feat: curation API for datasets

[ebezzi]

No description provided.

[ebezzi]

Ignore this for now

[ebezzi]

@Bento007 do you know what's going on here? Looks like the mock below isn't enough since this line:
x = assert_authorized_token(token, CorporaAuthConfig().curation_audience)
is calling the config in the parameters, which happens before the mock is called.

[ebezzi]

That should also exist in my function but it throws an error. I'll double check.

[ebezzi]

Here we need to decide if we want to always do a double lookup or if those methods only make sense when using either the canonical or the version id. I can look at the notebooks and see what makes sense, or just add the double lookup anywhere.

[Bento007]

I added this function

single-cell-data-portal/backend/portal/api/curation/v1/curation/collections/common.py

Lines 287 to 298 in 9e0a2cb

	def get_infered_collection_version_else_forbidden(collection_id: str) -> Optional[CollectionVersion]:
	"""
	Infer the collection version from either a CollectionId or a CollectionVersionId and return the CollectionVersion.
	:param collection_id: identifies the collection version
	:return: The CollectionVersion if it exists.
	"""
	version = get_business_logic().get_published_collection_version(CollectionId(collection_id))
	if version is None:
	version = get_business_logic().get_collection_version(CollectionVersionId(collection_id))
	if version is None:
	raise ForbiddenHTTPException()
	return version

We can punt on the double look-up decision and be inefficient for now for the sake of time.

[ebezzi]

I agree. I will use that function.

[Bento007]

a few tings to address

[Bento007]

How can this happen?

[ebezzi]

I think this isn't supposed to happen, but I added it anyway. For what it's worth, get_collection_version returns an Optional so pylance will force you to add the if is not None condition, or you'll get a warning.

[Bento007]

Is this a condition we can test for?

[ebezzi]

I don't think it's necessary. Also the test should rely on a database state that's impossible to achieve. If you prefer, we can add a TODO to double check or specify that the check is only to prevent a possible edge case that should not be possible in the system.

[ebezzi]

Actually nvm, this was a comment taken from the portal API. In this case, this will simply be raised if a wrong collection_id is specified (i.e., a collection_id that doesn't contain the specified dataset_id). I removed the comment.

[Bento007]

Lin 130-136 is duplicated logic from a the delete handler. We should factor is out.