search

onemoredata / bagger

Massive log storage in PostgreSQL
BSD 2-Clause "Simplified" License
12 stars 1 forks source link

Explore JSON pointers for indexing #8

Closed einhverfr closed 1 year ago

einhverfr commented 1 year ago 
          We don't need to do this now, but I'd like to expose as little postgresql internal surface as possible. Therefore I'd like to have index fields as json pointers: https://datatracker.ietf.org/doc/rfc6901/

I haven't made my mind up on how we'd solve indexes on expressions.

Originally posted by @65278 in https://github.com/onemoredata/bagger/issues/4#issuecomment-1763509219

einhverfr commented 1 year ago

My thinking here is that the tooling should probably wrap PostgreSQL internal surface stuff into Lenkwerk for maximum flexibility, and then the Lenkwerk -> storage side should do stuff with more internal surface exposed.

My reasoning here is that this provides an ability to add new functionality on the tooling side and the index creation side doesn't have to be aware of what the tooling "compiles" into Postgres index expressions.