stream()

• Core
  • m
  • m.render
  • m.mount
  • m.route
  • m.request
  • m.jsonp
  • m.parseQueryString
  • m.buildQueryString
  • m.withAttr
  • m.trust
  • m.fragment
  • m.redraw
  • m.version
  • Promise
• Optional
  • Stream
    • Description
    • Signature
      • Static members
        • Stream.combine
        • Stream.merge
        • Stream.scan
        • Stream.scanMerge
        • Stream.HALT
        • Stream["fantasy-land/of"]
      • Instance members
        • stream.map
        • stream.end
        • stream["fantasy-land/of"]
        • stream["fantasy-land/map"]
        • stream["fantasy-land/ap"]
    • Basic usage
      • Streams as variables
      • Bidirectional bindings
      • Computed properties
    • Chaining streams
    • Combining streams
    • Stream states
    • Serializing streams
    • Streams do not trigger rendering
    • What is Fantasy Land
• Tooling
  • Ospec

Description

A Stream is a reactive data structure, similar to cells in spreadsheet applications.

For example, in a spreadsheet, if A1 = B1 + C1, then changing the value of B1 or C1 automatically changes the value of A1.

Similarly, you can make a stream depend on other streams so that changing the value of one automatically updates the other. This is useful when you have very expensive computations and want to only run them when necessary, as opposed to, say, on every redraw.

Streams are NOT bundled with Mithril's core distribution. To include the Streams module, use:

var Stream = require("mithril/stream")

You can also download the module directly if your environment does not support a bundling toolchain:

<script src="https://unpkg.com/mithril-stream"></script>

When loaded directly with a <script> tag (rather than required), the stream library will be exposed as window.m.stream. If window.m is already defined (e.g. because you also use the main Mithril script), it will attach itself to the existing object. Otherwise it creates a new window.m. If you want to use streams in conjunction with Mithril as raw script tags, you should include Mithril in your page before mithril-stream, because mithril will otherwise overwrite the window.m object defined by mithril-stream. This is not a concern when the libraries are consumed as CommonJS modules (using require(...)).

Signature

Creates a stream

stream = Stream(value)

How to read signatures

Static members

Stream.combine

Creates a computed stream that reactively updates if any of its upstreams are updated. See combining streams

stream = Stream.combine(combiner, streams)

How to read signatures

combiner

Specifies how the value of a computed stream is generated. See combining streams

any = combiner(streams..., changed)

How to read signatures

Stream.merge

Creates a stream whose value is the array of values from an array of streams

stream = Stream.merge(streams)

How to read signatures

Stream.scan

Creates a new stream with the results of calling the function on every value in the stream with an accumulator and the incoming value.

stream = Stream.scan(fn, accumulator, stream)

How to read signatures

Stream.scanMerge

Takes an array of pairs of streams and scan functions and merges all those streams using the given functions into a single stream.

stream = Stream.scanMerge(pairs, accumulator)

How to read signatures

Stream.HALT

A special value that can be returned to stream callbacks to halt execution of downstreams

Stream["fantasy-land/of"]

This method is functionally identical to stream. It exists to conform to Fantasy Land's Applicative specification. For more information, see the What is Fantasy Land section.

stream = Stream["fantasy-land/of"](value)

Instance members

stream.map

Creates a dependent stream whose value is set to the result of the callback function. This method is an alias of stream["fantasy-land/map"].

dependentStream = stream().map(callback)

How to read signatures

stream.end

A co-dependent stream that unregisters dependent streams when set to true. See ended state.

endStream = stream().end

stream["fantasy-land/of"]

This method is functionally identical to stream. It exists to conform to Fantasy Land's Applicative specification. For more information, see the What is Fantasy Land section.

stream = stream()["fantasy-land/of"](value)

stream["fantasy-land/map"]

Creates a dependent stream whose value is set to the result of the callback function. See chaining streams

This method exists to conform to Fantasy Land's Applicative specification. For more information, see the What is Fantasy Land section.

dependentStream = stream()["fantasy-land/map"](callback)

How to read signatures

stream["fantasy-land/ap"]

The name of this method stands for apply. If a stream a has a function as its value, another stream b can use it as the argument to b.ap(a). Calling ap will call the function with the value of stream b as its argument, and it will return another stream whose value is the result of the function call. This method exists to conform to Fantasy Land's Applicative specification. For more information, see the What is Fantasy Land section.

stream = stream()["fantasy-land/ap"](apply)

Basic usage

Streams are not part of the core Mithril distribution. To include them in a project, require its module:

var stream = require("mithril/stream")

Streams as variables

stream() returns a stream. At its most basic level, a stream works similar to a variable or a getter-setter property: it can hold state, which can be modified.

var username = stream("John")
console.log(username()) // logs "John"

username("John Doe")
console.log(username()) // logs "John Doe"

The main difference is that a stream is a function, and therefore can be composed into higher order functions.

var users = stream()

// request users from a server using the fetch API
fetch("/api/users")
    .then(function(response) {return response.json()})
    .then(users)

In the example above, the users stream is populated with the response data when the request resolves.

Bidirectional bindings

Streams can also be populated from other higher order functions, such as m.withAttr

// a stream
var user = stream("")

// a bi-directional binding to the stream
m("input", {
    oninput: m.withAttr("value", user),
    value: user()
})

In the example above, when the user types in the input, the user stream is updated to the value of the input field.

Computed properties

Streams are useful for implementing computed properties:

var title = stream("")
var slug = title.map(function(value) {
    return value.toLowerCase().replace(/\W/g, "-")
})

title("Hello world")
console.log(slug()) // logs "hello-world"

In the example above, the value of slug is computed when title is updated, not when slug is read.

It's of course also possible to compute properties based on multiple streams:

var firstName = stream("John")
var lastName = stream("Doe")
var fullName = stream.merge([firstName, lastName]).map(function(values) {
    return values.join(" ")
})

console.log(fullName()) // logs "John Doe"

firstName("Mary")

console.log(fullName()) // logs "Mary Doe"

Computed properties in Mithril are updated atomically: streams that depend on multiple streams will never be called more than once per value update, no matter how complex the computed property's dependency graph is.

Chaining streams

Streams can be chained using the map method. A chained stream is also known as a dependent stream.

// parent stream
var value = stream(1)

// dependent stream
var doubled = value.map(function(value) {
    return value * 2
})

console.log(doubled()) // logs 2

Dependent streams are reactive: their values are updated any time the value of their parent stream is updated. This happens regardless of whether the dependent stream was created before or after the value of the parent stream was set.

You can prevent dependent streams from being updated by returning the special value stream.HALT

var halted = stream(1).map(function(value) {
    return stream.HALT
})

halted.map(function() {
    // never runs
})

Combining streams

Streams can depend on more than one parent stream. These kinds of streams can be created via stream.merge()

var a = stream("hello")
var b = stream("world")

var greeting = stream.merge([a, b]).map(function(values) {
    return values.join(" ")
})

console.log(greeting()) // logs "hello world"

There's also a lower level method called stream.combine() that exposes the stream themselves in the reactive computations for more advanced use cases

var a = stream(5)
var b = stream(7)

var added = stream.combine(function(a, b) {
    return a() + b()
}, [a, b])

console.log(added()) // logs 12

A stream can depend on any number of streams and it's guaranteed to update atomically. For example, if a stream A has two dependent streams B and C, and a fourth stream D is dependent on both B and C, the stream D will only update once if the value of A changes. This guarantees that the callback for stream D is never called with unstable values such as when B has a new value but C has the old value. Atomicity also brings the performance benefits of not recomputing downstreams unnecessarily.

You can prevent dependent streams from being updated by returning the special value stream.HALT

var halted = stream.combine(function(stream) {
    return stream.HALT
}, [stream(1)])

halted.map(function() {
    // never runs
})

Stream states

At any given time, a stream can be in one of three states: pending, active, and ended.

Pending state

Pending streams can be created by calling stream() with no parameters.

var pending = stream()

If a stream is dependent on more than one stream, any of its parent streams is in a pending state, the dependent streams is also in a pending state, and does not update its value.

var a = stream(5)
var b = stream() // pending stream

var added = stream.combine(function(a, b) {
    return a() + b()
}, [a, b])

console.log(added()) // logs undefined

In the example above, added is a pending stream, because its parent b is also pending.

This also applies to dependent streams created via stream.map:

var value = stream()
var doubled = value.map(function(value) {return value * 2})

console.log(doubled()) // logs undefined because `doubled` is pending

Active state

When a stream receives a value, it becomes active (unless the stream is ended).

var stream1 = stream("hello") // stream1 is active

var stream2 = stream() // stream2 starts off pending
stream2("world") // then becomes active

A dependent stream with multiple parents becomes active if all of its parents are active.

var a = stream("hello")
var b = stream()

var greeting = stream.merge([a, b]).map(function(values) {
    return values.join(" ")
})

In the example above, the a stream is active, but b is pending. setting b("world") would cause b to become active, and therefore greeting would also become active, and be updated to have the value "hello world"

Ended state

A stream can stop affecting its dependent streams by calling stream.end(true). This effectively removes the connection between a stream and its dependent streams.

var value = stream()
var doubled = value.map(function(value) {return value * 2})

value.end(true) // set to ended state

value(5)

console.log(doubled())
// logs undefined because `doubled` no longer depends on `value`

Ended streams still have state container semantics, i.e. you can still use them as getter-setters, even after they are ended.

var value = stream(1)
value.end(true) // set to ended state

console.log(value(1)) // logs 1

value(2)
console.log(value()) // logs 2

Ending a stream can be useful in cases where a stream has a limited lifetime (for example, reacting to mousemove events only while a DOM element is being dragged, but not after it's been dropped).

Serializing streams

Streams implement a .toJSON() method. When a stream is passed as the argument to JSON.stringify(), the value of the stream is serialized.

var value = stream(123)
var serialized = JSON.stringify(value)
console.log(serialized) // logs 123

Streams also implement a valueOf method that returns the value of the stream.

var value = stream(123)
console.log("test " + value) // logs "test 123"

Streams do not trigger rendering

Unlike libraries like Knockout, Mithril streams do not trigger re-rendering of templates. Redrawing happens in response to event handlers defined in Mithril component views, route changes, or after m.request calls resolve.

If redrawing is desired in response to other asynchronous events (e.g. setTimeout/setInterval, websocket subscription, 3rd party library event handler, etc), you should manually call m.redraw()

What is Fantasy Land

Fantasy Land specifies interoperability of common algebraic structures. In plain english, that means that libraries that conform to Fantasy Land specs can be used to write generic functional style code that works regardless of how these libraries implement the constructs.

For example, say we want to create a generic function called plusOne. The naive implementation would look like this:

function plusOne(a) {
    return a + 1
}

The problem with this implementation is that it can only be used with a number. However it's possible that whatever logic produces a value for a could also produce an error state (wrapped in a Maybe or an Either from a library like Sanctuary or Ramda-Fantasy), or it could be a Mithril stream, or a flyd stream, etc. Ideally, we wouldn't want to write a similar version of the same function for every possible type that a could have and we wouldn't want to be writing wrapping/unwrapping/error handling code repeatedly.

This is where Fantasy Land can help. Let's rewrite that function in terms of a Fantasy Land algebra:

var fl = require("fantasy-land")

function plusOne(a) {
    return a[fl.map](function(value) {return value + 1})
}

Now this method works with any Fantasy Land compliant Functor, such as R.Maybe, S.Either, stream, etc.

This example may seem convoluted, but it's a trade-off in complexity: the naive plusOne implementation makes sense if you have a simple system and only ever increment numbers, but the Fantasy Land implementation becomes more powerful if you have a large system with many wrapper abstractions and reused algorithms.

When deciding whether you should adopt Fantasy Land, you should consider your team's familiarity with functional programming, and be realistic regarding the level of discipline that your team can commit to maintaining code quality (vs the pressure of writing new features and meeting deadlines). Functional style programming heavily depends on compiling, curating and mastering a large set of small, precisely defined functions, and therefore it's not suitable for teams who do not have solid documentation practices, and/or lack experience in functional oriented languages.