WhereUniqueInputSchema validation behaviour

[binary-birthday]

I have a User model with two unique fields, id and phoneNumber.

When validating the input using the UserWhereUniqueInputSchema, validation succeeds in the scenario where:

phoneNumber = undefined

UserWhereUniqueInputSchema.parse({
      phoneNumber,
});

I would expect that validating using the WhereUniqueInputSchema should only succeed if at least one of the required unique fields is a valid input.

[chrishoermann]

@binary-birthday Since the generator just takes the information that is present in the prisma dmmf it creates the schemas according to the prisma types. In these types all unique keys are optional.

model User {
  id         String      @id @default(cuid()) 
  email      String      @unique 
  //...rest
}

// Type in prisma index.d.ts

  export type UserWhereUniqueInput = {
    id?: string
    email?: string
  }

// created schema

export const UserWhereUniqueInputSchema: z.ZodType<Prisma.UserWhereUniqueInput> =
  z
    .object({
      id: z.string().optional(),
      email: z.string().optional(),
    })
    .strict();

export default UserWhereUniqueInputSchema;

So I think currenly you need to check if the fields are undefined before hitting the schema. Another way could be to add a custom refine to the Input schema like

// I'm sure this check can be refined a bit :)
UserWhereUniqueInputSchema.refine((data) => {
  return Object.entries(data).some(([_key, value]) => {
    return value !== undefined;
  });
});

This is something that could also be generated automatically by the generator, since all the information should be there, but I would suggest one of the manual options from above until I can look into this in more detail.

[chrishoermann]

I couldn't get it off my mind so I implemented it straight away. Should work now in the latest version as you described it

[binary-birthday]

Oh awesome. I'll check it out now. For a workaround I thinking about using:

UserSchema.pick({ phoneNumber: true}).parse({phoneNumber})

But this will be much better.

[chrishoermann]

closing since it's now implemented

[binary-birthday]

@chrishoermann I've encountered an issue with this feature:

when UserWhereUniqueInputSchema is used in a connectOrCreate block the original behavior is the expected behavior and I think it makes this a bug rather than a feature.

I apologise for not recognising this earlier.

[chrishoermann]

Thanks for pointing this out and no problem @binary-birthday. I'll revert back to the previous implementation in the next release.