Group TripleMaps for Performance

[dachafra]

issue: running each source row through each TripleMap may be slow because it means "rewinding" the database and processing each row multiple times

suggestion: process all TripleMaps with the same (or equivalent source) for each input row

link: SDM-TIB/SDM-RDFizer#24 (comment)

[marioscrock]

I think this is more a tool-related optimization, however, I would suggest considering also joins conditions in the process of grouping TripleMaps for performance. This type of analysis (considering join "dependencies" among TripleMaps) could also help in enabling efficient concurrency in processing groups of not-constrained TripleMaps.

[dachafra]

This issue is out of the scope of the CG, as it's specific for the optimizations of each tool. Closing it

[bjdmeest]

Discussions like this would be nice for a 'Best Practices' note or smth, do you think we should keep tabs somewhere? eg Yet Another Repo? 😅 (bc I remember implementation considerations at our slack channel too)

[dachafra]

I agree that we can generate a Best practices repository where we collaboratively post well-known optimizations or techniques to speed up the materialization process. @bjdmeest would you mind to create that repo under the organization?

[bjdmeest]

Done at https://github.com/kg-construct/best-practices, with a first issue! :D

[dachafra]

re-open it in the correct repo