Bug: Exception Error trying to use Page_Rank algo!?

[wingraver]

Ladybug version

v0.18.0

What operating system are you using?

Windows 10

What happened?

Going through the "Getting started -> Run graph algorithms" documentation, and after loading the "algo" extension successfully... Calling the page_rank algo on the projected subgraph via the connection in python results in a "RuntimeError: Caught an unknown exception!". At this point the nodes and their page_ranks should have been passed to a results (res) object. The error appears to come from lines 546 and 460 of the connection.py file of the ladybug source???

Are there known steps to reproduce?

Following the steps outlined in the documentation (Getting started) for running graph algorithms... the section "Run PageRank" fails with this unknown exception.

[adsharma]

There is another (preferred) way to run pagerank:

https://huggingface.co/datasets/ladybugdb/wikidata-20260401/blob/main/pagerank.py

It involves using nodejs or python bindings. I understand it'd be convenient to use the built-in algo variant. We could support an icebug UDF in the future and/or fix these types of issues. Just be aware that right now icebug via bindings is a higher priority.

[wingraver]

There is another (preferred) way to run pagerank:

https://huggingface.co/datasets/ladybugdb/wikidata-20260401/blob/main/pagerank.py

It involves using nodejs or python bindings. I understand it'd be convenient to use the built-in algo variant. We could support an icebug UDF in the future and/or fix these types of issues. Just be aware that right now icebug via bindings is a higher priority.

Hi @adsharma, thanks for the quick feedback, appreciated. The alternative approach is appreciated however it also runs counter to what is effectively a broken feature of the solution... I can use networkx to do the pagerank as outlined in the documentation. So I was raising this as other users might also experience the issue with similar expectation of using the inbuilt capability.

[adsharma]

For large graphs, icebug is significantly faster and more efficient than networkx due to the use of Apache Arrow memory for representing the graph and a fast single copy path from CSR rel table on disk -> CSR in arrow memory. In fact the performance increase is so significant that "built-in" provides no advantage.