This repository was archived by the owner on Oct 18, 2023. It is now read-only.

Description
After each query, we collect information about its rows read and rows written, and use it to bump our global metrics. We should also consider extending our protocol so that we return those values to the user as well. That would help users optimize queries, since they'd be able to compare which queries cost how many units.
That would make responses larger though, so it can affect throughput, esp. for small responses, so perhaps this mode should be opt-in, used only for query planning and debugging.