Versioning and Migrating Game Save Files

In this post I briefly described how I handle saving the game state in a JSON file. One thing that I knew I had to address is migration of game save files from one version of the game to the next. During alpha and beta it may not be as important to be forward compatible, but when the game is released I’ll have to accommodate for this. Otherwise if any of the models in the game change, an old game save file may load in an inconsistent state, or worse, fail to load altogether.

I put together a simple migration mechanism that allows me to directly manipulate the JSON before it’s deserialized into the game entities/objects. This is not too dissimilar to many libraries used for migrating databases:

Continue reading

Creating Fog of War in Games


Update: It has been brought to my attention that what I have implemented below is essentially the Marching Squares algorithm. Still it might be useful for those who don’t want to deal with too much theory.

One of the core systems in Startup Freak is feature building. The player is presented with a grid which I call “Market Space”. At any given time only some parts of the market is visible and known to player, and this is the area he/she can build features on. In order to build more features the player needs to perform market research. Visually it’s not too dissimilar to an RTS map, hence the title “fog of war”.

Initially this visible area was simply one circle that increased in radius. Recently I made a change so that the player could explicitly pick an area they want to research. As such, the fog of war could have a lot more variations and non-symmetric shapes. I wanted to have a smooth transition into the dark areas, but this turned out to be non-trivial:

Continue reading

Art and Dealing with Multiple Resolutions in Unity

The big news with Startup Freak is that I am now working with an amazing artist to help with the game art. I found Rob Hayes on Indie Gamer Forums and his professionalism and  quality of work have been superb. You can see some of his work over on his portfolio site:


Hayes Art & Design


Continue reading

Saving and Loading Games in Unity with Json.NET

As I expected game persistence turned out to be a bit of beast and has taken up all of my development time for the past 2 weeks. The result is not perfect and in retrospect I would have designed for persistence earlier. Nevertheless here are a few tips and ideas that may help you on your projects:

Not Persisting MonoBehaviours

The decision to move all game state out of MonoBehaviour classes was the right one when it came to persistence. I don’t have to worry about Unity’s inner workings. I persist and load my own entity classes, and inject them into MonoBehaviours which in turn instantiate all the needed graphics and UI. I had to do some minor tweaks to these components but for the most part, it just worked.

Persistence and Dependency Injection

I’m using Zenject for dependency injection so the persistence had to fit into that. After several false starts, I ended up with something like this:


  • All top level objects, like Product, Company, and Finances have a persistence model associated with them (ProductPersistModel, CompanyPersistModel, etc), and a corresponding IPersister implementation.
  • When saving, we resolve all top-level objects from the container, create their persistence models using IPersisters, and put them in a final GameSave object. The majority of this is automated with the use of the container and marker interfaces so we don’t have to explicitly instantiate any classes.
  • Finally the GameSave object is serialized to a file using Json.NET.


  • We load the GameSave object from file.
  • All persistence model objects are registered with container during initialization.
  • The persistence models are (optionally) injected into the constructor of the relevant class. So ProductPersistModel is injected into Product which can use it to hydrate itself.
  • The rest of the game is initialized per normal.

Now it might seem overkill to have specialized persistence classes. However it does prevent me from having to sprinkle the [JsonIgnore] and other persistence specific code everywhere. Do keep in mind that the persistence model only exists for the top level objects (aggregate roots) and not any “inner classes”. For those I fully serialize and deserialize them, which bring me to my next point.

Serializing Child Entities

As I mentioned, only the aggregate roots have a special persistence model and are resolved from the container. Other, inner classes are persisted fully and instantiated using Json.NET. That means there are certain restrictions on these which I actually consider to be good (DDD) coding practice anyway:

  • Any dependencies (like services) should not be passed to the constructor and stored in a member variable. Instead the entity should either inform outer objects via events, or dependencies should be passed to the methods that need them. For example the Feature class informs its parent (Product) when it has completed using an event called FeatureCompleted rather than explicitly holding an instance of Product.
  • I use a custom ContractResolver to get Json.NET to serialize all private and public fields and properties. It looks something like this:

public class GameContractResolver : CamelCasePropertyNamesContractResolver
  protected override JsonProperty CreateProperty(MemberInfo member, MemberSerialization memberSerialization)
    var property = base.CreateProperty(member, memberSerialization);
    property.Readable = true;
    property.Writable = true;

    return property;

  protected override List<MemberInfo> GetSerializableMembers(Type objectType)
    const BindingFlags bindingFlags = BindingFlags.Instance | BindingFlags.NonPublic | BindingFlags.Public;
    var properties = objectType.GetProperties(bindingFlags).Where(p => p.CanWrite);
    var fields = objectType.GetFields(bindingFlags)
      .Where(f => !f.GetCustomAttributes(typeof(CompilerGeneratedAttribute), false).Any())
      .Where(f => !typeof(Delegate).IsAssignableFrom(f.FieldType));

    var allMembers = properties.Cast<MemberInfo>().Union(fields.Cast<MemberInfo>()).ToList();
    return allMembers;

Types and References in Json

I have enabled 2 features in Json.NET that greatly simplify persistence. Under normal circumstances where you are using Json for communication between two systems this would be code smell and not advisable. However here I’m just using it for persistence and so it’s less of an issue:

  • TypeNameHandling.Auto – This option adds a $type property to all Json objects that have an ambiguous type (ie interfaces and generics) so that during desensitization the engine knows what type to instantiate. The downside of course, is that if a type gets renamed or moved you’ll have trouble.
  • PreserveReferencesHandling.Objects – This is an awesome little option. I was originally going to implement it myself before finding out about it. In short when this is enabled if a particular object is referenced from multiple places (which will invariably happen), Json.NET will serialize it once with a $id property and for all other references simply uses $ref property to refer to that instance. So during deserilization you won’t get two instances of something that is only meant to have one instance.
  • ReferenceLoopHandling.Serialize – This goes hand in hand with the previous option, essentially allowing you to serialize a circular dependency (which while might indicate poor design, is sometimes unavoidable).

Dealing with Database Items

Like any game some assets like text, images, definitions, etc. are loaded into the game from a data store (in my case files). For these it made sense not to persist them with the rest of the game but rather load them again from the database. This way any updates to those assets would be correctly applied to a game that was previously saved.

Doing this is easy enough with Json.NET. I created a new interface to be put on any database class, called IDataStoreItemWithId which contains an Id property. I then implemented a JsonConverter class that uses the database to save/load such objects:

public class DataStoreItemConverter : JsonConverter
  readonly IDataStore _dataStore;

  public DataStoreItemConverter(IDataStore dataStore)
    _dataStore = dataStore;

  public override void WriteJson(JsonWriter writer, object value, JsonSerializer serializer)
    var item = (IDataStoreItemWithId) value;

    var jObj = new JObject { { "id", item.Id } };

  public override object ReadJson(JsonReader reader, Type objectType, object existingValue, JsonSerializer serializer)
    var jObj = JToken.ReadFrom(reader);
    return _dataStore.GetById(objectType, jObj.Value<string>("id"));

  public override bool CanConvert(Type objectType)
    return objectType.GetInterfaces().Any(i => i == typeof(IDataStoreItemWithId));

What’s Next

I still have a bunch more to do on persistence but I think at this point I have hit all the pain points and now it’s just a mechanical process of getting everything into the persistence pipeline and tested. There is probably another week of that left to do.

I have also been moving forward on the art front with some exciting new progress, but that’s a topic for another post…

Practical Unit Tests for Games

A little while ago I wrote an answer to a unit testing related question here on reddit which seemed to resonate with many. One of the points I made there was that testing, more than anything else, improves code design. Now it’s all good and well stating theories like that but an example goes a long way to show how this process might unfold. Here is a situation I ran into a couple of days ago which I think is worth sharing.

Continue reading

A Random Note


I know of very few games that don’t have a randomized element of some sort, be it randomized levels, items, AI behavior, and timing, among others. There are a lot of great articles online about random numbers in computing, including this one with a fantastic discussion of using random numbers in procedural generation. For Startup Freak I also have quite a few random elements, so I have formalized it a little with an interface: Continue reading

Unity Coding Conventions


Certain coding patterns have emerged in the game’s codebase, many of them influenced by my time spent writing web and client applications. I know some “experts” will get up in arms about how games are different and shouldn’t be written like other types of apps but I tend to disregard philosophical arguments in favor of pragmatism. On a related note, in this reddit thread a couple of people commented on a previous post on using a messaging system in Unity, mentioning it could suffer from performance problem. I must explain: I am not advocating that the patterns I’m using are either the best way, or even the only way to do things in the same code base. Indeed I only use the messenger/mediator pattern where the dependencies are very loose and there isn’t high traffic communication. Additionally I think some of these patterns lend themselves more to the type of game I’m making: turn-based, heavy on events / UI and light on real-time mechanics. With that, let’s have a look at the solution structure:

Continue reading

A Simple Message Bus in Unity


The idea of having many decoupled components in a system, with some sort of message bus to facilitate communication is nothing new. It gets used all the way from small applications to complex and distributed software systems. It has become especially useful in Startup Freak because of how interconnected many of the game systems are. Hiring an employee, for example, causes multiple UI components as well as the simulation to change.

In this post I’ll go over my implementation of a messaging broker in Unity.

Continue reading

Unity 3D Editor Workflow for UI-Heavy Games

Working on Startup Freak game has been a unique challenge in terms of the amount of UI it requires. As I have previously mentioned, the new Unity UI really does make life a lot easier than the previous alternatives. I have also started to find a better workflow which works for me, though to be fair I haven’t tested this long enough to know if I’ll hit some limitations.

Continue reading