Unity Pattern: Stateless Scenes
Unity does not provide much direction on how to transfer game state between scenes. Some community suggestions are singletons, static variables or DontDestroyOnLoad. But these have other effects that you might not want. Here, I will share a pattern inspired by functional programming I call Stateless Scenes. This pattern, inspired by asynchronous APIs, leads to a much cleaner design.
We consider a scene to be: a long running operation that, given its inputs, eventually produces an outcome. We can rephrase that as: a function that accepts a callback (a.k.a. Action) in its argument list:
To actually implement this, we have to make a call to SceneManager.LoadScene and transfer the MySceneParams to a MonoBehaviour within the scene, along with the callback. It’s simpler to include the callback in the Params, so we just transfer a single variable across the LoadScene divide. To transfer this state, we use the lesser evil of a single static variable for a short period of time.
The nice thing is that the caller of the scene now gets to decide what comes afterwards (inversion of control). This makes testing much easier, as the test harness can use the scene in a very different way to how the main game might.
Another side benefit is that if the scene is run directly from the editor, the params will not be overwritten and will keep their values to what is defined in the scene. This lets you test individual scenes without scene switching.
I hope you enjoy this pattern, its brought sanity to my game and I hope it will help yours too!
Discussion