Эти материалы помечены как "окончательные". Показать Ответов: 2
Could you please send a small re-producible test case to this email address alan dot wu at oracle dot com?
I tried a small case on my own and could not reproduce (maybe it is because I am using a newer version).
Zhe, thanks for all your help tracking this down.
It looks like there is indeed a problem with 10g, where triples with blank-node objects don't get deleted. This is fixed in 11g.
The children of the blank-nodes not getting deleted is the correct behavior and is the way that other triplestores behave. I tested this using Jena with memory, Hypersonic and OracleRDBMS backends, Sesame 2.0, and AllegroGraph. One of the RDF files I was using to test this lacked the children of the blank nodes, so it appeared that some triplestores deleted them automatically.