Stream topology endpoint

[soerenreichardt]

No description provided.

[soerenreichardt]

[Topology] Add stream topology arrow endpoint

[knutwalker]

Overall, looks good.
I have one remark about the default type, Github says there is a conflict, and TC says

graphdatascience/query_runner/arrow_query_runner.py:123: error: Argument 3 to "_run_arrow_property_get" of "ArrowQueryRunner" has incompatible type "Set[Any]"; expected "Dict[str, Any]"

[knutwalker]

Using mutable values as default is a common trap in Python, as the materialized default value is the same shared instance, not a new list. The common pattern is to use something like

Suggested change

	def stream(self, G: Graph, relationship_types: Strings = ["*"], **config: Any) -> DataFrame:
	def stream(self, G: Graph, relationship_types: Strings = None, **config: Any) -> DataFrame:
	if relationship_types is None:
	relationship_types = ["*"]

though maybe that needs an update to the type.

[soerenreichardt]

Good point, didn't know that. I think that pattern is used across the codebase. Maybe we should create a card to fix all locations

[knutwalker]

Approving, waiting for whatever CI complains about :)

[FlorentinD]

we did not depend on the neo4j ids before. As we are testing multiple version combinations on the same neo4j db, these fail here.

@soerenreichardt Instead we should get the neo-ids via a Cypher query and uses these for the expected ids

at least thats an idea

[FlorentinD]

waiting for fix on server-side to be merged

[soerenreichardt]

Yes that should be the reason for the flaky tests here