A place for developers to talk about building on Farcaster.

/dev

Anyone have experience running their own time series database for exposing analytics back to customers (i.e. not purely for internal use). 

Thinking about giving frame developers much more detailed analytics on their frames, and exploring one of a few options.

Building @warpcast and @farcaster. Previously Coinbase, Brigade, Causes.

I've had a solid experience with Timescale so far

Biggest gotchas come from ensuring the data model is correct. Time series works especially well when data is inserted, never updated (event stream)

For exposing to others though, materialized views with well defined query patterns work great (where Timescale shines)

Building Data 3.0 @ indexing.co; @orangedao W24; ex @coinbase; would rather be /running

Yes! One of my favorite parts is Timescale is just a bunch of extensions atop Postgres. So at the core, you get all the fun (and relational performance) of Postgres

It's been far, far better for us than trying to mix something like ClickHouse with a relational dataset

Are you storing your OLTP (e.g. relational) data in Timescale as well? Or only using it for these kinds of materialized view queries (which do look awesome). 

Curious if it becomes essential to colocate the data so you can perform more interesting joins.