Alex Chi Z.
Alex Chi Z.
+1 for this. PR's welcomed!
... this also reminds me of my time at AgateDB, https://github.com/tikv/agatedb/blob/7f6603b1f1641ddbb6eb1265bbcd608368bf105c/Cargo.toml#L42-L50 Maybe we can also apply the above settings to our bench profile.
I suspect it's caused by the optimizer: ``` PhysicalProjection: InputRef #0 PhysicalSimpleAgg: count(InputRef #0) -> INT Dummy: ``` The dummy executor should not return anything. However, ``` impl DummyScanExecutor {...
Another problem is that, ``` > explain select 1 PhysicalProjection: 1 Dummy: ``` Projection is relying on Dummy executor to return a value... Maybe we should use `ValuesExecutor` for queries...
> We need #593! 🥵 Next time for sure!
I'm working on a new project https://github.com/risinglightdb/sqlplannertest-rs. With planner test, we can merge such optimizer PRs with more confidence!
Hopefully I can finish it this weekend :)
It's here! https://github.com/risinglightdb/risinglight/pull/661/
Please add planner test cases for affected queries after #661 gets merged :)
Exactly! Block cache in LSM systems is the same as buffer pool. But it's hard to implement the buffer pool in Rust using the same way as cpp (like bustub)....