Operations

In this guide, we’ll explore some essential operations you can perform on streams. These operations allow you to manipulate and interact with stream elements in various ways.

Tapping

The Stream.tap operation allows you to run an effect on each element emitted by the stream, observing or performing side effects without altering the elements or return type. This can be useful for logging, monitoring, or triggering additional actions with each emission.

Example (Logging with Stream.tap)

For example, Stream.tap can be used to log each element before and after a mapping operation:

import { Stream, Console, Effect } from "effect"

const stream = Stream.make(1, 2, 3).pipe(
  Stream.tap((n) => Console.log(`before mapping: ${n}`)),
  Stream.map((n) => n * 2),
  Stream.tap((n) => Console.log(`after mapping: ${n}`))
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
before mapping: 1
after mapping: 2
before mapping: 2
after mapping: 4
before mapping: 3
after mapping: 6
{ _id: 'Chunk', values: [ 2, 4, 6 ] }
*/

Taking Elements

The “taking” operations in streams let you extract a specific set of elements, either by a fixed number, condition, or position within the stream. Here are a few ways to apply these operations:

API	Description
take	Extracts a fixed number of elements.
takeWhile	Extracts elements while a certain condition is met.
takeUntil	Extracts elements until a certain condition is met.
takeRight	Extracts a specified number of elements from the end.

Example (Extracting Elements in Different Ways)

import { Stream, Effect } from "effect"

const stream = Stream.iterate(0, (n) => n + 1)

// Using `take` to extract a fixed number of elements:
const s1 = Stream.take(stream, 5)
Effect.runPromise(Stream.runCollect(s1)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ 0, 1, 2, 3, 4 ] }
*/

// Using `takeWhile` to extract elements while a condition is met:
const s2 = Stream.takeWhile(stream, (n) => n < 5)
Effect.runPromise(Stream.runCollect(s2)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ 0, 1, 2, 3, 4 ] }
*/

// Using `takeUntil` to extract elements until a condition is met:
const s3 = Stream.takeUntil(stream, (n) => n === 5)
Effect.runPromise(Stream.runCollect(s3)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ 0, 1, 2, 3, 4, 5 ] }
*/

// Using `takeRight` to take elements from the end of the stream:
const s4 = Stream.takeRight(s3, 3)
Effect.runPromise(Stream.runCollect(s4)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ 3, 4, 5 ] }
*/

Streams as an Alternative to Async Iterables

When working with asynchronous data sources, such as async iterables, you often need to consume data in a loop until a certain condition is met. Streams provide a similar approach and offer additional flexibility.

With async iterables, data is processed in a loop until a break or return statement is encountered. To replicate this behavior with Streams, consider these options:

API	Description
takeUntil	Takes elements from a stream until a specified condition is met, similar to breaking out of a loop.
toPull	Returns an effect that continuously pulls data chunks from the stream. This effect can fail with None when the stream is finished or with Some error if it fails.

Example (Using Stream.toPull)

import { Stream, Effect } from "effect"

// Simulate a chunked stream
const stream = Stream.fromIterable([1, 2, 3, 4, 5]).pipe(
  Stream.rechunk(2)
)

const program = Effect.gen(function* () {
  // Create an effect to get data chunks from the stream
  const getChunk = yield* Stream.toPull(stream)

  // Continuously fetch and process chunks
  while (true) {
    const chunk = yield* getChunk
    console.log(chunk)
  }
})

Effect.runPromise(Effect.scoped(program)).then(console.log, console.error)
/*
Output:
{ _id: 'Chunk', values: [ 1, 2 ] }
{ _id: 'Chunk', values: [ 3, 4 ] }
{ _id: 'Chunk', values: [ 5 ] }
(FiberFailure) Error: {
  "_id": "Option",
  "_tag": "None"
}
*/

Mapping

Basic Mapping

The Stream.map operation applies a specified function to each element in a stream, creating a new stream with the transformed values.

Example (Incrementing Each Element by 1)

import { Stream, Effect } from "effect"

const stream = Stream.make(1, 2, 3).pipe(
  Stream.map((n) => n + 1) // Increment each element by 1
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ 2, 3, 4 ] }
*/

Mapping to a Constant Value

The Stream.as method allows you to replace each success value in a stream with a specified constant value. This can be useful when you want all elements in the stream to emit a uniform value, regardless of the original data.

Example (Mapping to null)

import { Stream, Effect } from "effect"

const stream = Stream.range(1, 5).pipe(Stream.as(null))

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ null, null, null, null, null ] }
*/

Effectful Mapping

For transformations involving effects, use Stream.mapEffect. This function applies an effectful operation to each element in the stream, producing a new stream with effectful results.

Example (Random Number Generation)

import { Stream, Random, Effect } from "effect"

const stream = Stream.make(10, 20, 30).pipe(
  // Generate a random number between 0 and each element
  Stream.mapEffect((n) => Random.nextIntBetween(0, n))
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Example Output:
{ _id: 'Chunk', values: [ 5, 9, 22 ] }
*/

To handle multiple effectful transformations concurrently, you can use the concurrency option. This option allows a specified number of effects to run concurrently, with results emitted downstream in their original order.

Example (Fetching URLs Concurrently)

import { Stream, Effect } from "effect"

const fetchUrl = (url: string) =>
  Effect.gen(function* (_) {
    console.log(`Fetching ${url}`)
    yield* Effect.sleep("100 millis")
    console.log(`Fetching ${url} done`)
    return [`Resource 0-${url}`, `Resource 1-${url}`, `Resource 2-${url}`]
  })

const stream = Stream.make("url1", "url2", "url3").pipe(
  // Fetch each URL concurrently with a limit of 2
  Stream.mapEffect(fetchUrl, { concurrency: 2 })
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
Fetching url1
Fetching url2
Fetching url1 done
Fetching url3
Fetching url2 done
Fetching url3 done
{
  _id: 'Chunk',
  values: [
    [ 'Resource 0-url1', 'Resource 1-url1', 'Resource 2-url1' ],
    [ 'Resource 0-url2', 'Resource 1-url2', 'Resource 2-url2' ],
    [ 'Resource 0-url3', 'Resource 1-url3', 'Resource 2-url3' ]
  ]
}
*/

Stateful Mapping

Stream.mapAccum is similar to Stream.map, but it applies a transformation with state tracking, allowing you to map and accumulate values within a single operation. This is useful for tasks like calculating a running total in a stream.

Example (Calculating a Running Total)

import { Stream, Effect } from "effect"

const stream = Stream.range(1, 5).pipe(
  //                                  ┌─── next state
  //                                  │          ┌─── emitted value
  //                                  ▼          ▼
  Stream.mapAccum(0, (state, n) => [state + n, state + n])
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ 1, 3, 6, 10, 15 ] }
*/

Mapping and Flattening

The Stream.mapConcat operation is similar to Stream.map, but it goes further by mapping each element to zero or more elements (as an Iterable) and then flattening the entire stream. This is particularly useful for transforming each element into multiple values.

Example (Splitting and Flattening a Stream)

import { Stream, Effect } from "effect"

const numbers = Stream.make("1-2-3", "4-5", "6").pipe(
  Stream.mapConcat((s) => s.split("-"))
)

Effect.runPromise(Stream.runCollect(numbers)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ '1', '2', '3', '4', '5', '6' ] }
*/

Filtering

The Stream.filter operation allows you to pass through only elements that meet a specific condition. It’s a way to retain elements in a stream that satisfy a particular criteria while discarding the rest.

Example (Filtering Even Numbers)

import { Stream, Effect } from "effect"

const stream = Stream.range(1, 11).pipe(Stream.filter((n) => n % 2 === 0))

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ 2, 4, 6, 8, 10 ] }
*/

Scanning

Stream scanning allows you to apply a function cumulatively to each element in the stream, emitting every intermediate result. Unlike reduce, which only provides a final result, scan offers a step-by-step view of the accumulation process.

Example (Cumulative Addition)

import { Stream, Effect } from "effect"

const stream = Stream.range(1, 5).pipe(Stream.scan(0, (a, b) => a + b))

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ 0, 1, 3, 6, 10, 15 ] }
*/

If you need only the final accumulated value, you can use Stream.runFold:

Example (Final Accumulated Result)

import { Stream, Effect } from "effect"

const fold = Stream.range(1, 5).pipe(Stream.runFold(0, (a, b) => a + b))

Effect.runPromise(fold).then(console.log) // Output: 15

Draining

Stream draining lets you execute effectful operations within a stream while discarding the resulting values. This can be useful when you need to run actions or perform side effects but don’t require the emitted values. The Stream.drain function achieves this by ignoring all elements in the stream and producing an empty output stream.

Example (Executing Effectful Operations without Collecting Values)

import { Stream, Effect, Random } from "effect"

const stream = Stream.repeatEffect(
  Effect.gen(function* () {
    const nextInt = yield* Random.nextInt
    const number = Math.abs(nextInt % 10)
    console.log(`random number: ${number}`)
    return number
  })
).pipe(Stream.take(3))

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Example Output:
random number: 7
random number: 5
random number: 0
{ _id: 'Chunk', values: [ 7, 5, 0 ] }
*/

const drained = Stream.drain(stream)

Effect.runPromise(Stream.runCollect(drained)).then(console.log)
/*
Example Output:
random number: 0
random number: 1
random number: 7
{ _id: 'Chunk', values: [] }
*/

Detecting Changes in a Stream

The Stream.changes operation detects and emits elements that differ from their preceding elements within a stream. This can be useful for tracking changes or deduplicating consecutive values.

Example (Emitting Distinct Consecutive Elements)

import { Stream, Effect } from "effect"

const stream = Stream.make(1, 1, 1, 2, 2, 3, 4).pipe(Stream.changes)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ 1, 2, 3, 4 ] }
*/

Zipping

Zipping combines elements from two streams into a new stream, pairing elements from each input stream. This can be achieved with Stream.zip or Stream.zipWith, allowing for custom pairing logic.

Example (Basic Zipping)

In this example, elements from the two streams are paired sequentially. The resulting stream ends when one of the streams is exhausted.

import { Stream, Effect } from "effect"

// Zip two streams together
const stream = Stream.zip(
  Stream.make(1, 2, 3, 4, 5, 6),
  Stream.make("a", "b", "c")
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ [ 1, 'a' ], [ 2, 'b' ], [ 3, 'c' ] ] }
*/

Example (Custom Zipping Logic)

Here, Stream.zipWith applies custom logic to each pair, combining elements in a user-defined way.

import { Stream, Effect } from "effect"

// Zip two streams with custom pairing logic
const stream = Stream.zipWith(
  Stream.make(1, 2, 3, 4, 5, 6),
  Stream.make("a", "b", "c"),
  (n, s) => [n + 10, s + "!"]
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ [ 11, 'a!' ], [ 12, 'b!' ], [ 13, 'c!' ] ] }
*/

Handling Stream Endings

If one input stream ends before the other, you might want to zip with default values to avoid missing pairs. The Stream.zipAll and Stream.zipAllWith operators provide this functionality, allowing you to specify defaults for either stream.

Example (Zipping with Default Values)

In this example, when the second stream completes, the first stream continues with “x” as a default value for the second stream.

import { Stream, Effect } from "effect"

const stream = Stream.zipAll(Stream.make(1, 2, 3, 4, 5, 6), {
  other: Stream.make("a", "b", "c"),
  defaultSelf: -1,
  defaultOther: "x"
})

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{
  _id: 'Chunk',
  values: [
    [ 1, 'a' ],
    [ 2, 'b' ],
    [ 3, 'c' ],
    [ 4, 'x' ],
    [ 5, 'x' ],
    [ 6, 'x' ]
  ]
}
*/

Example (Custom Logic with zipAllWith)

With Stream.zipAllWith, custom logic determines how to combine elements when either stream runs out, offering flexibility to handle these cases.

import { Stream, Effect } from "effect"

const stream = Stream.zipAllWith(Stream.make(1, 2, 3, 4, 5, 6), {
  other: Stream.make("a", "b", "c"),
  onSelf: (n) => [n, "x"],
  onOther: (s) => [-1, s],
  onBoth: (n, s) => [n + 10, s + "!"]
})

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{
  _id: 'Chunk',
  values: [
    [ 11, 'a!' ],
    [ 12, 'b!' ],
    [ 13, 'c!' ],
    [ 4, 'x' ],
    [ 5, 'x' ],
    [ 6, 'x' ]
  ]
}
*/

Zipping Streams at Different Rates

When combining streams that emit elements at different speeds, you may not want to wait for the slower stream to emit. Using Stream.zipLatest or Stream.zipLatestWith, you can zip elements as soon as either stream produces a new value. These functions use the most recent element from the slower stream whenever a new value arrives from the faster stream.

Example (Combining Streams with Different Emission Rates)

import { Stream, Schedule, Effect } from "effect"

const s1 = Stream.make(1, 2, 3).pipe(
  Stream.schedule(Schedule.spaced("1 second"))
)

const s2 = Stream.make("a", "b", "c", "d").pipe(
  Stream.schedule(Schedule.spaced("500 millis"))
)

const stream = Stream.zipLatest(s1, s2)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{
  _id: 'Chunk',
  values: [
    [ 1, 'a' ],    // s1 emits 1 and pairs with the latest value from s2
    [ 1, 'b' ],    // s2 emits 'b', pairs with the latest value from s1
    [ 2, 'b' ],    // s1 emits 2, pairs with the latest value from s2
    [ 2, 'c' ],    // s2 emits 'c', pairs with the latest value from s1
    [ 2, 'd' ],    // s2 emits 'd', pairs with the latest value from s1
    [ 3, 'd' ]     // s1 emits 3, pairs with the latest value from s2
  ]
}
*/

Pairing with Previous and Next Elements

API	Description
zipWithPrevious	Pairs each element of a stream with its previous element.
zipWithNext	Pairs each element of a stream with its next element.
zipWithPreviousAndNext	Pairs each element with both its previous and next.

Example (Pairing Stream Elements with Next)

import { Stream, Effect } from "effect"

const stream = Stream.zipWithNext(Stream.make(1, 2, 3, 4))

Effect.runPromise(Stream.runCollect(stream)).then((chunks) =>
  console.log("%o", chunks)
)
/*
Output:
{
  _id: 'Chunk',
  values: [
    [ 1, { _id: 'Option', _tag: 'Some', value: 2 }, [length]: 2 ],
    [ 2, { _id: 'Option', _tag: 'Some', value: 3 }, [length]: 2 ],
    [ 3, { _id: 'Option', _tag: 'Some', value: 4 }, [length]: 2 ],
    [ 4, { _id: 'Option', _tag: 'None' }, [length]: 2 ],
    [length]: 4
  ]
}
*/

Indexing Stream Elements

The Stream.zipWithIndex operator is a helpful tool for indexing each element in a stream, pairing each item with its respective position in the sequence. This is particularly useful when you want to keep track of the order of elements within a stream.

Example (Indexing Each Element in a Stream)

import { Stream, Effect } from "effect"

const stream = Stream.zipWithIndex(
  Stream.make("Mary", "James", "Robert", "Patricia")
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{
  _id: 'Chunk',
  values: [
    [ 'Mary', 0 ],
    [ 'James', 1 ],
    [ 'Robert', 2 ],
    [ 'Patricia', 3 ]
  ]
}
*/

Cartesian Product of Streams

The Stream module includes a feature for computing the Cartesian Product of two streams, allowing you to create combinations of elements from two different streams. This is helpful when you need to pair each element from one set with every element of another.

In simple terms, imagine you have two collections and want to form all possible pairs by picking one item from each. This pairing process is the Cartesian Product. In streams, this operation generates a new stream that includes every possible pairing of elements from the two input streams.

To create a Cartesian Product of two streams, the Stream.cross operator is available, along with similar variants. These operators combine two streams into a new stream of all possible element combinations.

Example (Creating a Cartesian Product of Two Streams)

import { Stream, Effect, Console } from "effect"

const s1 = Stream.make(1, 2, 3).pipe(Stream.tap(Console.log))
const s2 = Stream.make("a", "b").pipe(Stream.tap(Console.log))

const cartesianProduct = Stream.cross(s1, s2)

Effect.runPromise(Stream.runCollect(cartesianProduct)).then(console.log)
/*
Output:
1
a
b
2
a
b
3
a
b
{
  _id: 'Chunk',
  values: [
    [ 1, 'a' ],
    [ 1, 'b' ],
    [ 2, 'a' ],
    [ 2, 'b' ],
    [ 3, 'a' ],
    [ 3, 'b' ]
  ]
}
*/

Multiple Iterations of Right Stream

Note that the right-hand stream (s2 in this example) will be iterated over multiple times, once for each element in the left-hand stream (s1). If the right-hand stream involves expensive or side-effect-producing operations, those will be executed repeatedly.

Partitioning

Partitioning a stream involves dividing it into two distinct streams based on a specified condition. The Stream module offers two functions for this purpose: Stream.partition and Stream.partitionEither. Let’s look at how these functions work and the best scenarios for their use.

partition

The Stream.partition function takes a predicate (a condition) as input and divides the original stream into two substreams. One substream will contain elements that meet the condition, while the other contains those that do not. Both resulting substreams are wrapped in a Scope type.

Example (Partitioning a Stream into Even and Odd Numbers)

import { Stream, Effect } from "effect"

//      ┌─── Effect<[Stream<number>, Stream<number>], never, Scope>
//      ▼
const program = Stream.range(1, 9).pipe(
  Stream.partition((n) => n % 2 === 0, { bufferSize: 5 })
)

Effect.runPromise(
  Effect.scoped(
    Effect.gen(function* () {
      const [evens, odds] = yield* program
      console.log(yield* Stream.runCollect(evens))
      console.log(yield* Stream.runCollect(odds))
    })
  )
)
/*
Output:
{ _id: 'Chunk', values: [ 2, 4, 6, 8 ] }
{ _id: 'Chunk', values: [ 1, 3, 5, 7, 9 ] }
*/

partitionEither

In some cases, you might need to partition a stream using a condition that involves an effect. For this, the Stream.partitionEither function is ideal. This function uses an effectful predicate to split the stream into two substreams: one for elements that produce Either.left values and another for elements that produce Either.right values.

Example (Partitioning a Stream with an Effectful Predicate)

import { Stream, Effect, Either } from "effect"

//      ┌─── Effect<[Stream<number>, Stream<number>], never, Scope>
//      ▼
const program = Stream.range(1, 9).pipe(
  Stream.partitionEither(
    // Simulate an effectful computation
    (n) => Effect.succeed(n % 2 === 0 ? Either.left(n) : Either.right(n)),
    { bufferSize: 5 }
  )
)

Effect.runPromise(
  Effect.scoped(
    Effect.gen(function* () {
      const [evens, odds] = yield* program
      console.log(yield* Stream.runCollect(evens))
      console.log(yield* Stream.runCollect(odds))
    })
  )
)
/*
Output:
{ _id: 'Chunk', values: [ 2, 4, 6, 8 ] }
{ _id: 'Chunk', values: [ 1, 3, 5, 7, 9 ] }
*/

Grouping

When processing streams of data, you may need to group elements based on specific criteria. The Stream module provides two functions for this purpose: groupByKey, groupBy, grouped and groupedWithin. Let’s review how these functions work and when to use each one.

groupByKey

The Stream.groupByKey function partitions a stream based on a key function of type (a: A) => K, where A is the type of elements in the stream, and K represents the keys for grouping. This function is non-effectful and groups elements by simply applying the provided key function.

The result of Stream.groupByKey is a GroupBy data type, representing the grouped stream. To process each group, you can use GroupBy.evaluate, which takes a function of type (key: K, stream: Stream<V, E>) => Stream.Stream<...>. This function operates across all groups and merges them together in a non-deterministic order.

Example (Grouping by Tens Place in Exam Scores)

In the following example, we use Stream.groupByKey to group exam scores by the tens place and count the number of scores in each group:

import { Stream, GroupBy, Effect, Chunk } from "effect"

class Exam {
  constructor(readonly person: string, readonly score: number) {}
}

// Define a list of exam results
const examResults = [
  new Exam("Alex", 64),
  new Exam("Michael", 97),
  new Exam("Bill", 77),
  new Exam("John", 78),
  new Exam("Bobby", 71)
]

// Group exam results by the tens place in the score
const groupByKeyResult = Stream.fromIterable(examResults).pipe(
  Stream.groupByKey((exam) => Math.floor(exam.score / 10) * 10)
)

// Count the number of exam results in each group
const stream = GroupBy.evaluate(groupByKeyResult, (key, stream) =>
  Stream.fromEffect(
    Stream.runCollect(stream).pipe(
      Effect.andThen((chunk) => [key, Chunk.size(chunk)] as const)
    )
  )
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ [ 60, 1 ], [ 90, 1 ], [ 70, 3 ] ] }
*/

groupBy

For more complex grouping requirements where partitioning involves effects, you can use the Stream.groupBy function. This function accepts an effectful partitioning function and returns a GroupBy data type, representing the grouped stream. You can then process each group by using GroupBy.evaluate, similar to Stream.groupByKey.

Example (Grouping Names by First Letter)

In the following example, we group names by their first letter and count the number of names in each group. Here, the partitioning operation is set up as an effectful operation:

import { Stream, GroupBy, Effect, Chunk } from "effect"

// Group names by their first letter
const groupByKeyResult = Stream.fromIterable([
  "Mary",
  "James",
  "Robert",
  "Patricia",
  "John",
  "Jennifer",
  "Rebecca",
  "Peter"
]).pipe(
  // Simulate an effectful groupBy operation
  Stream.groupBy((name) => Effect.succeed([name.substring(0, 1), name]))
)

// Count the number of names in each group and display results
const stream = GroupBy.evaluate(groupByKeyResult, (key, stream) =>
  Stream.fromEffect(
    Stream.runCollect(stream).pipe(
      Effect.andThen((chunk) => [key, Chunk.size(chunk)] as const)
    )
  )
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{
  _id: 'Chunk',
  values: [ [ 'M', 1 ], [ 'J', 3 ], [ 'R', 2 ], [ 'P', 2 ] ]
}
*/

grouped

The Stream.grouped function is ideal for dividing a stream into chunks of a specified size, making it easier to handle data in smaller, organized segments. This is particularly helpful when processing or displaying data in batches.

Example (Dividing a Stream into Chunks of 3 Elements)

import { Stream, Effect } from "effect"

// Create a stream of numbers and group them into chunks of 3
const stream = Stream.range(0, 8).pipe(Stream.grouped(3))

Effect.runPromise(Stream.runCollect(stream)).then((chunks) =>
  console.log("%o", chunks)
)
/*
Output:
{
  _id: 'Chunk',
  values: [
    { _id: 'Chunk', values: [ 0, 1, 2, [length]: 3 ] },
    { _id: 'Chunk', values: [ 3, 4, 5, [length]: 3 ] },
    { _id: 'Chunk', values: [ 6, 7, 8, [length]: 3 ] },
    [length]: 3
  ]
}
*/

groupedWithin

The Stream.groupedWithin function allows for flexible grouping by creating chunks based on either a specified maximum size or a time interval, whichever condition is met first. This is especially useful for working with data where timing constraints are involved.

Example (Grouping by Size or Time Interval)

In this example, Stream.groupedWithin(18, "1.5 seconds") groups the stream into chunks whenever either 18 elements accumulate or 1.5 seconds elapse since the last chunk was created.

import { Stream, Schedule, Effect, Chunk } from "effect"

// Create a stream that repeats every second and group by size or time
const stream = Stream.range(0, 9).pipe(
  Stream.repeat(Schedule.spaced("1 second")),
  Stream.groupedWithin(18, "1.5 seconds"),
  Stream.take(3)
)

Effect.runPromise(Stream.runCollect(stream)).then((chunks) =>
  console.log(Chunk.toArray(chunks))
)
/*
Output:
[
  {
    _id: 'Chunk',
    values: [
      0, 1, 2, 3, 4, 5, 6,
      7, 8, 9, 0, 1, 2, 3,
      4, 5, 6, 7
    ]
  },
  {
    _id: 'Chunk',
    values: [
      8, 9, 0, 1, 2,
      3, 4, 5, 6, 7,
      8, 9
    ]
  },
  {
    _id: 'Chunk',
    values: [
      0, 1, 2, 3, 4, 5, 6,
      7, 8, 9, 0, 1, 2, 3,
      4, 5, 6, 7
    ]
  }
]
*/

Concatenation

In stream processing, you may need to combine the contents of multiple streams. The Stream module offers several operators to achieve this, including Stream.concat, Stream.concatAll, and Stream.flatMap. Let’s look at how each of these operators works.

Simple Concatenation

The Stream.concat operator is a straightforward method for joining two streams. It returns a new stream that emits elements from the first stream (left-hand) followed by elements from the second stream (right-hand). This is helpful when you want to combine two streams in a specific sequence.

Example (Concatenating Two Streams Sequentially)

import { Stream, Effect } from "effect"

const stream = Stream.concat(Stream.make(1, 2, 3), Stream.make("a", "b"))

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ 1, 2, 3, 'a', 'b' ] }
*/

Concatenating Multiple Streams

If you have multiple streams to concatenate, Stream.concatAll provides an efficient way to combine them without manually chaining multiple Stream.concat operations. This function takes a Chunk of streams and returns a single stream containing the elements of each stream in sequence.

Example (Concatenating Multiple Streams)

import { Stream, Effect, Chunk } from "effect"

const s1 = Stream.make(1, 2, 3)
const s2 = Stream.make("a", "b")
const s3 = Stream.make(true, false, false)

const stream = Stream.concatAll<number | string | boolean, never, never>(
  Chunk.make(s1, s2, s3)
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{
  _id: 'Chunk',
  values: [
    1,     2,     3,
    'a',   'b',   true,
    false, false
  ]
}
*/

Advanced Concatenation with flatMap

The Stream.flatMap operator allows for advanced concatenation by creating a stream where each element is generated by applying a function of type (a: A) => Stream<...> to each output of the source stream. This operator then concatenates all the resulting streams, effectively flattening them.

Example (Generating Repeated Elements with Stream.flatMap)

import { Stream, Effect } from "effect"

// Create a stream where each element is repeated 4 times
const stream = Stream.make(1, 2, 3).pipe(
  Stream.flatMap((a) => Stream.repeatValue(a).pipe(Stream.take(4)))
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{
  _id: 'Chunk',
  values: [
    1, 1, 1, 1, 2,
    2, 2, 2, 3, 3,
    3, 3
  ]
}
*/

If you need to perform the flatMap operation concurrently, you can use the concurrency option to control how many inner streams run simultaneously. Additionally, if the order of concatenation is not important, you can use the switch option.

Merging

Sometimes, you may want to interleave elements from two streams and create a single output stream. In such cases, Stream.concat isn’t suitable because it waits for the first stream to complete before consuming the second. For interleaving elements as they become available, Stream.merge and its variants are designed for this purpose.

merge

The Stream.merge operation combines elements from two source streams into a single stream, interleaving elements as they are produced. Unlike Stream.concat, Stream.merge does not wait for one stream to finish before starting the other.

Example (Interleaving Two Streams with Stream.merge)

import { Schedule, Stream, Effect } from "effect"

// Create two streams with different emission intervals
const s1 = Stream.make(1, 2, 3).pipe(
  Stream.schedule(Schedule.spaced("100 millis"))
)
const s2 = Stream.make(4, 5, 6).pipe(
  Stream.schedule(Schedule.spaced("200 millis"))
)

// Merge s1 and s2 into a single stream that interleaves their values
const merged = Stream.merge(s1, s2)

Effect.runPromise(Stream.runCollect(merged)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ 1, 4, 2, 3, 5, 6 ] }
*/

Termination Strategy

When merging two streams, it’s important to consider the termination strategy, especially if each stream has a different lifetime. By default, Stream.merge waits for both streams to terminate before ending the merged stream. However, you can modify this behavior with haltStrategy, selecting from four termination strategies:

Termination Strategy	Description
"left"	The merged stream terminates when the left-hand stream terminates.
"right"	The merged stream terminates when the right-hand stream terminates.
"both" (default)	The merged stream terminates only when both streams have terminated.
"either"	The merged stream terminates as soon as either stream terminates.

Example (Using haltStrategy: "left" to Control Stream Termination)

import { Stream, Schedule, Effect } from "effect"

const s1 = Stream.range(1, 5).pipe(
  Stream.schedule(Schedule.spaced("100 millis"))
)
const s2 = Stream.repeatValue(0).pipe(
  Stream.schedule(Schedule.spaced("200 millis"))
)

const merged = Stream.merge(s1, s2, { haltStrategy: "left" })

Effect.runPromise(Stream.runCollect(merged)).then(console.log)
/*
Output:
{
  _id: 'Chunk',
  values: [
    1, 0, 2, 3,
    0, 4, 5
  ]
}
*/

mergeWith

In some cases, you may want to merge two streams while transforming their elements into a unified type. Stream.mergeWith is designed for this purpose, allowing you to specify transformation functions for each source stream.

Example (Merging and Transforming Two Streams)

import { Schedule, Stream, Effect } from "effect"

const s1 = Stream.make("1", "2", "3").pipe(
  Stream.schedule(Schedule.spaced("100 millis"))
)
const s2 = Stream.make(4.1, 5.3, 6.2).pipe(
  Stream.schedule(Schedule.spaced("200 millis"))
)

const merged = Stream.mergeWith(s1, s2, {
  // Convert string elements from `s1` to integers
  onSelf: (s) => parseInt(s),
  // Round down decimal elements from `s2`
  onOther: (n) => Math.floor(n)
})

Effect.runPromise(Stream.runCollect(merged)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ 1, 4, 2, 3, 5, 6 ] }
*/

Interleaving

interleave

The Stream.interleave operator lets you pull one element at a time from each of two streams, creating a new interleaved stream. If one stream finishes first, the remaining elements from the other stream continue to be pulled until both streams are exhausted.

Example (Basic Interleaving of Two Streams)

import { Stream, Effect } from "effect"

const s1 = Stream.make(1, 2, 3)
const s2 = Stream.make(4, 5, 6)

const stream = Stream.interleave(s1, s2)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{ _id: 'Chunk', values: [ 1, 4, 2, 5, 3, 6 ] }
*/

interleaveWith

For more complex interleaving, Stream.interleaveWith provides additional control by using a third stream of boolean values to dictate the interleaving pattern. When this stream emits true, an element is taken from the left-hand stream; otherwise, an element is taken from the right-hand stream.

Example (Custom Interleaving Logic Using Stream.interleaveWith)

import { Stream, Effect } from "effect"

const s1 = Stream.make(1, 3, 5, 7, 9)
const s2 = Stream.make(2, 4, 6, 8, 10)

// Define a boolean stream to control interleaving
const booleanStream = Stream.make(true, false, false).pipe(Stream.forever)

const stream = Stream.interleaveWith(s1, s2, booleanStream)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{
  _id: 'Chunk',
  values: [
    1, 2,  4, 3, 6,
    8, 5, 10, 7, 9
  ]
}
*/

Interspersing

Interspersing adds separators or affixes in a stream, useful for formatting or structuring data in streams.

intersperse

The Stream.intersperse operator inserts a specified delimiter element between each pair of elements in a stream. This delimiter can be any chosen value and is added between each consecutive pair.

Example (Inserting Delimiters Between Stream Elements)

import { Stream, Effect } from "effect"

// Create a stream of numbers and intersperse `0` between them
const stream = Stream.make(1, 2, 3, 4, 5).pipe(Stream.intersperse(0))

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{
  _id: 'Chunk',
  values: [
    1, 0, 2, 0, 3,
    0, 4, 0, 5
  ]
}
*/

intersperseAffixes

For more complex needs, Stream.intersperseAffixes provides control over different affixes at the start, between elements, and at the end of the stream.

Example (Adding Affixes to a Stream)

import { Stream, Effect } from "effect"

// Create a stream and add affixes:
// - `[` at the start
// - `|` between elements
// - `]` at the end
const stream = Stream.make(1, 2, 3, 4, 5).pipe(
  Stream.intersperseAffixes({
    start: "[",
    middle: "|",
    end: "]"
  })
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
{
  _id: 'Chunk',
  values: [
    '[', 1,   '|', 2,   '|',
    3,   '|', 4,   '|', 5,
    ']'
  ]
}
*/

Broadcasting

Broadcasting a stream creates multiple downstream streams that each receive the same elements from the source stream. This is useful when you want to send each element to multiple consumers simultaneously. The upstream stream has a maximumLag parameter that sets the limit for how much it can get ahead before slowing down to match the speed of the slowest downstream stream.

Example (Broadcasting to Multiple Downstream Streams)

In the following example, we broadcast a stream of numbers to two downstream consumers. The first calculates the maximum value in the stream, while the second logs each number with a delay. The upstream stream’s speed adjusts based on the slower logging stream:

import { Effect, Stream, Console, Schedule, Fiber } from "effect"

const numbers = Effect.scoped(
  Stream.range(1, 20).pipe(
    Stream.tap((n) =>
      Console.log(`Emit ${n} element before broadcasting`)
    ),
    // Broadcast to 2 downstream consumers with max lag of 5
    Stream.broadcast(2, 5),
    Stream.flatMap(([first, second]) =>
      Effect.gen(function* () {
        // First downstream stream: calculates maximum
        const fiber1 = yield* Stream.runFold(first, 0, (acc, e) =>
          Math.max(acc, e)
        ).pipe(
          Effect.andThen((max) => Console.log(`Maximum: ${max}`)),
          Effect.fork
        )

        // Second downstream stream: logs each element with a delay
        const fiber2 = yield* second.pipe(
          Stream.schedule(Schedule.spaced("1 second")),
          Stream.runForEach((n) =>
            Console.log(`Logging to the Console: ${n}`)
          ),
          Effect.fork
        )

        // Wait for both fibers to complete
        yield* Fiber.join(fiber1).pipe(
          Effect.zip(Fiber.join(fiber2), { concurrent: true })
        )
      })
    ),
    Stream.runCollect
  )
)

Effect.runPromise(numbers).then(console.log)
/*
Output:
Emit 1 element before broadcasting
Emit 2 element before broadcasting
Emit 3 element before broadcasting
Emit 4 element before broadcasting
Emit 5 element before broadcasting
Emit 6 element before broadcasting
Emit 7 element before broadcasting
Emit 8 element before broadcasting
Emit 9 element before broadcasting
Emit 10 element before broadcasting
Emit 11 element before broadcasting
Logging to the Console: 1
Logging to the Console: 2
Logging to the Console: 3
Logging to the Console: 4
Logging to the Console: 5
Emit 12 element before broadcasting
Emit 13 element before broadcasting
Emit 14 element before broadcasting
Emit 15 element before broadcasting
Emit 16 element before broadcasting
Logging to the Console: 6
Logging to the Console: 7
Logging to the Console: 8
Logging to the Console: 9
Logging to the Console: 10
Emit 17 element before broadcasting
Emit 18 element before broadcasting
Emit 19 element before broadcasting
Emit 20 element before broadcasting
Logging to the Console: 11
Logging to the Console: 12
Logging to the Console: 13
Logging to the Console: 14
Logging to the Console: 15
Maximum: 20
Logging to the Console: 16
Logging to the Console: 17
Logging to the Console: 18
Logging to the Console: 19
Logging to the Console: 20
{ _id: 'Chunk', values: [ undefined ] }
*/

Buffering

Effect streams use a pull-based model, allowing downstream consumers to control the rate at which they request elements. However, when there’s a mismatch in the speed between the producer and the consumer, buffering can help balance their interaction. The Stream.buffer operator is designed to manage this, allowing the producer to keep working even if the consumer is slower. You can set a maximum buffer capacity using the capacity option.

buffer

The Stream.buffer operator queues elements to allow the producer to work independently from the consumer, up to a specified capacity. This helps when a faster producer and a slower consumer need to operate smoothly without blocking each other.

Example (Using a Buffer to Handle Speed Mismatch)

import { Stream, Console, Schedule, Effect } from "effect"

const stream = Stream.range(1, 10).pipe(
  // Log each element before buffering
  Stream.tap((n) => Console.log(`before buffering: ${n}`)),
  // Buffer with a capacity of 4 elements
  Stream.buffer({ capacity: 4 }),
  // Log each element after buffering
  Stream.tap((n) => Console.log(`after buffering: ${n}`)),
  // Add a 5-second delay between each emission
  Stream.schedule(Schedule.spaced("5 seconds"))
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
before buffering: 1
before buffering: 2
before buffering: 3
before buffering: 4
before buffering: 5
before buffering: 6
after buffering: 1
after buffering: 2
before buffering: 7
after buffering: 3
before buffering: 8
after buffering: 4
before buffering: 9
after buffering: 5
before buffering: 10
...
*/

Different buffering options let you tailor the buffering strategy based on your use case:

Buffering Type	Configuration	Description
Bounded Queue	{ capacity: number }	Limits the queue to a fixed size.
Unbounded Queue	{ capacity: "unbounded" }	Allows an unlimited number of buffered items.
Sliding Queue	{ capacity: number, strategy: "sliding" }	Keeps the most recent items, discarding older ones when full.
Dropping Queue	{ capacity: number, strategy: "dropping" }	Keeps the earliest items, discarding new ones when full.

Debouncing

Debouncing is a technique used to prevent a function from firing too frequently, which is particularly useful when a stream emits values rapidly but only the last value after a pause is needed.

The Stream.debounce function achieves this by delaying the emission of values until a specified time period has passed without any new values. If a new value arrives during the waiting period, the timer resets, and only the latest value will eventually be emitted after a pause.

Example (Debouncing a Stream of Rapidly Emitted Values)

import { Stream, Effect } from "effect"

// Helper function to log with elapsed time since the last log
let last = Date.now()
const log = (message: string) =>
  Effect.sync(() => {
    const end = Date.now()
    console.log(`${message} after ${end - last}ms`)
    last = end
  })

const stream = Stream.make(1, 2, 3).pipe(
  // Emit the value 4 after 200 ms
  Stream.concat(
    Stream.fromEffect(Effect.sleep("200 millis").pipe(Effect.as(4)))
  ),
  // Continue with more rapid values
  Stream.concat(Stream.make(5, 6)),
  // Emit 7 after 150 ms
  Stream.concat(
    Stream.fromEffect(Effect.sleep("150 millis").pipe(Effect.as(7)))
  ),
  Stream.concat(Stream.make(8)),
  Stream.tap((n) => log(`Received ${n}`)),
  // Only emit values after a pause of at least 100 milliseconds
  Stream.debounce("100 millis"),
  Stream.tap((n) => log(`> Emitted ${n}`))
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Example Output:
Received 1 after 5ms
Received 2 after 2ms
Received 3 after 0ms
> Emitted 3 after 104ms
Received 4 after 99ms
Received 5 after 1ms
Received 6 after 0ms
> Emitted 6 after 101ms
Received 7 after 50ms
Received 8 after 1ms
> Emitted 8 after 101ms
{ _id: 'Chunk', values: [ 3, 6, 8 ] }
*/

Throttling

Throttling is a technique for regulating the rate at which elements are emitted from a stream. It helps maintain a steady data output pace, which is valuable in situations where data processing needs to occur at a consistent rate.

The Stream.throttle function uses the token bucket algorithm to control the rate of stream emissions.

Example (Throttle Configuration)

Stream.throttle({
  cost: () => 1,
  duration: "100 millis",
  units: 1
})

In this configuration:

• Each chunk processed uses one token (cost = () => 1).
• Tokens are replenished at a rate of one token (units: 1) every 100 milliseconds (duration: "100 millis").

Throttling Applies to Chunks, Not Elements

Note that throttling operates on chunks rather than individual elements. The cost function sets the token cost for each chunk.

Shape Strategy (Default)

The “shape” strategy moderates data flow by delaying chunk emissions until they comply with specified bandwidth constraints. This strategy ensures that data throughput does not exceed defined limits, allowing for steady and controlled data emission.

Example (Applying Throttling with the Shape Strategy)

import { Stream, Effect, Schedule, Chunk } from "effect"

// Helper function to log with elapsed time since last log
let last = Date.now()
const log = (message: string) =>
  Effect.sync(() => {
    const end = Date.now()
    console.log(`${message} after ${end - last}ms`)
    last = end
  })

const stream = Stream.fromSchedule(Schedule.spaced("50 millis")).pipe(
  Stream.take(6),
  Stream.tap((n) => log(`Received ${n}`)),
  Stream.throttle({
    cost: Chunk.size,
    duration: "100 millis",
    units: 1
  }),
  Stream.tap((n) => log(`> Emitted ${n}`))
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Example Output:
Received 0 after 56ms
> Emitted 0 after 0ms
Received 1 after 52ms
> Emitted 1 after 48ms
Received 2 after 52ms
> Emitted 2 after 49ms
Received 3 after 52ms
> Emitted 3 after 48ms
Received 4 after 52ms
> Emitted 4 after 47ms
Received 5 after 52ms
> Emitted 5 after 49ms
{ _id: 'Chunk', values: [ 0, 1, 2, 3, 4, 5 ] }
*/

Enforce Strategy

The “enforce” strategy strictly regulates data flow by discarding chunks that exceed bandwidth constraints.

Example (Throttling with the Enforce Strategy)

import { Stream, Effect, Schedule, Chunk } from "effect"

// Helper function to log with elapsed time since last log
let last = Date.now()
const log = (message: string) =>
  Effect.sync(() => {
    const end = Date.now()
    console.log(`${message} after ${end - last}ms`)
    last = end
  })

const stream = Stream.make(1, 2, 3, 4, 5, 6).pipe(
  Stream.schedule(Schedule.exponential("100 millis")),
  Stream.tap((n) => log(`Received ${n}`)),
  Stream.throttle({
    cost: Chunk.size,
    duration: "1 second",
    units: 1,
    strategy: "enforce"
  }),
  Stream.tap((n) => log(`> Emitted ${n}`))
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Example Output:
Received 1 after 106ms
> Emitted 1 after 1ms
Received 2 after 200ms
Received 3 after 402ms
Received 4 after 801ms
> Emitted 4 after 1ms
Received 5 after 1601ms
> Emitted 5 after 1ms
Received 6 after 3201ms
> Emitted 6 after 0ms
{ _id: 'Chunk', values: [ 1, 4, 5, 6 ] }
*/

burst option

The Stream.throttle function offers a burst option that allows for temporary increases in data throughput beyond the set rate limits. This option is set to greater than 0 to activate burst capability (default is 0, indicating no burst support). The burst capacity provides additional tokens in the token bucket, enabling the stream to momentarily exceed its configured rate when bursts of data occur.

Example (Throttling with Burst Capacity)

import { Effect, Schedule, Stream, Chunk } from "effect"

// Helper function to log with elapsed time since last log
let last = Date.now()
const log = (message: string) =>
  Effect.sync(() => {
    const end = Date.now()
    console.log(`${message} after ${end - last}ms`)
    last = end
  })

const stream = Stream.fromSchedule(Schedule.spaced("10 millis")).pipe(
  Stream.take(20),
  Stream.tap((n) => log(`Received ${n}`)),
  Stream.throttle({
    cost: Chunk.size,
    duration: "200 millis",
    units: 5,
    strategy: "enforce",
    burst: 2
  }),
  Stream.tap((n) => log(`> Emitted ${n}`))
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Example Output:
Received 0 after 16ms
> Emitted 0 after 0ms
Received 1 after 12ms
> Emitted 1 after 0ms
Received 2 after 11ms
> Emitted 2 after 0ms
Received 3 after 11ms
> Emitted 3 after 0ms
Received 4 after 11ms
> Emitted 4 after 1ms
Received 5 after 11ms
> Emitted 5 after 0ms
Received 6 after 12ms
> Emitted 6 after 0ms
Received 7 after 11ms
Received 8 after 12ms
Received 9 after 11ms
Received 10 after 11ms
> Emitted 10 after 0ms
Received 11 after 11ms
Received 12 after 11ms
Received 13 after 12ms
> Emitted 13 after 0ms
Received 14 after 11ms
Received 15 after 12ms
Received 16 after 11ms
Received 17 after 11ms
> Emitted 17 after 0ms
Received 18 after 12ms
Received 19 after 10ms
{
  _id: 'Chunk',
  values: [
    0, 1,  2,  3,  4,
    5, 6, 10, 13, 17
  ]
}
*/

In this setup, the stream starts with a bucket containing 5 tokens, allowing the first five chunks to be emitted instantly. The additional burst capacity of 2 accommodates further emissions momentarily, allowing for handling of subsequent data more flexibly. Over time, as the bucket refills according to the throttle configuration, additional elements are emitted, demonstrating how the burst capability can manage uneven data flows effectively.

Scheduling

When working with streams, you may need to introduce specific time intervals between each element’s emission. The Stream.schedule combinator allows you to set these intervals.

Example (Adding a Delay Between Stream Emissions)

import { Stream, Schedule, Console, Effect } from "effect"

// Create a stream that emits values with a 1-second delay between each
const stream = Stream.make(1, 2, 3, 4, 5).pipe(
  Stream.schedule(Schedule.spaced("1 second")),
  Stream.tap(Console.log)
)

Effect.runPromise(Stream.runCollect(stream)).then(console.log)
/*
Output:
1
2
3
4
5
{
  _id: "Chunk",
  values: [ 1, 2, 3, 4, 5 ]
}
*/

In this example, we’ve used the Schedule.spaced("1 second") schedule to introduce a one-second gap between each emission in the stream.