GraphQLThis is the third post in our series of articles where we explore a JVM based stack comprising of Kotlin, Vert.X and jOOQ for development of GraphQL APIs.
In the previous post we integrated a dagger based DI system to simplify wiring up of components. In this post we will add some more GraphQL resolvers which connect to the database using the jooq based DAO layer.
Before we do that, let's add an AppConfig component to read configuration from a properties file in a type safe manner so that our application becomes runtime configurable:
Our configuration file (src/resources/app-config.properties) looks like this:
db.port = 5432
db.host = localhost
db.database = jooq_graphql_sample
db.user = lorefnonWe will write an AppConfigModule which will make the parsed configuration available to other services through dagger.
package me.lorefnon.sample.module
3 lines collapsed (Imports)
@Module
class AppConfigModule {
@Provides
fun provideAppConfig(): AppConfig {
val properties = Properties()
// Enable application consumer to inject a properties file through environment variable
val fsPropertiesPath = System.getenv()["APP_CONFIG_PATH"]
val propertiesStream = fsPropertiesPath
?.let { FileInputStream(fsPropertiesPath) }
// Otherwise use the one in class path
?: javaClass.classLoader.getResourceAsStream("app-config.properties")
properties.load(propertiesStream)
return AppConfig(properties)
}
}
/**
* A type safe wrapper over the untyped Properties instance
*/
class AppConfig(private val properties: Properties) {
val appPort get(): Int = properties.getProperty("server.http.port")?.toInt() ?: 8888
val dbPort get(): Int? = properties.getProperty("db.port")?.toInt()
val dbHost get(): String = properties.getProperty("db.host")!!
val dbName get(): String = properties.getProperty("db.database")!!
val dbUser get(): String? = properties.getProperty("db.user")
val dbPassword get(): String? = properties.getProperty("db.password")
}
Let us add a DBAccessModule which uses this configuration to wire up database access:
package me.lorefnon.sample.module
6 lines collapsed (Imports)
@Module
class DBAccessModule {
/**
* Provides an database pool client powered by Vert.X asynchronous postgres driver
*/
@Provides
@Singleton
fun providePGClient(connectOptions: PgConnectOptions, poolOptions: PoolOptions): SqlClient =
PgPool.pool(connectOptions, poolOptions)
@Provides
@Singleton
fun providePGPoolOptions(config: AppConfig) =
PoolOptions()
@Provides
@Singleton
fun providePGConnectOptions(config: AppConfig) =
PgConnectOptions()
.apply {
port = config.dbPort
host = config.dbHost
database = config.dbName
config.dbUser?.let { user = it }
config.dbPassword?.let { password = it }
}
}
Thanks to jklingsporn's ClassicReactiveVertxGenerator, our DAO classes are compatible with the Vert.X postgres driver pool we configured above.
So we can simply instantiate these DAO classes providing passing this postgres client to them. Let us write a module that instantiates our generated DAO classes:
package me.lorefnon.sample.module
8 lines collapsed (Imports)
@Module
class DAOModule {
@Provides
@Singleton
fun provideJooQConfiguration(): Configuration =
DefaultConfiguration().apply {
setSQLDialect(SQLDialect.POSTGRES)
}
@Provides
@Singleton
fun provideUsersDao(jooqConfig: Configuration, sqlClient: SqlClient) =
UsersDao(jooqConfig, sqlClient)
// Expose other DAO classes here
}
We will also need to make our MainVerticleComponent aware of these modules:
package me.lorefnon.sample
6 lines collapsed (Imports)
@Component(modules = [
AppConfigModule::class,
DBAccessModule::class,
DAOModule::class,
GraphQLModule::class,
RouterModule::class
])
@Singleton
interface MainVerticleComponent {
fun getRouter(): Router
}
And then inject these instances in MainVerticle:
class MainVerticle : AbstractVerticle() {
val components by lazy {
DaggerMainVerticleComponent
.builder()
.appConfigModule(AppConfigModule())
.dBAccessModule(DBAccessModule())
.dAOModule(DAOModule())
.graphQLModule(GraphQLModule())
.routerModule(RouterModule(vertx))
.build()
}
override fun start(startPromise: Promise<Void>) {
val appPort = component.getAppConfig().appPort
vertx
.createHttpServer()
.requestHandler(component.getRouter())
.listen(appPort) { http ->
5 lines collapsed }
}
}
Our API still works, but obviously our DBAccessModule & DAOModule are still unused. Let's address that.
Here is the simplest UserRegistrationService implementation, that receives incoming credentials, and saves it in DB after hashing the password:
package me.lorefnon.sample.service
4 lines collapsed
class UserRegistrationService @Inject constructor(
private val userDao: UserDao,
private val passwordEncryptionService: PasswordEncryptionService
) {
fun registerUser(name: String, email: String, password: String): Future<RegistrationResult> {
return userDao.insert(User().apply {
this.name = name
this.password = passwordEncryptionService.encrypt(password)
this.email = email
}).map { insertCount ->
RegistrationResult(true)
}.otherwise { throwable ->
val comments =
if (throwable.message?.contains("violates unique constraint") == true)
listOf(
"A user is already registered for this email/username",
"Do you want to sign in instead ?"
)
else null
RegistrationResult(false, comments)
}
}
}
data class RegistrationResult(
val success: Boolean,
val comments: List<String>? = null
)
package me.lorefnon.sample.service
import at.favre.lib.crypto.bcrypt.BCrypt
import javax.inject.Inject
private const val LOGARITHMIC_COST_FACTOR = 12
class PasswordEncryptionService @Inject constructor() {
fun encrypt(password: String) =
BCrypt.withDefaults().hashToString(LOGARITHMIC_COST_FACTOR, password.toCharArray())
fun verify(password: String, encryptedPassword: String) =
BCrypt.verifyer().verify(password.toCharArray(), encryptedPassword).verified
}
A production ready solution will also do email verification, time-limit number of signups from an IP, etc. but for our tutorial we will not go into the intricacies of user registration best practices.
Our service, at this point, is a plain kotlin service that can be tested in isolation. It is not, however, GraphQL aware.
We need to next expose this service through a resolver:
package me.lorefnon.sample.module
9 lines collapsed (Imports)
@Module
class GraphQLModule {
@Provides
@Singleton
fun provideGraphQL(
userRegistrationService: UserRegistrationService
) = GraphQLBuilder(
userRegistrationService
).build()
inner class GraphQLBuilder(
private val userRegistrationService: UserRegistrationService
) {
fun build(): GraphQL = GraphQL
.newGraphQL(buildExecutableSchema())
.build()
private val rawSchema = """
type Query {
hello: String
}
type Mutation {
registerUser(username: String!, email: String!, password: String!): RegistrationResult!
}
type RegistrationResult {
success: Boolean!
}
""".trimIndent()
private fun buildRuntimeWiring() = newRuntimeWiring()
.type("Query") {
it.dataFetcher("hello") {
"world"
}
}
.type("Mutation") {
it.associateRegisterUserMutation()
}
.build()
// Local extension method for graphql-java's TypeRuntimeWiring Builder
// to associate the registerUser resolver
private fun TypeRuntimeWiring.Builder.associateRegisterUserMutation() =
dataFetcher("registerUser", VertxDataFetcher.create { env ->
val username = env.getArgument<String>("username")
val email = env.getArgument<String>("email")
val password = env.getArgument<String>("password")
userRegistrationService.registerUser(username, email, password)
})
private fun buildExecutableSchema() =
SchemaGenerator().makeExecutableSchema(parseSchema(), buildRuntimeWiring())
private fun parseSchema() =
SchemaParser().parse(rawSchema)
}
}
We have updated our GraphQL schema SDL to include a registerUser mutation field and we have updated our runtime wiring to resolve this field through the registerUser method of UserRegistrationService.
In an extension method (associateRegisterUserMutation), we are bridging the non-typesafe arguments API of GraphQL java to the type safe UserRegistrationService.registerUser arguments.
One interesting thing to note here is the use of VertxDataFetcher.create. In case of our static resolver before (Query.hello field), we had passed a lambda to TypeRuntimeWiring.Builder.dataFetcher. This works perfectly fine for simple values or java objects. If our userRegistrationService.registerUser returned a RegistrationResult instance directly, we would not have to use VertxDataFetcher.create.
But what it actually returns is a Vert.X Future instance that resolves to a RegistrationResult.
graphql-java library is not aware of this Future class. It does understand java.util.concurrent.CompletionStage (Read more about asynchronous execution in graphql-java), but the Future being returned here is not java.util.concurrent.CompletableFuture (which implements java.util.concurrent.Future and java.util.concurrent.CompletionStage) but rather io.vertx.core.Future, which is used by the vert.x libraries.
So we need to convert the vert.x future to a CompletionStage. We already have a .toCompletionStage() method in vert.x Futures for this, but the VertxDataFetcher fetcher is an added convenience on top of that which is Vert.x context aware.
We can now restart our service and try out our mutation through Altair, or other GraphQL Client:
mutation {
registerUser(
username: "lorefnon",
email: "lorefnon@gmail.com",
password: "password"
) {
success
comments
}
}If it's all wired up correctly, we should get a successful response:
{
"data": {
"registerUser": {
"success": true,
"comments": null
}
}
}If we try to create another user with the same login, we should receive a clear error:
{
"data": {
"registerUser": {
"success": false,
"comments": [
"A user is already registered for this email/username",
"Do you want to sign in instead ?"
]
}
}
}We also need a mechanism to allow users to login. So let us add another mutation where we can pass in the credentials are retrieve a JWT token:
package me.lorefnon.sample.service
import io.vertx.core.Future
import me.lorefnon.sample.generated.tables.daos.UserDao
import javax.inject.Inject
class UserLoginService @Inject constructor(
private val userDao: UserDao,
private val passwordEncryptionService: PasswordEncryptionService,
private val authTokenService: AuthTokenService
) {
fun login(name: String, password: String): Future<String?> =
userDao.findOneByName(name).map {
it?.let { user ->
if (passwordEncryptionService.verify(password, user.password)) user
else null
}?.let { user ->
authTokenService.getToken(user.name)
}
}
}Let's add that to our GraphQL module similar to the registration field:
package me.lorefnon.sample.module
11 lines collapsed (Imports)
@Module
class GraphQLModule {
@Provides
@Singleton
fun provideGraphQL(
userRegistrationService: UserRegistrationService,
userLoginService: UserLoginService
) = GraphQLBuilder(
userRegistrationService,
userLoginService
).build()
inner class GraphQLBuilder(
private val userRegistrationService: UserRegistrationService,
private val userLoginService: UserLoginService
) {
fun build(): GraphQL = GraphQL
.newGraphQL(buildExecutableSchema())
.build()
private val rawSchema = """
type Query {
hello: String
}
type Mutation {
registerUser(
username: String!,
email: String!,
password: String!
): RegistrationResult!
login(username: String!, password: String!): String
}
4 lines collapsed """.trimIndent()
private fun buildRuntimeWiring() = newRuntimeWiring()
.type("Query") {
it.dataFetcher("hello") { "world" }
}
.type("Mutation") {
it.associateRegisterUserMutation()
it.associateLoginMutation()
}
.build()
6 lines collapsed
private fun TypeRuntimeWiring.Builder.associateLoginMutation() =
dataFetcher("login", VertxDataFetcher.create { env ->
val username = env.getArgument<String>("username")
val password = env.getArgument<String>("password")
userLoginService.login(username, password)
})
private fun buildExecutableSchema() =
SchemaGenerator().makeExecutableSchema(parseSchema(), buildRuntimeWiring())
private fun parseSchema() =
SchemaParser().parse(rawSchema)
}
}
An invocation of this API looks like this:
mutation {
login(username: "lorefnon", password: "password")
}{
"data": {
"login": "eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpc3MiOiJtZS5sb3JlZm5vbi5zYW1wbGUiLCJ1c2VyIjoibG9yZWZub24ifQ.vJNjffx0dMAqoORfXlf-dUF6cAhwMfTPLeZzxB632_k"
}
}In the next post we will utilize this API to secure other fields in our GraphQL API.