Immutability is a useful tool. Think of it as an extension of your favourite editor or ReSharper. Using immutable values helps eliminate some of the easy to make mistakes when programming. Even if your language doesn’t fully support immutability you can still take advantage of the benefits.
What Is Immutability?
Immutability is something which does not change. You set a value and it doesn’t change through its lifetime. There are absolutely no mutations of that value. Unexpected mutations are a very common cause of issues in a program. Mutations can happen seemingly anywhere. They can happen concurrently from another thread causing all sorts of issues which are difficult to track down. During my talk I said “there is nothing wrong with global state, the problem is when global state changes.” I was trying to make them think about what the real danger was in their applications. A value which cannot change is immutable.
The main benefit with using immutable values is the removal of unexpected changes. This leads to easier to reason about code. Future readers will have confidence that the values set don’t change later in the function or by any function. We limit the mutation or changes to specific areas of the code. This in turn let’s us define locations as being dangerous areas. Solutions using immutable values require less effort to move to a concurrent paradigm. Mutation is the bane of concurrent programming. Limiting mutation lowers the complexity of understanding the solution.
The main disadvantage of immutable values is always having to make a copy of the values or object to apply the changes. This is not a big disadvantage if the language supports immutability. It is tedious in a language that doesn’t as this would have to be implemented by the developer. There are also some performance issues that may be noticed in mutations of many values in a collection. This is due to all the copying, updating, and garbage collection. This again can be handled by the language if supported. Mutations are a form of optimization.
Levels of Immutability
There are many functional languages that have rigid no mutations allowed. Some notable ones are Haskell and Erlang. If you come from a language that heavily relies on mutating variables, then these rigid languages can really force you to think of different ways to solve the task at hand. There are middle ground languages that support both immutable and mutable values. F# and Clojure are good languages that are immutable by default. It’s up to the developer to mark the values as mutable and accept the responsibility of that decision. Using an immutable by default language will allow you to design your solution using immutability and later modify to using mutations if the performance is required.
What’s nice about Clojure and F# is you get visual cues as to the dangerous areas in your code. F# uses the mutable keyword and a left arrow <- to define mutation.
let mutable x = 5 x <- 6
Clojure uses the earmuffs operator.
(def ^:dynamic *there-be* "dragons")
Immutability is a tool we can use to help avoid easy to make mistakes of unintentional mutations. This can lead to solutions that are easier to reason about as the code is not littered with mutations. Limiting mutations to specific areas allowing the reader to focus on those dangerous areas instead of worrying about unexpected changes. Like any tool, there are times when it doesn't fit. If performance is paramount to the solution, using mutation may be the way to go. In saying that, it's generally a good practice to make the code correct and then add mutations to make it fast.