Multi Modules codegen
For multi-modules projects, Apollo Kotlin enables you to define queries in a feature module and reuse fragments and types from another module dependency. This helps with better separation of concerns and build times.
.graphql
operations in different modules. If all your .graphql
files are in a single module, you can use apollo-runtime
like any other Kotlin dependency without any of this.Setup
Multi-modules requires that one and only one module contains a schema. This is the schema that all other modules can reuse. In this document, we'll refer to this module as the "schema module".
In your schema module, opt-in multi-module by generating metadata for use by downstream feature modules:
1// schema/build.gradle.kts
2apollo {
3 service("service") {
4 packageName.set("schema")
5
6 // Enable generation of metadata for use by downstream modules
7 generateApolloMetadata.set(true)
8
9 // If you need to specify the location of your schema files
10 schemaFiles.from(/*...*/)
11
12 // Scalar mappings and generateDataBuilders must be defined in the schema module
13 mapScalar(/*...*/)
14 generateDataBuilders.set(true)
15
16 // Other options can be different between modules.
17 // If you want those options to be applied to all modules, use a convention plugin and shared build logic.
18 useSemanticNaming.set(true)
19 generateFragmentImplementations.set(true)
20 }
21}
In your feature module, declare your schema module as a dependency:
1// feature/build.gradle.kts
2apollo {
3 // Service names must match
4 service("service") {
5 packageName.set("feature")
6
7 // The 'feature' depends on the 'schema' module
8 dependsOn(project(":schema"))
9 }
10}
Auto-detection of used types
By default, Apollo Kotlin generates all the types in your schema module. This is because there is no way to know in advance what types are going to be used by feature modules.
For large schemas, this can generate a lot of code and increase your build time significantly. In this case, you can opt in auto-detection of used types.
This works by splitting the codegen task in different steps so that feature modules can let the schema module know what types are used before actually generating the models. To opt in, call isADependencyOf()
to establish the bidirectional dependency between your feature module and your schema module:
1// schema/build.gradle.kts
2apollo {
3 service("service") {
4 packageName.set("schema")
5 generateApolloMetadata.set(true)
6
7 isADependencyOf(project(":feature"))
8 // list all your feature modules here
9 isADependencyOf(project(":feature2"))
10 // ...
11 }
12}
Once you opt in auto-detection of used types, it's important that all modules are doubly linked like above. If not the feature modules will fail to compile due to some missing schema classes.
bidirectional parameter (Experimental)
Using isADependencyOf()
requires listing all your feature modules in your schema module which duplicates code. To keep things minimal and configure the dependency automatically, use bidirectional = true
:
1// feature/build.gradle.kts
2apollo {
3 service("service") {
4 packageName.set("feature")
5
6 // Use `bidirectional` to have the schema module get the used types from this module
7 dependsOn(dependencyNotation = project(":schema"), bidirectional = true)
8 }
9}
Also disable generating types in the schema module:
1// schema/build.gradle.kts
2apollo {
3 service("service") {
4 packageName.set("schema")
5 generateApolloMetadata.set(true)
6
7 // Disable generating all types
8 alwaysGenerateTypesMatching.set(emptyList())
9 }
10}
bidirectional
parameter is experimental and not compatible with Gradle project isolation.Optional optimizations
In Android projects, to further improve build times you can consider making your schema module a JVM module instead of an Android library. Since Apollo Kotlin is multiplatform, and Android is capable of depending on JVM-only modules, you can benefit from reducing the amount of work needed to compile the schema module. This is true because JVM modules build faster than Android libraries even with identical source code.
This is often a good idea because Apollo often happens to be on the critical path of compiling your app. This happens when all the feature modules that use Apollo need to wait for all the type-safe code to be generated before they themselves can start building too.