What's going on here?

[pydantic]

What's going on here?

[PrettyWood]

Hi and thanks for the invitation!

IMO keeping it private for now is for the best: less pressure, not too much feedback, ...

First and most importantly I feel like it's a wonderful idea and already a great start! Splitting the code base of pydantic between pure validation and serializers, json schema, settings... is needed and I'm glad to see rust used for pydantic-core like so many great high performance libraries (orjson, tokenizers...)

I was having a look and I really love this idea!
I see you're already implementing it 🚀
I started working a while ago on a POC of tydantic (it was actually the name I took too 😆) but it was not based on a rust + json core, which feels way better.

I will take some time to review the code you already wrote and maybe write some kind of UML diagram to have a better vision of the whole implementation. Maybe I missed something but I was surprised to see nothing on strict checks. IMO it could be great to have something a bit like this to allow also strictness and custom coercion! (with pydantic declaring its own coercers)

[samuelcolvin]

Thanks @PrettyWood, great to hear you think it's a good idea.

currently the implementation of #10 won't allow removal of python since only the input types are generalised, not the output types, but hopefully that can be change in future.

The problem I have right now is that parsing JSON inside pydantic-core is currently a lot slower than parsing json with ujson.loads or even json.loads then passing the output to pydantic-core - I'm looking into why that is now.

[samuelcolvin]

Forgot to say, strict types should be easy to implement. I've already altered some of the allowed types to make checks stricter should should improve usage, e.g.:

• floats are only allowed as inputs to int fields if f % 1 == 0
• lists of pairs won't be allowed as input to dict types