feat: support more validation libraries

[juliusmarminge]

Instead of

createEnv({
  server: {
    KEY: z.string(),
  }
});

it should be

createEnv({
  server: z.object({
    KEY: z.string(),
  })
});

and it should not be tied to zod, meaning something like scale-ts could work for example: import * as $ from 'scale-codec';

createEnv({
  server: $.object(
    $.field('text', $.str)
  ),
});

this would require some more considerations for typings as we'd need to pull out the shape of each to get access to the raw keys to compare against the client prefix

Should probably implement #169 first in some half-generic way 🤔

[chungweileong94]

Probably need an adapter for each validator, for couple of reasons:

• To tell the core how to extract the typing, mainly for validating client prefix, and type infer in general
• To tell the core how to parse the env values. As we call .safeParse in zod, but we need to do it in different way for other validator.

[juliusmarminge]

We support a bunch of validators in tRPC:

https://github.com/trpc/trpc/blob/main/packages/server/src/core/parser.ts

https://github.com/trpc/trpc/blob/main/packages/server/src/core/internals/getParseFn.ts

But separate entrypoints could also work i guess:

import { createEnv } from "@t3-oss/env-core/zod";
const env = createEnv({
  // ...
  server: {
    FOO: z.string(),
  }
})

import { createEnv } from "@t3-oss/env-core/scale";
const env = createEnv({
  server: $.object(
    $.field('FOO', $.str)
  ),
});

[taylorallen0913]

Can I look into integrating Yup as an option alongside Zod?

[juliusmarminge]

sure!

[aalakoski]

Will it also export zod / other validators from within t3-env or would you still need to add zod as a dependency? If not, could it be done?

Something like:

import { createEnv, z } from "@t3-oss/env-core/zod";

const env = createEnv({
  server: {
    FOO: z.string(),
  }
})

[chungweileong94]

@aalakoski Most probably no.

[SeanCassiere]

Tanner's pattern of validating the router search params, could be something to consider here.

The process in t3-env would look like this.

1. Make the client and server keys in the init of createEnv accept validation functions, that receive the values of process.env (or import.meta, ... etc.) as its argument.
2. Then call the user-defined client and server validation functions to the ensure the values are adhere to the schema.

Code-example with zod:

import { createEnv } from "@t3-oss/env-core";
import { z } from "zod";

const env = createEnv({
  server: z.object({ FOO: z.string().min(1) }).parse,
  client: z.object({ BAR: z.string().min(1) }).parse
});

Code example with a custom validator

import { createEnv } from "@t3-oss/env-core";

// custom validator which could be swapped for Yup, Joi, etc....
type ServerVars = { FOO: string }
function serverValidator(values: any): ServerVars {
  // validation magic
  return { FOO: "FOOOOOOOOO" }
};

type ClientVars = { BAR: string }
function clientValidator(values: any): ClientVars {
  // validation magic
  return { BAR: "BOOOOOOOOO" }
};

const env = createEnv({
  server: serverValidator, // (values) => serverValidator(values)
  client: clientValidator // (values) => clientValidator(values)
});

This pattern completely opens it up to the user to bring their own validation libraries and logic into the picture, with t3-env making use of the type-safety returned by the defined validators as well as the libs' client-server separation stuff.

[juliusmarminge]

Will it also export zod / other validators from within t3-env or would you still need to add zod as a dependency? If not, could it be done?

Something like:

import { createEnv, z } from "@t3-oss/env-core/zod";

const env = createEnv({
  server: {
    FOO: z.string(),
  }
})

No - we won't bundle the validator

[Bekacru]

why does it need validation library tho isn't is simple enough task to do plainly? what am I missing?

[mattddean]

Did a very naiive valibot replacement. Obviously it doesn't accomplish the goal here of not shipping a validation library, but it might be useful to someone. https://github.com/mattddean/t3-env-valibot/tree/feature/valibot

[kaptinlin]

any plan support more validation libraies? i want use valibot instead of zod

[estubmo]

Did a very naiive valibot replacement. Obviously it doesn't accomplish the goal here of not shipping a validation library, but it might be useful to someone. https://github.com/mattddean/t3-env-valibot/tree/feature/valibot

This seems to be no longer working, as parse is no longer a function of schemas.

[stunaz]

why not using the patterns as in https://github.com/react-hook-form/resolvers/tree/master ?

[chungweileong94]

why not using the patterns as in https://github.com/react-hook-form/resolvers/tree/master ?

Yeah, maybe we have to, as Valibot provides a separate parse method instead of from the scheme. Of course I also aware that they do have method from the scheme called _parse, but it's for internal use only, and I heard that it only provides some raw behavior.

[AlexisWalravens]

I just stumbled upon this issue, and maybe this could be useful https://typeschema.com ? next-safe-action just implemented it in it's last version, allowing people to use almost every schema validator.

[Talent30]

I just stumbled upon this issue, and maybe this could be useful https://typeschema.com ? next-safe-action just implemented it in it's last version, allowing people to use almost every schema validator.

but it only supports async parse

[decs]

would you guys be open to making createEnv an async function? there are some considerations regarding top-level awaits, but maye we can figure something out

[juliusmarminge]

I think we'd make a createEnvAsync in that case

[mwskwong]

why not using the patterns as in https://github.com/react-hook-form/resolvers/tree/master ?

Yeah, maybe we have to, as Valibot provides a separate parse method instead of from the scheme. Of course I also aware that they do have method from the scheme called _parse, but it's for internal use only, and I heard that it only provides some raw behavior.

+1 for this pattern. It also opens for opportunity for contributors to contribute to more resolvers for various schema validation libraries, once the API of the resolver has been standardized.

BTW, is this the next big thing the team is planning? Is there any rough target date for making it happen? I would be happy to contribute (I really want a Valibot resolver) if this is the route the team is going.