WebSockets in Scala: Part 2 - Integrating Redis and PostgreSQL

Herbert Kateu

Herbert Kateu

37 min read  • 

postgresq... redis scala

Share on:

Introduction

This article is a follow-up to the websocket article that was published previously. To recap, we created an in-memory chat application using WebSockets with the help of the http4s library. The chat application had a variety of features implemented through commands directly in the chat window such as the ability to create users, create chat rooms, and switch between chat rooms.

In this iteration, we’ll be integrating Redis to keep track of the users and rooms and we’ll also be persisting messages in Postgres so that new users can have access to previous conversations. Finally, We’ll get rid of chatState and create a new protocol that interacts with Postgres and Redis.

Since this tutorial builds on the previous article, to follow along, we’ll need to clone that GitHub repo where we’ll be making the necessary updates to build this new version.

Setup

We’ll be using skunk and redis4Cats in our application so let’s add them to our build.sbt file.

...
val RedisVersion = "1.5.2"
val SkunkVersion = "0.6.3"

lazy val root = (project in file("."))
  .settings(
    organization := "rockthejvm",
    name := "websockets",
    version := "0.0.1-SNAPSHOT",
    scalaVersion := "3.3.3",
    libraryDependencies ++= Seq(
      ...
      "dev.profunktor" %% "redis4cats-effects" % RedisVersion,
      "dev.profunktor" %% "redis4cats-streams" % RedisVersion,
      "org.tpolecat" %% "skunk-core" % SkunkVersion
    ),
    libraryDependencies ++= Seq(
      "io.circe" %% "circe-core",
      "io.circe" %% "circe-generic",
      "io.circe" %% "circe-parser"
    ).map(_ % CirceVersion)
  )

We also add circe core, generic and parser to build.sbt.

Creating the Domain

In this version, we’ll add a UUID to the User and Room case classes and move the validateutility to its own file. Let’s create a validateutility.scala in the following path, src/main/scala/rockthejvm/websockets/domain, and add the following code:

package rockthejvm.websockets.domain

import cats.data.Validated

object validateutility{
  def validateItem[F](
      value: String,
      userORRoom: F,
      name: String
  ): Validated[String, F] = {
    Validated.cond(
      (value.length >= 2 && value.length <= 10),
      userORRoom,
      s"$name must be between 2 and 10 characters"
    )
  }
}

Nothing has changed from the previous version. Next, we’ll create a user.scala file in the following path, src/main/scala/rockthejvm/websockets/domain. This file will contain the new User case class.

package rockthejvm.websockets.domain

import java.util.UUID
import cats.Applicative
import cats.data.Validated
import rockthejvm.websockets.domain.validateutility.*
import cats.syntax.all.*

object user {
  case class UserId(id: UUID)
  case class UserName(name: String)
  case class User(id: UserId, name: UserName)
  object User {
    def apply[F[_]: Applicative](
        id: UUID,
        name: String
    ): F[Validated[String, User]] =
      validateItem(name, new User(UserId(id), UserName(name)), "User name")
        .pure[F]
  }
}

Here we create a User case class that contains the UserId and UserName that take a UUID and String respectively. In the apply method we still use validateItem() to create the F[Validated[String, User]].

Let’s also do the same for room.scala in the same path and add the following code:

package rockthejvm.websockets.domain

import java.util.UUID
import cats.data.Validated
import cats.syntax.all.*
import cats.*
import rockthejvm.websockets.domain.validateutility.*

object room {
  case class RoomId(id: UUID)
  case class RoomName(name: String)
  case class Room(id: RoomId, name: RoomName)
  object Room {
    def apply[F[_]: Applicative](
        id: UUID,
        name: String
    ): F[Validated[String, Room]] =
      validateItem(name, new Room(RoomId(id), RoomName(name)), "Room").pure[F]
  }
}

The code here is very similar to user.scala. We’ll also need to insert messages into our database, for this we’ll need another case class, create a message.scala file under domain, and add the following code:

package rockthejvm.websockets.domain

import java.util.UUID
import java.time.LocalDateTime
import rockthejvm.websockets.domain.user.{UserId, User}
import rockthejvm.websockets.domain.room.RoomId

object message {
  case class MessageId(id: UUID)
  case class MessageText(value: String)
  case class MessageTime(time: LocalDateTime)
  case class InsertMessage(
      id: MessageId,
      value: MessageText,
      time: MessageTime,
      userId: UserId,
      roomId: RoomId
  )
}

The values in InsertMessage match the columns we’ll have when we create the messages table in Postgres. We’ll also need another case class that will hold the messages fetched from the database:

package rockthejvm.websockets.domain

...

object message {
  ...
  case class FetchMessage(value: MessageText, from: User)
}

The FetchMessage case class contains a MessageText and a User from whom the message was sent. The reason we need another case class is that only want to fetch two columns from Postgres.

Docker for Redis and PostgreSQL

We’ll be using Docker images for Redis and Postgres. To follow along, you’ll need Docker and Docker Compose installed. We can install them by installing Docker desktop on your system.

After installation, we can check if we have everything installed by running the following:

$ docker -v
Docker version 25.0.3, build 4debf41

$ docker-compose -v
Docker Compose version v2.24.5

Next, we’ll create the SQL commands to create the database and necessary tables for Postgres. In the root folder create setup.sql and add the following:

CREATE DATABASE websocket;

\c websocket;

CREATE TABLE users (
    id UUID PRIMARY KEY,
    name VARCHAR(255) NOT NULL
);

CREATE TABLE rooms (
    id UUID PRIMARY KEY,
    name VARCHAR(255) NOT NULL
);

CREATE TABLE messages (
    id UUID PRIMARY KEY,
    message TEXT NOT NULL,
    time TIMESTAMP DEFAULT NOW(),
    user_id UUID REFERENCES users (id),
    room_id UUID REFERENCES rooms (id)
);

The first command creates a database called websocket, \c websocket connects to it, then we create a users and rooms table each with an id of type UUID and name of type VARCHAR(255), and finally we create the messages table with an id, message and time columns and it also references the users and rooms table id.

Then we need manage our docker stack using docker-compose. Let’s create a docker-compose.yaml file in the root folder of our application and add the following:

services:
  postgres:
    image: postgres:alpine
    container_name: postgres-server
    restart: always
    environment:
      - POSTGRES_USER=postgres
      - POSTGRES_PASSWORD=password
    ports:
      - 5432:5432
    volumes:
      - ./postgres_volume:/var/lib/postresql/data
      - ./setup.sql:/docker-entrypoint-initdb.d/setup.sql

  redis:
    image: redis:alpine
    container_name: redis-server
    restart: always
    ports:
      - 6379:6379

volumes:
  postgres_volume:
    driver: local

Here we define, two services, postgres and redis. Under postgres, we specify the image name as postgres:alpine, we also specify the container name as postgres-server and set it to restart always.

Postgres containers need a username and password which we specify as environment variables under the environment segment. Next, we give it a port number of 5432 for both internal and external and lastly, we specify the docker volume for the container as postgres_volume.

A docker volume helps keep the state of the container, in our case all the databases and tables will be backed up in case the container is destroyed.

Finally, to run our setup.sql file when the container is initialized, we add ./setup.sql:/docker-entrypoint-initdb.d/setup.sql under volumes.

Under the Redis section, we only specify the image name, container name, restart policy, and port number.

Moving from the services to volumes section, we set the postgres_volume driver as local so that docker writes to the external disk.

Finally, we can start our docker containers by running the following command in the root of our application.:

docker compose up

We can confirm that our containers have been created and are running with the following command:

docker ps

We can also confirm that our database and tables have been created by running the following:

docker exec -it postgres-server psql -U postgres

Then connect to the database and finally list the tables by running the following:

\c websocket
\d

Skunk for PostgreSQL Integration

In this section, we’ll implement the protocols necessary for interacting with Postgres in our application using Skunk.

First, we’ll need to implement Codecs for the types in our domain. Create a codecs.scala file in the following path, src/main/scala/rockthejvm/websockets/codecs/codecs.scala and add the following code:

package rockthejvm.websockets.codecs

import skunk.*
import skunk.codec.all.*
import rockthejvm.websockets.domain.user.{UserId, UserName}
import rockthejvm.websockets.domain.room.{RoomId, RoomName}
import rockthejvm.websockets.domain.message.{
  MessageId,
  MessageText,
  MessageTime
}

object codecs {
  val userId: Codec[UserId] = uuid.imap[UserId](UserId(_))(_.id)
  val userName: Codec[UserName] =
    varchar(255).imap[UserName](UserName(_))(_.name)
  val roomId: Codec[RoomId] = uuid.imap[RoomId](RoomId(_))(_.id)
  val roomName: Codec[RoomName] =
    varchar(255).imap[RoomName](RoomName(_))(_.name)
  val messageId: Codec[MessageId] = uuid.imap[MessageId](MessageId(_))(_.id)
  val messageText: Codec[MessageText] =
    text.imap[MessageText](MessageText(_))(_.value)
  val messageTime: Codec[MessageTime] =
    timestamp.imap[MessageTime](MessageTime(_))(_.time)
}

Skunk provides several important codecs through the skunk.codec.all.* import such as uuid, varchar, text, and timestamp. We use the imap method to provide the encode and decode methods for each case class, here’s the signature:

  def imap[B](f: A => B)(g: B => A): Codec[B] =
    Codec(b => encode(g(b)), decode(_, _).map(f), types)

Here we provide the methods from case class to postgres and the reverse. Here we provide codecs for case classes that take native types, however, later we shall see how to join codecs for types such as InsertMessage.

Now we’ll create various methods that will interact with the Postgres database through Skunk. Let’s create a PostgresProtocol.scala file under the websockets and add the following code:

package rockthejvm.websockets

import fs2.Stream
import java.time.LocalDateTime
import rockthejvm.websockets.domain.user.*
import rockthejvm.websockets.domain.room.*
import rockthejvm.websockets.domain.message.*

trait PostgresProtocol[F[_]] {
  def createUser(name: String): F[Either[String, User]]
  def createRoom(name: String): F[Either[String, Room]]
  def deleteRoom(roomId: RoomId): F[Unit]
  def deleteUser(userId: UserId): F[Unit]
  def saveMessage(
      message: String,
      time: LocalDateTime,
      userId: UserId,
      roomId: RoomId
  ): F[Unit]
  def fetchMessages(roomId: RoomId): F[Stream[F, FetchMessage]]
  def deleteRoomMessages(roomId: RoomId): F[Unit]
}

First, we create a PostgresProtocol trait with a number of methods that will later be used by our chat protocol:

...
import cats.effect.std.UUIDGen
import cats.effect.*
import skunk.*

trait PostgresProtocol[F[_]] {
  ...
}

object PostgresProtocol {
  def make[F[_]: UUIDGen: Concurrent](
      postgres: Resource[F, Session[F]]
      ): F[PostgresProtocol[F]] =
          postgres.use {session =>
            new PostgresProtocol[F] {
            ???
            }.pure[F]
          }
}
private object SqlCommands {
  ???
}

We also provide a companion object with a make() method where we’ll implement all the methods defined in the trait, below that we have the SqlCommands object which will contain other Codecs used in our implementations. Here we call postgres.use() to access the session which represents a live connection to the database, and inside that, we’ll implement an F[PostgresProtocol[F]].

First, we start with createUser(), which inserts a user into the database, to do this, we’ll need a Codec[User] for this to work.

import rockthejvm.websockets.codecs.codecs.*
...
private object SqlCommands {
    val usercodec: Codec[User] =
        (userId ~ userName).imap {
            case i ~ n => User(i,n)
        }(u => (u.id, u.name))
}

Here we create a twiddle list, (userId ~ userName) which produces a Codec[(UserId, UserName)], we then call imap and provide the two functions from userId ~ userName to User and the reverse

Let’s also go ahead and implement the Codec[Room] since its very similar to Codec[User]:

private object SqlCommands {
  ...
  val roomcodec: Codec[Room] =
      (roomId ~ roomName).imap {
          case i ~ n => Room(i,n)
      }(u => (u.id, u.name))
}

Next, we’ll need to create its associate Command for the INSERT statement. In Skunk, Select statements are prepared with a Query while Insert and Update statements are prepared with a Command since it return no rows.

private object SqlCommands {
  ...
  val insertUser: Command[User] =
      sql"""
          INSERT INTO users
          VALUES($usercodec)
      """
      .command
}

Here we prepare an insert statement, and pass the usercodec to VALUES(), then call the command method giving us a Command[User]. A Command[User], means we need to pass a User inorder to execute the query:

Let’s also create Command[Room] since its also similar to the above:

private object SqlCommands {
  ...
  val insertRoom: Command[Room] =
    sql"""
            INSERT INTO rooms
            VALUES($roomcodec)
        """.command
}

Now we can implement createUser():

...
new PostgresProtocol[F] {
  override def createUser(name: String): F[Either[String, User]] =
    session.prepare(insertUser).flatMap { cmd =>
      UUIDGen.randomUUID.flatMap { id =>
        User(id, name).flatMap {
          case Valid(u) =>
            cmd.execute(u).as(Right(u))
          case Invalid(err) =>
            Left(err).pure[F]
        }
      }
    }
}.pure[F]

The checkUser() function takes a name of type String and returns an F[Either[String, User]]. Here we call the prepare() on session passing it the insertUser Command, this caches and prepares the query.

The prepare() method produces a F[PreparedCommand[F, User]], which we flatMap on to create our User. We use the UUIDGen.randomUUID method from cats.effect to create our UUID which we pass to User along with the name.

If User creation succeeds, we call cmd.execute(u) which runs the INSERT query, and finally, we return a Right(u). If the creation fails, we pass the error as Left(err).pure[F].

Let’s also implement createRoom() since it’s similar:

...
new PostgresProtocol[F] {
  override def createRoom(name: String): F[Either[String, Room]] =
    session.prepare(insertRoom).flatMap { cmd =>
      UUIDGen.randomUUID.flatMap { id =>
        Room(id, name).flatMap {
          case Valid(r) =>
            cmd.execute(r).as(Right(r))
          case Invalid(err) =>
            Left(err).pure[F]
        }
      }
    }
}.pure[F]

Next is deleteUser() and deleteRoom(), like before, we’ll need to first create sql statements of type Command for both scenarios. The Command[UserId] means that a UserId is needed as an argument:

private object SqlCommands {
  ...
  val delUser: Command[UserId] =
      sql"""
          UPDATE users
          SET name = "deletedUser"
          WHERE id = $userId
      """
      .command
}

In our case instead of completely deleting the user, we replace the name with deletedUser with the help of an UPDATE statement, so that we keep the messages in the database even if the user doesn’t exist:

However, for the case of Rooms, we’ll be completely deleting a room that matches a particular RoomId using a DELETE statement:

private object SqlCommands {
  ...
  val delRoom: Command[RoomId] =
    sql"""
        DELETE FROM rooms
        WHERE id = $roomId
    """
    .command
}

Now let’s define the deleteUserOrRoom() private function:

new PostgresProtocol[F] {
  ...
  private def deleteUserOrRoom[A](id: A, command: Command[A]): F[Unit] =
    session.prepare(command).flatMap { cmd =>
      cmd.execute(id).void
    }
}.pure[F]

This function takes a command of type Command[A] and an id of type A that we pass to prepare and execute methods respectively to produce an F[Unit]. Here’s how we’d use it

...
new PostgresProtocol[F] {
  ...
  override def deleteUser(uId: UserId): F[Unit] =
    deleteUserOrRoom[UserId](uId, delUser)

  override def deleteRoom(rId: RoomId): F[Unit] =
    deleteUserOrRoom[RoomId](rId, delRoom)
}.pure[F]

Both these functions take either a UserId or a RoomId, which we pass to deleteUserOrRoom() along with it’s respective Command.

Next, we have saveMessage(), which will be used to insert new messages into the database, this will require a message codec to decode and encode messages to and from Postgres:

private object SqlCommands {
  ...
  val messagecodec: Codec[InsertMessage] =
      (messageId ~ messageText ~ messageTime ~ userId ~ roomId).imap {
          case mi ~ mt ~ mtm ~ ui ~ ri => InsertMessage(mi,mt,mtm,ui,ri)
      }(m => (((((m.id), m.value), m.time), m.userId), m.roomId))
}

Here we a Codec[InsertMessage], the process is similar to how we created the usercodec, except that the decode function has to return a nested tuple as shown above.

Now we can create a Command[InsertMessage] to INSERT a message:

private object SqlCommands {
  ...
  val insertMessage: Command[InsertMessage] =
      sql"""
          INSERT INTO messages
          VALUES($messagecodec)
      """
      .command
}

At this point, we can now implement the save message function as follows:

...
new PostgresProtocol[F] {
  ...
  override def saveMessage(
      message: String,
      time: LocalDateTime,
      userId: UserId,
      roomId: RoomId
  ): F[Unit] =
    session.prepare(insertMessage).flatMap { cmd =>
      UUIDGen.randomUUID.flatMap { id =>
        cmd
          .execute(
            InsertMessage(
              MessageId(id),
              MessageText(message),
              MessageTime(time),
              userId,
              roomId
            )
          )
          .void
      }
    }
}.pure[F]

Just like before, we create a random UUID then pass all the required values to InsertMessage within the execute method and finally call void to return an F[Unit].

The fetchMessages() function is interesting since it requires a JOIN sql statement. This is the reason we needed a FetchMessage case class to retrieve messages. Like before, we’ll need a codec:

private object SqlCommands {
  ...
  val fetchmessagecodec: Codec[FetchMessage] =
    (messageText ~ userId ~ userName).imap { case mt ~ ui ~ un =>
      FetchMessage(mt, User(ui, un))
    }(m => (((m.value), m.from.id), m.from.name))
}

What’s unique about this codec, is we are directly passing userId and userName to User within FetchMessage, it also produces a nested tuple in the decode function. To use our codec we’ll create a SELECT statement, which means we’ll be using a Query.

private object SqlCommands {
  ...
  val getMessage: Query[RoomId, FetchMessage] =
    sql"""
            SELECT messages.message, users.id, users.name
            FROM messages
            INNER JOIN users
            ON messages.user_id = users.id
            WHERE messages.room_id = $roomId;
        """.query(fetchmessagecodec)
}

In this SELECT statement, we are selecting the message from the messages table and id and name from the users table by using an INNER JOIN with the id values. The results are filtered by RoomId using the WHERE clause to produce a Query[RoomId, FetchMessage].

We can now use this in our function:

...
new PostgresProtocol[F] {
  ...
  override def fetchMessages(roomId: RoomId): F[Stream[F, FetchMessage]] =
    session.prepare(getMessage).map { cmd =>
      cmd.stream(roomId, 32)
    }
}.pure[F]

Here we use the stream method on cmd, passing it the roomId and chunckSize of 32, this returns an F[Stream[F, FetchMessage]]. The stream method repeatedly emits chunks of data from Postgres based on our query.

Finally, let’s implement the last function of PostgresProtocol. The deleteRoomMessages(), is a function that deletes messages by RoomId, this requires a Command[RoomId].

private object SqlCommands {
  ...
    val delMessages: Command[RoomId] =
      sql"""
              DELETE FROM messages
              WHERE room_id = $roomId
          """.command
}

The usage is similar to deleteUserOrRoom() with the help of the execute method:

...
new PostgresProtocol[F] {
  ...
  override def deleteRoomMessages(roomId: RoomId): F[Unit] =
    session.prepare(delMessages).flatMap { cmd =>
      cmd.execute(roomId).void
    }
}.pure[F]

Redis and Chat Protocols

Before we dive into the Redis implementation, we’ll need an overview of how the schema will be. A Redis hash is a record type structured as a collection of field-value pairs, we’ll need the following hashes for our application:

The users hash will be a collection of UUID string fields mapped to user name values. The rooms hash will be a collection of UUID string fields mapped to room name values. The userroomid hash will be a collection of user UUID string fields mapped to and room UUID string values.

A Redis set is a collection of unique string values, we’ll need a Redis set for each room created in our application: The sets will be named in the format, room:<room id>, and will be a collection of user UUID strings.

We’ll be creating both the Redis Algebra and Chat Algebra in this section so that we can follow why each Redis function is needed.

First, we’ll start by creating the Redis Algebra for our application, in the websockets folder create a RedisPotocol.scala file with the following contents:

package rockthejvm.websockets
import rockthejvm.websockets.domain.user.*
import rockthejvm.websockets.domain.room.*

trait RedisProtocol[F[_]] {
  def createUser(user: User): F[Unit]
  def createRoom(room: Room): F[Unit]
  def getRoomFromName(roomname: String): F[Option[Room]]
  def getUsersRoomId(user: User): F[Option[RoomId]]
  def usernameExists(username: String): F[Boolean]
  def listUserIds(roomid: RoomId): F[Set[UserId]]
  def listRooms: F[List[String]]
  def roomExists(roomid: RoomId): F[Boolean]
  def mapUserToRoom(userid: UserId, roomid: RoomId): F[Unit]
  def addUserToRoom(userid: UserId, roomid: RoomId): F[Unit]
  def removeUserFromRoom(roomid: RoomId, userid: UserId): F[Unit]
  def deleteUserRoomMapping(userid: UserId): F[Unit]
  def deleteRoom(roomid: RoomId): F[Long]
  def deleteUser(userid: UserId): F[Long]
  def getSelectedUsers(
      userid: UserId,
      rest: List[UserId]
  ): F[Option[List[User]]]
  def chatState: F[String]
}

Just like with Postgres, we’ll be implementing each of these functions, and by the end of this section, we will have learned alot about Redis and the redis4cats package.

We’ll also need to create the companion object for RedisProtocol along with a make() function.

import dev.profunktor.redis4cats.RedisCommands
 ...

object RedisProtocol {
  def make[F[_]](
      redis: RedisCommands[F, String, String]
  ): F[RedisProtocol[F]] = {
    new RedisProtocol[F] {
      ???
    }.pure[F]
  }
}

The make() function takes an argument, redis of type RedisCommands[F, String, String] as an argument. This contains all the methods for interacting with Redis, and returns an F[RedisProtocol[F]].

Let’s also create ChatProtocol.scala in the same folder and add the following code:

package rockthejvm.websockets

import rockthejvm.websockets.domain.user.*

trait ChatProtocol[F[_]] {
  def register(name: String): F[OutputMessage]
  def enterRoom(user: User, room: String): F[List[OutputMessage]]
  def chat(user: User, text: String): F[List[OutputMessage]]
  def help(user: User): F[OutputMessage]
  def listRooms(user: User): F[List[OutputMessage]]
  def listMembers(user: User): F[List[OutputMessage]]
  def disconnect(maybeuser: Option[User]): F[List[OutputMessage]]
  def chatState: F[String]
}

Similarly, we’ll also need a companion object:

object ChatProtocol {

  def make[F[_]](
      redisP: RedisProtocol[F],
      postgresP: PostgresProtocol[F]
  ): F[ChatProtocol[F]] = {
    new ChatProtocol[F] {
      ???
    }.pure[F]
  }
}

Registering Users

Let’s start with the register() function in ChatProtocol. To register someone, we’ll need to first check if the user name exists in Redis, then add it to both the Redis and Postgres databases:

new RedisProtocol[F] {
  ...
  override def usernameExists(username: String): F[Boolean] = {
    redis.hVals("users").map { nameslist =>
      nameslist.exists(u => u.toLowerCase == username.toLowerCase)
    }
  }
}.pure[F]

The usernameExists() function takes a username of type String and returns an F[Boolean]. Here we use the hVals() function, which takes the name of the hash, and returns an F[List[String]] containing the list of values. We map on this value and call the exists() method which returns a Boolean depending on if a match is found or not.

Next we’ll implement the createUser(), and createRoom() functions in RedisProtocol[F] since they are very similar:

import cats.syntax.all.*
import cats.Monad

object RedisProtocol {
  def make[F[_]: Monad](
      redis: RedisCommands[F, String, String]
  ): F[RedisProtocol[F]] = {
    new RedisProtocol[F] {
      override def createUser(user: User): F[Unit] = {
        redis.hSet("users", user.toMap).void
      }

      override def createRoom(room: Room): F[Unit] = {
        redis.hSet("rooms", room.toMap).void
      }
    }.pure[F]
  }
}

Here we use the hSet() function to create the users and rooms Redis hashes. The method takes a key value which we supplied as users and rooms, these are the names of the hashes, the second argument is the field which is of type Map[String, String].

Notice the new toMap() function on Room and User, that we used to create the Map. Let’s implement them next:

  //in user.scala
  case class User(id: UserId, name: UserName) {
    def toMap = Map((id.id.toString, name.name))
  }

  //in room.scala
  case class Room(id: RoomId, name: RoomName) {
    def toMap = Map((id.id.toString, name.name))
  }

The toMap method creates a Map with a tuple of the UUID string and the name from UserName and RoomName respectively. We update these methods in user.scala and room.scala.

Now back the ChatProtocol[F] register() function:

import cats.Monad
import cats.syntax.all.*
import cats.effect.std.UUIDGen
import cats.effect.kernel.Concurrent

object ChatProtocol {
  def make[F[_]: Monad: UUIDGen: Concurrent](
      redisP: RedisProtocol[F],
      postgresP: PostgresProtocol[F]
  ): F[ChatProtocol[F]] = {
    new ChatProtocol[F] {
      override def register(name: String): F[OutputMessage] = {
        for {
          userExists <- redisP.usernameExists(name)
          maybeUser <-
            if(userExists == true) {
              Left("User name exists").pure[F]
            } else {
              postgresP.createUser(name)
            }
          om <- maybeUser match {
            case Right(u: User) =>
              redisP.createUser(u) *>
                SuccessfulRegistration(u).pure[F]
            case Left(err: String) =>
              ParsingError(None, err)
                .pure[F]
          }
        } yield om
      }
    }.pure[F]
  }
}

We start by checking if the user name exists in Redis, if this is true we return a Left("User name exists").pure[F] other wise we create the user in Postgres by calling postgresP.createUser(name).

If the creation was successful, we also create the user in Redis by calling redisP.createUser(u) which returns SuccessfulRegistration(u).pure[F]. In case of an error we return, ParsingError(None, err).pure[F].

Entering a Room

To enter a room, we’ll need to first get the user’s current room, and then make the transfer to the new room. We’ll receive the room name inform of a String, which we’ll use to check against the Redis database:

import java.util.UUID
...

new RedisProtocol[F] {
  ...
  override def getRoomFromName(roomname: String): F[Option[Room]] = {
    redis.hGetAll("rooms").flatMap { rmap =>
      rmap.find(_._2 == roomname) match {
        case Some((i, n)) =>
          Room(
            UUID
              .fromString(i),
            n
          )
            .map(_.toOption)
        case None => None.pure[F]
      }
    }
  }
}.pure[F]

Here the getRoomFromName() function that takes a roomname of type String and returns an F[Option[Room]], we use the hGetAll("rooms") method that returns a Map of all the key-values pairs in the rooms hash of type F[Map[String, String]]. We flatMap of this value and call find() on the Map, which returns an Option based on the predicate.

If we get a Some, we return an F[Validated[String, Room]] from the apply method on Room. We map this value and convert it to an Option with the toOption method. We can’t get an Invalid since these values were verified before we put them into Redis. If we get a None were return a None.pure[F].

We’ll also need the user’s current roomid, so that we transfer the user from there:

new RedisProtocol[F] {
  ...
  override def getUsersRoomId(user: User): F[Option[RoomId]] = {
    redis.hGet("userroomid", user.id.id.toString).map {
      case Some(roomid) => Some(RoomId(UUID.fromString(roomid)))
      case None         => None
    }
  }
}.pure[F]

The getUsersRoomId() function takes a User and returns the RoomId of the Room a User is in. Here we use the hGet() function, which we supply with the user’s id to return the room id from the userroomid hash. If we get a Some, we return a Some of RoomId. Otherwise, we return a None.

Now we can implement the enterroom() function in ChatProtocol:

new ChatProtocol[F] {
  ...
  override def enterRoom(
      user: User,
      room: String
  ): F[List[OutputMessage]] = {
    redisP.getRoomFromName(room).flatMap {
      case Some(r) =>
        redisP.getUsersRoomId(user).flatMap {
          case Some(usersroomid) =>
            if (usersroomid == r.id) {
              List(
                SendToUser(
                  user,
                  s"You are already in the ${r.name.name} room"
                )
              ).pure[F]
            } else {
              transferUserToRoom(redisP, postgresP, user, r)
            }
          case None =>
            addToRoom(redisP, postgresP, user, r)
        }
      case None =>
        createRoom(redisP, postgresP, room).flatMap {
          case Right(r) =>
            transferUserToRoom(redisP, postgresP, user, r)
          case Left(err) =>
            List(
              ParsingError(
                Some(user),
                err
              )
            ).pure[F]
        }

    }
  }
}.pure[F]

The enterroom() function takes a user of type User and a room of type String and returns an F[List[OutputMessage]]. There are several new functions here which we’ll implement later.

We start by calling redisP.getRoomFromName(room), if the room exists, we compare that RoomId with the user’s RoomId which we get from redisP.getUsersRoomId(user). If they are the same we inform the user, otherwise we transfer the user by calling transferUserToRoom(redisP, postgresP, user, r).

If the user doesn’t have a RoomId, we add the user to the requested room by calling addToRoom(redisP, postgresP, user, r). Now, if the room name doesn’t exist, we create that room by calling createRoom(redisP, postgresP, room), then transfer the user to the new room.

Moving to Another Room

This is a private function, that transfers the user to a new room:

object ChatProtocol {
  ...
  private def transferUserToRoom[F[_]: Monad: UUIDGen: Concurrent](
      redisP: RedisProtocol[F],
      postgresP: PostgresProtocol[F],
      user: User,
      room: Room
  ): F[List[OutputMessage]] =
    val leaveMessages = removeFromCurrentRoom(redisP, postgresP, user)
    val enterMessages = addToRoom(redisP, postgresP, user, room)
    for {
      leave <- leaveMessages
      enter <- enterMessages
    } yield leave ++ enter
}

To do this we remove the user from the current room by calling removeFromCurrentRoom(), then add the user to the requested room by calling addToRoom(redisP, postgresP, user, room). This happens both in Redis and Postgres to keep everything in sync.

Removing a User from a Room

To remove a user from the current room, we’ll need to remove the user from the Redis room:<roomid> set and remove the entry from the userroomid hash.

In Redis, when the last member is deleted from a set, the entire set is deleted, therefore if this occurs, we’ll also need to update the rooms hash:

new RedisProtocol[F] {
  ...
  override def removeUserFromRoom(
      roomid: RoomId,
      userid: UserId
  ): F[Unit] = {
    redis.sRem(s"room:${roomid.id.toString}", userid.id.toString).void
  }
}.pure[F]

The removeUserFromRoom() uses the sRem function to delete the user from the room:<roomid> set. It requires the room name and the user id:

new RedisProtocol[F] {
  ...
  override def deleteUserRoomMapping(userid: UserId): F[Unit] = {
    redis.hDel("userroomid", userid.id.toString).void
  }

  override def deleteRoom(roomid: RoomId): F[Long] = {
    redis.hDel("rooms", roomid.id.toString)
  }
}.pure[F]

The deleteUserRoomMapping() and deleteRoom() uses the hDel function to delete an entry from the userroomid and rooms hashes, and only requires the user id and room id respectively.

Since Redis automatically deletes empty sets, we’ll have to check if the room:<roomid> set still exists:

new RedisProtocol[F] {
  ...
  override def roomExists(roomid: RoomId): F[Boolean] = {
    redis.exists(s"room:${roomid.id.toString}")
  }
}.pure[F]

We use the exists function which checks if a key exists within redis.

Finally, we can implement the removeFromCurrentRoom() function:

object ChatProtocol {
  ...
  private def removeFromCurrentRoom[F[_]: Monad: UUIDGen](
        redisP: RedisProtocol[F],
        postgresP: PostgresProtocol[F],
        user: User
    ): F[List[OutputMessage]] = {
      redisP.getUsersRoomId(user).flatMap {
        case Some(roomid) =>
          for {
            _ <- redisP.removeUserFromRoom(roomid, user.id)
            _ <- redisP.roomExists(roomid).flatMap { b =>
              println(b)
              if (b == true) { ().pure[F] }
              else {
                redisP.deleteRoom(roomid) *>
                postgresP.deleteRoomMessages(roomid) *>
                  postgresP.deleteRoom(roomid)
              }
            }
            _ <- redisP.deleteUserRoomMapping(user.id)
            om <- broadcastMessage(
              redisP,
              roomid,
              SendToUser(user, s"${user.name.name} has left the room")
            )
          } yield om
        case None =>
          List.empty[OutputMessage].pure[F]
      }
    }
}

Here we start by getting the user’s current roomid by calling redisP.getUsersRoomId(user), if the room doesn’t exist we return List.empty[OutputMessage].pure[F] otherwise we use a for comprehension to do the following.

  1. Remove the user from the room:<roomid> set by calling redisP.removeUserFromRoom(roomid, user.id)
  2. Check if the room still exists by calling redisP.roomExists(roomid).
  3. If it still exists, we return ().pure[F] otherwise we delete the entry from the rooms hash, then delete those room messages and the room from Postgres by calling postgresP.deleteRoomMessages(roomid) and postgresP.deleteRoom(roomid) respectively.
  4. We also delete the entry from userroomid by calling redisP.deleteUserRoomMapping(user.id)
  5. Finally we tell the old room member that the user has left the room.

Broadcasting a Message

To broadcast messages to members in a room, we first need to retrieve a list of user ids from a room:<roomid> set:

new RedisProtocol[F] {
  ...
  override def listUserIds(roomid: RoomId): F[Set[UserId]] = {
    redis.sMembers(s"room:${roomid.id.toString}").map { set =>
      set.map(id => UserId(UUID.fromString(id)))
    }
  }
}.pure[F]

The listUserIds() function takes a RoomId and returns an F[Set[UserId]]. Here we use the sMembers() function which returns all the members in a Redis set, it takes the name of the set, which we supplied as room:<roomid>. This returns an F[Set[String]] which we convert to a Set[UserId] using the map method.

Next we’ll need to somehow convert those UserIds into Users:

new RedisProtocol[F] {
  ...
  override def getSelectedUsers(
      userid: UserId,
      rest: List[UserId]
  ): F[Option[List[User]]] = {
    val idMap: F[Map[String, String]] =
      if (rest.isEmpty) {
        redis
          .hmGet(
            "users",
            userid.id.toString
          )
      } else {
        redis
          .hmGet(
            "users",
            userid.id.toString,
            rest.map(_.id.toString): _*
          )
      }

    idMap.flatMap { usermap =>
      usermap.toList.map { (id, name) =>
        User(UUID.fromString(id), name).map(_.toOption)
      }.sequence
    } map (_.sequence)
  }
}.pure[F]

Here we use the hmGet() function that returns multiple values associated with UserIds from the users hash. It takes the hash name, and one or more field values and returns an F[Map[String, String]]. If rest is empty we pass only userid, otherwise we pass both rest, and userid to hmGet(). We then flatMap() on idMap, convert it to a list and map each value to a User. This produces a List[F[Option[User]]], so we call sequence twice to convert it into an F[Option[List[User]]].

Finally, let’s implement the broadcastMessage() function.

object ChatProtocol {
  ...
  private def broadcastMessage[F[_]: Monad: UUIDGen](
      redisP: RedisProtocol[F],
      roomid: RoomId,
      om: OutputMessage
  ): F[List[OutputMessage]] = {
    redisP.listUserIds(roomid).flatMap { uset =>
      val userlist = uset.toList
      if(userlist.isEmpty){
        List.empty[OutputMessage].pure[F]
      } else{
        redisP.getSelectedUsers(userlist.head, userlist.tail).map { maybelist =>
          maybelist match {
            case Some(ulist) =>
              ulist.map { u =>
                om match {
                  case SendToUser(user, msg)  => SendToUser(u, msg)
                  case ChatMsg(from, to, msg) => ChatMsg(from, u, msg)
                  case _                      => DiscardMessage
                }
              }
            case None => List.empty[OutputMessage]
          }
        }
      }

    }
  }
}

We start by getting a list of user ids by calling redisP.listUserIds(roomid), if its empty, we return List.empty[OutputMessage].pure[F].

Otherwise, we retrieve the list of Users by calling redisP.getSelectedUsers(userlist.head, userlist.tail). The rest of the implementation hasn’t changed from before.

Adding a User to a Room

This function adds a user to a room, however, in Redis we’ll need to add the user id to the room:<roomid> set, and add the userid -> roomid pair to the userroomid hash:

new RedisProtocol[F] {
  ...
  override def addUserToRoom(userid: UserId, roomid: RoomId): F[Unit] = {
    redis.sAdd(s"room:${roomid.id.toString}", userid.id.toString).void
  }
}.pure[F]

The addUserToRoom() function takes a UserId, and RoomId and returns an F[Unit]. Here we use the sAdd function to add the userid to the room:<roomid> redis set.

Next, the mapUserToRoom() function adds a userid -> roomid pair to the userroomid redis hash:

new RedisProtocol[F] {
  ...
  override def mapUserToRoom(userid: UserId, roomid: RoomId): F[Unit] = {
    val urmap: Map[String, String] = Map(
      (userid.id.toString, roomid.id.toString)
    )
    redis.hSet("userroomid", urmap).void
  }
}.pure[F]

The mapUserToRoom() function takes a userid and roomid and returns an F[Unit]. First, we create a urmap, which is a Map of these String values, then call the hSet method to add this pair to the userroomid hash.

Finally, here’s the addToRoom() function:

object ChatProtocol {
  ...
  private def addToRoom[F[_]: Monad: UUIDGen: Concurrent](
      redisP: RedisProtocol[F],
      postgresP: PostgresProtocol[F],
      user: User,
      room: Room
  ): F[List[OutputMessage]] = {
    for {
      _ <- redisP.addUserToRoom(user.id, room.id)
      _ <- redisP.mapUserToRoom(user.id, room.id)
      previousMessages <- fetchRoomMessages(postgresP, room.id, user)
      om <- broadcastMessage(
        redisP,
        room.id,
        SendToUser(
          user,
          s"${user.name.name} has joined the ${room.name.name} room"
        )
      )
    } yield previousMessages ++ om
  }
}

Here we add the user to the room:<roomid> set, then add the pair to the userroomid hash, and finally call fetchRoomMessages(postgresP, room.id, user) to retrieve all the previous messages from postgres, we’ll be implementing this next.

We then inform all the room members that the new user has joined the room and pass all the previous messages to the new user.

Fetching Current Messages

object ChatProtocol {
  ...
  private def fetchRoomMessages[F[_]: FlatMap: Concurrent](
    postgresP: PostgresProtocol[F],
    roomid: RoomId,
    user: User
  ): F[List[OutputMessage]] =
    postgresP
      .fetchMessages(roomid)
      .flatMap {
        _.map { case FetchMessage(msg, from) =>
          ChatMsg(from, user, msg.value)
        }.compile.toList
      }
}

To fetch messages from Postgres, we call postgresP.fetchMessages(roomid), this returns an F[Stream[F, FetchMessage]] which we map on to convert to ChatMsg, we then compile to list to return an F[List[OutputMessage]].

Creating a Chat Room

We already looked at the createRoom() redis implementation in the register() function section, now let’s look at how to implement it in ChatProtocol:

object ChatProtocol {
  ...
  private def createRoom[F[_]: Monad](
      redisP: RedisProtocol[F],
      postgresP: PostgresProtocol[F],
      room: String
  ): F[Either[String, Room]] =
    postgresP.createRoom(room).flatMap {
      case v @ Right(r) =>
        redisP.createRoom(r) *>
          v.pure[F]
      case l @ Left(err) => l.pure[F]
    }
}

Here we start by calling postgresP.createRoom(room) which returns an F[Either[String, Room]], if the creation is successful, we call redisP.createRoom(r) then return a Right(r) otherwise we return a Left(err).

Sending Messages

When we receive a chat message, we’ll need to save it into Postgres and then broadcast it:

import java.time.LocalDateTime
...
new ChatProtocol[F] {
  ...
  override def chat(user: User, text: String): F[List[OutputMessage]] = {
    redisP.getUsersRoomId(user).flatMap {
      case Some(roomid) =>
        postgresP.saveMessage(text, LocalDateTime.now(), user.id, roomid) *>
          broadcastMessage(redisP, roomid, ChatMsg(user, user, text))
      case None =>
        List(SendToUser(user, "You are not currently in a room")).pure[F]
    }
  }
}.pure[F]

We start by getting the user’s roomid, then calling postgresP.saveMessage() followed by the broadcastMessage() function. In case the user has no room id, we inform the user by returning List(SendToUser(user, "You are not currently in a room")).pure[F].

The Help Prompt

The implementation of this function remains unchanged from before:

new ChatProtocol[F] {
  ...
  override def help(user: User): F[OutputMessage] = {
    val text = """Commands:
                    | /help             - Show this text
                    | /room <room name> - Change to specified room
                    | /rooms            - List all rooms
                    | /members          - List members in current room
                """.stripMargin
    SendToUser(user, text).pure[F]
  }
}.pure[F]

Listing Rooms

Here we’ll need to retrieve a list of rooms from Redis:

new RedisProtocol[F] {
  ...
  override def listRooms: F[List[String]] = {
    redis.hVals("rooms")
  }
}.pure[F]

We do this through the hVals method which returns all the values from the rooms hash:

new ChatProtocol[F] {
  ...
  override def listRooms(user: User): F[List[OutputMessage]] = {
    redisP.listRooms.map { rooms =>
      val roomList = rooms.toList.sorted.mkString("Rooms:\n\t", "\n\t", "")
      List(SendToUser(user, roomList))
    }
  }
}.pure[F]

The listRooms ChatProtocol[F] method starts by calling redisP.listRooms function, then we sort and make a String from the resulting list and finally pass this value to the SendToUser() apply method.

Listing Members

This function gets the list of users in the room the user is in.

new ChatProtocol[F] {
  ...
  override def listMembers(user: User): F[List[OutputMessage]] = {
    val membersList: F[String] = redisP.getUsersRoomId(user).flatMap {
      case Some(roomid) =>
        redisP.listUserIds(roomid).flatMap { u =>
          redisP.getSelectedUsers(u.toList.head, u.toList.tail).map {
            case Some(lu) =>
              lu.map(_.name.name)
                .sorted
                .mkString("Room Members:\n\t", "\n\t", "")
            case None => ""
          }
        }
      case None => "You are not currently in a room".pure[F]
    }
    membersList.map(mlist => List(SendToUser(user, mlist)))
  }
}.pure[F]

We start by calling redisP.getUsersRoomId(user) to get the roomid of the user, if the user has no roomid, we return You are not currently in a room".pure[F].

Otherwise, we pass the roomid to redisP.listUserIds(roomid) to get a list of member ids which we pass to redisP.getSelectedUsers(u.toList.head, u.toList.tail) and convert to Users.

Finally, we produce a string of user names which we sequentially pass to the SendToUser() apply method

Disconnecting

We’ll need to be able to remove a user from the users hash in Redis before we implement disconnect:

new RedisProtocol[F] {
  ...
  override def deleteUser(userid: UserId): F[Long] = {
    redis.hDel("users", userid.id.toString)
  }
}.pure[F]

Here we use the hDel function similarly to deleteRoom() that we implemented previously:

new ChatProtocol[F] {
  ...
  override def disconnect(
      maybeuser: Option[User]
  ): F[List[OutputMessage]] = {
    maybeuser match {
      case Some(user) =>
        postgresP.deleteUser(user.id) *>
          redisP.deleteUser(user.id) *>
          removeFromCurrentRoom(redisP, postgresP, user)
      case None => List.empty[OutputMessage].pure[F]
    }
  }
}.pure[F]

We first delete the user from Postgres and Redis by calling postgresP.deleteUser(user.id) and redisP.deleteUser(user.id) then we finally call removeFromCurrentRoom() to remove the user from the current room.

Getting State

This is the last function to implement, however, to get an overview from Redis, quite a lot has to be done:

new RedisProtocol[F] {
  ...
  override def chatState: F[String] = {
    for {
      maybeusers <- redis.hLen("users")
      mayberooms <- redis.hLen("rooms")
      roomIds <- redis.hKeys("rooms")
      usersPerRoom <- roomIds.traverse { id =>
        redis.sMembers(s"room:$id").flatMap { uSet =>
          redis.hGet("rooms", id).flatMap {
            case Some(r) =>
              val uids = uSet.map(id => UserId(UUID.fromString(id))).toList
              getSelectedUsers(uids.head, uids.tail)
                .map {
                  _.getOrElse(List.empty[User])
                    .map(_.name.name)
                    .mkString(s"$r Room Members:\n\t", "\n\t", "")
                }
            case None =>
              s"An error occured while fetching rooms".pure[F]
          }
        }
      }
    } yield s"""
                          |<!Doctype html>
                          |<title>Chat Server State</title>
                          |<body>
                          |<pre>Users: ${maybeusers.getOrElse("")}</pre>
                          |<pre>Rooms: ${mayberooms.getOrElse("")}</pre>
                          |<pre>Overview:
                          |${usersPerRoom.mkString}
                          |</pre>
                          |</body>
                          |</html>
                    """.stripMargin
  }
}.pure[F]

From the top:

  1. We call hLen("users") to get the number of users currently, returning maybeusers as an Option[Long]
  2. We call hLen("rooms") to get the number of rooms currently, returning mayberooms as an Option[Long]
  3. We call hKeys("rooms") to get a list of all the room ids and return roomIds as a List[String]
  4. We then traverse on roomIds and for each roomid we get a Set[String], (uSet) containing user ids
  5. Next, we convert these user ids to UserIds forming a List[UserId] as uids
  6. Then we call getSelectedUsers(uids.head, uids.tail) to retrieve a list of User’s and convert that to a String.
  7. This forms a List[String] of users per room, usersPerRoom
  8. In the yield section, we get the number of users and rooms by calling maybeusers.getOrElse("") and mayberooms.getOrElse("") respectively
  9. Finally we convert usersPerRoom into a string as well

This whole function returns an F[String]:

new ChatProtocol[F] {
  ...
  override def chatState: F[String] = redisP.chatState
}.pure[F]

Lastly in ChatProtocol, we simply call redisP.chatState

Getting Input

Let’s create an InputMessage.scala file in the websocket folder and add the following contents:

package rockthejvm.websockets

import cats.effect.kernel.Ref
import rockthejvm.websockets.domain.user.*

trait InputMessage[F[_]] {
  def parse(
      userRef: Ref[F, Option[User]],
      text: String
  ): F[List[OutputMessage]]
}

Here the InputMessage[F[_]], trait contains a parse method whose signature remains unchanged, however now we went ahead and removed defaultRoom:

import cats.Monad
import cats.syntax.all.*

...
case class TextCommand(left: String, right: Option[String])

object InputMessage {
  def make[F[_]: Monad](
      chatP: ChatProtocol[F]
  ): F[InputMessage[F]] = {
    new InputMessage[F] {
      override def parse(
          userRef: Ref[F, Option[User]],
          text: String
      ): F[List[OutputMessage]] = {
        text.trim match {
          case "" => List(DiscardMessage).pure[F]
          case txt =>
            userRef.get.flatMap {
              case Some(user) => procesText4Reg(user, txt, chatP)
              case None => processText4UnReg(txt, chatP, userRef)
            }
        }
      }
    }.pure[F]
  }
}

The main difference here is that we have replaced Protocol[F] with ChatProtocol[F], and since we don’t have default room anymore we directly flatMap on userRef.get to produce either a Some or a None:

...
import cats.parse.Parser
import cats.parse.Parser.char
import cats.parse.Rfc5234.{alpha, sp, wsp}

...
object InputMessage {
  ...
  private def commandParser: Parser[TextCommand] = {
    val leftSide = (char('/').string ~ alpha.rep.string).string
    val rightSide: Parser[(Unit, String)] = sp ~ alpha.rep.string
    ((leftSide ~ rightSide.?) <* wsp.rep.?).map((left, right) =>
      TextCommand(left, right.map((_, s) => s))
    )
  }

  private def parseToTextCommand(
      value: String
  ): Either[Parser.Error, TextCommand] = {
    commandParser.parseAll(value)
  }
}

These two functions remain the same as before.

object InputMessage {
  ...
  private def processText4UnReg[F[_]: Monad](
      text: String,
      chatP: ChatProtocol[F],
      userRef: Ref[F, Option[User]]
  ): F[List[OutputMessage]] = {
    if (text.charAt(0) == '/') {
      parseToTextCommand(text).fold(
        _ =>
          List(
            ParsingError(
              None,
              "Characters after '/' must be between A-Z or a-z"
            )
          ).pure[F],
        v =>
          v match {
            case TextCommand("/name", Some(n)) =>
              chatP.register(n).flatMap{
                case sr @ SuccessfulRegistration(u,_) =>
                  for {
                    _ <- userRef.update(_ => Some(u))
                    om <- chatP.enterRoom(u, "Default")
                    help <- chatP.help(u)
                  } yield sr :: (help :: om)
                case ParsingError(None, err) =>
                  List(ParsingError(None, err)).pure[F]
                case _ =>  List(DiscardMessage).pure[F]
              }
            case _ =>
              List(UnsupportedCommand(None)).pure[F]
          }
      )
    } else {
      List(Register(None)).pure[F]
    }
  }
}

Here the changes in the processText4UnReg() function occur under TextCommand("/name", Some(n)), since we now call chatP.register(n) passing it the user name.

In the case of SuccessfulRegistration we update the userRef, and call chatP.enterRoom(u, "Default") to enter the Default room and send the user the welcome message and help menu. The rest of the logic remains unchanged:

...
import cats.Applicative

object InputMessage {
  ...
  private def procesText4Reg[F[_]: Applicative](
      user: User,
      text: String,
      chatP: ChatProtocol[F],
  ): F[List[OutputMessage]] = {
    if (text.charAt(0) == '/') {
      parseToTextCommand(text).fold(
        _ =>
          List(
            ParsingError(
              None,
              "Characters after '/' must be between A-Z or a-z"
            )
          ).pure[F],
        v =>
          v match {
            case TextCommand("/name", Some(n)) =>
              List(ParsingError(Some(user), "You can't register again")).pure[F]
            case TextCommand("/room", Some(r)) =>
              chatP.enterRoom(user, r)
            case TextCommand("/help", None) =>
              chatP.help(user).map(List(_))
            case TextCommand("/rooms", None) =>
              chatP.listRooms(user)
            case TextCommand("/members", None) =>
              chatP.listMembers(user)
            case _ => List(UnsupportedCommand(Some(user))).pure[F]
          }
      )
    } else {
      chatP.chat(user, text)
    }
  }
}

The procesText4Reg() function also remains unchanged except for the new ChatProtocol[F] methods

The Web App Routes

In this section we’ll continue to upgrade our application to the new ChatProtocol[F]:

package rockthejvm.websockets

import fs2.io.file.Files
import cats.effect.Temporal
import org.http4s.dsl.Http4sDsl
import org.http4s.server.websocket.WebSocketBuilder2
import cats.effect.std.Queue
import fs2.concurrent.Topic
import org.http4s.{HttpApp, HttpRoutes}
import org.http4s.StaticFile
import cats.syntax.all.*
import org.http4s.headers.`Content-Type`
import org.http4s.MediaType

class Routes[F[_]: Files: Temporal] extends Http4sDsl[F] {
  def service (
      wsb: WebSocketBuilder2[F],
      q: Queue[F, OutputMessage],
      t: Topic[F, OutputMessage],
      im: InputMessage[F],
      chatP: ChatProtocol[F]
  ): HttpApp[F] = {
    HttpRoutes.of[F] {
      case request @ GET -> Root / "chat.html" =>
        StaticFile
          .fromPath(
            fs2.io.file.Path("src/main/resources/chat.html"),
            Some(request)
          )
          .getOrElseF(NotFound())

      case GET -> Root / "metrics" =>
        def currentState: F[String] = {
          chatP.chatState
        }

        currentState.flatMap { currState =>
          Ok(currState, `Content-Type`(MediaType.text.html))
        }
    }
  }
}

The /chat.html route remains unchanged, however, the /metrics now produces the state of the application from chatP.chatState:

import cats.effect.kernel.Ref
import rockthejvm.websockets.domain.user.User
...
    HttpRoutes.of[F] {
      ...
      case GET -> Root / "ws" =>
        for {
          uRef <- Ref.of[F, Option[User]](None)
          uQueue <- Queue.unbounded[F, OutputMessage]
          ws <- wsb.build(
            send(t, uQueue, uRef),
            receive(chatP, im, uRef, q, uQueue)
          )
        } yield ws
    }

Under the /ws once again we replace protocol with chatP in the recieve() handle. Everything else remains unchanged.

...
import fs2.Stream
import org.http4s.websocket.WebSocketFrame
import scala.concurrent.duration.*

class Routes[F[_]: Files: Temporal] extends Http4sDsl[F] {
  ...
  private def handleWebSocketStream(
      wsf: Stream[F, WebSocketFrame],
      im: InputMessage[F],
      chatP: ChatProtocol[F],
      uRef: Ref[F, Option[User]]
  ): Stream[F, OutputMessage] = {
    for {
      sf <- wsf
      maybeuser <- Stream.eval(uRef.get)
      om <- Stream.evalSeq(
        sf match {
          case WebSocketFrame.Text(text, _) =>
            im.parse(uRef, text)
          case WebSocketFrame.Close(_) =>
            chatP.disconnect(maybeuser)
        }
      )
    } yield om
  }

  private def receive(
      chatP: ChatProtocol[F],
      im: InputMessage[F],
      uRef: Ref[F, Option[User]],
      q: Queue[F, OutputMessage],
      uQueue: Queue[F, OutputMessage]
  ): Pipe[F, WebSocketFrame, Unit] = { wsf =>
    handleWebSocketStream(wsf, im, chatP, uRef)
      .evalMap { m =>
        uRef.get.flatMap {
          case Some(_) =>
            q.offer(m)
          case None =>
            uQueue.offer(m)
        }
      }
      .concurrently {
        Stream
          .awakeEvery(30.seconds)
          .map(_ => KeepAlive)
          .foreach(uQueue.offer)
      }
  }
}

Here we now use the new ChatProtocol[F], for both handleWebSocketStream() and receive() functions:

import io.circe.generic.auto.*
import io.circe.syntax.*

class Routes[F[_]: Files: Temporal] extends Http4sDsl[F] {
  ...
  private def filterMsg(
      msg: OutputMessage,
      userRef: Ref[F, Option[User]]
  ): F[Boolean] = {
    msg match {
      case DiscardMessage => false.pure[F]
      case sendtouser @ SendToUser(_, _) =>
        userRef.get.map { _.fold(false)(u => sendtouser.forUser(u)) }
      case chatmsg @ ChatMsg(_, _, _) =>
        userRef.get.map { _.fold(false)(u => chatmsg.forUser(u)) }
      case _ => true.pure[F]
    }
  }

  private def processMsg(msg: OutputMessage): WebSocketFrame = {
    msg match {
      case KeepAlive => WebSocketFrame.Ping()
      case msg @ _   => WebSocketFrame.Text(msg.asJson.noSpaces)
    }
  }

  private def send(
      t: Topic[F, OutputMessage],
      uQueue: Queue[F, OutputMessage],
      uRef: Ref[F, Option[User]]
  ): Stream[F, WebSocketFrame] = {
    def uStream =
      Stream
        .fromQueueUnterminated(uQueue)
        .filter {
          case DiscardMessage => false
          case _              => true
        }
        .map(processMsg)

    def mainStream =
      t.subscribe(maxQueued = 1000)
        .evalFilter(filterMsg(_, uRef))
        .map(processMsg)

    Stream(uStream, mainStream).parJoinUnbounded
  }
}

The rest of the above 3 functions also remain unchanged.

The Server

The server function is also now uses ChatProtocol[F]:

package rockthejvm.websockets

import cats.effect.kernel.Async
import fs2.io.file.Files
import fs2.io.net.Network
import cats.effect.std.Queue
import fs2.concurrent.Topic
import com.comcast.ip4s.*
import cats.syntax.all.*
import org.http4s.ember.server.EmberServerBuilder

object Server {
  def server[F[_]: Async: Files: Network](
      q: Queue[F, OutputMessage],
      t: Topic[F, OutputMessage],
      im: InputMessage[F],
      chatP: ChatProtocol[F]
  ): F[Unit] = {
    val host = host"0.0.0.0"
    val port = port"8080"
    EmberServerBuilder
      .default[F]
      .withHost(host)
      .withPort(port)
      .withHttpWebSocketApp(wsb =>
        new Routes().service(wsb, q, t, im, chatP)
      )
      .build
      .useForever
      .void
  }
}

The Main Program

In this section, we have several updates that involve initializing Redis and Postgres:

package rockthejvm.websockets

import cats.effect.IOApp
import cats.effect.kernel.Resource
import cats.effect.IO
import dev.profunktor.redis4cats.RedisCommands
import dev.profunktor.redis4cats.Redis
import dev.profunktor.redis4cats.effect.Log.Stdout.given

object Program extends IOApp.Simple {
  private def makeRedis: Resource[IO, RedisCommands[IO, String, String]] =
    Redis[IO].utf8("redis://127.0.0.1:6379")
}

Here we create the makeRedis function by calling Redis[IO].utf8() and passing it a redis uri value of redis://127.0.0.1:6379. This returns a Resource that we’ll use in our program.

import skunk.Session
import natchez.Trace.Implicits.noop
...
object Program extends IOApp.Simple {
  ...
  private def makePostgres: Resource[IO, Resource[IO, Session[IO]]] =
    Session
      .pooled[IO](
        host = "localhost",
        port = 5432,
        user = "postgres",
        password = Some("password"),
        database = "websocket",
        max = 10
      )
}

To create the Postgres Resource, we use the Session.pooled[IO] function where we provide the host as localhost, a port, user, password, and database as 5432, postgres, Some("password") and websocket all tallying with what we provided in the docker container.

Lastly, we provide a 10 as the number of concurrent sessions for our Resource:

import cats.effect.std.Queue
import fs2.concurrent.Topic
import fs2.Stream
import scala.concurrent.duration.*
import Server.server
...

object Program extends IOApp.Simple {
  ...
  def program: IO[Unit] = {
    makeRedis.use { redis =>
      makePostgres.use { session =>
        for {
          postgresprotocol <- PostgresProtocol.make[IO](session)
          redisprotocol <- RedisProtocol.make[IO](redis)
          chatprotocol <- ChatProtocol.make[IO](redisprotocol, postgresprotocol)
          im <- InputMessage.make[IO](chatprotocol)
          q <- Queue.unbounded[IO, OutputMessage]
          t <- Topic[IO, OutputMessage]
          s <- Stream(
            Stream.fromQueueUnterminated(q).through(t.publish),
            Stream
              .awakeEvery[IO](30.seconds)
              .map(_ => KeepAlive)
              .through(t.publish),
            Stream.eval(server[IO](q, t, im, chatprotocol))
          ).parJoinUnbounded.compile.drain
        } yield s
      }
    }
  }

  override def run: IO[Unit] = program
}

Finally, in the program function, we took out ChatState and Protocol and added the PostgresProtocol and RedisProtocol which we provide as arguments to the ChatProtocol make() function.

Serving HTML

Since we upgraded the User case class, we’ll also need to make changes to chat.html:

        socket.onmessage = function (event) {
            ...
            } else if (obj.ChatMsg) {
                output.append(colorText(obj.ChatMsg.from.name.name + " : " + obj.ChatMsg.msg, "purple"))
            }
            ...
        };

In the obj.ChatMsg branch we now add obj.ChatMsg.from.name.name to access the user name. Everything else remains the same.

Now to run our application, we first need to start our Redis and Postgres Docker containers using Docker-Compose, and finally our application server. The application should function closely to the original.

Conclusion

In conclusion, this article has gone in-depth on how to implement Redis and Postgres in a Scala application using the redis4cats and skunk libraries. Now we can persist our messages, and rip all the benefits of storing our information in Redis such as high availability and persistence.

In this version we simply dump all the previous messages to the new user but this should be done progressively whenever the user scrolls up, however, this was beyound the scope of this tutorial.

Share on:

Herbert Kateu

Herbert Kateu

Experienced software developer that is very passionate about functional programming in Scala, and always keeps tabs on whats trending in the Scala ecosystem. When he is not programming he plays jazz guitar and takes his dog on walks.