GatedMiddleware

Turn SwiftRex middlewares on or off dynamically

Gated middleware is a middleware that holds an inner middleware that could be either active or not.

There are two gate variations that can be used: by action or by state.

GatedMiddleware by action:

It holds an internal state, called gate state, that determines whether or not the inner middleware should be in active or bypass mode. This can be changed dynamically.

It starts with an initial gate state, called "default gate state". From that point on, it will evaluate all incoming actions to detect a "control action", which is an action for switching on or off the gate state. This control action is detected thanks to a control action map closure or a control action map KeyPath configured in the GatedMiddleware's init, which from a given input action allows the user to inform either or not this is a control action, by returning an Optional instance of that ControlAction (or nil in case it's a regular action).

The init also requires some comparison values, for turnOn or turnOff the gate. If it's a control action, and it's equals to turn on, it will set the inner middleware to active. If it's a control action, and it's equals to turn off, it will set the inner middleware to bypass. If it's not a control action, or it's not equals to any of the comparison values, the gate will remain untouched.

The gated middleware by action will ALWAYS forward control actions to inner middlewares, regardless of their gate state (active or bypass) and regardless of the turn on/turn off comparison result. This will allow important actions like disabling or enabling the inner middleware for control actions, so for example, even for when we close the gate we still want to tell the inner middleware that it's gonna be bypassed and it should kill all of its timers or async side-effects.

GatedMiddleware by state:

It won't hold any internal state, instead, it will use some state from your App Global State. You're responsible for mutating this state from your own reducers. At any point that the state tells that this middleware is active, it's gonna handle actions and be able to dispatch new actions. However, whenever the state is set to bypass, this middleware will ignore incoming actions and won't be able to dispatch any new action.

When handling actions, the state is evaluated before reducers, so whatever state is set BEFORE reducer, will define if the inner middleware will be called before and after the reducer, even if the reducer changes that value. An action that changes the state from active to bypass, will trigger inner middleware before and after the reducer, and after the reducer that value will be already set to bypass, so you can stop timers and async tasks. An action that changes the state from bypass to active, will not trigger the inner middleware before the reducer nor after it, so you may want to send a second action to start the middleware timers again, because the gated middleware can't do that for you.

Examples:

1. Gated by action, simple example

// sourcery: Prism
enum AppAction {
    case something
    case anotherSomething
    case dynamicMiddlewares(DynamicMiddlewareAction)
}

// sourcery: Prism
enum DynamicMiddlewareAction: Equatable {
    case toggleCrashReportsMiddleware(enable: Bool)
}

let gatedCrashReportsMiddleware =
    CrashReportsMiddleware
        .init()
        .gated(
            controlAction: \AppAction.dynamicMiddlewares?.toggleCrashReportsMiddleware?.enable,
            default: .active            
        )

2. Gated by action, with custom comparison:

// sourcery: Prism
enum AppAction {
    case something
    case anotherSomething
    case dynamicMiddlewares(DynamicMiddlewareAction)
}

// sourcery: Prism
enum DynamicMiddlewareAction: Equatable {
    case controlCrashReportsMiddleware(controlAction: MiddlewareControlAction)
}

// sourcery: Prism
enum MiddlewareControlAction: Equatable {
    case activate
    case bypass
    case sayHello
}

let gatedCrashReportsMiddleware =
    CrashReportsMiddleware
        .init()
        .gated(
            controlAction: \AppAction.dynamicMiddlewares?.controlCrashReportsMiddleware,
            turnOn: MiddlewareControlAction.activate,
            turnOff: MiddlewareControlAction.bypass,
            default: .active
        )
        // in this example, MiddlewareControlAction.activate will activate the crash reports,
        // MiddlewareControlAction.bypass will disable the crash reports, and
        // MiddlewareControlAction.sayHello won't change the gate state, but will be forwarded to the
        // crash reports middleware regardless of its current state.

3. Gated by state, simple example

struct AppState {
    var isCrashReportEnabled: Bool
}

let gatedCrashReportsMiddleware =
    CrashReportsMiddleware
        .init()
        .gated(
            state: \AppState.isCrashReportEnabled
        )

Lift, Composed and other higher-order middlewares

You can also lift the inner middleware before gating it, in case the AppActions or AppStates don't match. Evidently lift can also be done after the gated middleware if this is what you need.

Gating composed middlewares will disable or enable all of them at once, and the control action will be the same and be forwarded to all the inner middlewares all the times. If this is not what you need, you need disabling them individually, you can first gate them and with the gated collection you compose them.

This has no interference on Reducers and this doesn't change the AppState in any way. GatedMiddleware only matches AppState with its inner middleware to allow proxying the getState context.

About the examples and code generation

All examples above use Sourcery Prism templates to simplify traversing action trees, but the GatedMiddleware offers closures in case you prefer switch/case approach or other custom functions.