Haolin Wang
Haolin Wang
Both option A and B introduce a different behavior(or impact) comparing with the original design, although seems to be harmless. Looks like option C is between them, which limits the...
@t1mursadykov and @Sasasu Seems this PR stays for months, is it still actionable or OK to close if not ?
Thanks @soumyadeep2007 for following up this, I rebased it and will try to catchup your later comments.
> > Another problem we encountered was the poor random access > > performance of the visibility map. So we added a bitmap to > > record the tuple id...
> > > > Another problem we encountered was the poor random access > > > > performance of the visibility map. So we added a bitmap to > >...
Need to add a case to test a question from @soumyadeep2007: if the first insert operation was aborted after creating table, currently no data in visimap but has dead tuples...
> > Commit [258ec96](https://github.com/greenplum-db/gpdb/commit/258ec966b26929430fc5dc9f6e6fe09854644302) changed the format version to `AORelationVersion_PG12`, so only tables with this format version should enable index-only scan using the above logic. So, tables in-place upgraded from...
> (1) Please defer the segno optimization to a completely separate PR and remove from this PR entirely. Seg files don't have to be consulted for index only scans. Separate...