Edit Page

Android Frameworks Using Annotation Processing

This tutorial describes the usage of popular Android frameworks and libraries that use annotation processing with Kotlin.

The Android world has many popular frameworks simplifying development. You can use the same frameworks if you develop in Kotlin, often as easily as you'd use them in Java. This tutorial provides examples and highlights the differences in settings.

We'll look at Dagger, Butterknife, Auto-parcel and DBFlow (other frameworks can be set up similarly). All these frameworks work through annotation processing: you annotate the code to have the boiler-plate code generated for you. Annotations allow to hide all the verbosity and keep your code simple, and if you need to understand what actually happens at runtime, you can look at the generated code. Note that all these frameworks generate source code in Java, not Kotlin.

In Kotlin you specify the dependencies in a similar to Java way using Kotlin Annotation processing tool (kapt) instead of annotationProcessor.


Dagger is a dependency injection framework. If you're not familiar with it yet, you can read its user's guide. We've converted The coffee example described in this guide into Kotlin, and you can find the result here. The Kotlin code looks pretty much the same; you can browse the whole example in one file.

As in Java, you use @Inject to annotate the constructor used by Dagger to create instances of a class. Kotlin has concise syntax for declaring a property and a constructor parameter at the same time. To annotate the constructor, use the constructor keyword explicitly and put the @Inject annotation before it:

class Thermosiphon 
@Inject constructor(
        private val heater: Heater
) : Pump {
    // ...

@Module annotates classes defining how to provide different objects. When you pass an annotation argument as a vararg argument, you have to explicitly wrap it into arrayOf, like in @Module(includes = arrayOf(PumpModule::class)) below:

abstract class PumpModule {
    abstract fun providePump(pump: Thermosiphon): Pump

@Module(includes = arrayOf(PumpModule::class))
class DripCoffeeModule {
    @Provides @Singleton
    fun provideHeater(): Heater = ElectricHeater()

Annotating methods looks absolutely the same. In the example above @Binds determines that a Thermosiphon object is used whenever a Pump is required, @Provides specifies the way to build a Heater, and @Singleton says that the same Heater should be used all over the place.

To have a dependency-injected implementation generated for the type, annotate it with @Component. The generated class will have the name of this type prepended with Dagger, like DaggerCoffeeShop below:

@Component(modules = arrayOf(DripCoffeeModule::class))
interface CoffeeShop {
    fun maker(): CoffeeMaker

fun main(args: Array<String>) {
    val coffee = DaggerCoffeeShop.builder().build()

Dagger generates an implementation of CoffeeShop that allows you to get a fully-injected CoffeeMaker. You can navigate and see the implementation of DaggerCoffeeShop if you open the project in IDE.

We observed that annotating your code almost hasn't changed when you switched to Kotlin. Now let's see what changes should be made to the build script.

In Java you specify Dagger as annotationProcessor (or apt) dependency:

dependencies {
  annotationProcessor "com.google.dagger:dagger-compiler:$dagger-version"

In Kotlin you have to add the kotlin-kapt plugin to enable kapt, and then replace annotationProcessor with kapt:

apply plugin: 'kotlin-kapt'
dependencies {
    kapt "com.google.dagger:dagger-compiler:$dagger-version"

That's all! Note that kapt takes care of your Java files as well, so you don't need to keep the apt dependency.

The full build script for the sample project can be found here. You can also look at the converted code for the Android sample.


ButterKnife allows to bind views to fields directly instead of calling findViewById.

Note that Kotlin Android Extensions plugin (automatically bundled into the Kotlin plugin in Android Studio) solves the same issue: replacing findViewById with a concise and straightforward code. Consider using it unless you're already using ButterKnife and don't want to migrate.

You use ButterKnife with Kotlin in the same way as you use it with Java. Let's see first the changes in the Gradle build script, and then highlight some of the differences in the code.

In the Gradle dependencies you use add the kotlin-kapt plugin and replace annotationProcessor with kapt:

apply plugin: 'kotlin-kapt'

dependencies {
    compile "com.jakewharton:butterknife:$butterknife-version"
    kapt "com.jakewharton:butterknife-compiler:$butterknife-version"

We've converted the ButterKnife sample to Kotlin. The resulting code can be found here.

Let's look over it to spot what has changed. In Java you annotated the field, binding it with the corresponding view:

@BindView(R2.id.title) TextView title;

In Kotlin you can't work with fields directly, you work with properties. You annotate the property:

lateinit var title: TextView

The @BindView annotation can be applied to the fields only, but the Kotlin compiler understands that and annotates the corresponding field under the hood when you apply the annotation to the whole property.

Note how the lateinit modifier allows to declare a non-null type initialized after the object is created (after the constructor call). Without lateinit you'd have to declare a nullable type and add additional nullability checks.

You can also configure methods as listeners, using ButterKnife annotations:

internal fun sayHello() {
    Toast.makeText(this, "Hello, views!", LENGTH_SHORT).show()

This code specifies an action to be performed on the "hello" button click. Note that with the Anko library and Kotlin Android Extensions, the same Kotlin code looks rather concise and can be written directly:

hello.onClick {
    toast("Hello, views!")


DBFlow is a SQLite library that simplifies interaction with databases. It heavily relies on annotation processing.

To use it with a project you configure annotation processing dependency using kapt:

apply plugin: 'kotlin-kapt'

kapt "com.github.raizlabs.dbflow:dbflow-processor:$dbflow_version"
compile "com.github.raizlabs.dbflow:dbflow-core:$dbflow_version"
compile "com.github.raizlabs.dbflow:dbflow:$dbflow_version"
compile "com.github.raizlabs.dbflow:dbflow-kotlinextensions:$dbflow_version"

Here is a detailed guide how to add DBFlow to your project.

If your application already uses DBFlow, you can safely add Kotlin to your project. You can gradually convert existing code to Kotlin (ensuring that everything compiles along the way). The converted code doesn't differ much from Java. For instance, declaring a table looks similar to Java with the small difference that default values for properties must be specified explicitly:

@Table(name="users", database = AppDatabase::class)
class User: BaseModel() {

    @PrimaryKey(autoincrement = true)
    @Column(name = "id")
    var id: Long = 0

    var name: String? = null

Besides being able to convert existing functionality to Kotlin, you can also enjoy the Kotlin specific module. DBFlow defines a bunch of extensions to make its usage in Kotlin more idiomatic. Let's highlight some of the supported features.

You can declare tables as data classes:

@Table(database = KotlinDatabase::class)
data class User(@PrimaryKey var id: Long = 0, @Column var name: String? = null)

Queries might be expressed via C#-like LINQ syntax. Thus the Java code below can be either converted directly or rewritten into the following style:

/* java */
List<Result> = SQLite.select()
    .and(Result_Table.column2.in("5", "6", "9")).queryList()
/* kotlin */                   
val results = (select
      from Result::class
      where (column eq 6)
      and (column2 `in`("5", "6", "9"))
      groupBy column).list

Having lambdas allows to write much simpler code for asynchronous computations:

var items = (select from TestModel::class).list

// delete all these items.
items.processInTransactionAsync { it, databaseWrapper -> it.delete(databaseWrapper) }

More details can be found here. You can also browse the converted sample application.


Auto-Parcel allows to generate Parcelable values for classes annotated with @AutoValue.

When you specify the dependency you again use kapt as annotation processor to take care of Kotlin files:

apply plugin: 'kotlin-kapt'

dependencies {
    kapt "frankiesardo:auto-parcel:$latest-version"

The converted sample can be found here.

You can safely annotate Kotlin classes with @AutoValue. Let's look at the converted Address class for which the Parcelable implementation will be generated:

abstract class Address : Parcelable {
    abstract fun coordinates(): DoubleArray
    abstract fun cityName(): String

    companion object {
        fun create(coordinates: DoubleArray, cityName: String): Address {
            return builder().coordinates(coordinates).cityName(cityName).build()
        fun builder(): Builder = `$AutoValue_Address`.Builder()
    interface Builder {
        fun coordinates(x: DoubleArray): Builder
        fun cityName(x: String): Builder
        fun build(): Address

Kotlin doesn't have static methods, so they should be place inside a companion object. If you still want to use them from Java code, annotate them with @JvmStatic.

If you need to access a Java class or method with a name that is not a valid identifier in Kotlin, you can escape the name with the backtick (`) character, like in accessing the generated class `$AutoValue_Address`.

Overall the converted code looks very similar to the original Java code.