VIVO-1058: Does the SimpleReasoner remove any inferred triple that you later assert?

[chenejac]

Jim Blake (Migrated from VIVO-1058) said:

Joe McInerney says that this happens: He asserts s => p => o and o => pi => s, where pi is the inverse of p. Afterwards, he finds that o => pi => s exists both in kb-2 and in kb-inf.

I have not run an experiment to verify that this happens, but I wonder whether this scenario is reasonable:

• the first triple is asserted in kb-2, the reasoner infers the second and writes it to kb-inf
• the second triple is asserted in kb-2, the reasoner and infers the first. However, it does not write it to kb-inf, since it already exists in kb-2. But does it also check to see whether the second triple is already in kb-inf, and delete it as redundant?

[chenejac]

Brian Lowe said:

Yes, it's supposed to remove any redundant triple from kb-inf if the same thing is later asserted in another graph. If it isn't doing this, there's a bug.

I thought Stella's unit tests covered this case, but perhaps not for inverses, as that feature was added some time later.