Crate rumpy 

rumpy — a numpy-feature-compatible Python module implemented in Rust on top of ndarray, exposed to rustpython_vm as the module numpy.

Supports the full numpy numeric dtype set:

• bool
• signed integers: int8 / int16 / int32 / int64
• unsigned integers: uint8 / uint16 / uint32 / uint64
• floats: float16 / float32 / float64
• complex: complex64 / complex128

Element type promotion follows numpy.result_type (see promote.rs). The on-array data is a plain ndarray::ArrayD<T> per dtype — there is no internal locking; synchronization is the embedder’s responsibility.

Re-exports

pub use dtype::ArrayElement;

pub use dtype::ArraysD;

pub use dtype::CoerceArray;

pub use dtype::DType;

Modules

convert

Python ↔ ArraysD conversion.

create

Array constructors: zeros, ones, full, eye, arange, linspace.

dtype

Dtype machinery: the DType enum, the ArraysD storage enum that holds one ndarray::ArrayD<T> per numpy element type, and the dispatch macros and helpers used by every operation.

einsum

numpy.einsum — the subscript-string variant.

extras

Second-tier numpy operations: logical/bitwise/finite predicates, clip, cumulative ops, where/nonzero, sort/argsort/unique, stack/squeeze/expand, repeat/tile, median/ptp.

extras2

Additional creation helpers, trig auxiliaries, and stats — the “obvious-numpy” surface that wasn’t in the first cut.

fft

numpy.fft — 1-D forward/inverse FFT, 2-D variants, and real-input transforms, all routed through rustfft. Frequencies follow numpy’s fftfreq/rfftfreq conventions.

fmt

repr() and str() formatting for ArraysD.

index

Indexing & slicing.

internal

Small helpers for “this should not happen, but if it does, raise a clean Python error instead of panicking.”

linalg

dot / matmul (numpy semantics for 1-D and 2-D cases).

linalg_extra

numpy.linalg algorithms: norm, det, inv, solve, matrix_rank. Pure-Rust implementations on f64 (input promoted to f64 first).

more_ops

Third-tier numpy operations: flip/roll/rot90, column_stack/dstack, diag/triu/tril/diagflat, atleast_Nd, count_nonzero/bincount/histogram, nan-aware reductions, searchsorted, meshgrid, interp/trapz/gradient.

npy

Reader/writer for the numpy .npy format, version 1.0.

npz

Reader/writer for the numpy .npz format.

ops

Elementwise binary arithmetic with broadcasting + unary ufuncs.

poly

Polynomial helpers: polyval, roots, polyfit.

promote

Numpy-compatible type promotion (numpy.result_type).

random

numpy.random — pseudo-random generators and distribution sampling.

reduce

Reductions: sum / prod / mean / min / max / std / var.

textio

numpy.savetxt / loadtxt (text format) and tofile / fromfile (raw binary).

Macros

dispatch_numeric

Dispatch that ONLY covers the 14 numeric variants. Non-numeric variants (Object, Str, …) take the wildcard arm $other, useful for ops that only make sense on numeric data.

dispatch_view

“View-style” dispatch — runs $body once per arm with $a bound to a reference to the inner array. Useful for read-only inspection where the element type does not appear in the result.

Structs

PyNdArray

Python-visible ndarray. Wraps an ArraysD — a tagged union over all numpy element types.

Functions

module_def

Return the numpy module definition for embedding into a rustpython_vm::Interpreter.