Bifocal

Install

$ npm install --save bifocal

Motivation

This isn't intended to be a thorough explanation of what lenses are, but here's one way to think about them:

Many functions can be thought of as extracting some value from another value. For example, we might have the function fst (first) which takes the first element of a pair:

const fst = ([a, _]) => a;
fst([1, 2]); // => 1

This function captures the idea that we can take a pair and take the first thing out of it. What it doesn't let us do is put the first thing back. I can't, with this one function, take out the first thing, add 1 to it, and then re-insert it into the pair. For this we can use a lens. To make a lens, we need both a way to look at the value (fst already does this) and put it back. To put it back we can use putFst:

const putFst = ([_, b], a) => [a, b];
putFst([1, 2], 3); // => [3, 2]

Note that these are all pure functions, so we never mutate the values passed in. With fst and putFst, we can create a lens:

const fstLens = createLens(fst, putFst);

and with this, we can take out a value, change it, and put it back, all in one action:

over(fstLens, x => x + 1, [1, 3]); // => [2, 3]

over takes out the value, applies a function to it, and then puts it back. In this case, we took out the value 1, applied x => x + 1 to get 2, and then re-inserted it to [1, 3] to get [2, 3].

A lens, with respect to this library, is a function which takes either one or two arguments. When given one, it returns the result of focusing on that value (we'll call this "getting"). When given two, it returns the result of inserting the second argument into the first (we'll call this "putting"). createLens is just a helper that takes two separate functions and makes them satisfy this interface.

Usage

import {
  createLens,
  over,
  lift
} from 'bifocal';

// A lens which gets and puts the `a` property of some object
let aGet = obj => obj.a;
let aPut = (obj, a) => ({...obj, a});
let aLens = createLens(aGet, aPut);

aLens({ a: 1 }); // => 1
aLens({ a: 1 }, 2); // => { a: 2 }

const sqr = x => x * x;

over(aLens, sqr, { a: 2 }); // => { a: 4 }

const sqrA = lift(aLens, sqr)
sqrA({ a: 2 }); // => { a: 4 }

See test.js for more detailed examples.

API

`createLens(get, put)`

A convenience helper to create a lens. If you want to you can just create the lenses directly, but this lets you avoid the logic of checking the number of arguments.

Example: A lens which looks at an object's a property

let aLens = createLens(
  obj => obj.a,
  (obj, a) => ({...obj, a})
);

`compose(...lenses)`

Compose a sequence of lenses from right to left.

Example:

let abLens = compose(
  bLens, // accesses an object's 'b' property
  aLens  // accesses an object's 'a' property
);

abLens({ a: { b: 3 }}); // => 3
abLens({ a: { b: 3 }}, 4); // => { a: { b: 4 }}

`fromPath(...path)`

Create a lens which looks into a deeply nested object.

Example:

let abLens = fromPath('a', 'b');

abLens({ a: { b: 3 }}); // => 3
abLens({ a: { b: 3 }}, 4); // => { a: { b: 4 }}

You might notice that these go in reverse order to composition. This is to mimic property access (x.a.b) which goes in reverse order to function application (b(a(x))).

`lift(lens, f)`

"Lift" f into the world of lens. If L is a lens from A to B, and F is a function from B -> B, then lift(L, F) is a function from A -> A.

Example:

const sqr = x => x * x;
const sqrA = lift(aLens, sqr)
sqrA({ a: 2 }); // => { a: 4 }

`over(lens, f, value)`

Apply a function over a lens. Equivalent to lifting and then applying. Conceptually similar to map.

Example:

over(aLens, sqr, { a: 2 }); // => { a: 4 }

`combineLenses({ <name>: lens, ... })`

Given an object of lenses, creates a lens which gives an object with the same structure, with the value of each lens as its values. Note that the lenses should commute (not overlap) for this to work in a reasonable way.

Example:

const lens = combineLenses({
  a: fromPath('hello'),
  b: fromPath('goodbye', 'farewell')
});

lens({
  hello: 1,
  goodbye: {
    farewell: 2
  }
}); // => { a: 1, b: 2 }

lens({
  hello: 1,
  goodbye: {
    farewell: 2
  }
}, { a: 3, b: 4 });
=> {
  hello: 3,
  goodbye: {
    farewell: 4
  }
}

`fromPathImmutable(...path)`

Like fromPath, but for collections from ImmutableJS.

Extra Stuff

You don't need to read this, but it might be interesting/helpful.

Lenses that make sense

You might notice that we can't use any old functions with createLens to get a lens that always works properly. For example, what if our putter put the value back in a different place than the getter looked?

const what = createLens(
  ([a, _]) => a,
  ([a, _], b) => [a, b]
);

This doesn't really work how we want:

over(what, x => x + 1, [1, 100]); // => [1, 2]

To avoid this, there are 3 rules that if followed, guarantee a lens will work in a way that is reasonable.

1: Put-Put

If we put something into a value, and then put something else, it should be as if we only put the second thing.

That is,

lens(lens(x, a), b) === lens(x, b)

2: Put-Get

If we put something into a value using a lens, and then read from the result, we should get back what we put.

Meaning,

lens(lens(x, a)) === a

3: Get-Put

If we read something from a value using a lens, and then put that thing back, nothing should change.

lens(x, lens(x)) === x

(this is the one that the lens what above, breaks)

More lenses

The most obvious kind of lens is one that looks at a small part of a large value, like fstLens does. You can do this basically any time you have a function whose operation can be reversed given the original value.

In particular it works with any bijection (or, function which can be inverted):

const negationLens = createLens(
  x => -x,
  (_, x) => -x
);

over(negationLens, x => x + 1, 5); // => 4

In this case, this diagram might look familiar (but no worries if it doesn't):

  A    -- f ->    B
  
  |               ^
  |               |
  h             h^{-1}
  |               |
  v               |

  A'   -- g ->    B'

You can operate on strings as if they are arrays of characters:

const arrayLens = createLens(
  s => s.split(''),
  (_, a) => a.join('')
);

over(arrayLens, x => x.slice().reverse(), "hello"); // => "olleh"

It should be noted that negationLens and arrayLens aren't using all of the power available to lenses, however, since they ignore the original value (this is because they're bijections).

Another sort of weird example: the division algorithm says that given numbers m and n, there exists a unique quotient q and remainder r such that

n = qm + r

where r < m.

So for example, if m = 102 and n = 5, then q = 20 and r = 2.

We can create a lens that focuses on the quotient or remainder for a given number.

// represents division by `m`
function division(m) {
  const quotient = n => Math.floor(n / m);
  const remainder = n => n % m;
  const quotientLens = createLens(
    n => quotient(n),
    (n, q) => q * m + remainder(n)
  );
  const remainderLens = createLens(
    n => remainder(n),
    (n, r) => quotient(n) * m + r
  );
  return {
    quotientLens,
    remainderLens
  };
}

const { quotientLens, remainderLens } = division(100);

// 105 = 1 * 100 + 5
quotientLens(105); // => 1
remainderLens(105); // => 5

quotientLens(105, 2); // => 205
remainderLens(105, 27); // => 127

justinj/bifocal

Bifocal

Install

Motivation

Usage

API

createLens(get, put)

compose(...lenses)

fromPath(...path)

lift(lens, f)

over(lens, f, value)

combineLenses({ <name>: lens, ... })

fromPathImmutable(...path)

Extra Stuff

Lenses that make sense

1: Put-Put

2: Put-Get

3: Get-Put

More lenses

`createLens(get, put)`

`compose(...lenses)`

`fromPath(...path)`

`lift(lens, f)`

`over(lens, f, value)`

`combineLenses({ <name>: lens, ... })`

`fromPathImmutable(...path)`