numpy-scientific

name	numpy-scientific
description	NumPy for scientific Python: dtypes and casting, creation, reshaping, broadcasting, indexing vs views, ufuncs and reductions, linear algebra and einsum, random Generator, I/O, structured arrays, performance and pandas boundaries. Use when writing or reviewing ndarray code. Triggers on numpy, ndarray, broadcasting, dtype, ufunc, einsum, vectorization.

NumPy (scientific computing)

Quick start

1. Shape and dtype first: decide shape, dtype, and memory order (C vs F) before filling arrays; avoid silent astype widening. See reference-arrays-dtypes.md.
2. Vectorize: prefer ufuncs and whole-array expressions over Python loops on large data; push loops to compiled layers when needed. See reference-ufuncs-reductions.md and reference-interop-performance.md.
3. Know view vs copy: basic slicing is usually a view; fancy indexing is a copy; reshape may view or copy. Mutations alias bugs are common. See reference-indexing-views.md.
4. Broadcasting: align trailing dimensions; use None / newaxis and np.broadcast_to deliberately. See reference-broadcasting-shape.md.
5. Reductions: always pass axis and use keepdims=True when broadcasting the result back; prefer np.nanmean etc. when NaNs appear. Mind float summation order for reproducibility. See reference-ufuncs-reductions.md.
6. Random: use np.random.Generator (PCG64), not legacy global RandomState, for reproducible science. See reference-random-io-structured.md.

Stack synergy

Resource	Role
general-python	uv, ruff, ty, vectorization policy, python-reviewer numerics footguns
dataframes	Table handoff at boundaries
numpy-docstrings	Public API docstrings for array-heavy modules
matplotlib-scientific	Plotting array results
lab-instrumentation	Binary waveform I/O, instrument-side buffers
python-reviewer	dtype/shape/reproducibility in review

Reference index

Topic	File
Creation, dtype, casting, order, strides	reference-arrays-dtypes.md
Slicing, fancy index, views, copies, reshape	reference-indexing-views.md
Broadcasting, stacking, meshgrid / ogrid	reference-broadcasting-shape.md
Ufuncs, reductions, axis, where, NaNs	reference-ufuncs-reductions.md
linalg, @, einsum, tensordot	reference-linalg-einsum.md
Generator, .npy / .npz, structured dtypes	reference-random-io-structured.md
Pandas/Polars handoff, memory, threads	reference-interop-performance.md

Official documentation

• NumPy user guide
• Absolute basics
• Broadcasting
• Indexing
• Copies and views
• NumPy reference

Dependency

Add or upgrade with uv add numpy per project policy; do not hand-edit pins in pyproject.toml.