Julian Hyde

For what it's worth, I asked some questions on druid-dev and several people suggested docker. By that time I had invested effort in vagrant. We wouldn't be able to use...

Getting the software is easy. The hard part is loading the data (i.e. starting an indexing task). The off-the-shelf version doesn't have the data.

I don't feel strongly about it. The VM is working fine for my purposes (albeit using a little more memory than I'd like, due to Druid) but moving to Docker...

I feel your pain. I can use the VM on my new laptop with 8GB RAM but it was not possible on my old laptop with 4GB RAM.

+1 to remove `@Beta`. Apache Calcite, as of [release 1.26](https://issues.apache.org/jira/browse/CALCITE-4173), relies on `RangeSet` (and it appears in the Calcite's public API in one place, `class Sarg`). We are prepared to...

-1 This PR is just a test case (that currently fails) and should not be checked in as it stands. The added tests should be in a method named `testIsNull`...

Is there a corresponding JIRA case?

I am not sure we need this. I am not denying that the fix works. But something similar could be accomplished in client code without any changes to Calcite. Without...

> I'm sorry that the data type in Calcite is not pluggable now I’ll modify that statement slightly. I’m sorry that data types are not *fully* pluggable. But we’ve done...

Has this issue shown up empirically? The stack is implemented as an `ArrayDeque`, so push and pop should be very few machine instructions. `RelShuttleImpl` is not the only possible implementation...