Are you looking for feedback yet?

Theres also a needed rebase/merge.

@rukai , I am still trying to ensure that the changes pass the original tests or change the tests in expected ways. I will be reporting that before this gets merged. Yes, I know it need to rebase but I wanted to limit the problems to things I changed from the original code. Rebasing just adds complexity but will be done soon.

Code is complete -- or at least as complete as the original main code was.

I added cache documentation as cache.md. There is significant change to caching as it had the deepest hooks into the old SQL based parsing. All original tests pass, though some of the expected results are different.

Cache keys now comprise the fully qualified table name and the primary key. The Redis structure is now a Hash with the hash keys being the base column names (not aliases) from the query as well as any additional WHERE clause components. The items stored in the table is the serialized CassandraResult.

Cache matching then is simply looking up the Redis key table_name:key_seg1:key_seg2:...:key_segN and extracting the hash key col1 col2 col3 ... colN WHERE clause1 AND clause2 AND ... AND clauseN and then deserializing the result as the CassandraResult, setting the id and other tracking info from the result and returning it. The redis key and hash lookup occur in a single Redis call.

1 benchmark regressed. 0 benchmark improved. Please check the benchmark workflow logs for full details: https://github.com/shotover/shotover-proxy/actions/runs/2318666400

                        time:   [2.8589 us 2.8602 us 2.8613 us]
                        change: [-0.3537% -0.1421% +0.0571%] (p = 0.18 > 0.05)
                        No change in performance detected.
transform/cassandra_rewrite_peers_passthrough
                        time:   [13.910 us 13.919 us 13.927 us]
                        change: [+69.280% +69.464% +69.653%] (p = 0.00 < 0.05)
                        Performance has regressed.

transforms.md still describes the cache as a "write through cache", it should be changed to describe it as a "read behind cache"

Everything from this PR has been landed in #661