Home

With great power, comes great responsibility

If you had to choose what is better - short and concise code or abstractions visible at the first glance - what would you choose?

Got your answer? Fine, let’s look at the examples.

Iterables

Let’s look at the standard abstraction of iterable collections, available in Java and Scala (example written in Scala):

1 trait Iterable[T] {
2   def iterator(): Iterator[T]
3 }
4 
5 trait Iterator[T] {
6   def hasNext: Boolean
7   def next(): T
8 }

Iterable is an abstraction for all collections which can be stepped through one-by-one.

Iterator is a data structure (again, an abstraction) that allows to iterate over a sequence of elements. It has two methods - first for checking if there is a next element available, and second which returns the next element.

So far, so good - we can explicitly see an intent and usage for such abstractions. It is readable at the first sight.

Observables

Things can get complicated if we are dealing with asynchronous operations, which return a collection as well. Iterables define only a collection of existing elements, not possible results.

But we can introduce new abstractions. Let’s look at the example (again, written in Scala):

 1 trait Observable[T] {
 2   def subscribe(observer: Observer[T]): Subscription
 3 }
 4 
 5 trait Observer[T] {
 6   def onNext(value: T): Unit
 7   def onError(error: Throwable): Unit
 8   def onCompleted(): Unit
 9 }
10 
11 trait Subscription {
12   def unsubscribe(): Unit
13 }

We can compare these two concepts and definitely see similarities, such as that Observable is similar to Iterable. The same for Observer - Iterator pair. Also, we can easily spot an intent and usage here.

But when we add a new trait called Subscription, there is a slight and subtle change in approach for Observable.

It receives an Observer instance as an argument, instead of returning it as presented in the Iterable / Iterator approach.

This subtle change has huge consequences in modelling and understanding these two abstractions. We trade the possibility of pulling things out from an Iterable collection to pushing a model for the Observable which will notify an Observer that something appears, but not for free - we receive an ability to model an asynchronous operation which returns a collection of elements (in other words: we transformed from pulling mode to the pushing mode).

Power of expressiveness

What if this Observable / Iterable approach was to be presented in one line as a composite type?

Ready? Superb, but first we will bring up two definitions.

Try[T]

1 val sumTry = for {
2   int1 <- Try(Integer.parseInt("1"))   // int1 is Success(1).
3   int2 <- Try(Integer.parseInt("a"))   // int2 is Failure(...).
4 } yield {
5   int1 + int2
6 }

Try is an analog for the try...catch block. Instead of having an exception thrown and having to deal with it immediately in the same thread, it disconnects the error handling and recovery.

Option[T]

1 val optionResult = Option(null)   // optionResult is None.
2 val optionResult2 = Option(1)     // optionResult2 is Some(1).

Option represents optional values, returning an instance of Some(A) if A exists, or None if it does not.

Composition

The same abstraction, of Observable can be built on top of Iterable and a simple type, which is a composite of two previously explained elements:

val collection: Iterable[Try[Option[T]]]

It’s a sequence which introduces the same abstraction, but definitely loses the intent. Usage is not visible at the first glance as well (how are we communicating that something changed?).

Summary

The whole point of this article is to illustrate that intent and usage can be provided by certain abstractions. Of course, these abstractions can be verbose and seem unnecessary at first sight, but if we chase expressiveness, and introduce it out of hand we will loose details and initial intent.

The post is strongly inspired by Coursera course “Principles of Reactive Programming” provided by courtesy of Erik Meijer, Martin Odersky and Roland Kuhn.

References

  1. Coursera - Principles of Reactive Programming
Home