How does columnar = json? json isn't colunar at all... If you just want to have a schema in json instead of sql, use a no-sql db, postgres nosql features are strong, but the db features are actually much stronger.

json isn't necessarily columnar, but it is a natrual fit for stuff that is columnar that's otherwise harder to model in a traditional relational db

here's my usecase:

- we have a bunch of attributes (all different names by customer, and many different values for each record that a customer stores)

- it's a fairly natural fit for a json value with only one level of key: value mapping

- we use mysql on GCP (no columnar plugins, too hard to switch to postgres)

Someone could go back in time and correctly model it as columns and not json but that ship has sorta sailed. While it's not impossible to change, it would be pretty hard, time will tell if that ends up happening.

I would love to be able to tell mysql "put this column in a collumnar engine and use that when I query on it" (AlloyDB is this for postgres on GCP)

Too bad mysql doesn't even have materialized views - its not a good combination with a json only situation, but you clearly are suffering through that, good luck :) probably etl it to clickhouse and move on.

Could always try and dump to GCP/Parquet/json and use duckdb/bq to just query it all.

great point.

with pg_mooncake v0.2 (launching in ~couple weeks), you'll be able to get a columnar copy of your Postgres that's always synced (<s freshness).

Keep your write path unchanged, and keep your Postgres where it is. Deploy Mooncake as a replica for the columnar queries.

I wish, someone unfortunately picked mysql