`ValueError: No Zarr data type found that matches dtype 'dtype('uint32')'` on Windows

[radoering]

Zarr version

3.15

Numcodecs version

0.14.1

Python Version

3.13

Operating System

Windows

Installation

using poetry into virtual environment

Description

On Windows, the following snippet, which worked with zarr 2, fails with zarr 3 with

ValueError: No Zarr data type found that matches dtype 'dtype('uint32')'

Repro:

import zarr

a = np.array([1, 2], dtype=np.uint32) & 1
zarr.array(a)

The issue seems to be that the dtype is UIntDtype instead of UInt32Dtype.

When running the following snippet:

dtypes = [
    (np.array([1, 2], dtype=np.uint16) & 1).dtype,
    (np.array([1, 2], dtype=np.uint32) & 1).dtype,
    (np.array([1, 2], dtype=np.uint64) & 1).dtype,
]
for dtype in dtypes:
    print(f"{dtype} {type(dtype)}")

we get

uint16 <class 'numpy.dtypes.UInt16DType'>
uint32 <class 'numpy.dtypes.UIntDtype'>
uint64 <class 'numpy.dtypes.UInt64DType'>

Steps to reproduce

import zarr

a = np.array([1, 2], dtype=np.uint32) & 1
zarr.array(a)

Additional output

No response

[d-v-b]

thanks for this report! The fix is probably to robustly figure out which numpy int data type is actually a 32 bit integer.