Module numpy_compat 

NumPy compatibility layer for scirs2-core

This module ensures that scirs2_core::ndarray types are fully compatible with PyO3’s numpy crate, enabling seamless Python integration.

Problem Statement

The PyO3 numpy crate (v0.27) expects types from ndarray v0.16. When scirs2-core uses ndarray v0.17 by default, there’s a type incompatibility. Python integration modules must explicitly use ndarray16 types to ensure compatibility with numpy.

Solution

This module provides:

1. Explicit re-exports of ndarray types that numpy needs
2. Type aliases that guarantee compatibility
3. Conversion utilities for zero-copy operations

Usage in Python Bindings

use pyo3::prelude::*;
use scirs2_core::python::numpy_compat::*;

#[pyfunction]
fn process_array(array: PyReadonlyArrayDyn<f32>) -> PyResult<Py<PyArrayDyn<f32>>> {
    // Convert from NumPy to scirs2 array
    let scirs_array = numpy_to_scirs_arrayd(array)?;

    // Process with scirs2-core
    let result = scirs_array.map(|x| x * 2.0);

    // Convert back to NumPy
    scirs_to_numpy_arrayd(result, array.py())
}

Macros

array

Create an Array with one, two, three, four, five, or six dimensions.

s

Slice argument constructor.

Structs

ArrayBase

An n-dimensional array.

Axis

An axis index.

Dim

Dimension description.

IxDynImpl

Dynamic dimension or index type.

OwnedRepr

Array’s representation.

ViewRepr

Array view’s representation.

Zip

Lock step function application across several arrays or other producers.

Enums

MemoryLayout

Get the memory layout of a NumPy array

Traits

Data

Array representation trait.

DataMut

Array representation trait.

DataOwned

Array representation trait.

Dimension

Array shape and index trait.

RawData

Array representation trait.

Functions

Dim

Create a new dimension value.

Ix0

Create a zero-dimensional index

Ix1

Create a one-dimensional index

Ix2

Create a two-dimensional index

Ix3

Create a three-dimensional index

Ix4

Create a four-dimensional index

IxDyn

Create a dynamic-dimensional index

arr1

Create a one-dimensional array with elements from xs.

arr2

Create a two-dimensional array with elements from xs.

get_numpy_layout

is_numpy_compatible

Check if a NumPy array is compatible with scirs2 operations

numpy_batch_to_scirs

Convert a vector of NumPy arrays to scirs2 ArrayD arrays

numpy_readonly_to_scirs_arrayd

Convert NumPy readonly array to scirs2 ArrayD

numpy_readonly_to_scirs_view

Convert NumPy readonly array to scirs2 ArrayView (zero-copy)

numpy_to_scirs_arrayd

Convert NumPy array to scirs2 ArrayD (zero-copy when possible)

scirs_batch_to_numpy

Convert a vector of scirs2 arrays to NumPy arrays

scirs_to_numpy_array1

Convert scirs2 Array1 to NumPy array

scirs_to_numpy_array2

Convert scirs2 Array2 to NumPy array

scirs_to_numpy_arrayd

Convert scirs2 ArrayD to NumPy array

Type Aliases

Array

An array that owns its data uniquely.

Array0

zero-dimensional array

Array1

one-dimensional array

Array2

two-dimensional array

Array3

three-dimensional array

Array4

four-dimensional array

ArrayD

dynamic-dimensional array

ArrayView

A read-only array view.

ArrayView1

one-dimensional array view

ArrayView2

two-dimensional array view

ArrayViewD

dynamic-dimensional array view

ArrayViewMut

A read-write array view.

ArrayViewMut1

one-dimensional read-write array view

ArrayViewMut2

two-dimensional read-write array view

ArrayViewMutD

dynamic-dimensional read-write array view

Ix0

zero-dimensionial

Ix1

one-dimensional

Ix2

two-dimensional

Ix3

three-dimensional

Ix4

four-dimensional

IxDyn

dynamic-dimensional

NumpyCompatArray1

Type alias for numpy-compatible 1D arrays

NumpyCompatArray2

Type alias for numpy-compatible 2D arrays

NumpyCompatArrayD

Type alias for numpy-compatible dynamic arrays