Scott Carey

Is it possible too use the fork/join pool to do the cleanup? Probably not, WeakReferenceQueue blocks. There are some other ways to do the bi-modal behavior, you can have both...

Try as I might, I can not get push to work. I have tried setting all timeout values to 900s, and disabling buffering using the fix in #125 and rpardini/docker-registry-proxy@sha256:9c3df0f61f4fe377328a2111ee72f291513f7723ceb6fa943e534a62ba1d0d7f...

Is there going to be any replacement for the sort of Unsafe ByteBuffer manipulations done here? VarHandles perhaps?

The two stack traces in the failed build make no sense at all. A byte array's length field is being passed in to another byte array initializer, which gives a...

I take that back, I can see that these tests have a ~ 4% or so chance of failing due to the test set-up.

What about compressing the .index and tombstone files? If I understand correctly, these are not actively read by queries, sequentially during compaction and initialization. Index data is a large space...

> index and tombstone files are not major factor of disk space consuming. Ok, well I've got one data set right here with a 870MB Index files and a 904MB...

Would you accept a new feature that helps other use cases if it doesn't help yours that much? FWIW, HaloDB is killing every other DB engine except in: 1. disk...

@wangtao724 After more in depth review of the code I have several optimizations I'll discuss in other issues.

I have a PR available now (PR #49) The solution I came to for the SegmentWithMemoryPool differs from my ideas above. I used the existing chain structure within the slots...