A place for developers to talk about building on Farcaster.

/dev

Ponder

Should Ponder replace SQLite with PGlite (pglite.dev) as the dev database? 

Pros:
1) Less magic/astonishment ("wait, what database am I using rn?")
2) Simpler schema definition API (currently stuck with the subset of stuff that both PG and SQLite support)

Going all-in on Postgres was the single best decision we made in 2024.

Much simpler, with no regression to the quickstart / local dev experience (thanks to PGlite). Obvious in retrospect.

Building ponder.sh. EVM, TypeScript, open-source enjoyer.

🦀 /foundry @paradigm | founding engineer /ithaca

Big fan of ClickHouse, it's unbeatable for some use cases like structured logs.

It would be a bad fit for Ponder's user-facing database, though. Indexing workloads can be OLTP-y and do many key-value / unique index lookups, which are anathema for a column-oriented database.

Reads: Most people use Ponder for use cases where queries are expected to be fast and represent the latest state. OLAP databases are typically too slow.

Writes: Ponder apps convert raw EVM data (logs, storage, traces) into an app-specific data model. It's like a big incremental materialized view. To achieve the "incremental" part, it's convenient to author bits of logic that read and write individual rows at a time. OLAP databases are generally bad at this.

(complete data noob here) I've been curious about this. My impression was usually that indexed data is more or less read-only, and an OLAP db might fit well as a data source. Can you explain more what you mean by the OLTP-ness of indexing workloads?

Good point, I was wondering because I had read that Nansen and Goldsky are using it in their infrastructure but seemingly in an OLAP way

Good point, I was wondering because I had read that Nansen and Goldsky are using it in their infrastructure but seemingly in an OLAP way

https://clickhouse.com/blog/clickhouse-redpanda-architecture-with-goldsky

https://clickhouse.com/blog/unlocking-the-power-of-onchain-analytics-how-nansen-transformed-their-data-infrastructure-with-clickhouse-cloud