refactor: Rename exposed fixtures and pytest options

[FBruzzesi]

Description

This PR addresses the section from #3552:

In light of the CI failures for downstream projects, I also realized that having fixtures called constructor and constructor_eager is not really intuitive. I am happy to brainstorm better names for downstream users. And similarly the flags --all-cpu-constructors, --constructors. (*)

with the following renaming:

• fixtures: nw_frame_constructor (eager + lazy), nw_eager_constructor (eager only), nw_pandas_like_constructor (pandas like only).
• pytest options: --nw-backends, --all-nw-backends, --use-external-nw-backend

---

I know that nw.from_native(nw_frame_constructor(data)) is extremely verbose, but bear with me - I am also proposing to directly return nw.LazyFrame or nw.DataFrame from the nw_frame_constructor fixture.

From #3552:

If we find a way to achieve the following points, then I personally think we should do so, as we almost always run nw.from_native right after.

• Distinguish between native frame and nw.from_native kwargs -> For this it's enough to provide it as different proper arguments in __call__, e.g. nw_kwargs, backend_kwargs.
• How do we know which version a user would use? Should we require to pass either the Version.X or the namespace {nw, nw_v1, nw_v2} ? If we do so, I would argue that we must enforce it and never assume for the user. In my opinion nw_frame_constructor(data, version=nw) would still look better than nw.from_native(nw_frame_constructor(data))

---

TODO:

• Update Docs

What type of PR is this? (check all applicable)

• 💾 Refactor
• ✨ Feature
• 🐛 Bug Fix
• 🔧 Optimization
• 📝 Documentation
• ✅ Test
• 🐳 Other

[dangotbanned]

I know that nw.from_native(nw_frame_constructor(data)) is extremely verbose, but bear with me

Agreed, IMO this was already too much ceremony.
Like this blows up to almost double the size 😰

result = pl.DataFrame({"a": [1, 2, 3]}).select()
result = nw.from_native(constructor_eager({"a": [1, 2, 3]}), eager_only=True).select() 

I am also proposing to directly return nw.LazyFrame or nw.DataFrame from the nw_frame_constructor fixture.

👍 from me on skipping from_native.

Note

you probably already knew I was on board, I just wanted to make it more visible

---

To bikeshed on the naming some more:

1. Is the nw_ prefix to avoid clobbering the fixture namespace?
   If so, pytest is pretty flexible in Overriding fixtures on various levels.
   Just seems like a problem that pytest would have a solution for hidden somewhere 😄
2. If all of our fixtures are suffixed by _constructor, is that part not redundant?

Combining those two points, you'd get something like this, where the concepts are embedded in the types, rather than the names:

Mandatory plug for tests.plan

Made you click 😉

I'm still missing a frame fixture, but that would come eventually

narwhals/tests/plan/conftest.py

Lines 22 to 27 in a09be54

	@pytest.fixture
	def lazyframe() -> LazyFrame: ...
	@pytest.fixture
	def dataframe() -> DataFrame: ...
	@pytest.fixture
	def series() -> Series: ...

narwhals/tests/plan/utils.py

Lines 565 to 572 in a09be54

	LazyFrame: TypeAlias = Constructor[nwp.LazyFrame[Any]]
	"""The type of the `lazyframe` fixture."""
	DataFrame: TypeAlias = Constructor[nwp.DataFrame[Any, Any]]
	"""The type of the `dataframe` fixture."""
	Series: TypeAlias = Constructor[nwp.Series[Any]]
	"""The type of the `series` fixture."""

You can still see more detail in an IDE if you want it:

Show IDE screenshot

Although another option is using a suffix for the type of the fixture, e.g.:

lazyframe: LazyFrame
lazyframe: LazyFrameFixture

Then you can alias the import if you want something shorter 😎