[flashgordon]

I really like the idea of it. My dream has always been to work with "types" first and foremost across any and all languages (yep it is a dream). And little tools like these are really nice to see push that boundary.

One feedback - if you are truly comparing with "other" tools - you should be looking at grpc and protoc plugins. I have used to great effect for things like:

1. Generating wasm bindings for grpc services

2. Generating "data access layer" types so you can choose how a api proto is transformed to a data domain type and vice versa

3. MCP bindings for APIs

4. GraphQL/BFF bindings with multiple services

5. All of the above "across" langauges.

The tooling is fantastic and extensible - if you are ok to start with a proto view of your world - it sounds wierd and like an N+1 problem but once you are used to it it is surprisingly fun (ok we may have different ideas of fun)

[emyrk]

I totally agree a proto first approach to your types can pay back in dividends if you need to serialize over different wires.

This project admittedly was developed to solve a specific need in an existing codebase with a lot existing types.

The codebase is also mostly maintain by the backend Golang engineers. Letting them use their native type system increases adoption and buy in.

[flashgordon]

Totally - The other really nice thing about Golang "type-system" ecosystem is their native ast in the stdlib. You can do so many amazing things from there. Infact if you pledge your life to Go (which I think I have atleast for now) starting from Go and generating everything outwards is not necessarily a bad strategy.

[goldenCeasar]

On this “types first across languages”, I’ve been hacking on something in that vein called Kumi (typed calculation schemas compiled to Ruby/JS). Tiny demo here https://kumi-play-web.fly.dev

[flashgordon]

Hot damn. Id love to hear the origin story of this.

[c-hendricks]

GraphQL has been my holy grail for this. Easy to grok, easy to build types for various languages and keep everything in sync.

[ashishb]

GraphQL is awesome if you are a frontend engineer. GraphQL is terrible if you are a backend engineer - https://bessey.dev/blog/2024/05/24/why-im-over-graphql

[chrisweekly]

Type-first is cool. But I think I'll always aim to avoid gRPC, at least in part because grpc-web was so completely broken. I also have an instinctive aversion to binary formats. YMMV, just my PoV.

[jsunderland323]

I’ve had a lot of success with grpc web. Had to patch a couple of things along the way. My biggest misgiving is thinking having bigints would be a good idea (it is not a good idea). Aside from that though, I’ve been happy with it. What felt broken to you?

[flashgordon]

+1 Couple of things I really hate about proto - No generics/templates. No composition of services or mixins (you do have composition in messages but that feels very limited). Also the clunkiness around having to declare more things (try a repeated map field or a map of repeated fields).

My comment about protos was just the spec (and was seperating the binary formats as a different concern). But your concerns are pretty valid.