After a long time learning and working with object-oriented programming, I took a step back to think about system complexity.
“Complexity is anything that makes software hard to understand or to modify." — John Outerhout
Doing some research, I found functional programming concepts like immutability and pure function. Those concepts are big advantages to build side-effect-free functions, so it is easier to maintain systems — with some other benefits.
In this post, I will tell you more about functional programming, and some important concepts, with a lot of code examples. In Javascript!
What is functional programming?
Functional programming is a programming paradigm — a style of building the structure and elements of computer programs — that treats computation as the evaluation of mathematical functions and avoids changing-state and mutable data — Wikipedia
Pure functions
The first fundamental concept we learn when we want to understand functional programming is pure functions. But what does that really mean? What makes a function pure?
So how do we know if a function is pure or not? Here is a very strict definition of purity:
It returns the same result if given the same arguments (it is also referred as
deterministic)It does not cause any observable side effects
It returns the same result if given the same arguments
Imagine we want to implement a function that calculates the area of a circle. An impure function would receive radius as the parameter, and then calculate radius * radius * PI:
const PI = 3.14;
function calculateArea(radius) {
return radius * radius * PI;
}
calculateArea(10); // returns 314.0
Why is this an impure function? Simply because it uses a global object that was not passed as a parameter to the function.
Now imagine some mathematicians argue that the PI value is actually 42and change the value of the global object.
Our impure function will now result in 10 * 10 * 42 = 4200. For the same parameter (radius = 10), we have a different result. Let's fix it!
const PI = 3.14;
function calculateArea(radius, pi) {
return radius * radius * pi;
}
calculateArea(10, PI); // returns 314.0
TA-DA 🎉! Now we’ll always pass thePI value as a parameter to the function. So now we are just accessing parameters passed to the function. No external object.
For the parameters
radius = 10&PI = 3.14, we will always have the same the result:314.0For the parameters
radius = 10&PI = 42, we will always have the same the result:4200
Reading Files
If our function reads external files, it’s not a pure function — the file’s contents can change.
function charactersCounter(text) {
return `Character count: ${text.length}`;
}
function analyzeFile(filename) {
let fileContent = open(filename);
return charactersCounter(fileContent);
}
Random number generation
Any function that relies on a random number generator cannot be pure.
function yearEndEvaluation() {
if (Math.random() > 0.5) {
return "You get a raise!";
} else {
return "Better luck next year!";
}
}
It does not cause any observable side effects
Examples of observable side effects include modifying a global object or a parameter passed by reference.
Now we want to implement a function to receive an integer value and return the value increased by 1.
let counter = 1;
function increaseCounter(value) {
counter = value + 1;
}
increaseCounter(counter);
console.log(counter); // 2
We have the counter value. Our impure function receives that value and re-assigns the counter with the value increased by 1.
Observation: mutability is discouraged in functional programming.
We are modifying the global object. But how would we make it pure? Just return the value increased by 1. Simple as that.
let counter = 1;
function increaseCounter(value) {
return value + 1;
}
increaseCounter(counter); // 2
console.log(counter); // 1
See that our pure function increaseCounter returns 2, but the counter value is still the same. The function returns the incremented value without altering the value of the variable.
If we follow these two simple rules, it gets easier to understand our programs. Now every function is isolated and unable to impact other parts of our system.
Pure functions are stable, consistent, and predictable. Given the same parameters, pure functions will always return the same result. We don’t need to think of situations when the same parameter has different results — because it will never happen.
Pure functions benefits
The code’s definitely easier to test. We don’t need to mock anything. So we can unit test pure functions with different contexts:
Given a parameter
A→ expect the function to return valueBGiven a parameter
C→ expect the function to return valueD
A simple example would be a function to receive a collection of numbers and expect it to increment each element of this collection.
let list = [1, 2, 3, 4, 5];
function incrementNumbers(list) {
return list.map(number => number + 1);
}
We receive the numbers array, use map incrementing each number, and return a new list of incremented numbers.
incrementNumbers(list); // [2, 3, 4, 5, 6]
For the input [1, 2, 3, 4, 5], the expected output would be [2, 3, 4, 5, 6].
Immutability
Unchanging over time or unable to be changed.
after it’s created. **If you want to change an immutable object, you can’t. Instead, you create a new object with the new value.**
In Javascript we commonly use the for loop. This next for statement has some mutable variables.
var values = [1, 2, 3, 4, 5];
var sumOfValues = 0;
for (var i = 0; i < values.length; i++) {
sumOfValues += values[i];
}
sumOfValues // 15
For each iteration, we are changing the i and the sumOfValue state. But how do we handle mutability in iteration? Recursion!
let list = [1, 2, 3, 4, 5];
let accumulator = 0;
function sum(list, accumulator) {
if (list.length == 0) {
return accumulator;
}
return sum(list.slice(1), accumulator + list[0]);
}
sum(list, accumulator); // 15
list; // [1, 2, 3, 4, 5]
accumulator; // 0
So here we have the sum function that receives a vector of numerical values. The function calls itself until we get the list empty (our recursion base case). For each "iteration" we will add the value to the total accumulator.
immutable. The list and the accumulator variables are not changed. It keeps the same value.
Observation: Yes! We can use reduce to implement this function. We will cover this in the Higher Order Functions topic.
It is also very common to build up the final state of an object. Imagine we have a string, and we want to transform this string into a url slug.
In OOP in Ruby, we would create a class, let’s say, UrlSlugify. And this class will have a slugify! method to transform the string input into a url slug.
class UrlSlugify
attr_reader :text
def initialize(text)
@text = text
end
def slugify!
text.downcase!
text.strip!
text.gsub!(' ', '-')
end
end
UrlSlugify.new(' I will be a url slug ').slugify! # "i-will-be-a-url-slug"
Beautiful! It’s implemented! Here we have imperative programming saying exactly what we want to do in each slugify process — first lower case, then remove useless white spaces and, finally, replace remaining white spaces with hyphens.
But we are mutating the input state in this process.
We can handle this mutation by doing function composition, or function chaining. In other words, the result of a function will be used as an input for the next function, without modifying the original input string.
let string = " I will be a url slug ";
function slugify(string) {
return string.toLowerCase()
.trim()
.split(" ")
.join("-");
}
slugify(string); // i-will-be-a-url-slug
Here we have:
toLowerCase: converts the string to all lower casetrim: removes whitespace from both ends of a stringsplitandjoin: replaces all instances of match with replacement in a given string
We combine all these 4 functions and we can "slugify" our string.
Referential transparency
Let’s implement a square function:
function square(n) {
return n * n;
}
This pure function will always have the same output, given the same input.
square(2); // 4
square(2); // 4
square(2); // 4
// ...
Passing 2 as a parameter of the square function will always returns 4. So now we can replace the square(2) with 4. That's it! Our function is referentially transparent.
Basically, if a function consistently yields the same result for the same input, it is referentially transparent.
pure functions + immutable data = referential transparency
With this concept, a cool thing we can do is to memoize the function. Imagine we have this function:
function sum(a, b) {
return a + b;
}
And we call it with these parameters:
sum(3, sum(5, 8));
The sum(5, 8) equals 13. This function will always result in 13. So we can do this:
sum(3, 13);
And this expression will always result in 16. We can replace the entire expression with a numerical constant and memoize it.
Functions as first-class entities
The idea of functions as first-class entities is that functions are also treated as values and used as data.
Functions as first-class entities can:
refer to it from constants and variables
pass it as a parameter to other functions
return it as result from other functions
The idea is to treat functions as values and pass functions like data. This way we can combine different functions to create new functions with new behavior.
Imagine we have a function that sums two values and then doubles the value. Something like this:
function doubleSum(a, b) {
return (a + b) * 2;
}
Now a function that subtracts values and the returns the double:
function doubleSubtraction(a, b) {
return (a - b) * 2;
}
These functions have similar logic, but the difference is the operators functions. If we can treat functions as values and pass these as arguments, we can build a function that receives the operator function and use it inside our function. Let’s build it!
function sum(a, b) {
return a + b;
}
function subtraction(a, b) {
return a - b;
}
function doubleOperator(f, a, b) {
return f(a, b) * 2;
}
doubleOperator(sum, 3, 1); // 8
doubleOperator(subtraction, 3, 1); // 4
Done! Now we have an f argument, and use it to process a and b. We passed the sum and subtraction functions to compose with the doubleOperator function and create a new behavior.
Higher-order functions
When we talk about higher-order functions, we mean a function that either:
takes one or more functions as arguments, or
returns a function as its result
The doubleOperator function we implemented above is a higher-order function because it takes an operator function as an argument and uses it.
You’ve probably already heard about filter, map, and reduce. Let's take a look at these.
Filter
Given a collection, we want to filter by an attribute. The filter function expects a true or false value to determine if the element should or should not be included in the result collection. Basically, if the callback expression is true, the filter function will include the element in the result collection. Otherwise, it will not.
A simple example is when we have a collection of integers and we want only the even numbers.
Imperative approach
An imperative way to do it with Javascript is to:
create an empty array
evenNumbersiterate over the
numbersarraypush the even numbers to the
evenNumbersarray
var numbers = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10];
var evenNumbers = [];
for (var i = 0; i < numbers.length; i++) {
if (numbers[i] % 2 == 0) {
evenNumbers.push(numbers[i]);
}
}
console.log(evenNumbers); // (6) [0, 2, 4, 6, 8, 10]
We can also use the filter higher order function to receive the even function, and return a list of even numbers:
function even(number) {
return number % 2 == 0;
}
let listOfNumbers = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10];
listOfNumbers.filter(even); // [0, 2, 4, 6, 8, 10]
One interesting problem I solved on Hacker Rank FP Path was the Filter Array problem. The problem idea is to filter a given array of integers and output only those values that are less than a specified value X.
An imperative Javascript solution to this problem is something like:
var filterArray = function(x, coll) {
var resultArray = [];
for (var i = 0; i < coll.length; i++) {
if (coll[i] < x) {
resultArray.push(coll[i]);
}
}
return resultArray;
}
console.log(filterArray(3, [10, 9, 8, 2, 7, 5, 1, 3, 0])); // (3) [2, 1, 0]
We say exactly what our function needs to do — iterate over the collection, compare the collection current item with x, and push this element to the resultArray if it pass the condition.
Declarative approach
But we want a more declarative way to solve this problem, and using the filter higher order function as well.
A declarative Javascript solution would be something like this:
function smaller(number) {
return number < this;
}
function filterArray(x, listOfNumbers) {
return listOfNumbers.filter(smaller, x);
}
let numbers = [10, 9, 8, 2, 7, 5, 1, 3, 0];
filterArray(3, numbers); // [2, 1, 0]
Using this in the smaller function seems a bit strange in the first place, but is easy to understand.
this will be the second parameter in the filter function. In this case, 3 (the x) is represented by this. That's it.
We can also do this with maps. Imagine we have a map of people with their name and age.
let people = [
{ name: "TK", age: 26 },
{ name: "Kaio", age: 10 },
{ name: "Kazumi", age: 30 }
];
And we want to filter only people over a specified value of age, in this example people who are more than 21 years old.
function olderThan21(person) {
return person.age > 21;
}
function overAge(people) {
return people.filter(olderThan21);
}
overAge(people); // [{ name: 'TK', age: 26 }, { name: 'Kazumi', age: 30 }]
Summary of code:
we have a list of people (with
nameandage).we have a function
olderThan21. In this case, for each person in people array, we want to access theageand see if it is older than 21.we filter all people based on this function.
Map
The idea of map is to transform a collection.
The map method transforms a collection by applying a function to all of its elements and building a new collection from the returned values.
Let’s get the same people collection above. We don't want to filter by “over age” now. We just want a list of strings, something like TK is 26 years old. So the final string might be :name is :age years old where :name and :age are attributes from each element in the people collection.
In a imperative Javascript way, it would be:
var people = [
{ name: "TK", age: 26 },
{ name: "Kaio", age: 10 },
{ name: "Kazumi", age: 30 }
];
var peopleSentences = [];
for (var i = 0; i < people.length; i++) {
var sentence = people[i].name + " is " + people[i].age + " years old";
peopleSentences.push(sentence);
}
console.log(peopleSentences); // ['TK is 26 years old', 'Kaio is 10 years old', 'Kazumi is 30 years old']
In a declarative Javascript way, it would be:
function makeSentence(person) {
return `${person.name} is ${person.age} years old`;
}
function peopleSentences(people) {
return people.map(makeSentence);
}
peopleSentences(people); // ['TK is 26 years old', 'Kaio is 10 years old', 'Kazumi is 30 years old']
The whole idea is to transform a given array into a new array.
Another interesting Hacker Rank problem was the update list problem. We just want to update the values of a given array with their absolute values.
For example, the input [1, 2, 3, -4, 5]needs the output to be [1, 2, 3, 4, 5]. The absolute value of -4 is 4.
A simple solution would be an in-place update for each collection value.
var values = [1, 2, 3, -4, 5];
for (var i = 0; i < values.length; i++) {
values[i] = Math.abs(values[i]);
}
console.log(values); // [1, 2, 3, 4, 5]
We use the Math.abs function to transform the value into its absolute value, and do the in-place update.
This is not a functional way to implement this solution.
First, we learned about immutability. We know how immutability is important to make our functions more consistent and predictable. The idea is to build a new collection with all absolute values.
Second, why not use map here to "transform" all data?
My first idea was to test the Math.abs function to handle only one value.
Math.abs(-1); // 1
Math.abs(1); // 1
Math.abs(-2); // 2
Math.abs(2); // 2
We want to transform each value into a positive value (the absolute value).
Now that we know how to do absolute for one value, we can use this function to pass as an argument to the map function. Do you remember that a higher order function can receive a function as an argument and use it? Yes, map can do it!
let values = [1, 2, 3, -4, 5];
function updateListMap(values) {
return values.map(Math.abs);
}
updateListMap(values); // [1, 2, 3, 4, 5]
Wow. So beautiful! 😍
Reduce
The idea of reduce is to receive a function and a collection, and return a value created by combining the items.
A common example people talk about is to get the total amount of an order. Imagine you were at a shopping website. You’ve added Product 1, Product 2, Product 3, and Product 4 to your shopping cart (order). Now we want to calculate the total amount of the shopping cart.
In imperative way, we would iterate the order list and sum each product amount to the total amount.
var orders = [
{ productTitle: "Product 1", amount: 10 },
{ productTitle: "Product 2", amount: 30 },
{ productTitle: "Product 3", amount: 20 },
{ productTitle: "Product 4", amount: 60 }
];
var totalAmount = 0;
for (var i = 0; i < orders.length; i++) {
totalAmount += orders[i].amount;
}
console.log(totalAmount); // 120
Using reduce, we can build a function to handle the amount sum and pass it as an argument to the reduce function.
let shoppingCart = [
{ productTitle: "Product 1", amount: 10 },
{ productTitle: "Product 2", amount: 30 },
{ productTitle: "Product 3", amount: 20 },
{ productTitle: "Product 4", amount: 60 }
];
const sumAmount = (currentTotalAmount, order) => currentTotalAmount + order.amount;
function getTotalAmount(shoppingCart) {
return shoppingCart.reduce(sumAmount, 0);
}
getTotalAmount(shoppingCart); // 120
Here we have shoppingCart, the function sumAmount that receives the current currentTotalAmount , and the order object to sum them.
The getTotalAmount function is used to reduce the shoppingCart by using the sumAmount and starting from 0.
Another way to get the total amount is to compose map and reduce. What do I mean by that? We can use map to transform the shoppingCart into a collection of amount values, and then just use the reduce function with sumAmount function.
const getAmount = (order) => order.amount;
const sumAmount = (acc, amount) => acc + amount;
function getTotalAmount(shoppingCart) {
return shoppingCart
.map(getAmount)
.reduce(sumAmount, 0);
}
getTotalAmount(shoppingCart); // 120
The getAmount receives the product object and returns only the amount value. So what we have here is [10, 30, 20, 60]. And then the reduce combines all items by adding up. Beautiful!
We took a look at how each higher order function works. I want to show you an example of how we can compose all three functions in a simple example.
Talking about shopping cart, imagine we have this list of products in our order:
let shoppingCart = [
{ productTitle: "Functional Programming", type: "books", amount: 10 },
{ productTitle: "Kindle", type: "eletronics", amount: 30 },
{ productTitle: "Shoes", type: "fashion", amount: 20 },
{ productTitle: "Clean Code", type: "books", amount: 60 }
]
We want the total amount of all books in our shopping cart. Simple as that. The algorithm?
filter by book type
transform the shopping cart into a collection of amount using map
combine all items by adding them up with reduce
let shoppingCart = [
{ productTitle: "Functional Programming", type: "books", amount: 10 },
{ productTitle: "Kindle", type: "eletronics", amount: 30 },
{ productTitle: "Shoes", type: "fashion", amount: 20 },
{ productTitle: "Clean Code", type: "books", amount: 60 }
]
const byBooks = (order) => order.type == "books";
const getAmount = (order) => order.amount;
const sumAmount = (acc, amount) => acc + amount;
function getTotalAmount(shoppingCart) {
return shoppingCart
.filter(byBooks)
.map(getAmount)
.reduce(sumAmount, 0);
}
getTotalAmount(shoppingCart); // 70
Done! 🎉
Resources
I’ve organised some resources I read and studied. I’m sharing the ones that I found really interesting. For more resources, visit my Functional Programming Github repository.
Intros
Pure functions
Immutable data
Higher-order functions
Declarative Programming
That’s it!
Hey people, I hope you had fun reading this post, and I hope you learned a lot here! This was my attempt to share what I’m learning.
Here is the repository with all codes from this article.
Come learn with me. I’m sharing resources and my code in this Learning Functional Programming repository.
I also wrote an FP post but using mainly Clojure ❤.
I hope you saw something useful to you here. And see you next time! :)
My Twitter, Medium & Github. ☺
TK.
Top comments (11)
This is one of the better posts I've seen on here so far. Thank you so much for this! Lots of useful resources and very well written and explained examples.
Thank you! I'm glad you liked!
I hope I could help you in any way!
Hi TK,
I really like the fp style in JS, but how can I (or you) justify it, if, when I refer to your last shoppingCart example...
imperativeGetTotalAmount, also gives me the correct value and will be much quicker to execute, particularly when the "shoppingCart" array length is length 10,000 or greater? Additionally, the code length of imperativeGetTotalAmount is shorter.
Once again, I must reiterate, I am a huge fp fan and I am playing devil's advocate here in posing this question, but I feel that one must be able to defend a particular style,(oop, imperative, fp), if it is to be accepted by the community in general.
Thanks for your post and it's huge list of useful resources.
This solution returns an amazingly fast execution time for 100,000 generated items of approx. 2.5ms
or, alternatively ...
visit runkit link
I thought
caculateAreais a pure function.Am I understand wrongly?It will return the same value if you call it with same variable, because
PIis a constant.If you use replace
PIwith3.14159orMath.PI, it should still be a pure function.Yes, Maple, maybe my example was a bit confusing. My attempt was to show that the "constant"
PI(an external variable) could change (in my example, I changed it to42).If the
PIvalue changes, ourcalculateAreafunction will change the output for the same input (radius). The whole idea is to show that the function to be pure, it needs to not change the output, for the same input.I will try give another example with a variable instead of a constant.
For the input
100, we will have120as a result. But if our tax changes, our output will change, for the same input (100).To make our function pure, we can pass the
taxvariable to the function as a parameterFor the inputs 100 and 20, it will always be 120
For the inputs 100 and 100, it will always be 200
With this example it is a bit less confusing I think.
Thanks for the feedback!
This is gold, fantastic! I'm a new programmer trying to wrap my head around functional programming and this has been extremely helpful!
This is Great post! thank you so much.
Great post ! I strongly recommend "JavaScript Allongé" to learn functional programming in JS : leanpub.com/javascriptallongesix !
Awesome resource, thanks for sharing!
Great article. Thank you
I'm a little concerned to find this is still a "new concept" in 2018. SOLID, SRP, separation of concerns, functions without side-effects, please use these.
Some comments have been hidden by the post's author - find out more