I'm sorry but I've had to revert this PR in preparation for release. I noticed documentation generation failures for sparql-gremlin which pointed to a problem with AND. When you had said that AND wasn't optimized I didn't take that to mean that it wasn't working at all and clearly we didn't have tests in place for that. I've added a test along with the reverting of the change:
https://github.com/apache/tinkerpop/commit/d67f7c9c757bfeeaedec4d6650640c1cc4b0aed2 I did try to fix the problem first prior to reverting, but given how close we are to release (Monday) I didn't think I was going to easily solve the problem so it was simply safer to revert. In doing that investigation though I do now better understand what these changes were doing. As I wrote in the commits on the above commit, It made me wonder if we should try to simply optimize `match()` better so that it gives providers a better chance at using an index...in that way we get improvements to `match()` for native usage and to sparql-gremlin and other language compilers. As I mentioned previously, we'll create new JIRA issues as needed for these improvements going forward. [ Full content available at: https://github.com/apache/tinkerpop/pull/1237 ] This message was relayed via gitbox.apache.org for [email protected]
