One of the things I dislike about JavaScript is how it intertwines object-oriented paradigm with functional programming. Ironically, that’s also the thing that I do like about JS. I like that I have a choice and JS actually does allow you to write functional code that looks natural (as opposed to Ruby that doesn’t even have actual functions per se).

However, the standard library in JavaScript is written in a OOP manner, meaning that you need to call methods on objects instead of using functions. This creates inconsistency when you write in functional style because you end up with code that uses both functions end methods: getUsers(ids).map(userToJson).

I’d much rather prefer to have a consistent function-oriented code. Lodash, defining a lot of commonly-used functions, basically provides functional alternative to JS methods from standard library.

What is lodash?

Basically, lodash is just a library that provides a lot of utility functions. Years ago JavaScript was in much worse shape than it is now so lodash (which was a fork of Underscore.js) filled in a lot of gaps in standard library.

Nowadays a lot of functions in it seem obsolete with JS evolved so much over the years. But the reality is, the language is still a mess (for instance, are you aware of which version of the language your most recent webpack project uses?) and lodash is quite stable.

Plus, even tho it gets better and better, JS still lacks a lot of useful functions, e.g. orderBy. Yes, in most cases those functions are really simple and anyone can implement them on their own. But the truth is, it’s good when you don’t have to write them by yourself all the time. Lodash collects them in one place; it’s stable; and it’s, most likely, already in your project even if you don’t use it directly.

What is lodash/fp?

Lodash/fp provides the most of the original lodash functions but modified to promote functional style.

For example, some functions have been aliased to more common (I guess?) names, e.g. flow was aliased to pipe (like pipe operator in Elixir, for example).

Another thing is many (or all?) functions allow currying. This is very nice because it allows you to write shorter and more expressive code.

Also parameter order is different in lodash/fp. Now functional arguments come first and data arguments come last. For example, compare original lodash map(arr, f) with lodash/fp map(f, arr).

This was done to get the most out of currying. Compare:

// lodash
const ageFilter = users => filter(users, u => u.age > 18);

// lodash/fp
const ageFilter = filter(u => u.age > 18);

Some common cases

Pipe operator

Pipe operator is a great way to express a computation because instead of a complicated expression involving lots of nested functions, you can write your computation as a list of transformations (functions) that are being called in order.

Lodash doesn’t really define it however it has compose functions flow and flowRight:

// Simple
const getBankByUserId = id => getUserBank(getUser(id));
const bank = getBankByUserId(userId);

// flowRight - most common function composition h(x) = f*g(x) = f(g(x))
// it's also aliased to `compose` in lodash/fp
const getBankByUserId = flowRight([getUserBank, getUser])
const bank = getBankByUserId(userId);

// flow - same as flowRight but in a reverse order
const getBankByUserId = flow([getUser, getUserBank])
const bank = getBankByUserId(userId);

Now, this is, basically, what a pipe operator would do, it composes functions but also calls the result with a value. So all we have to do is to call our composition:

// sendEmail(message, getContacts(getUser(getUserId(transaction))).email)

flow([
  getUserId,
  getUser,
  getContacts,
  c => sendEmail(message, c.email),
])(transaction);

I, personally, hate that the value we are transforming comes in the very end. What if there’re a lot of transformations? It’s not very nice to have the actual value so far down. I came up with a simple hack: call the function without arguments but the first function should return the value itself. Take a look at the final snippet:

import pipe from 'lodash/fp/pipe'

pipe([
  () => transaction,
  getUserId,
  getUser,
  getContacts,
  c => sendEmail(message, c.email),
])();

Here’s another hack for you. You can inject a console.log anywhere in the pipeline to see what’s going on after some step. Check the diff:

  import pipe from 'lodash/fp/pipe'

  pipe([
    () => transaction,
    getUserId,
    getUser,
+   x => { console.log('USER:', x); return x },
    getContacts,
    c => sendEmail(message, c.email),
  ])();

Collection processing (map, reduce, filter, etc…)

As I’ve mentioned before, they are all curried and their parameters have a different order from their original lodash counterparts. However, if you have experience with functional languages (e.g. Clojure, Haskell, Elm) you will feel right at home.

Just take look at the snippet:

import pipe from 'lodash/fp/pipe'
import map from 'lodash/fp/map'
import reduce from 'lodash/fp/reduce'
import toPairs from 'lodash/fp/toPairs'
import sortBy from 'lodash/fp/sortBy'
import take from 'lodash/fp/take'
import last from 'lodash/fp/last'

const topFiveUserArtists = pipe([
  () => getUser(userId),
  getUserSongs,
  map(song => [song.artist, getNumberOfPlays(song, userId)]),
  reduce((obj, ([artist, plays])) => ({ ...obj, [artist]: (obj[artist] || 0) + plays }), {}),
  toPairs,
  sortBy(last),
  take(5),
])();

Another approach:

import pipe from 'lodash/fp/pipe'
import groupBy from 'lodash/fp/groupBy'
import sumBy from 'lodash/fp/sumBy'
import mapValues from 'lodash/fp/mapValues'
import toPairs from 'lodash/fp/toPairs'
import sortBy from 'lodash/fp/sortBy'
import take from 'lodash/fp/take'
import last from 'lodash/fp/last'

const topFiveUserArtists = pipe([
  () => getUser(userId),
  getUserSongs,
  groupBy(s => s.artist),
  mapValues(sumBy(song => getNumberOfPlays(song, userId))),
  toPairs,
  sortBy(last),
  take(5),
])();

I, personally, prefer the first one as it uses less specific functions and just relies on the essentials. But yeah, lodash has a lot of cool utility functions like mapValues. Also, I import each function separately. This is my personal preference and it, supposedly, reduces the bundle size when webpack has finished.

Some misc. utilities (with JS analogs)

range(a, b) = [...Array(b - a).keys()].map(i => a + i)
take(n, arr) = arr.slice(0, n)
isEqual(a, b). JS doesn’t have any alternatives included. It performs deep equality checks (meaning that it will compare objects and arrays value by value, not by reference).
isEmpty(coll) = coll.length == 0 for arrays but isEmpty is more powerful as it works with arrays, objects, sets, etc.
uniq(arr) = [...new Set(arr)] but uniq also preserves order.
sortBy(f, arr) = [...arr].sort((a, b) => f(a) < f(b) ? -1 : 1) ([...arr] is for cloning).
tail(arr) (careful, don’t confuse with rest) = arr.slice(1)

Conclusion

By using this DSL we are able to get rid of the most object-oriented code. It becomes better-structured, more consistent, and even a bit safer (bc some standard JS methods are actually mutating data, e.g. Array#sort).

For me personally the result code looks a lot like Clojure which I adore.

Another advantage is that using a third-party library instead of the standard library frees you from runtime/language headache. You don’t have to think about your runtime or webpack having stuff implemented anymore.

Plus, it provides solutions for many common use-cases. I, actually, get annoyed when I want to use some small utility function I or my teammate wrote but only to find out that it’s defined in a different project and I have to copy-paste it from there.

If you get your team consistently use a DSL like that, your whole codebase should become healthier (I think). Apparently, there’re other options apart from lodash.