An extremely intuitive, small, and fast
functional reactive stream library for JavaScript

  • Only 26 core operators and factories
  • Only “hot” streams
  • Written in TypeScript
  • Approximately 30 kB in size, when minified
  • On average, faster than RxJS 4, Kefir, Bacon.js, as fast as RxJS 5, and slower than most.js
  • Tailored for Cycle.js, or applications with limited use of subscribe

ComVer Gitter

Example

import xs from 'xstream'

// Tick every second incremental numbers,
// only pass even numbers, then map them to their square,
// and stop after 5 seconds has passed

var stream = xs.periodic(1000)
  .filter(i => i % 2 === 0)
  .map(i => i * i)
  .endWhen(xs.periodic(5000).take(1))

// So far, the stream is idle.
// As soon as it gets its first listener, it starts executing.

stream.addListener({
  next: i => console.log(i),
  error: err => console.error(err),
  complete: () => console.log('completed'),
})

Installation

npm install xstream

Usage

ES2015 or TypeScript

import xs from 'xstream'

CommonJS

var xs = require('xstream').default

API

Factories

Methods and Operators

Overview

XStream has four fundamental types: Stream, Listener, Producer, and MemoryStream.

Stream

A Stream is an event emitter with multiple Listeners. When an event happens on the Stream, it is broadcast to all its Listeners at the same time.

Streams have methods attached to them called operators, such as map, filter, fold, take, etc. When called, an operator creates and returns another Stream. Once the first Stream broadcasts an event, the event will pass through the operator logic and the output Stream may perhaps broadcast its own event based on the source one.

You can also trigger an event to happen on a Stream with the shamefullySend* methods. But you don’t want to do that. Really, avoid doing that because it’s not the reactive way and you’ll be missing the point of this library. Ok?

Listener

A Listener is an object with one to three functions attached to it: next, error, and complete. There is usually one function for each type of event a Stream may emit but only next is always required.

  • next events are the typical type, they deliver a value.
  • error events abort (stop) the execution of the Stream, and happen when something goes wrong in the Stream (or upstream somewhere in the chain of operators)
  • complete events signal the peaceful stop of the execution of the Stream.

This is an example of a typical listener:

var listener = {
  next: (value) => {
    console.log('The Stream gave me a value: ', value);
  },
  error: (err) => {
    console.error('The Stream gave me an error: ', err);
  },
  complete: () => {
    console.log('The Stream told me it is done.');
  },
}

And this is how you would attach that Listener to a Stream:

stream.addListener(listener)

And when you think the Listener is done, you can remove it from the Stream:

stream.removeListener(listener)

Producer

A Producer is like a machine that produces events to be broadcast on a Stream.

Events from a Stream must come from somewhere, right? That’s why we need Producers. They are objects with two functions attached: start(listener) and stop(). Once you call start with a listener, the Producer will start generating events and it will send those to the listener. When you call stop(), the Producer should quit doing its own thing.

Streams are also Listeners (actually they are “InternalListeners”, not Listeners, but that’s a detail you can ignore), so you can theoretically give a Stream as the listener in producer.start(streamAsListener). Then, essentially the Producer is now generating events that will be broadcast on the Stream. Nice, huh? Now a bunch of listeners can be attached to the Stream and they can all get those events originally coming from the Producer. That’s why xs.create(producer) receives a Producer to be the heart of a new Stream. Check this out:

var producer = {
  start: function (listener) {
    this.id = setInterval(() => listener.next('yo'), 1000)
  },

  stop: function () {
    clearInterval(this.id)
  },

  id: 0,
}

// This fellow delivers a 'yo' next event every 1 second
var stream = xs.create(producer)

But remember, a Producer has only one listener, but a Stream may have many listeners.

You may wonder “when is start and stop called”, and that’s actually a fairly tricky topic, so let’s get back to that soon. First let me tell you about MemoryStreams.

MemoryStream

A MemoryStream is just like a Stream: it has operators, it can have listeners attached, you can shamefully send events to it, blabla. But it has one special property: it has memory. It remembers the most recent (but just one) next event that it sent to its listeners.

Why is that useful? If a new Listener is added after that next event was sent, the MemoryStream will get its value stored in memory and will send it to the newly attached Listener.

This is important so MemoryStreams can represent values or pieces of state which are relevant even after some time has passed. You don’t want to lose those, you want to keep them and send them to Listeners that arrive late, after the event was originally created.

How a Stream starts and stops

A Stream controls its Producer according to its number of Listeners, using reference counting with a synchronous start and a cancelable asynchronous stop. That’s how a Stream starts and stops, basically. Usually this part of XStream is not so relevant to remember when building applications, but if you want to understand it for debugging or curiosity, it’s explained in plain English below.

When you create a Stream with xs.create(producer), the start() function of the Producer is not yet called. The Stream is still “idle”. It has the Producer, but the Producer was not turned on.

Once the first Listener is added to the Stream, the number of Listeners attached suddenly changed from 0 to 1. That’s when the Stream calls start, because after all there is at least one Listener interested in this Stream.

More Listeners may be added in the future, but they don’t affect whether the Producer will continue working or stop. Just the first Listener dictates when the Stream starts.

What matters for stopping the Producer is stream.removeListener. When the last Listener leaves (or in other words, when the number of Listeners suddenly changes from 1 to 0), the Stream schedules producer.stop() to happen on the next event loop. That is, asynchronously. If, however, a new Listener is added (number goes from 0 to 1) before that scheduled moment, the producer.stop() will be cancelled, and the Producer will continue generating events for its Stream normally.

The reason the Producer is not suddenly (synchronously) stopped, is that it is often necessary to swap the single listener of a Stream, but still keep its ongoing execution. For instance:

var listenerA = {/* ... */}
var listenerB = {/* ... */}

// number goes from 0 to 1, so the Stream's Producer starts
stream.addListener(listenerA)

// ...

// number goes from 1 to 0, but then immediately goes back
// to 1, because listenerB was added
stream.removeListener(listenerA)
stream.addListener(listenerB)

// Stream's Producer does not stop, everything continues as before

It’s still useful to eventually (asynchronously) stop a Stream’s internal Producer, because you don’t want useless computation lying around producing gibberish. At least I don’t.

Factories

Factories are functions that create Streams, such as xs.create(), xs.periodic(), etc.

create(producer)

Creates a new Stream given a Producer.

Arguments:

  • producer: Producer An optional Producer that dictates how to start, generate events, and stop the Stream.

Returns: Stream


createWithMemory(producer)

Creates a new MemoryStream given a Producer.

Arguments:

  • producer: Producer An optional Producer that dictates how to start, generate events, and stop the Stream.

Returns: MemoryStream


never()

Creates a Stream that does nothing when started. It never emits any event.

Marble diagram:

         never
-----------------------

Returns: Stream


empty()

Creates a Stream that immediately emits the “complete” notification when started, and that’s it.

Marble diagram:

empty
-|

Returns: Stream


throw(error)

Creates a Stream that immediately emits an “error” notification with the value you passed as the error argument when the stream starts, and that’s it.

Marble diagram:

throw(X)
-X

Arguments:

  • error The error event to emit on the created stream.

Returns: Stream


from(input)

Creates a stream from an Array, Promise, or an Observable.

Arguments:

  • input: Array|Promise|Observable The input to make a stream from.

Returns: Stream


of(a, b)

Creates a Stream that immediately emits the arguments that you give to of, then completes.

Marble diagram:

of(1,2,3)
123|

Arguments:

  • a The first value you want to emit as an event on the stream.
  • b The second value you want to emit as an event on the stream. One or more of these values may be given as arguments.

Returns: Stream


fromArray(array)

Converts an array to a stream. The returned stream will emit synchronously all the items in the array, and then complete.

Marble diagram:

fromArray([1,2,3])
123|

Arguments:

  • array: Array The array to be converted as a stream.

Returns: Stream


fromPromise(promise)

Converts a promise to a stream. The returned stream will emit the resolved value of the promise, and then complete. However, if the promise is rejected, the stream will emit the corresponding error.

Marble diagram:

fromPromise( ----42 )
-----------------42|

Arguments:

  • promise: Promise The promise to be converted as a stream.

Returns: Stream


fromObservable(observable)

Converts an Observable into a Stream.

Arguments:

  • observable: any The observable to be converted as a stream.

Returns: Stream


periodic(period)

Creates a stream that periodically emits incremental numbers, every period milliseconds.

Marble diagram:

    periodic(1000)
---0---1---2---3---4---...

Arguments:

  • period: number The interval in milliseconds to use as a rate of emission.

Returns: Stream


merge(stream1, stream2)

Blends multiple streams together, emitting events from all of them concurrently.

merge takes multiple streams as arguments, and creates a stream that behaves like each of the argument streams, in parallel.

Marble diagram:

--1----2-----3--------4---
----a-----b----c---d------
           merge
--1-a--2--b--3-c---d--4---

Arguments:

  • stream1: Stream A stream to merge together with other streams.
  • stream2: Stream A stream to merge together with other streams. Two or more streams may be given as arguments.

Returns: Stream


combine(stream1, stream2)

Combines multiple input streams together to return a stream whose events are arrays that collect the latest events from each input stream.

combine internally remembers the most recent event from each of the input streams. When any of the input streams emits an event, that event together with all the other saved events are combined into an array. That array will be emitted on the output stream. It’s essentially a way of joining together the events from multiple streams.

Marble diagram:

--1----2-----3--------4---
----a-----b-----c--d------
         combine
----1a-2a-2b-3b-3c-3d-4d--

Note: to minimize garbage collection, combine uses the same array instance for each emission. If you need to compare emissions over time, cache the values with map first:

import pairwise from 'xstream/extra/pairwise'

const stream1 = xs.of(1);
const stream2 = xs.of(2);

xs.combine(stream1, stream2).map(
  combinedEmissions => ([ ...combinedEmissions ])
).compose(pairwise)

Arguments:

  • stream1: Stream A stream to combine together with other streams.
  • stream2: Stream A stream to combine together with other streams. Multiple streams, not just two, may be given as arguments.

Returns: Stream


Methods and Operators

Methods are functions attached to a Stream instance, like stream.addListener(). Operators are also methods, but return a new Stream, leaving the existing Stream unmodified, except for the fact that it has a child Stream attached as Listener.

addListener(listener)

Adds a Listener to the Stream.

Arguments:

  • listener: Listener

removeListener(listener)

Removes a Listener from the Stream, assuming the Listener was added to it.

Arguments:

  • listener: Listener\<T>

subscribe(listener)

Adds a Listener to the Stream returning a Subscription to remove that listener.

Arguments:

  • listener: Listener

Returns: Subscription


map(project)

Transforms each event from the input Stream through a project function, to get a Stream that emits those transformed events.

Marble diagram:

--1---3--5-----7------
   map(i => i * 10)
--10--30-50----70-----

Arguments:

  • project: Function A function of type (t: T) => U that takes event t of type T from the input Stream and produces an event of type U, to be emitted on the output Stream.

Returns: Stream


mapTo(projectedValue)

It’s like map, but transforms each input event to always the same constant value on the output Stream.

Marble diagram:

--1---3--5-----7-----
      mapTo(10)
--10--10-10----10----

Arguments:

  • projectedValue A value to emit on the output Stream whenever the input Stream emits any value.

Returns: Stream


filter(passes)

Only allows events that pass the test given by the passes argument.

Each event from the input stream is given to the passes function. If the function returns true, the event is forwarded to the output stream, otherwise it is ignored and not forwarded.

Marble diagram:

--1---2--3-----4-----5---6--7-8--
    filter(i => i % 2 === 0)
------2--------4---------6----8--

Arguments:

  • passes: Function A function of type (t: T) +> boolean that takes an event from the input stream and checks if it passes, by returning a boolean.

Returns: Stream


take(amount)

Lets the first amount many events from the input stream pass to the output stream, then makes the output stream complete.

Marble diagram:

--a---b--c----d---e--
   take(3)
--a---b--c|

Arguments:

  • amount: number How many events to allow from the input stream before completing the output stream.

Returns: Stream


drop(amount)

Ignores the first amount many events from the input stream, and then after that starts forwarding events from the input stream to the output stream.

Marble diagram:

--a---b--c----d---e--
      drop(3)
--------------d---e--

Arguments:

  • amount: number How many events to ignore from the input stream before forwarding all events from the input stream to the output stream.

Returns: Stream


last()

When the input stream completes, the output stream will emit the last event emitted by the input stream, and then will also complete.

Marble diagram:

--a---b--c--d----|
      last()
-----------------d|

Returns: Stream


startWith(initial)

Prepends the given initial value to the sequence of events emitted by the input stream. The returned stream is a MemoryStream, which means it is already remember()‘d.

Marble diagram:

---1---2-----3---
  startWith(0)
0--1---2-----3---

Arguments:

  • initial The value or event to prepend.

Returns: MemoryStream


endWhen(other)

Uses another stream to determine when to complete the current stream.

When the given other stream emits an event or completes, the output stream will complete. Before that happens, the output stream will behaves like the input stream.

Marble diagram:

---1---2-----3--4----5----6---
  endWhen( --------a--b--| )
---1---2-----3--4--|

Arguments:

  • other Some other stream that is used to know when should the output stream of this operator complete.

Returns: Stream


fold(accumulate, seed)

“Folds” the stream onto itself.

Combines events from the past throughout the entire execution of the input stream, allowing you to accumulate them together. It’s essentially like Array.prototype.reduce. The returned stream is a MemoryStream, which means it is already remember()‘d.

The output stream starts by emitting the seed which you give as argument. Then, when an event happens on the input stream, it is combined with that seed value through the accumulate function, and the output value is emitted on the output stream. fold remembers that output value as acc (“accumulator”), and then when a new input event t happens, acc will be combined with that to produce the new acc and so forth.

Marble diagram:

------1-----1--2----1----1------
  fold((acc, x) => acc + x, 3)
3-----4-----5--7----8----9------

Arguments:

  • accumulate: Function A function of type (acc: R, t: T) => R that takes the previous accumulated value acc and the incoming event from the input stream and produces the new accumulated value.
  • seed The initial accumulated value, of type R.

Returns: MemoryStream


replaceError(replace)

Replaces an error with another stream.

When (and if) an error happens on the input stream, instead of forwarding that error to the output stream, replaceError will call the replace function which returns the stream that the output stream will replicate. And, in case that new stream also emits an error, replace will be called again to get another stream to start replicating.

Marble diagram:

--1---2-----3--4-----X
  replaceError( () => --10--| )
--1---2-----3--4--------10--|

Arguments:

  • replace: Function A function of type (err) => Stream that takes the error that occurred on the input stream or on the previous replacement stream and returns a new stream. The output stream will behave like the stream that this function returns.

Returns: Stream


flatten()

Flattens a “stream of streams”, handling only one nested stream at a time (no concurrency).

If the input stream is a stream that emits streams, then this operator will return an output stream which is a flat stream: emits regular events. The flattening happens without concurrency. It works like this: when the input stream emits a nested stream, flatten will start imitating that nested one. However, as soon as the next nested stream is emitted on the input stream, flatten will forget the previous nested one it was imitating, and will start imitating the new nested one.

Marble diagram:

--+--------+---------------
  \        \
   \       ----1----2---3--
   --a--b----c----d--------
          flatten
-----a--b------1----2---3--

Returns: Stream


compose(operator)

Passes the input stream to a custom operator, to produce an output stream.

compose is a handy way of using an existing function in a chained style. Instead of writing outStream = f(inStream) you can write outStream = inStream.compose(f).

Arguments:

  • operator: function A function that takes a stream as input and returns a stream as well.

Returns: Stream


remember()

Returns an output stream that behaves like the input stream, but also remembers the most recent event that happens on the input stream, so that a newly added listener will immediately receive that memorised event.

Returns: MemoryStream


debug(labelOrSpy)

Returns an output stream that identically behaves like the input stream, but also runs a spy function fo each event, to help you debug your app.

debug takes a spy function as argument, and runs that for each event happening on the input stream. If you don’t provide the spy argument, then debug will just console.log each event. This helps you to understand the flow of events through some operator chain.

Please note that if the output stream has no listeners, then it will not start, which means spy will never run because no actual event happens in that case.

Marble diagram:

--1----2-----3-----4--
        debug
--1----2-----3-----4--

Arguments:

  • labelOrSpy: function A string to use as the label when printing debug information on the console, or a ‘spy’ function that takes an event as argument, and does not need to return anything.

Returns: Stream


imitate(target)

imitate changes this current Stream to emit the same events that the other given Stream does. This method returns nothing.

This method exists to allow one thing: circular dependency of streams. For instance, let’s imagine that for some reason you need to create a circular dependency where stream first$ depends on stream second$ which in turn depends on first$:

import delay from 'xstream/extra/delay'

var first$ = second$.map(x => x * 10).take(3);
var second$ = first$.map(x => x + 1).startWith(1).compose(delay(100));

However, that is invalid JavaScript, because second$ is undefined on the first line. This is how imitate can help solve it:

import delay from 'xstream/extra/delay'

var secondProxy$ = xs.create();
var first$ = secondProxy$.map(x => x * 10).take(3);
var second$ = first$.map(x => x + 1).startWith(1).compose(delay(100));
secondProxy$.imitate(second$);

We create secondProxy$ before the others, so it can be used in the declaration of first$. Then, after both first$ and second$ are defined, we hook secondProxy$ with second$ with imitate() to tell that they are “the same”. imitate will not trigger the start of any stream, it just binds secondProxy$ and second$ together.

The following is an example where imitate() is important in Cycle.js applications. A parent component contains some child components. A child has an action stream which is given to the parent to define its state:

const childActionProxy$ = xs.create();
const parent = Parent({...sources, childAction$: childActionProxy$});
const childAction$ = parent.state$.map(s => s.child.action$).flatten();
childActionProxy$.imitate(childAction$);

Note, though, that imitate() does not support MemoryStreams. If we would attempt to imitate a MemoryStream in a circular dependency, we would either get a race condition (where the symptom would be “nothing happens”) or an infinite cyclic emission of values. It’s useful to think about MemoryStreams as cells in a spreadsheet. It doesn’t make any sense to define a spreadsheet cell A1 with a formula that depends on B1 and cell B1 defined with a formula that depends on A1.

If you find yourself wanting to use imitate() with a MemoryStream, you should rework your code around imitate() to use a Stream instead. Look for the stream in the circular dependency that represents an event stream, and that would be a candidate for creating a proxy Stream which then imitates the target Stream.

Arguments:

  • target: Stream The other stream to imitate on the current one. Must not be a MemoryStream.

shamefullySendNext(value)

Forces the Stream to emit the given value to its listeners.

As the name indicates, if you use this, you are most likely doing something The Wrong Way. Please try to understand the reactive way before using this method. Use it only when you know what you are doing.

Arguments:

  • value The “next” value you want to broadcast to all listeners of this Stream.

shamefullySendError(error)

Forces the Stream to emit the given error to its listeners.

As the name indicates, if you use this, you are most likely doing something The Wrong Way. Please try to understand the reactive way before using this method. Use it only when you know what you are doing.

Arguments:

  • error: any The error you want to broadcast to all the listeners of this Stream.

shamefullySendComplete()

Forces the Stream to emit the “completed” event to its listeners.

As the name indicates, if you use this, you are most likely doing something The Wrong Way. Please try to understand the reactive way before using this method. Use it only when you know what you are doing.


setDebugListener(listener)

Adds a “debug” listener to the stream. There can only be one debug listener, that’s why this is ‘setDebugListener’. To remove the debug listener, just call setDebugListener(null).

A debug listener is like any other listener. The only difference is that a debug listener is “stealthy”: its presence/absence does not trigger the start/stop of the stream (or the producer inside the stream). This is useful so you can inspect what is going on without changing the behavior of the program. If you have an idle stream and you add a normal listener to it, the stream will start executing. But if you set a debug listener on an idle stream, it won’t start executing (not until the first normal listener is added).

As the name indicates, we don’t recommend using this method to build app logic. In fact, in most cases the debug operator works just fine. Only use this one if you know what you’re doing.

Arguments:

  • listener: Listener\<T>

Extra operators and factories

The operators and factories listed above are the core functions. xstream has plenty of extra operators, documented here.

FAQ

Q: Why does imitate() support a Stream but not a MemoryStream?

A: MemoryStreams are meant for representing “values over time” (your age), while Streams represent simply events (your birthdays). MemoryStreams are usually initialized with a value, and imitate() is meant for creating circular dependencies of streams. If we would attempt to imitate a MemoryStream in a circular dependency, we would either get a race condition (where the symptom would be “nothing happens”) or an infinite cyclic emission of values.

If you find yourself wanting to use imitate() with a MemoryStream, you should rework your code around imitate() to use a Stream instead. Look for the stream in the circular dependency that represents an event stream, and that would be a candidate for creating a MimicStream which then imitates the real event stream.

Q: What’s the difference between xstream and RxJS?

A: Read this blog post on the topic. In short:

  • xstream Streams are multicast always.
  • RxJS Observables are unicast by default, and opt-in multicast.
  • xstream has few operators as a feature (helps against decision paralysis).
  • RxJS has many operators as a feature (helps for flexibility and power).

Q: What is the equivalent of withLatestFrom in xstream?

A: withLatestFrom is implemented as an extra named sampleCombine.


Misc.

Acknowledgements

xstream is built by staltz and TylorS.

CHANGELOG

Read the CHANGELOG for release notes of all versions of xstream.

License

MIT