Tracking Game Scores with Event Sourcing (Event Sourcing Part 3)

Introduction

In this post we’ll use our knowledge from the previous articles (Part 1 and Part 2) to build a simple Event Sourcing solution.

We will be building a Scoring System for a Computer Game e.g. Candy Crush. It’s main purpose is to record player scores, and then to expose the top 10 scores with the corresponding player names.

Logical Flow

The main flow of this system is that when a player finishes the game, a GameTerminated event is saved to the datastore. This event looks like:

Copycopy code to clipboard
{
  PlayerId: 1,
  PlayerName: "MajorLazer",
  GameScore: 999
}

Once the GameTerminated event is saved to the datastore, a process is triggered to recalculate a materialised view. This materialised view is the Top 10 Players of the game.

A Http Rest Api exposes the “Top 10 Materialised View” at:

Copycopy code to clipboard
GET /api/toptenscores

With a response:

Copycopy code to clipboard
[
  {
    PlayerName: "MajorLazer",
    GameScore: 999
  },
  {
    PlayerName: "MinorLazer",
    GameScore: 888
  }
  ...
]

Components

1. Game Events Api for GameTerminated events, this will be hosted on an Azure Function.
2. A Game Events Datastore, this will be Cosmos DB.
3. A Top 10 Materialised View of the games scores, this will be an Azure Function with a Cosmos DB Trigger.
4. A Scores Api that will expose the Top 10 Players of the Game, this will be another Azure Function.

Game Event Api

An Azure Function which has the following endpoint.

Copycopy code to clipboard
POST /api/gameterminated 

{
  PlayerId: 1,
  PlayerName: "MajorLazer",
  GameScore: 999
}

Response: 201 Created, 400 Bad Request

Process:

1. Validates incoming data
2. Maps data to database entity
3. Saves entity to Game Events Datastore

Game Events Datastore

A Cosmos DB which saves all the Game Events. In this case we have only implemented the GameTerminated event. If we had more than one event, we would have an abstract parent event class for the standard properties and inherit it into concrete class such as GameTerminated.

The Cosmos DB will have 1 collection called Event, in which all the events are stored. A GameTerminated event will map to this type:

Copycopy code to clipboard
public abstract record GameEvent(long PlayerId, string PlayerName);

public record GameTerminated(long PlayerId, string PlayerName, long GameScore) : GameEvent(PlayerId, PlayerName);

Top 10 Materialised View

The materialised view we are building is to quickly view the top 10 players of the game. This avoids aggregating over potentially millions of events and causing an expensive query, both in terms of money and computation. In this case, the Materialised View doesn't provide much benefit over calling MAX(10)on the data, but I believe it is a good demonstration of how to implement event sourcing and a materialised view.

To set up this materialised view we will create an Azure Function with a Cosmos DB trigger. This trigger will start the function every time an event is saved to the Cosmos DB Collection and call it with the latest event.

Process:

1. Validate incoming data
2. Map data to PlayerScore entity
3. If PlayerScore is higher than any of the scores in the TopTenScores collection, then insert it in the correct position.
4. Else if the Player is already in the top 10 and it is a higher score, remove the old PlayerScore and insert it in the correct position.
5. Else if, there are less than 10 PlayerScores and less than the existing PlayerScores, insert it in the end.
6. Else, ignore PlayerScore

Scores Api

An Azure Function which has the following endpoint.

Copycopy code to clipboard
GET /api/toptenscores
 
[
  {
    PlayerName: "MajorLazer",
    GameScore: 999
  },
  {
    PlayerName: "MinorLazer",
    GameScore: 888
  }
  ...
]


Response: 200 Ok, 500 Server Error

Process:

1. Get all data from TopTenScores collection in Cosmos DB
2. Maps data to view model, to discard redundant properties
3. Returns Top Ten Scores as a list

Conclusion

In this post we have outlined the components needed to implement a simple Event Sourcing solution for a Game Scoring System. Given time I will update this article with a link to a Github Repository containing the C# code and Terraform to set it up.