Is this supposed to prevent type errors?

[gtamas]

I added this to my project, but I'm confused.
I can see that the JSON fields are now typed correctly in the generated code @prisma/client.

However, if I call the prisma method:

someModel.create(d) where d is not the expected JSON data but some other type (or some different JSON structure), the operation still succeeds! It creates the record in the DB using the wrong data.

Probably I don't understand something. Isn't this supposed to prevent wrong data types from being used with prisma APIs?

Give these types:

declare global {
    namespace PrismaJson {
      type SomeType = { name: string; some: string }
      type OtherType = {a: number, b: string}
      type AnotherType = Record<string, unknown>
    }
  }

And this schema:

model  Some {
  /// [SomeType]
 prop1 Json
  /// [OtherType]
 prop2 Json
  /// [AnotherType]
 prop3 Json[] 
}

Then this call works:

prisma.someModel.create({
     prop1: {
        name: 'bar',
       some: 'foo'
      },
     prop2: "2" as unknown as PrismaJson.OtherType,
     prop3: "string"
    })
  }) 

What am I doing wrong?

I know Typing info doesn't exist at runtime, types live in the TS world only. So is this only for generating types for Prisma?

[arthurfiorette]

This package only connect existing types in your codebase into your prisma schema, as the readme says: ⚠️ It just changes the declaration files of your generated client, no runtime code is affected! no runtime code is actually changed.

You still need to validate it manually before inserting into your database.

This is fine because previously you should have an unknown type (which is not compatible) and after your validation you should have the correct type to insert into.