Zac Blanco

@ThePantsThief I think if you wanted to graph a summation you can try this: [https://www.desmos.com/calculator](https://www.desmos.com/calculator)

Don't forget to squash your commits

> I think we need to document the serialized binary format, similar to T-Digest and QDigest. The serialized binary format is specified by the Apache Datasketches library. This is already...

The binary format is not documented unfortunately. However, all versions (cpp, java, go, etc) of the datasketches libraries can serialize and deserialize the sketches. e.g. java can deserialize cpp serialized...

Just for some continuity on the discussion of sketch serialized format compatibility, we brought this to their community in hopes to clarify the guarantees that the maintainers provide. This resulted...

I'm not familiar with the code for `approx_distinct`, but if the KLL sketch answers the same question as the Q-digest (e.g. what percentile does value X fall? or the inverse),...

A few TODOs just to give an idea of where I want to improve this PR: 1. ~Add a configuration property to choose to use histogram calculation over the previous...

The header check is failing on `presto-benchto-benchmarks/src/test/resources/sql/presto/tpcds/q85.plan.txt`. This file never had a license header and is only used for checking test output. I think the check can be safely ignored.

I ran a quick experiment using TPC-DS SF1 on a real cluster to verify the performance does not degrade.  I've also provided the raw data for these plots [tpcds-sf1-results.tar.gz](https://github.com/prestodb/presto/files/14149532/tpcds-sf1-results.tar.gz)

> ... I think the [header check](https://app.circleci.com/pipelines/github/prestodb/presto/12906/workflows/4640dbd8-6c09-4256-9ec4-6a535fd71310/jobs/49822) is failing again btw I reverted the commit that originally fixed the header check. #21875 was filed for it. Deepak said he would...