decomposition
  • Add QR decomposition
  • Add Cholesky decomposition
  • Add SVD (Singular Value Decomposition)
  • Add eigenvalue/eigenvector computation
  • Add linear system solver Ax = b
  • Add least-squares solver

Phase 5: Performance & Optimization

  • Optimize matrix multiplication (Strassen or blocked algorithms)
  • Add optional NumPy fallback mode for performance comparisons
  • Add lazy evaluation for chained operations (optional)
  • Profile and optimize hotspots
  • Add benchmarking utilities and performance docs

Phase 6: Advanced Features

  • Add sparse matrix support (CSR/CSC formats)
  • Add immutability mode (frozen=True)
  • Add matrix views (no-copy slicing where possible)
  • Add .copy() and copy-on-write semantics
  • Add stacking operations (vstack, hstack, block)
  • Add advanced indexing (fancy indexing, boolean masks)
  • Add matrix condition number calculation
  • Add pseudo-inverse (Moore-Penrose)

Phase 7: Documentation & Polish

  • Expand README with comprehensive examples
  • Add Sphinx docs site
  • Add example notebooks (Jupyter)
  • Add comparison guide: TinyMatrix vs NumPy
  • Add migration guide for NumPy users
  • Complete PyPI metadata
  • Add mathematical notation in docstrings
  • Create logo and branding

Stretch Goals

  • Add matrix calculus operations (gradient, Jacobian, Hessian)
  • Add graph/network matrices (adjacency, Laplacian)
  • Add special matrix generators (Toeplitz, Hankel, Vandermonde)
  • Add matrix equation solvers (Sylvester, Lyapunov)
  • Add tensor product operations
  • Consider Cython/C extensions for critical paths (while keeping pure-Python as default)