Erik Welch

Looking more closely at this, and I think the current behavior as used by `extract_subgraph` and `edge_props_to_graph` might be correct (at least per my understanding). If there are multiple edges...

Renamed `do_expensive_check` to `check_multi_edges`. I just noticed `allow_multi_edges` is still used in `cugraph_service`. Should I take a look at this and try to figure out what to do?

@alexbarghi-nv can you comment on the test failure here? It is failing on cugraph/tests/test_pyg_extensions.py:test_get_subgraph The edge data from the PropertyGraph is: ``` _EDGE_ID_ _SRC_ _DST_ _TYPE_ 0 0 3 1...

pyg tests updated based on feedback from @alexbarghi-nv 🤞

@alexbarghi-nv I fixed test_pyg_extensions.py:test_get_subgraph, but now test_pyg_extensions.py:test_neighbor_sample_multi_vertex is failing. In this test, for `cugraph_edge_type`, the data from `pG.get_edge_data` is: ``` _EDGE_ID_ _SRC_ _DST_ _TYPE_ edge_type 0 0 3 1 cow...

Thanks all for the reviews and help. Now we're waiting on CI...

The fix to `sampling_utils_impl.cuh` in this PR appears to also fix https://github.com/rapidsai/cugraph/issues/2770

Thanks @dschult! I think this is good to go in. It shouldn't hurt anything, and it gives backends a chance to improve UX (if they return False for `should_run`, then...

> One problem is that 3.2 now introduced a new code pattern that makes type annotating much more verbose and complicated [python/typeshed#10544 (comment)](https://github.com/python/typeshed/pull/10544#issuecomment-1776036652) I created #7185 as a possible way...

> There is one signicant downside though to not having inline annotations at all. You cannot do this: > > ```python > g = nx.Graph[str]() > ``` To support that...