Concatenate to support Ellipsis argument

[Daraan]

Fixes #110.

This PR adds support for Ellipsis in Concatenate for the Python 3.8-3.10 backports.

Does not fix:

• Concatenate implementation generates runtime exception if parameterized by ... #48 : Using Ellipsis as argument for Callable with a generic..
  • <= 3.9: because of typing.Callable limitations
  • For 3.10 needs modification of _ConcatenateGenericAlias

Changes:
- Changed _concatenate_getitem to reflect the Python3.11 typing module that allows the Ellipsis parameter.
- For Python 3.10 uses typing._ConcatenateGenericAlias and _concatenate_getitem
- For <=Python 3.9 the typing_extensions._ConcatenateGenericAlias and _concatenate_getitem like before
- For <3.9.2 uses a Dummy class that is temporarily swapped with an Ellipsis input
- added tests and updated tests

[Daraan]

I am sorry I had to recreate #442.

Latest change since then:

-        if not TYPING_3_11_0:
-            with self.assertRaisesRegex(
-                TypeError,
-                'each arg must be a type',
-           ):
-              Concatenate[1, P]
# different location
+    def test_invalid_use(self):
+       # Assure that `_type_check` is called.
+        P = ParamSpec('P')
+        with self.assertRaisesRegex(
+            TypeError,
+            "each arg must be a type",
+        ):
+            Concatenate[(str,), P]

[JelleZijlstra]

Suggested change

	- Extended the Concatenate backport for Python 3.8-3.10 to now accept
	ellipsis as an argument. Patch by [Daraan](https://github.com/Daraan).
	- Extended the `Concatenate` backport for Python 3.8-3.10 to now accept
	`Ellipsis` as an argument. Patch by [Daraan](https://github.com/Daraan).

[JelleZijlstra]

In this case I think I'd want to block it on all 3.10 versions. It's a bad experience if something works on early 3.10 and breaks if people upgrade to a later bugfix release. Better to allow it only on 3.11+.

[JelleZijlstra]

Also, the version check seems wrong here, the change was made in 3.10.3 not 3.10.2.

[JelleZijlstra]

If you make this inherit from ParamSpec, can it fix 3.10.3+?

[Daraan]

I might be missing something here. To what case are you referring here?
This backport of Concatenate[...] also works for all 3.10, as the critical part is in _concatenate_getitem

Is it possibly related to the following test I jointly introduced also in #479?

@skipUnless(TYPING_3_11_0,  # <-- Needs PR #479 to backport for 3.10
                "Cannot be backported to <=3.9. See issue #48"
                "Cannot use ... with typing._ConcatenateGenericAlias after 3.10.2")
def test_alias_subscription_with_ellipsis(self):
    P = ParamSpec('P')
    X = Callable[Concatenate[int, P], Any]
    C1 = X[...]  # <-- Needs PR #479 for 3.10.3+
    self.assertEqual(get_args(C1), (Concatenate[int, ...], Any))  # <-- needs this PR

---

That aside; It is possible to do _EllipsisDummy = ParamSpec("_EllipsisDummy") if needed as subclassing ParamSpec is (currently) not possible. I thought using a new class might have less side-effects than using a ParamSpec instance.

[JelleZijlstra]

I'm referring to the comment right above here, about the test that works only in 3.10.0-3.10.2 but not 3.10.3+. Making _EllipsisDummy an instance of ParamSpec might fix that test for later versions of 3.10.

[Daraan]

I am sorry for the confusion. I do think we talk about the same location :), if not correct me please. I made the necessary edits above.

#479 allows to backport that test to all versions of 3.10, however for #479 the test self.assertEqual(get_args(C1), (Concatenate[int, ...], Any)) would need this PR to work without an error on 3.10 as well.
Both PRs are needed make this test work for the 3.10 version range.

[JelleZijlstra]

I see, thanks. #479 is merged now, so would you mind updating this PR? There's a merge conflict right now.

[Daraan]

Sure. Unified the code and the test we talked about, aside from that added one more test for invalid usage.