EJ Song

@andrei-ionescu could you share the plan of the query? With the current version of Hyperspace, you could create these indexes for your join query: * ts => indexed column (ns,...

@imback82 After some investigation, I think it’s a little bit difficult to support the serialization of complete `LogicalPlan` across spark versions seamlessly. Here’s my analysis based on the current code...

> If we do not support, I think we should keep both Spark / Scala version since Spark 2.4 supports both Scala 2.11 and 2.12? Yes, Scala version also should...

@clee704 No, this issue is about the source plan. Currently, we only allow "Relation" type of logical plan, and it's only one relation. However, similar to "view" in db, a...

Yea, because of the compatibility issue, I replaced plan de/serialization with "Relation" approach in v0.2.0. Unused utility code is removed by #325. As there's no feasible solution for plan de/serialization...

Relation approach is not covering an arbitrary query plan (e.g. with filter conditions or .. etc) = index view. I'll update the description when I have the time.

>Whether we vacuum log files or not, we should update getLatestStableLog to handle the last vacuum operation correctly? Created a quick fix #398

> Is there a mechanism to trigger the build once more on this PR? Something like commenting a keyword to trigger the build again? You can trigger via "re-run" in...

Actually I'm waiting for #393 because of the conflict. @andrei-ionescu Could you review the change again? Since SchemaUtil is removed in that change, I wonder if it's ok or not.

@andrei-ionescu I made an issue for build failure - https://github.com/actions/virtual-environments/issues/3146 > I don't think we need another one. Or, is there a specific behaviour that you want to test? I...