Introduce DNDarray.__array__() method

[ClaudiaComito]

Description

This method returns a view of the underlying process-local array as numpy ndarray. If the DNDarray is on GPU, the array is copied to CPU first.

With this change, passing a DNDarray to np.asarray() returns a valid np.ndarray. This is needed for interoperability with other libraries in the numpy ecosystem.

Basic example with xarray:

>>> import xarray as xr
>>> import heat as ht
>>> a = ht.arange(10).reshape(5,2)
>>> a
DNDarray([[0, 1],
          [2, 3],
          [4, 5],
          [6, 7],
          [8, 9]], dtype=ht.int32, device=cpu:0, split=None)
>>> xr_a = xr.DataArray(a)
>>> xr_a
<xarray.DataArray (dim_0: 5, dim_1: 2)>
array([[0, 1],
       [2, 3],
       [4, 5],
       [6, 7],
       [8, 9]], dtype=int32)
Dimensions without coordinates: dim_0, dim_1

If the DNDarray is distributed, the example above will return an xarray of the local slice of data on each MPI process.

Related: #1031

Issue/s resolved: #1153

Changes proposed:

• Implement DNDarray.__array__() method
• Update documentation of DNDarray.numpy() method
• Implement tests

Type of change

• New feature (non-breaking change which adds functionality)

Memory requirements

NA

Performance

NA

Due Diligence

• All split configurations tested
• Multiple dtypes tested in relevant functions
• Documentation updated (if needed)
• Title of PR is suitable for corresponding CHANGELOG entry

Does this change modify the behaviour of other functions? If so, which?

no

[ghost]

👇 Click on the image for a new way to code review

Legend

[github-actions]

Thank you for the PR!

[github-actions]

Thank you for the PR!

[mrfh92]

From my point of view the changes look fine 👍
Therefore I recommend merging if the CI succeeds.

[github-actions]

Thank you for the PR!

[bhagemeier]

Looks good to me.