Dask arrays are not recognized as array in is_ndarray_like

[whatnick]

Minimal, reproducible code sample, a copy-pastable example if possible

from numcodecs.ndarray_like import is_ndarray_like
import numpy as np
import dask.array as da
np_array = np.zeros((10,10))
da_array = da.zeros((10,10))
cast_da_array = np.array(da_array)
print(is_ndarray_like(da_array))
print(is_ndarray_like(np_array))
print(is_ndarray_like(cast_da_array))

False
True
True

Problem description

Not supporting compatibility with dask arrays as being array like for especially for Zarr persistence prevents chaining large computations and saving them gradually to zarr.

Version and installation information

Please provide the following:

• Value of numcodecs.__version__ : 0.13.1
• Version of Python interpreter : 3.12.3 (main, Sep 11 2024, 14:17:37) [GCC 13.2.0]
• Operating system (Linux/Windows/Mac) : Linux
• How NumCodecs was installed (e.g., "using pip into virtual environment", or "using conda") : using uv into virtualenv

Also, if you think it might be relevant, please provide the output from pip list or
conda list depending on which was used to install NumCodecs.

[martindurant]

dask.array is not numpy-like: it generally doesn't have data in memory. However, when you act on the array, the operation is broken up to act on its partitions, so that within a task, you see only a real numpy array. This is exactly how to_zarr works, including partition encoding via numcodecs.

Do you have an actual workflow where you have an issue?