Open chenejac opened 9 years ago
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.
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: