here's my small, less efficient, and likely incorrect implementation of numpy in C++, built without external dependencies, and from scratch.

hopefully, I learn some C++ through this.

brew install fmt

# Install pybind11 (macOS)
brew install pybind11

# Or using pip
pip install pybind11

mkdir -p build
cd build
cmake ..
make
cd ..

• basic array construction (1d and 2d)
• zeros(shape)
• ones(shape)
• eye(n)
• arange(start, stop, step)
• linspace(start, stop, num)
• full(shape, value)

• integer indexing
• bounds checking
• row slicing (e.g., arr(0) for first row)
• range slicing (start:stop:step)
• view-based slicing (no data copying)
• advanced indexing (arr({1, 7, 3}) to select reorder these rows)

• addition (+)
• subtraction (-)
• multiplication (*)
• division (/)
• scalar operations
• unary operations (-arr, abs(arr))

• sum()
• min(), max()
• mean()
• std()
• axis-based reductions (e.g., sum(axis=0))

• shape compatibility checking
• broadcasted arithmetic operations
• efficient memory handling for broadcasts

• reshape()
• transpose()
• flatten()
• ravel()

• shared pointer implementation
• view-based operations
• copy-on-write optimization

• boolean masking
• matrix multiplication (matmul/dot)
• simd optimizations
• parallel operations