Value types

It's common to reuse a GraphQL type across multiple subgraphs. For example, you might define a generic Error interface and an ErrorCode enum:

1interface Error {
2  code: ErrorCode!
3  message: String!
4}
5
6enum ErrorCode {
7  UNKNOWN_ERROR
8  NETWORK_ERROR
9  AUTHENTICATION_ERROR
10  SERVER_ERROR
11}

These value types don't "originate" in a particular subgraph. Instead, all of the subgraphs that define a value type share ownership.

Most importantly, a value type must be defined identically in every subgraph that defines it. Otherwise, your federated schema will fail composition. Not every subgraph needs to define a particular value type.

Any of the following can be a value type:

• Scalars

• Objects

• Interfaces

• Inputs

• Enums

• Unions

Specific considerations for each of these are listed below.

Type-specific considerations

Scalars

If you define a custom scalar as a value type, make sure all of your subgraphs use the exact same serialization and parsing logic for it. The schema composition process does not verify this.

Objects, interfaces, and inputs

For types with field definitions, all fields and their types must be identical (including nullability).

If an object type is an entity, it cannot be a value type.

Enums and unions

For enums and unions, all possible values must match across all defining subgraphs. Even if a particular subgraph doesn't use a particular value, that value still must be defined in the schema. Otherwise, your federated schema will fail composition.

Modifying a value type

As mentioned, a value type must be defined identically in every subgraph that defines it. This means that if you want to modify a value type's definition in one subgraph, you must make that same modification in each other defining subgraph. Otherwise, a composition error occurs.

While you are actively deploying your updated subgraphs, your value type's definition will temporarily differ between subgraphs. During this time, GraphQL operations might produce validation errors.

If you aren't using managed federation, you must wait until all of your defining subgraphs are deployed with the updated value type before you restart your gateway to compose the new supergraph schema.