[Refactoring] Add variables for skip decorators

[PokhodenkoSA]

It could be reused in other tests.
Branch based on #678.

[github-actions]

View rendered docs @ https://intelpython.github.io/dpctl/pulls/684/index.html

[oleksandr-pavlyk]

Why not move definitions of these decorators to _helper.py

This is where has_gpu, etc are defined anyway.

[PokhodenkoSA]

Moving skips variables to common place and reuse is the next step.

[PokhodenkoSA]

Also pytest can skip the whole module if some clause not satisfied (example):

pytestmark = pytest.mark.skipif(
    not shutil.which("gdb-oneapi"),
    reason="Intel® Distribution for GDB* is not available",
)

[PokhodenkoSA]

Why not move

If you agree that this style improves tests, then I will keep this style in my new code.

[oleksandr-pavlyk]

Saving decorators as variables makes it more readable, but we are using programmatic way to skip tests based on try/except constructs in newer code, e.g.:

try:
   q = dpctl.SyclQueue("cpu")
except dpctl.SyclQueueCreationError:
   pytest.skip("CPU queue could not be constructed")

[PokhodenkoSA]

we are using programmatic way to skip tests based on try/except constructs in newer code

This also could be checked in skip decorator. And this skip decorator could be transformed to variable.

[oleksandr-pavlyk]

This also could be checked in skip decorator. And this skip decorator could be transformed to variable.

It could, but use of decorator forces one to create a queue for the sake of checking in decorator, throw it away only to recreate it again in the body of the test.

[PokhodenkoSA]

throw it away only to recreate it again in the body of the test

It seems like a preliminary optimization. Why we need it?
We need to skip tests if we can not do some things. Moreover we will be able to do it on the level of module and only once. So it will be faster, just require reordering code in modules.

[oleksandr-pavlyk]

This PR has become mangled, I am rescinding my approval until this is resolved, or PR is split into smaller chunks.

[coveralls]

Coverage remained the same at 81.32% when pulling 444aa0b on spokhode/enh/tests into 439c2f4 on master.

[PokhodenkoSA]

Fixed PR mangling: merge w/ master and target to master.

[oleksandr-pavlyk]

Please squash all commits into 1, since all the changes affect single file.

[PokhodenkoSA]

Please squash all commits into 1

You mean squash and force push to the same branch?

[oleksandr-pavlyk]

Or squash when merging, net effect is the same

[PokhodenkoSA]

net effect is the same

I think it is not the same. I will squash to one commit and you apply Create a merge commit. It will create 2 commits on target branch.
squash when merging will create only 1 commit on target branch.

[github-actions]

Deleted rendered PR docs from intelpython.github.com/dpctl, latest should be updated shortly. 🤞