fact_store's lifecycle?

[guillaume-duong-bib]

I'm still trying to figure out #2930, and I noticed that every fact I ever created, directly through a Fact Source, or indirectly through a link's parser, is stored in fact_store (after some tests, only the facts that have been "loaded", e.g. through an operation, are in this case).
Even though 99% of the "sources" of these facts, i.e. either operations or Fact Sources, have been deleted since.

I see no call to _delete_fact from base_knowledge_svc.py, so it makes sense that they remain, but I would expect these facts to disappear when their source disappears, e.g. when I delete a Fact Source or an operation.

Is this a desired behavior, or an oversight? If this is desired, what are the reasons for it?

[clenk]

It looks like an oversight to me.

[guillaume-duong-bib]

Thanks for your answer.
So can I work on a PR for the issue on #2930 plus these, assuming that:

• deleting a Fact Source should delete all the facts originating from that Source from the fact_store? (I'll check again that they are indeed not removed already)
• deleting an operation should delete all the learned facts of that operation from the fact_store? (same note)

[guillaume-duong-bib]

(And while you're there, any idea on whether #2930 is a mistake on my end or indeed a bug of the same kind?)