Remote Functions Output Validation

[Oshimani]

Hey everyone,

I was just wondering if remote functions should have the option to validate their output just like the input.
TRPC offers the option to validate the output object https://trpc.io/docs/v9/output-validation.

Use case:
I have a remote function that makes a http call to a 3rd party backend. I now have to assume that the type matches my expected type or I have to validate by hand (which is not the worst thing but since we have input validation, we might as well have output validation).

Looks like this:

export const getSomeThirdPartyData = command(
	z.object({
		inputParam: z.string()
	}),
	async (payload) => {
		const { fetch } = getRequestEvent();
		const { inputParam } = payload;

		const response = await fetch(`https://my-third-party-app.domain/api/${inputParam}`);
		const data = await response.json();

		return MyExpectedSchema.parse(data); // z.object({...})
	}
);

Could look like this:

export const getSomeThirdPartyData = command(
	z.object({
		inputParam: z.string()
	}),
	async (payload) => {
		const { fetch } = getRequestEvent();
		const { inputParam } = payload;

		const response = await fetch(`https://my-third-party-app.domain/api/${inputParam}`);
		return await response.json();
	},
    MyExpectedSchema // z.object({...})
);

It would default to "unchecked" so when we interact with a orm or any typesafe source, we don't need to do any additional work.

[brunnerh]

Issue that contains a proposal regarding this:

• Remote functions: provide spec-driven response contracts and mandatory guards #14612