Robert Muir

dude, that's nothing compared to the cost of disabling bulk merge, which will also force all stored fields to be recompressed every merge as well. sorry, i'm completely against this...

it uses `instanceof`: https://github.com/apache/lucene/blob/main/lucene/core/src/java/org/apache/lucene/codecs/lucene90/compressing/Lucene90CompressingStoredFieldsWriter.java#L686 there's nothing to "do about it", except to say, don't wrap entire codec if you want to add a stats interface. Add a stats interface if...

> I like the idea of making `SegmentMerger` implement some sort of a Stats interface that can return a metrics object. We can populate the TimeMetric object in `SegmentMerger.merge()` as...

If the user has java 16 or above, and we pass `-XX:+UnlockDiagnosticVMOptions -XX:+DumpPerfMapAtExit -XX:+PreserveFramePointer` to the JVM, we should get java symbol resolution from `perf` subsystem? Will it help this...

quoting from Uwe: > All files are write-once so there are no updates to files which were written before. This is the key piece that I think you are missing....

personally I'm not a fan of the api change on LongValues and the additional complexity in DirectReader just to read blocks of 2 values at a time. I'd be more...

Might be easiest to prototype a new DocValuesFormat (I would fork the existing Lucene90 as a start)? That's how I would attack the problem. Then I'd try to remove (at...

This seems like it is worth investigating! We just try to speed up the "addressing" without dealing with all the rest of docvalues (which may be much more complex change).

> I like this change, it's a net win that is not only going to speed up binary doc values but also sorted set and sorted numeric doc values, which...

> Hmm, we added multiple JVMs long ago precisely because HotSpot was so unpredictable. I.e. we had clear examples where HotSpot would paint itself into a corner, compiling e.g. readVInt...