Composing Lenses

To cap things off, let's talk about how lenses compose. Spoiler alert: it looks backwards, but it's not. (8 min. read)

We'll cover the following

Let’s go back a few lessons and review our code from Use With Arrays. We wished to access an object’s third friend using lensIndex.

import { lensIndex, view } from 'ramda';
const person = {
  firstName: 'Bobo',
  lastName: 'Flakes',
  friends: [{
    firstName: 'Clark',
    lastName: 'Kent'
  }, {
    firstName: 'Bruce',
    lastName: 'Wayne'
  }, {
    firstName: 'Barry',
    lastName: 'Allen'
  }]
};
const getThirdFriend = lensIndex(2);
const result = view(getThirdFriend, person.friends);
console.log({ result });

A Bit Too Specific

It works fine, but look at how view's being used.

view(getThirdFriend, person.friends);
// person.friends?

Lenses help decouple your logic and data, so it’s counterintuitive to specify person.friends. The whole point’s to just pass in person and let the lens do the work for us!

But lensIndex only works on arrays, so how can it focus on friends before the index? Say it with me: function composition! We’ll just compose lensIndex with lensProp and get our result.

import { pipe, lensIndex, lensProp, view } from 'ramda';
const person = {
  firstName: 'Bobo',
  lastName: 'Flakes',
  friends: [{
    firstName: 'Clark',
    lastName: 'Kent'
  }, {
    firstName: 'Bruce',
    lastName: 'Wayne'
  }, {
    firstName: 'Barry',
    lastName: 'Allen'
  }]
};
const getThirdFriend = pipe(
  lensProp('friends'),
  lensIndex(2)
);
const result = view(getThirdFriend, person);
console.log({ result });

This Is Wrong

There ya go, nice and…wait. This returns undefined… Why?! The composition looks correct.

const getThirdFriend = pipe(
  lensProp('friends'),
  lensIndex(2)
);
  1. Get friends
  2. Get third one (index 2)

Believe it or not, we composed them backwards. Check this out.

import { pipe, lensIndex, lensProp, view } from 'ramda';
const person = {
  firstName: 'Bobo',
  lastName: 'Flakes',
  friends: [{
    firstName: 'Clark',
    lastName: 'Kent'
  }, {
    firstName: 'Bruce',
    lastName: 'Wayne'
  }, {
    firstName: 'Barry',
    lastName: 'Allen'
  }]
};
// Flip the composition
const getThirdFriend = pipe(
  lensIndex(2),
  lensProp('friends'),
);
const result = view(getThirdFriend, person);
console.log({ result });

This Is Correct

const getThirdFriend = pipe(
  lensIndex(2),
  lensProp('friends'),
);

Now it works. Why?

Back to the Functor

The last lesson introduced the relationship between functors and lenses.

After receiving a getter/setter, the lens requires a toFunctorFn–a function that turns a value into a functor.

This is because Ramda’s map relinquishes control to any functor carrying that special fantasy-land/map property, allowing view, set, and over to do their jobs.

Look again at our composition.

const getThirdFriend = pipe(
  lensIndex(2),
  lensProp('friends'),
);

So lensIndex(2) returns a function expecting its toFunctorFn, as does lensProp('friends').

Following the pipe sequence leads us to an interesting conclusion: lensIndex(2) is the toFunctorFn to lensProp('friends')! This is how they’re composing. We can prove it with some logs.

import { tap, pipe, lensIndex, lensProp, view } from 'ramda';
const person = {
  firstName: 'Bobo',
  lastName: 'Flakes',
  friends: [{
    firstName: 'Clark',
    lastName: 'Kent'
  }, {
    firstName: 'Bruce',
    lastName: 'Wayne'
  }, {
    firstName: 'Barry',
    lastName: 'Allen'
  }]
};
// Flip the composition
const getThirdFriend = pipe(
  tap((fn) => {
    console.log('lensIndex will be called with this\n');
    console.log(fn.toString());
    console.log('\n');
  }),
  lensIndex(2),
  tap((fn) => {
    console.log('lensProp will be called with this\n');
    console.log(fn.toString());
    console.log('\n');
  }),
  lensProp('friends'),
    tap((fn) => {
    console.log('The composition is this:\n');
    console.log(fn.toString());
    console.log('\n');
  }),
);
const result = view(getThirdFriend, person);
console.log({ result });

Unfold

Carefully read these logs. Passing view to the composition of lensIndex(2) and lensProp('friends') created the following sequence of events:

  1. view gave a toFunctorFn to lensIndex(2).

  2. Now lensIndex(2) awaits its data.

  3. lensIndex(2) becomes a toFunctorFn for lensProp('friends')

  4. lensProp('friends') now awaits its data.

// the composed lens
function (target) {
  return map(function (focus) {
    return setter(focus, target);
  }, toFunctorFn(getter(target)));
}

  1. view feeds it our person data.

  2. map fires like a rocket, rapidly unfolding the functor in the correct order, according to the getters and setters.

  3. The getter for lensProp('friends') is called first, so person.friends is retrieved

  4. That data is then passed to lensIndex(2), who grabs the third element, person.friends[2].

  5. You get your data back.

compose() Instead of pipe()

This takes some time to get used to. If you, like me, prefer pipe because it reads left-to-right, compose lets you write lenses left-to-right as well.

const getThirdFriend = compose(
  lensProp('friends'),
  lensIndex(2)
);
import { compose, lensIndex, lensProp, view } from 'ramda';
const person = {
  firstName: 'Bobo',
  lastName: 'Flakes',
  friends: [{
    firstName: 'Clark',
    lastName: 'Kent'
  }, {
    firstName: 'Bruce',
    lastName: 'Wayne'
  }, {
    firstName: 'Barry',
    lastName: 'Allen'
  }]
};
const getThirdFriend = compose(
  lensProp('friends'),
  lensIndex(2)
);
const result = view(getThirdFriend, person);
console.log({ result });

Summary

  • Compose lenses to handle objects and arrays at the same time (lensProp + lensIndex).

  • Lenses don’t compose backwards, one combines with the next by acting as its toFunctorFn.

  • Once built up and given the data, map takes that “giant” lens and calls its toFunctorFn, which is a composition of every lens that came before it.

  • This unfolding lets map drill all the way down through your data and return the property you’re after.

  • Use compose if you want to read lenses from left-to-right.