Rick Ratzel

Run release 22.08 benchmark notebook

1

The 22.08 release notebook needs to be done to capture official benchmark data.

Have `black` format cugraph code. Other RAPIDS repos have done this using a pre-commit hook (as seen [here](https://github.com/rapidsai/cudf/pull/2160)). Things to consider: * making sure `git blame` still works as expected...

Set up automated SG/MG/MNMG (based on scale and coverage type) PG tests and benchmarks

1

This will be needed for both the small and large-scale graph support in PG, and to ensure design changes produce the expected results (functional and performance) as early as possible....

The nightly MNMG test/benchmark runs include PG tests, but are not easily run by developers outside of the automated environment. This task is to get a developer set up to...

Use unit tests (initially disabled or calling stubbed out APIs in order to allow CI to pass) to define the cugraph python API for data masking.

setup.py files reference old paths, need cleanup

2

https://github.com/rapidsai/cugraph/blob/bc2242007c4b114733ca32fec68985d38bedddc9/python/pylibcugraph/setup.py#L111 is referencing `"../../thirdparty/cub"` which no longer exists. There may be others, and the `setup.py` files are in need of general cleanup.

[BUG] test_pagerank on asymmetric directed graph fails when comparing to NetX

2

From #317 : > page rank mostly works but needs some exploration. The email-Eu-core graph requires more iterations to converge. If I raise the number of iterations high enough everything...

[FIX] Triangle counting test on asymmetric directed graph produces different results than NetX

9

When the `email-EU-core.csv` dataset is used in the python triangle counting test (`python/cugraph/tests/test_triangle_count.py`), cugraph and NetX produce a different number of counted triangles and edge vals. A debug print of...

isolated vertices are not being captured properly when initializing a Graph with an adjacency matrix

4

Isolated vertices are lost when we initialize a Graph using an adj list: ``` >>> nxG = nx.from_numpy_array (np.array([[1,1,0], [1,0,0], [0,0,0]]), create_using=nx.DiGraph) >>> G = cugraph.from_numpy_array(np.array([[1,1,0], [1,0,0], [0,0,0]]), create_using=cugraph.DiGraph) >>>...

Update MG PropertyGraph tests to have coverage at least equal to SG PropertyGraph tests

1

The current [MG PropertyGraph test module](https://github.com/rapidsai/cugraph/blob/branch-22.08/python/cugraph/cugraph/tests/mg/test_mg_property_graph.py) contains only 3 tests and they're all being skipped.