remove values in runtime output like zod safeParse

[isimmons]

Is there a way with Arktype to remove values where they are omitted in the type or plans to implement a parse function similar to zod's safeParse?

If I use safeParse like this it removes the data completely from the parsed/output data In this example the content in the browser will not have the "private" field at all assuming I have omitted it in the schema .

import { userSchema } from 'schemas/user.ts'
import { safeParse } from 'zod'

//...

const parsed = userSchema.safeParse({
  name: "Fred",
  email: "[email protected]",
  private: "something private"
})

With Arktype I can omit the "private" field and get a typescript error in my IDE but if I run dev the actual data is still there in the browser.

The best I can find so far is to use the .narrow() method and my own doesObjectContainKey function to make it throw a runtime error if it receives a user with the "private" field included.

So do I need to create my own parse method to remove the field if it exists or is there something like that already baked in to Arktype?

Thanks

[TizzySaurus]

Hi @isimmons, it looks like what you're after is onUndeclaredKey:

const userSchema = type({
    name: "string",
    email: "email"
}).onUndeclaredKey("delete");

const parsed = useSchema({name: "fred", email: "[email protected]", private: "something private");
console.log(parsed); // doesn't have the "private" key

This can be found on the docs here: https://arktype.io/docs/type-api

There's also a shorthand where instead of .onUndeclaredKey("delete") you can do "+": "delete" within the schema itself, i.e.

const userSchema = type({
    name: "string",
    email": "email",
    "+": "delete"
});

And this can be configured per-type as shown above, or if you want this to apply to all created types by default, you can configure onUndeclaredKey globally, as per https://arktype.io/docs/configuration#onundeclaredkey:

import { configure } from "arktype/config";
configure({ onUndeclaredKey: "delete" });

// NB: The core arktype import must be AFTER the config has been set
import { type } from "arktype"'
const userSchema = type({
    name: "string",
    email: "email"
});

Hope this helps!

[isimmons]

Yes this helped a lot. I looked at onUndeclaredKey as a configure option but I was doing something wrong somewhere. I also missed that it can be a chain-able method on the schema. I wanted to do something inline rather than global but without a lot of code and this works.

Zod vs Arktype same result:

Zod

const getUser = async () => {
  return {
    id: 1,
    username: 'Zod User',
    private: 'My private data',
  }
}

// other code...

 const [user, setUser] = useState<ZodUserType | null>(null)

  useEffect(() => {
    const getTheUser = async () => {
      const data = await getUser()

      const user = zodUserSchema.omit({ private: true }).parse(data)

      setUser(user)
    }
    getTheUser()
  }, [])

  console.log('Zod User: ', user) // {id: 1, username: 'Zod User'}

Arktype

const getUser = async () => {
  return {
    id: 1,
    username: 'Arktype User',
    private: 'My private data',
  }
}

// other code...

  const [user, setUser] = useState<AtUserType | null>(null)

  useEffect(() => {
    const getTheUser = async () => {
      const user = atUserSchema.omit('private').onUndeclaredKey('delete')(
        await getUser(),
      )

      if (user instanceof type.errors) {
        console.error(user)
        return
      }

      setUser(user)
    }
    getTheUser()
  }, [])

  console.log('AT User: ', user) // {id: 1, username: 'Arktype User'}

Thanks

[ssalbdivad]

Thanks for the assist @TizzySaurus.

Quick note- creating a type is expensive, checking data is cheap.

I'd recommend moving atUserSchema.omit('private').onUndeclaredKey('delete') out of useEffect to initialization logic.