A place for developers to talk about building on Farcaster.

/dev

Do you prefer one, two, or a combination?

@kjs and I are designing a new TypeScript schema definition for Ponder to replace schema.graphql. This makes the TS lang server very snappy by eliminating a codegen step. Also unblocks direct database access.

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

https://github.com/o-az
—
https://x.com/awkroot

right side. Can you still `import type { Infer } ...` when using right side?

Making investing together better https://forumdaos.com | Variant Founder Fellow | Cambridge Alum | Formerly Ireland's #3 powerlifter (74kg)

Agree, one named export per table would be cleaner.

The reason we're considering this `schema` object, which isn't clear from the screenshot, is that the references (foreign key column names, enum names) have type-level validation and intellisense hints. That's not possible if each table is an isolated named export.

Left seems similar to drizzle which I have been enjoying quite a lot recently so I side with that 

Would probably also prefer even more like drizzle where you point to the schema file and each named export defines a table schema

Good idea! One hesitation is that IMO in framework design it's often better to offer only one way to do things. Makes it easier to move from one Ponder app to the next, share snippets, train LLMs 😉

Don't feel strongly that this is one of those cases, though

Why not both? Named exports for the example on the left, and a default export for the right - have your cake and eat it too 🍰🍽️

Types as strings always felt like a failure to me, right seems like the way to go