Reactive programming is a programming paradigm based on the concept of an asynchronous data flow. A data flow is like a river: it can be observed, filtered, manipulated, or merged with a second flow to create a new flow for a new consumer.
A key concept of reactive programming is events. Events can be awaited, can trigger procedures, and can trigger other events. Events are the only proper way to map our real world to our software: we open a window if it's too hot inside. In the same way, we want the Total
cell to update when we change some values in our spreadsheet (the propagation of the changes) or our robot to turn around the moment it reaches the wall (reaction to events).
Nowadays, one of the most common scenarios for reactive programming is UIs: we have mobile apps that have to react to network calls, user touch input, and system alerts. In this world, software has to be event-driven and reactive because real life is event-driven.
Functional reactive programming is an idea from the late 90s that inspired Erik Meijer, a computer scientist at Microsoft, to design and develop the Microsoft Rx library.
Rx is a reactive extension for Microsoft .NET. Rx provides an easy way to create asynchronous, event-driven programs using Observable sequences. A developer can model an asynchronous data stream using Observables, query the Observables using LINQ syntax, and easily manage concurrency with Schedulers.
Rx makes well-known concepts, such as the push approach, easy to implement and consume. In a reactive world, we can't just wait for a function result, a network call, or a database query to return and pretend that the user won't notice or even complain about it. Every moment we wait for something, we lose the opportunity to do other things in parallel, provide a better user experience, and free our software from the chains of sequential, blocking programming.
.NET Observable relates to .NET Enumerable according to the following table:
.NET Observable |
Single return value |
Multiple return values |
---|---|---|
Pull/Synchronous/Interactive |
T |
|
Push/Asynchronous/Reactive |
|
|
The push approach reverses the problem: instead of asking for a result and waiting, the developer simply asks for a result and gets notified when the result is available. The developer provides a clear sequence of reactions to the events that are going to happen. For every event, the developer provides a reaction; for example, the user is asked to log in and submit a form with his username and password. The application executes the network call for the login and states what is going to happen:
Show a success message and store the user's profile
Show an error message
As you can see with the push approach, the developer doesn't wait for the result. He will be notified when the result arrives. In the meantime, he can do whatever he wants:
Show a progress dialog
Store the username and password for future logins
Preload something he knows would take some time the moment the login succeeds
In 2012, at Netflix, they started to realize that their architecture was having a hard time to properly adapt to the amount of users they had. They decided to redesign their architecture to reduce the number of REST calls. Instead of having dozens of REST calls and letting the client process the data as needed, they decided to create a single optimized REST call based on the clients' needs.
To achieve this goal, they decided to go reactive, and they started to port .NET Rx to the JVM. They didn't want to target just Java; they decided to target the JVM instead, having the possibility to provide a new tool for every JVM language on the market: Java, Closure, Groovy, Scala, and so on.
With a post on the Netflix tech blog in February 2013, Ben Christensen and Jafar Husain showed RxJava to the world for the first time.
The key concepts were:
As for .NET, RxJava Observable is the push equivalent of Iterable, which is pull. The pull approach is a block-and-wait approach: the consumer pulls values from the source, blocking the thread until the producer provides new values.
The push approach works on subscription and reaction: the consumer subscribes to new values' emissions; the producer pushes these new values when they are available, and notifies the consumer. At this point, the consumer, well, consumes them. The push approach is clearly more flexible, because from a logical and practical point of view, the developer can simply ignore if the data he needs comes synchronously or asynchronously; his code will still work.
From a pure Java point of view, the RxJava Observable class extends the classic Gang of Four Observer pattern concept.
It adds three missing abilities:
The producer can now signal that there is no more data available: the
onCompleted()
eventThe producer can now signal that an error occurred: the
onError()
eventRxJava Observables can be composed instead of nested, saving the developer from the callback hell
Observables and Iterables share a similar API; lots of operations that we can perform on Iterable can be performed on Observable too. Of course, due to the "flow" nature of Observables, we don't have equivalents for methods such as Iterable.remove()
.
Pattern |
Single return value |
Multiple return values |
---|---|---|
Synchronous |
|
|
Asynchronous |
|
|
From a semantic point of view, RxJava is .NET Rx. From a syntactical point of view, Netflix takes care of porting every Rx method, keeping in mind Java code conventions and basic patterns.
In this chapter, we took the initial steps into the reactive world. We learned how Rx was born, from Microsoft .NET to Netflix RxJava. We learned the similarities and differences of the reactive approach compared with the classic approach.
In the next chapter, we will learn what Observables are, and how we can create them to bring reactive programming into our daily coding life.