Framework Overview
This document contains a high-level description of the different components within the ReactiveSwift framework, and an attempt to explain how they work together and divide responsibilities. This is meant to be a starting point for learning about new modules and finding more specific documentation.
For examples and help understanding how to use RAC, see the README or the Design Guidelines.
Events
An event, represented by the Event
type, is the formalized representation
of the fact that something has happened. In ReactiveSwift, events are the centerpiece
of communication. An event might represent the press of a button, a piece
of information received from an API, the occurrence of an error, or the completion
of a long-running operation. In any case, something generates the events and sends them over a
signal to any number of observers.
Event
is an enumerated type representing either a value or one of three
terminal events:
- The
value
event provides a new value from the source. - The
failed
event indicates that an error occurred before the signal could finish. Events are parameterized by anErrorType
, which determines the kind of failure that’s permitted to appear in the event. If a failure is not permitted, the event can use typeNever
to prevent any from being provided. - The
completed
event indicates that the signal finished successfully, and that no more values will be sent by the source. - The
interrupted
event indicates that the signal has terminated due to cancellation, meaning that the operation was neither successful nor unsuccessful.
Signals
A signal, represented by the Signal
type, is any series of events
over time that can be observed.
Signals are generally used to represent event streams that are already “in progress”, like notifications, user input, etc. As work is performed or data is received, events are sent on the signal, which pushes them out to any observers. All observers see the events at the same time.
Users must observe a signal in order to access its events. Observing a signal does not trigger any side effects. In other words, signals are entirely producer-driven and push-based, and consumers (observers) cannot have any effect on their lifetime. While observing a signal, the user can only evaluate the events in the same order as they are sent on the signal. There is no random access to values of a signal.
Signals can be manipulated by applying primitives to them.
Typical primitives to manipulate a single signal like filter
, map
and
reduce
are available, as well as primitives to manipulate multiple signals
at once (zip
). Primitives operate only on the value
events of a signal.
The lifetime of a signal consists of any number of value
events, followed by
one terminating event, which may be any one of failed
, completed
, or
interrupted
(but not a combination).
Terminating events are not included in the signal’s values—they must be
handled specially.
Pipes
A pipe, created by Signal.pipe()
, is a signal
that can be manually controlled.
The method returns a signal and an observer. The signal can be controlled by sending events to the observer. This can be extremely useful for bridging non-RAC code into the world of signals.
For example, instead of handling application logic in block callbacks, the blocks can simply send events to the observer. Meanwhile, the signal can be returned, hiding the implementation detail of the callbacks.
Signal Producers
A signal producer, represented by the SignalProducer
type, creates
signals and performs side effects.
They can be used to represent operations or tasks, like network
requests, where each invocation of start()
will create a new underlying
operation, and allow the caller to observe the result(s). The
startWithSignal()
variant gives access to the produced signal, allowing it to
be observed multiple times if desired.
Because of the behavior of start()
, each signal created from the same
producer may see a different ordering or version of events, or the stream might
even be completely different! Unlike a plain signal, no work is started (and
thus no events are generated) until an observer is attached, and the work is
restarted anew for each additional observer.
Starting a signal producer returns a disposable that can be used to interrupt/cancel the work associated with the produced signal.
Just like signals, signal producers can also be manipulated via primitives
like map
, filter
, etc.
Every signal primitive can be “lifted” to operate upon signal producers instead,
using the lift
method.
Furthermore, there are additional primitives that control when and how work
is started—for example, times
.
Observers
An observer is anything that is waiting or capable of waiting for events
from a signal. Within RAC, an observer is represented as an Observer
that accepts Event
values.
Observers can be implicitly created by using the callback-based versions of the
Signal.observe
or SignalProducer.start
methods.
Lifetimes
When observing a signal or starting a signal producer, it is important to consider how long the observation should last. For example, when observing a signal in order to update a UI component, it makes sense to stop observing it once the component is no longer on screen. This idea is expressed in ReactiveSwift by the Lifetime
type.
import Foundation
final class SettingsController {
// Define a lifetime for instances of this class. When an instance is
// deinitialized, the lifetime ends.
private let (lifetime, token) = Lifetime.make()
func observeDefaultsChanged(_ defaults: UserDefaults = .standard) {
// `take(during: lifetime)` ensures the observation ends when this object
// is deinitialized
NotificationCenter.default.reactive
.notifications(forName: UserDefaults.didChangeNotification, object: defaults)
.take(during: lifetime)
.observeValues { [weak self] _ in self?.defaultsChanged(defaults) }
}
private func defaultsChanged(_ defaults: UserDefaults) {
// perform some updates
}
}
The token
is a Lifetime.Token
, which we need to keep a strong reference to in order for the Lifetime
to work.
(Note: It’s crucial that there is only a single strong reference to token
, so that it is deinitialized at the same time as self
.)
Lifetime
is useful any time that an observation might outlive the observer:
- In the
NotificationCenter
example above, without aLifetime
the observation would never complete (leaking memory and wasting CPU cycles). - Consider a signal producer that fires a network request — incorporating a
Lifetime
might allow the request to be automatically cancelled if the observer is deinitialized.
Actions
An action, represented by the Action
type, will do some work when
executed with an input. While executing, zero or more output values and/or a
failure may be generated.
Actions are useful for performing side-effecting work upon user interaction, like when a button is clicked. Actions can also be automatically disabled based on a property, and this disabled state can be represented in a UI by disabling any controls associated with the action.
Properties
A property, represented by the PropertyProtocol
,
stores a value and notifies observers about future changes to that value.
The current value of a property can be obtained from the value
getter. The
producer
getter returns a signal producer that will send
the property’s current value, followed by all changes over time. The signal
getter returns a signal that will send all changes over time, but not the initial value.
The <~
operator can be used to bind properties in different ways. Note that in
all cases, the target has to be a binding target, represented by the BindingTargetProvider
. All mutable property types, represented by the MutablePropertyProtocol
, are inherently binding targets.
property <~ signal
binds a signal to the property, updating the property’s value to the latest value sent by the signal.property <~ producer
starts the given signal producer, and binds the property’s value to the latest value sent on the started signal.property <~ otherProperty
binds one property to another, so that the destination property’s value is updated whenever the source property is updated.
Properties provide a number of transformations like map
, combineLatest
or zip
for manipulation similar to signal and signal producer
Disposables
A disposable, represented by the Disposable
protocol, is a mechanism
for memory management and cancellation.
When starting a signal producer, a disposable will be returned.
This disposable can be used by the caller to cancel the work that has been started
(e.g. background processing, network requests, etc.), clean up all temporary
resources, then send a final interrupted
event upon the particular
signal that was created.
Observing a signal may also return a disposable. Disposing it will prevent the observer from receiving any future events from that signal, but it will not have any effect on the signal itself.
For more information about cancellation, see the RAC Design Guidelines.
Schedulers
A scheduler, represented by the Scheduler
protocol, is a
serial execution queue to perform work or deliver results upon.
Signals and signal producers can be ordered to deliver events on a specific scheduler. Signal producers can additionally be ordered to start their work on a specific scheduler.
Schedulers are similar to Grand Central Dispatch queues, but schedulers support
cancellation (via disposables), and always execute serially.
With the exception of the ImmediateScheduler
, schedulers do not
offer synchronous execution. This helps avoid deadlocks, and encourages the use
of signal and signal producer primitives instead of blocking work.
Schedulers are also somewhat similar to NSOperationQueue
, but schedulers
do not allow tasks to be reordered or depend on one another.