TST: revisit test for asarray copy= parameter

[crusaderky]

• Partially fixes Test errors with cupy==13.4.1 #324
• Dask now will raise ValueError instead of NotImplementedError when it fails asarray(..., copy=False).
• Align test to reflect changes in ENH: Simplify CuPy asarray and to_device #314

[Copilot]

Pull Request Overview

This PR updates the asarray(..., copy=…) behavior and its tests to always raise ValueError on unsupported no-copy conversions and simplifies test assertions.

• Switch Dask’s non-copy conversion error from NotImplementedError to ValueError.
• Align and simplify tests across all libraries to expect ValueError when copy=False is unsupported.
• Replace all(...) assertions with direct element checks.

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
tests/test_common.py	Removed conditional branches, unified exception expectations to ValueError, and simplified assertions.
array_api_compat/dask/array/_aliases.py	Changed the exception raised for copy=False on non‐Dask objects from NotImplementedError to ValueError.

Comments suppressed due to low confidence (1)

tests/test_common.py:355

• [nitpick] This comment is just above the array.array test—consider moving or duplicating it to the if library == 'dask.array' block to make the test intent clearer.

# dask changed behaviour of copy=None in 2024.12 to copy;

[Copilot]

[nitpick] Consider unifying the exception message style with the dtype‐change branch (e.g. both starting 'Unable to avoid copy when ...') for consistency.

Suggested change

	"Unable to avoid copy when converting a non-dask object to dask"
	"Unable to avoid copy when converting a non-dask object to dask array"

[ev-br]

Thanks Guido. Let's give it a shot, and test it locally together with other fixes.

[ev-br]

Can't say I'm happy about the need for branching here. It is what it is though

[crusaderky]

You're not happy that cupy can't create a GPU array as a view of host memory? 😕