Functional JavaScript: Learn by refactoring

Nov 10, 2016

In this blog post we'll learn some fundamentals of functional programing by comparing it with imperative code. We'll write a function in the imperative style, iteratively refactor it to a functional style, and compare the results.

Imperative style

Imperative code starts with data, takes some number of steps that mutate that data, and finally returns the mutated data.

Let's write a titleCase() function in the imperative style. Our function should take a string and return a new string with the first character of each word capitalized:

function titleCase(headlineString) {
  var headlineWords = headlineString.split(" ");
  var titleWords = [];
  for (var i = 0; i < headlineWords.length; i++) {
    var headlineWord = headlineWords[i];
    var titleWord = headlineWord[0].toUpperCase() + headlineWord.substring(1);
    titleWords.push(titleWord);
  }
  var title = titleWords.join(" ");
  return title;
}

Not bad! Pretty easy to understand, right?

Functional style

Functional code, by contrast, composes all business logic first, before the data is even available. Only after all the logic has been defined does it apply that logic to the data.

Logic first you say? Let's begin to refactor the imperative code we just wrote by extracting pieces of logic into functions. We will define those functions before the actual data is available. First, let's extract a words() function that splits a string into words, and a join() function that joins an array of words into a string:

var words = (str) => str.split(" ");
var join = (arr) => arr.join(" ");

function titleCase(headlineString) {
  var headlineWords = words(headlineString);
  var titleWords = [];
  for (var i = 0; i < headlineWords.length; i++) {
    var headlineWord = headlineWords[i];
    var titleWord = headlineWord[0].toUpperCase() + headlineWord.substring(1);
    titleWords.push(titleWord);
  }
  var title = join(titleWords);
  return title;
}

Now let's convert our for loop into a map:

var words = (str) => str.split(" ");
var join = (arr) => arr.join(" ");

function titleCase(headlineString) {
  var headlineWords = words(headlineString);
  var titleWords = headlineWords.map((headlineWord) => {
    return headlineWord[0].toUpperCase() + headlineWord.substring(1);
  });
  var title = join(titleWords);
  return title;
}

Map takes a function as an argument. Let's give that function a name and extract it out to before our data is available:

var words = (str) => str.split(" ");
var join = (arr) => arr.join(" ");
var capify = (word) => word[0].toUpperCase() + word.substring(1);

function titleCase(headlineString) {
  var headlineWords = words(headlineString);
  var titleWords = headlineWords.map(capify);
  var title = join(titleWords);
  return title;
}

Notice how the functions that we've extracted have more generic variable names. Since these functions are decoupled from the data on which they operate, they are more reusable and easier to reason about.

Now, even though we have most our our logic defined before we have our data, our titleCase function is still imperatively operating on our data one step at a time. Let's try nesting our functions and see how that removes some of the "statefullness" we associate with imperative programing. This will make the code a little hard to read for now. Don't worry, we'll come back and clean it up later.

var words = (str) => str.split(" ");
var join = (arr) => arr.join(" ");
var capify = (word) => word[0].toUpperCase() + word.substring(1);

function titleCase(headlineString) {
  return join(words(headlineString).map(capify));
}

This is already a lot less code than our original imperative attempt!

Higher order functions

A higher order function does at least one of the following two things:

Takes one or more functions as arguments
Returns a function as its result

You'll notice that many of the helper functions we've extracted are basically converting a method call to a function. Let's write a higher order function that converts Array.map() to a function. It will take a function, func, and return a new function. The new function will accept an array, and map func over that array:

var map = (func) => {
  return (arr) => {
    return arr.map(func);
  };
};

Taking advantage of the implicit return of ES6 arrow functions, we can rewrite this as:

var map = (func) => (arr) => arr.map(func);

Let's use our new map() function to clean up our titleCase() example:

var map = (func) => (arr) => arr.map(func);
var words = (str) => str.split(" ");
var join = (arr) => arr.join(" ");
var capify = (word) => word[0].toUpperCase() + word.substring(1);
var capifyWords = map(capify);

function titleCase(headlineString) {
  return join(capifyWords(words(headlineString)));
}

We still have a lump of nested functions, but now at least they are nested more simply.

Flow()

Earlier, I promised that we would clean up the confusing set of nested functions that remain in our titleCase() function. Introducing: flow().

Flow is a higher order function which takes an array of functions and returns a new function which takes a single value and successively applies the given functions to that value.

For example, the following two statements are equivalent:

flow([a, b, c]);

(value) => c(b(a(value)));

Flow can be found in lodash, but for the purposes of this exercise, we can write our own flow function by using Array.reduce():

var flow = (funcs) => {
  return (initialValue) => {
    return funcs.reduce((value, func) => {
      return func(value);
    }, initialValue);
  };
};

Using implicit returns:

var flow = (funcs) => (initialValue) =>
  funcs.reduce((value, func) => func(value), initialValue);

Let's try using flow in our titleCase() example:

var map = (func) => (arr) => arr.map(func);
var words = (str) => str.split(" ");
var join = (arr) => arr.join(" ");
var capify = (word) => word[0].toUpperCase() + word.substring(1);
var capifyWords = map(capify);

var titleCase = flow([words, capifyWords, join]);

Much better! It has turned our nest of unreadable nested functions into a readable "flow" of logic.

Method()

If you look for common patterns in our existing code, you'll see that many of our functions are just about converting method calls into functions. Let's abstract that pattern! We'll define a higher-order method() function which takes a method name and an argument. It will then return a function which takes an object and calls the given method on that object with the given argument:

var method = (func, arg) => {
  return (obj) => {
    return obj[func](arg);
  };
};

Now that we see how it works, we can clean it up with implicit return:

var method = (func, arg) => (obj) => obj[func](arg);

Let's use our new method() function to simplify our titleCase() example:

var method = (func, arg) => (obj) => obj[func](arg);
var map = (func) => (arr) => arr.map(func);
var words = method("split", " ");
var join = method("join", " ");
var capify = (word) => word[0].toUpperCase() + word.substring(1);
var capifyWords = map(capify);

var titleCase = flow([words, capifyWords, join]);

Notice how our word() and join() functions can now be defined without any variable names, which makes them even more generic and declarative. With functional programming we can focus on what we're doing rather than how we're doing it.

capifyWords is just a method call!

Since capifyWords is really just calling .map(), which is a method, with a specific argument, we can create it using our higher order method(), and get rid of our map() function:

var method = (func, arg) => (obj) => obj[func](arg);
var words = method("split", " ");
var join = method("join", " ");
var capify = (word) => word[0].toUpperCase() + word.substring(1);
var capifyWords = method("map", capify);

var titleCase = flow([words, capifyWords, join]);

We can read the that last line as:

titleCase is a function which splits its value into words, capifys those words, and then joins them back together.

That's pretty readable, and it clearly communicates the program's intent rather than its implementation details.

Compare the two

Before

function titleCase(headlineString) {
  var headlineWords = headlineString.split(" ");
  var titleWords = [];
  for (var i = 0; i < headlineWords.length; i++) {
    var headlineWord = headlineWords[i];
    var titleWord = headlineWord[0].toUpperCase() + headlineWord.substring(1);
    titleWords.push(titleWord);
  }
  var title = titleWords.join(" ");
  return title;
}

After

var method = (func, arg) => (obj) => obj[func](arg);
var words = method("split", " ");
var join = method("join", " ");
var capify = (word) => word[0].toUpperCase() + word.substring(1);
var capifyWords = method("map", capify);

var titleCase = flow([words, capifyWords, join]);

Our new code is six lines instead of eleven. It avoids having implementation-specific variable names, and reads in terms of the problem it's solving rather than in terms of implementation details. It's very declarative!

Prologue: Reusability

Another benefit of composing our business logic out of smaller, data-agnostic, functions is that we can reuse those small functions. Let's see how difficult it would be to create five other similar functions:

var lowerCase = method("toLowerCase");
var upperCase = method("toUpperCase");
var studlyCase = flow([words, capifyWords, method("join", "")]);
var kebabCase = flow([lowerCase, words, method("join", "-")]);
var snakeCase = flow([lowerCase, words, method("join", "_")]);

Take a second to read though these and figure out how they work. See how we are making novel use of our existing vocabulary of functions? Notice how we can even use our first new function, lowerCase() as a component of later functions like kebabCase!

Imagine doing this in an imperative style! We would have ended up with five more eleven line functions instead of five one line functions!

I hope this post as piqued your interest in functional programing in JavaScript! If you want to learn more, check out these links:

Tags: javascript