@kirby @dcc @fiore @snacks The JSON for Pleroma is stored as a binary format, there's no need for parsing it once it is in the DB. The overhead is at insertion-time. Once it's in the DB, Postgres has it's own representation of it, that it can understand without re-parsing it all the time. It does not store whitespace in the JSON itself, it does not store duplicate keys in the JSON.

To skim over the details, it's basically stored in the already parsed representation. To simplify even more, imagine Postgres ran jq on the JSON and stored the internal representation of the JSON made by jq.