DuckDB Support about electric

[thruflo]

Hey, thanks, this is a great idea. I think we have one major technical blocker atm which is that DuckDB doesn’t support triggers.

Are you aware of any mechanisms we can use to hook into writes / commits / transactions / a WAL or replication stream?

[KyleAMathews]

I'm not super familiar with DuckDB — they do have a WAL apparently https://duckdb.org/docs/sql/statements/checkpoint.html

Thinking about this more — DuckDB would perhaps be best run alongside SQLite as a read-only copy as it's intended for analytics not transactional. Potentially this could just be a userland concern — load duckdb as needed for a feature and sync data from sqlite in.

[thruflo]

Interesting

[AntonOfTheWoods]

I'm lazy... what are they using for storage with their wasm build? This is the key issue for OLAP with wasm sqlite - IDB-based storage layers are slow, and OLAP is for "big data" (right?). The sync file handle layers are super fast, but in the browser require having the sqlite engine in a web worker. That introduces quite a bit of extra complexity, particularly surrounding concurrent access, etc.

[KyleAMathews]

in 2021 at least on release they skipped on persistence — https://news.ycombinator.com/item?id=29039480 Not sure they've revisited it since

[balegas]

Turning this issue into a feature request discussion