fluture

Fluture

Futures are containers which represent some eventual value as a result of an asynchronous computation, much like Promises. Unlike Promises, however, Futures are lazy and logical by design. They have a predictable API governed by the Fantasy Land algebraic JavaScript specification.

npm install --save fluture ^{Requires a node 5.0.0 compatible environment like modern browsers, transpilers or Node 5+}

Table of contents

• Usage
• Motivation and Features
• Documentation
  • Type signatures
  • Constructors
  • Method API
  • Dispatcher API
  • Cancellation and resource disposal
  • Futurization
• Benchmarks
• The name

Usage

Using the low level, high performance method API:

const Future = require('fluture');
const program = file =>
  Future.node(done => fs.readFile(file, 'utf8', done))
  .chain(Future.encase(JSON.parse))
  .map(x => x.name)
  .fork(console.error, console.log);
program('package.json');
//> "fluture"

Or use the high level, fully curried, functional dispatch API for function composition using composers like S.pipe:

const {node, chain, encase, map, fork} = require('fluture');
const program = S.pipe([
  file => node(fs.readFile.bind(fs, file, 'utf8')),
  chain(encase(JSON.parse)),
  map(x => x.name),
  fork(console.error, console.log)
]);
program('package.json')
//> "fluture"

Motivation and Features

Existing implementations of Future are a pain to debug. This library was made in an effort to provide great error messages when something goes wrong. Some other features include:

• Plenty of async control utilities, like Future.parallel and Future#or.
• Process cancellation and automatic resource disposal.
• High performance.

For an overview of differences between Fluture and other Future implementations, have a look at this wiki article.

Documentation

Type signatures

Hindley-Milner type signatures are used to document functions. A list of all types used within these signatures follows:

• Forkable - Any Object with a fork method that takes at least two arguments. This includes instances of Fluture, instances of Task from data.task or instances of Future from ramda-fantasy.
• Future - Instances of Future provided by Fluture.
• Functor - Any object with a map method which satisfies the Fantasy Land Functor specification.
• Chain - Any object with a chain method which satisfies the Fantasy Land Chain specification.
• Apply - Any object with an ap method which satisfies the Fantasy Land Apply specification.

Constructors

Future :: ((a -> ()), (b -> ()) -> ?(() -> ())) -> Future a b

A (monadic) container which represents an eventual value. A lot like Promise but more principled in that it follows the Fantasy Land algebraic JavaScript specification.

const eventualThing = Future((reject, resolve) => {
  setTimeout(resolve, 500, 'world');
});

eventualThing.fork(
  console.error,
  thing => console.log(`Hello ${thing}!`)
);
//> "Hello world!"

of :: b -> Future a b

A constructor that creates a Future containing the given value.

const eventualThing = Future.of('world');
eventualThing.fork(
  console.error,
  thing => console.log(`Hello ${thing}!`)
);
//> "Hello world!"

after :: Number -> b -> Future a b

A constructor that creates a Future containing the given value after n milliseconds.

const eventualThing = Future.after(500, 'world');
eventualThing.fork(console.error, thing => console.log(`Hello ${thing}!`));
//> "Hello world!"

cast :: Forkable a b -> Future a b

Cast any Forkable to a Future.

Future.cast(require('data.task').of('hello')).value(console.log);
//> "hello"

encase :: (a -> !b | c) -> a -> Future b c

Creates a Future which resolves with the result of calling the given function with the given value, or rejects with the error thrown by the function.

const data = '{"foo" = "bar"}';
const parseJson = Future.encase(JSON.parse);
parseJson('a').fork(console.error, console.log)
//> [SyntaxError: Unexpected token =]

try :: (() -> !a | b) -> Future a b

A constructor that creates a Future which resolves with the result of calling the given function, or rejects with the error thrown by the given function.

Sugar for Future.encase(f, undefined).

const data = {foo: 'bar'}
Future.try(() => data.foo.bar.baz)
.fork(console.error, console.log)
//> [TypeError: Cannot read property 'baz' of undefined]

node :: ((a, b -> ()) -> ()) -> Future a b

A constructor that creates a Future which rejects with the first argument given to the function, or resolves with the second if the first is not present.

This is a convenience for NodeJS users who wish to easily obtain a Future from a node style callback API.

Future.node(done => fs.readFile('package.json', 'utf8', done))
.fork(console.error, console.log)
//> "{...}"

parallel :: PositiveInteger -> [Future a b] -> Future a [b]

Creates a Future which when forked runs all Futures in the given array in parallel, ensuring no more than limit Futures are running at once.

const tenFutures = Array.from(Array(10).keys()).map(Future.after(20));

//Runs all Futures in sequence:
Future.parallel(1, tenFutures).fork(console.error, console.log);
//after about 200ms:
//> [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]

//Runs upto five Futures in parallel:
Future.parallel(5, tenFutures).fork(console.error, console.log);
//after about 40ms:
//> [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]

//Runs all Futures in parallel:
Future.parallel(Infinity, tenFutures).fork(console.error, console.log);
//after about 20ms:
//> [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]

If you want to settle all Futures, even if some may fail, you can use this in combination with fold:

const fourInstableFutures = Array.from(Array(4).keys()).map(
  i => Future(
    (rej, res) => setTimeout(
      () => Math.random() > 0.8 ? rej('failed') : res(i),
      20
    )
  )
);

const stabalizedFutures = fourInstableFutures.map(Future.fold(S.Left, S.Right))

Future.parallel(2, stabalizedFutures).fork(console.error, console.log);
//after about 40ms:
//> [ Right(0), Left("failed"), Right(2), Right(3) ]

Method API

fork :: Future a b ~> (a -> ()), (b -> ()) -> () -> ()

Execute the Future (which up until now was merely a container for its computation), and get at the result or error.

It is the return from Fantasy Land to the real world. This function best shows the fundamental difference between Promises and Futures.

Future.of('world').fork(
  err => console.log(`Oh no! ${err.message}`),
  thing => console.log(`Hello ${thing}!`)
);
//> "Hello world!"

Future.reject(new Error('It broke!')).fork(
  err => console.log(`Oh no! ${err.message}`),
  thing => console.log(`Hello ${thing}!`)
);
//> "Oh no! It broke!"

map :: Future a b ~> (b -> c) -> Future a c

Map over the value inside the Future. If the Future is rejected, mapping is not performed.

Future.of(1)
.map(x => x + 1)
.fork(console.error, console.log);
//> 2

chain :: Future a b ~> (b -> Future a c) -> Future a c

FlatMap over the value inside the Future. If the Future is rejected, chaining is not performed.

Future.of(1)
.chain(x => Future.of(x + 1))
.fork(console.error, console.log);
//> 2

chainRej :: Future a b ~> (a -> Future a c) -> Future a c

FlatMap over the rejection value inside the Future. If the Future is resolved, chaining is not performed.

Future.reject(new Error('It broke!')).chainRej(err => {
  console.error(err);
  return Future.of('All is good')
})
.fork(console.error, console.log)
//> "All is good"

ap :: Future a (b -> c) ~> Future a b -> Future a c

Apply the value in the Future to the value in the given Future. If the Future is rejected, applying is not performed.

Future.of(x => x + 1)
.ap(Future.of(1))
.fork(console.error, console.log);
//> 2

race :: Future a b ~> Future a b -> Future a b

Race two Futures against eachother. Creates a new Future which resolves or rejects with the resolution or rejection value of the first Future to settle.

Future.after(100, 'hello')
.race(Future.after(50, 'bye'))
.fork(console.error, console.log)
//> "bye"

or :: Future a b ~> Future a b -> Future a b

Logical or for Futures.

Returns a new Future which either resolves with the first resolutation value, or rejects with the last rejection value once and if both Futures reject.

This behaves analogues to how JavaScript's or operator does, except both Futures run simultaneously, so it is not short-circuited. That means that if the second has side-effects, they will run even if the first resolves.

//An asynchronous version of:
//const result = planA() || planB();

const result = planA().or(planB());

In the example, assume both plans return Futures. Both plans are executed in parallel. If planA resolves, the returned Future will resolve with its value. If planA fails there is always planB. If both plans fail then the returned Future will also reject using the rejection reason of planB.

fold :: Future a b ~> (a -> c), (b -> c) -> Future _ c

Applies the left function to the rejection value, or the right function to the resolution value, depending on which is present, and resolves with the result.

This provides a convenient means to ensure a Future is always resolved. It can be used with other type constructors, like S.Either, to maintain a representataion of failures:

Future.of('hello')
.fold(S.Left, S.Right)
.value(console.log);
//> Right('hello')

Future.reject('it broke')
.fold(S.Left, S.Right)
.value(console.log);
//> Left('it broke')

cache :: Future a b ~> Future a b

Returns a cached version of the Future so that whenever it's forked, it can load the value from cache rather than reexecuting the chain. This means you can use the same Future in multiple chains, without having to worry that it's going to re-execute the computation every time.

Please note that cached Futures, and Futures derived from them do not automatically dispose of resources. When the cached Future is cancelled or disposed manually, it will clear its internal cache. The next time it's forked after that, it must re-execute the underlying Future to populate its cache again.

const eventualPackage = Future.node(done => {
  console.log('Reading some big data');
  fs.readFile('package.json', 'utf8', done)
})
.cache();

eventualPackage.fork(console.error, console.log);
//> "Reading some big data"
//> "{...}"

eventualPackage.fork(console.error, console.log);
//> "{...}"

value :: Future a b ~> (b -> ()) -> () -> ()

Extracts the value from a resolved Future by forking it. Only use this function if you are sure the Future is going to be resolved, for example; after using .fold(). If the Future rejects and value was used, an (likely uncatchable) Error will be thrown.

Future.reject(new Error('It broke'))
.fold(S.Left, S.Right)
.value(console.log)
//> Left([Error: It broke])

promise :: Future a b ~> Promise b a

An alternative way to fork the Future. This eagerly forks the Future and returns a Promise of the result. This is useful if some API wants you to give it a Promise. It's the only method which forks the Future without a forced way to handle the rejection branch, which means it's considered dangerous to use.

Future.of('Hello').promise().then(console.log);
//> "Hello"

Dispatcher API

fork :: (a -> ()) -> (b -> ()) -> Future a b -> () -> ()

Dispatches the first and second arguments to the fork method of the third argument.

const consoleFork = fork(console.error, console.log);
consoleFork(of('Hello'));
//> "Hello"

map :: Functor m => (a -> b) -> m a -> m b

Dispatches the first argument to the map method of the second argument. Has the same effect as R.map.

chain :: Chain m => (a -> m b) -> m a -> m b

Dispatches the first argument to the chain method of the second argument. Has the same effect as R.chain.

chainRej :: (a -> Future a c) -> Future a b -> Future a c

Dispatches the first argument to the chainRej method of the second argument.

ap :: Apply m => m (a -> b) -> m a -> m b

Dispatches the second argument to the ap method of the first argument. Has the same effect as R.ap.

race :: Future a b -> Future a b -> Future a b

Dispatches the first argument to the race method of the second argument.

const first = futures => futures.reduce(race);
first([
  after(100, 'hello'),
  after(50, 'bye'),
  Future(rej => setTimeout(rej, 25, 'nope'))
])
.fork(console.error, console.log)
//> [Error nope]

or :: Future a b -> Future a b -> Future a b

Dispatches the first argument to the or method of the second argument.

const program = S.pipe([
  reject,
  or(of('second chance')),
  value(console.log)
]);

program('first chance')
> "second chance"

fold :: (a -> c) -> (b -> c) -> Future a b -> Future _ c

Dispatches the first and second arguments to the fold method of the third argument.

cache :: Future a b -> Future a b

Dispatches to the cache method.

const cachedConnection = Future.cache(mysqlConnect());

value :: (b -> ()) -> Future a b -> () -> ()

Dispatches the first argument to the value method of the second argument.

promise :: Future a b -> Promise b a

Dispatches to the promise method.

Future.promise(Future.after(300, 'Hello')).then(console.log);
//> "Hello"

Cancellation and resource disposal

Automatic resource disposal

When a Future is created, it is given the fork function. This fork function sometimes creates resources, like setTimeouts in the event loop or database connections. In order to deal with the disposal of these resources, one may return a function from fork, which will be automatically called after the Future has forked. This function is expected to be idempotent.

It's the responsibility of this fork function to prevent rej or res from being called after its returned disposal function has been called. All internal Fluture functions, for example Future.after, play by these rules.

const createDatabaseConnection = settings => Future((rej, res) => {
  const conn = mysql.connect(settings, res);
  return () => conn.hasEnded || conn.end();
});

createDatabaseConnection()
.chain(conn => conn.query('SELECT 1 + 1 AS two'))

 //When we `fork`, all of the resource disposers will be automatically called.
.fork(console.error, console.log);

If you don't want a particular Future to automatically dispose of its resources, you can construct it with false passed as its second argument:

Future(
  rej => {
    const id = setTimeout(rej, 2000, 'timed out')
    return () => clearTimeout(id);
  },
  false //<---
)

This Future, and any Future's derived from it through map, chain, etc. will no longer automatically dispose of their resources after being forked. This allows resources to only be disposed manually.

Manual resource disposal and cancellation

Both .fork() and .value() return the disposal function and you can call it at any time to manually dispose of resources.

Besides resource disposal, this function can also be used to cancel running Futures. When called prematurely it will dispose the resources and cause the whole pipeline to come to a halt:

const loadPage = url => Future((rej, res) => {
  const req = new XMLHttpRequest();
  req.addEventListener('load', res);
  req.addEventListener('error', rej);
  req.open('GET', url);
  req.send();
  return () => (req.readyState < 4) && req.abort();
});

const cancel = loadPage('https://github.com').fork(onFailure, onSuccess);

//Cancel the Future immediately. Nothing will happen; `onFailure` nor
//`onSuccess` will ever be called anymore because `cancel` called `req.abort()`.
cancel();

Futurization

To reduce the boilerplate of making Node or Promise functions return Futures instead, one might use the Futurize library:

const Future = require('fluture');
const futurize = require('futurize').futurize(Future);
const readFile = futurize(require('fs').readFile);
readFile('README.md', 'utf8')
.map(text => text.split('\n'))
.map(lines => lines[0])
.fork(console.error, console.log);
//> "# Fluture"

Benchmarks

Simply run node ./bench/<file> to see how a specific method compares to implementations in data.task, ramda-fantasy.Future and Promise*.

* Promise is not included in all benchmarks because it tends to make the process run out of memory.

The name

A conjunction of the acronym to Fantasy Land (FL) and Future. Also "fluture" means butterfly in Romanian; A creature you might expect to see in Fantasy Land.

---

MIT licensed