Unidirectional Reactive Architecture. This is aReactiveSwiftcounterpart ofRxFeedback.
Requirements for iOS apps have become huge. Our code has to manage a lot of state e.g. server responses, cached data, UI state, routing etc. Some may say that Reactive Programming can help us a lot but, in the wrong hands, it can do even more harm to your code base.
The goal of this library is to provide a simple and intuitive approach to designing reactive state machines.
State
is the single source of truth. It represents a state of your system and is usually a plain Swift type (which doesn't contain any ReactiveSwift primitives). Your state is immutable. The only way to transition from oneState
to another is to emit anEvent
.
structResults<T:JSONSerializable>{
letpage:Int
lettotalResults:Int
lettotalPages:Int
letresults:[T]
staticfuncempty()->Results<T>{
returnResults<T>(page:0,totalResults:0,totalPages:0,results:[])
}
}
structContext{
varbatch:Results<Movie>
varmovies:[Movie]
staticvarempty:Context{
returnContext(batch:Results.empty(),movies:[])
}
}
enumState{
caseinitial
casepaging(context:Context)
caseloadedPage(context:Context)
caserefreshing(context:Context)
caserefreshed(context:Context)
caseerror(error:NSError,context:Context)
caseretry(context:Context)
}
Represents all possible events that can happen in your system which can cause a transition to a newState
.
enumEvent{
casestartLoadingNextPage
caseresponse(Results<Movie>)
casefailed(NSError)
caseretry
}
A Reducer is a pure function with a signature of(State, Event) -> State
.WhileEvent
represents an action that results in aState
change, it's actually not whatcausesthe change. AnEvent
is just that, a representation of the intention to transition from one state to another. What actually causes theState
to change, the embodiment of the correspondingEvent
,is a Reducer. A Reducer is the only place where aState
can be changed.
staticfuncreduce(state:State,event:Event)->State{
switch event{
case.startLoadingNextPage:
return.paging(context:state.context)
case.response(letbatch):
varcopy=state.context
copy.batch=batch
copy.movies+=batch.results
return.loadedPage(context:copy)
case.failed(leterror):
return.error(error:error,context:state.context)
case.retry:
return.retry(context:state.context)
}
}
WhileState
represents where the system is at a given time,Event
represents a trigger for state change, and aReducer
is the pure function that changes the state depending on current state and type of event received, there is not as of yet any type to emit events given a particular current state. That's the job of theFeedback
.It's essentially a "processing engine", listening to changes in the currentState
and emitting the corresponding next events to take place. It's represented by a pure function with a signature ofSignal<State, NoError> -> Signal<Event, NoError>
.Feedbacks don't directly mutate states. Instead, they only emit events which then cause states to change in reducers.
publicstructFeedback<State,Event>{
publicletevents:(Scheduler,Signal<State,NoError>)->Signal<Event,NoError>
}
funcloadNextFeedback(for nearBottomSignal:Signal<Void,NoError>)->Feedback<State,Event>{
returnFeedback(predicate:{!$0.paging}){_in
returnnearBottomSignal
.map{Event.startLoadingNextPage}
}
}
funcpagingFeedback()->Feedback<State,Event>{
returnFeedback<State,Event>(skippingRepeated:{$0.nextPage}){(nextPage)->SignalProducer<Event,NoError>in
returnURLSession.shared.fetchMovies(page:nextPage)
.map(Event.response)
.flatMapError{(error)->SignalProducer<Event,NoError>in
returnSignalProducer(value:Event.failed(error))
}
}
}
funcretryFeedback(for retrySignal:Signal<Void,NoError>)->Feedback<State,Event>{
returnFeedback<State,Event>(skippingRepeated:{$0.lastError}){_->Signal<Event,NoError>in
returnretrySignal.map{Event.retry}
}
}
funcretryPagingFeedback()->Feedback<State,Event>{
returnFeedback<State,Event>(skippingRepeated:{$0.retryPage}){(nextPage)->SignalProducer<Event,NoError>in
returnURLSession.shared.fetchMovies(page:nextPage)
.map(Event.response)
.flatMapError{(error)->SignalProducer<Event,NoError>in
returnSignalProducer(value:Event.failed(error))
}
}
}
- As you can see from the diagram above we always start with an initial state.
- Every change to the
State
will be then delivered to allFeedback
loops that were added to the system. Feedback
then decides whether any action should be performed with a subset of theState
(e.g calling API, observe UI events) by dispatching anEvent
,or ignoring it by returningSignalProducer.empty
.- Dispatched
Event
then goes to theReducer
which applies it and returns a new value of theState
. - And then cycle starts all over (see 2).
letincrement=Feedback<Int,Event>{_in
returnself.plusButton.reactive
.controlEvents(.touchUpInside)
.map{_inEvent.increment}
}
letdecrement=Feedback<Int,Event>{_in
returnself.minusButton.reactive
.controlEvents(.touchUpInside)
.map{_inEvent.decrement}
}
letsystem=SignalProducer<Int,NoError>.system(initial:0,
reduce:{(count,event)->Intin
switch event{
case.increment:
returncount+1
case.decrement:
returncount-1
}
},
feedbacks:[increment,decrement])
label.reactive.text<~system.map(String.init)
TODO
This is a community fork of theReactiveFeedbackproject (with the MIT license) from Babylon Health.