Discriminate Union with Zod and Prisma

discriminate-union-zod-prisma.md

Discriminated Union Arrays in Prisma Models

Prisma currently does not support union types for fields directly in its schema. For instance, this is not possible:

model Movie {
  id        String @id @default(uuid())
  title     String
  customers (Client | User)[] // <-- Not supported in Prisma
}

However, there are workarounds to achieve similar functionality:

---

1. Store as JSON Field

If you need a single field to hold polymorphic data, you can store it in a Json column. For example:

model Movie {
  id        String   @id @default(uuid())
  title     String
  customers Json?
}

In this case, you can store discriminated unions using libraries like Zod, but note that:

• Prisma treats it as untyped JSON.
• You lose relational features like foreign keys and constraints.

---

2. Use a Relational Approach with a Discriminator

Option 2A: Single Polymorphic Table

You can create a single table for "customers" with a type discriminator field:

model Movie {
  id        String @id @default(uuid())
  title     String
  customers PolymorphicCustomer[]
}

model PolymorphicCustomer {
  id          String @id @default(uuid())
  type        String // e.g., "Client" or "User"
  movieId     String
  movie       Movie  @relation(fields: [movieId], references: [id])
  // Additional fields for the specific types
}

Each row in PolymorphicCustomer behaves as either a "Client" or a "User," depending on the type value.

---

Option 2B: Separate Tables with a Union Table

Create separate models for Client and User, and use a bridging table for the union:

model Movie {
  id           String @id @default(uuid())
  title        String
  movieClients MovieClient[]
  movieUsers   MovieUser[]
}

// Bridging table for Clients
model MovieClient {
  movieId  String
  clientId String
  movie    Movie  @relation(fields: [movieId], references: [id])
  client   Client @relation(fields: [clientId], references: [id])
  @@id([movieId, clientId])
}

// Bridging table for Users
model MovieUser {
  movieId String
  userId  String
  movie   Movie @relation(fields: [movieId], references: [id])
  user    User  @relation(fields: [userId], references: [id])
  @@id([movieId, userId])
}

model Client {
  id     String @id @default(uuid())
  name   String
  movies MovieClient[]
}

model User {
  id     String @id @default(uuid())
  name   String
  movies MovieUser[]
}

With this approach:

• Use separate queries to fetch movieClients and movieUsers.
• Merge the results in application logic for a unified "customers" list.

---

Key Takeaways

Prisma does not directly support fields with union types (e.g., (Client | User)[]). To handle polymorphic data:

1. Use a Json field for flexibility, or
2. Use relational models with:
   • A single polymorphic table with a discriminator field.
   • Separate models and bridging tables for relational integrity.

You can then use libraries like Zod or application logic to enforce and handle discriminated unions at runtime.