Hey daft team! Crazy use case here: I was wondering what the interface or compatibility is between Daft and cuDF is? I'm looking at several of NVIDIA's recent AI libraries like morpheus and NVIngest which use cuDF extensively and also have a use case for networkx which has a cugraph implementation. cugraph supports construction with

G = cugraph.from_cudf_edgelist()

(see https://docs.rapids.ai/api/cugraph/stable/api_docs/api/cugraph/cugraph.graph.from_cudf_edgelist/) , and I know you guys have support for CUDA operations, but not sure what the workflow would look like. Perhaps using pandas as the intermediary?

Yeah you should be able to convert a Daft dataframe to pandas via

<http://df.to|df.to>_pandas

, and then do `from_pandas_edgelist`https://docs.rapids.ai/api/cugraph/stable/api_docs/api/cugraph/cugraph.graph.from_pandas_edgelist/ to convert the pandas df to a cugraph

Cool