Creating Streams

In this section, we’ll explore various methods for creating Effect Streams. These methods will help you generate streams tailored to your needs.

Common Constructors

make

You can create a pure stream by using the Stream.make constructor. This constructor accepts a variable list of values as its arguments.

1
import { Stream, Effect } from "effect"
2

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

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

empty

Sometimes, you may require a stream that doesn’t produce any values. In such cases, you can use Stream.empty. This constructor creates a stream that remains empty.

1
import { Stream, Effect } from "effect"
2

3
const stream = Stream.empty
4

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

void

If you need a stream that contains a single void value, you can use Stream.void. This constructor is handy when you want to represent a stream with a single event or signal.

1
import { Stream, Effect } from "effect"
2

3
const stream = Stream.void
4

5
Effect.runPromise(Stream.runCollect(stream)).then(console.log)
6
// { _id: 'Chunk', values: [ undefined ] }

range

To create a stream of integers within a specified range [min, max] (including both endpoints, min and max), you can use Stream.range. This is particularly useful for generating a stream of sequential numbers.

1
import { Stream, Effect } from "effect"
2

3
// Creating a stream of numbers from 1 to 5
4
const stream = Stream.range(1, 5)
5

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

iterate

With Stream.iterate, you can generate a stream by applying a function iteratively to an initial value. The initial value becomes the first element produced by the stream, followed by subsequent values produced by f(init), f(f(init)), and so on.

1
import { Stream, Effect } from "effect"
2

3
// Creating a stream of incrementing numbers
4
const stream = Stream.iterate(1, (n) => n + 1) // Produces 1, 2, 3, ...
5

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

scoped

Stream.scoped is used to create a single-valued stream from a scoped resource. It can be handy when dealing with resources that require explicit acquisition, usage, and release.

1
import { Stream, Effect, Console } from "effect"
2

3
// Creating a single-valued stream from a scoped resource
4
const stream = Stream.scoped(
5
  Effect.acquireUseRelease(
6
    Console.log("acquire"),
7
    () => Console.log("use"),
8
    () => Console.log("release"),
9
  ),
10
)
11

12
Effect.runPromise(Stream.runCollect(stream)).then(console.log)
13
/*
14
Output:
15
acquire
16
use
17
release
18
{ _id: 'Chunk', values: [ undefined ] }
19
*/

From Success and Failure

Much like the Effect data type, you can generate a Stream using the fail and succeed functions:

1
import { Stream, Effect } from "effect"
2

3
// Creating a stream that can emit errors
4
const streamWithError: Stream.Stream<never, string> = Stream.fail("Uh oh!")
5

6
Effect.runPromise(Stream.runCollect(streamWithError))
7
// throws Error: Uh oh!
8

9
// Creating a stream that emits a numeric value
10
const streamWithNumber: Stream.Stream<number> = Stream.succeed(5)
11

12
Effect.runPromise(Stream.runCollect(streamWithNumber)).then(console.log)
13
// { _id: 'Chunk', values: [ 5 ] }

From Chunks

You can construct a stream from a Chunk like this:

1
import { Stream, Chunk, Effect } from "effect"
2

3
// Creating a stream with values from a single Chunk
4
const stream = Stream.fromChunk(Chunk.make(1, 2, 3))
5

6
Effect.runPromise(Stream.runCollect(stream)).then(console.log)
7
// { _id: 'Chunk', values: [ 1, 2, 3 ] }

Moreover, you can create a stream from multiple Chunks as well:

1
import { Stream, Chunk, Effect } from "effect"
2

3
// Creating a stream with values from multiple Chunks
4
const stream = Stream.fromChunks(Chunk.make(1, 2, 3), Chunk.make(4, 5, 6))
5

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

From Effect

You can generate a stream from an Effect workflow by employing the Stream.fromEffect constructor. For instance, consider the following stream, which generates a single random number:

1
import { Stream, Random, Effect } from "effect"
2

3
const stream = Stream.fromEffect(Random.nextInt)
4

5
Effect.runPromise(Stream.runCollect(stream)).then(console.log)
6
// Example Output: { _id: 'Chunk', values: [ 1042302242 ] }

This method allows you to seamlessly transform the output of an Effect into a stream, providing a straightforward way to work with asynchronous operations within your streams.

From Asynchronous Callback

Imagine you have an asynchronous function that relies on callbacks. If you want to capture the results emitted by those callbacks as a stream, you can use the Stream.async function. This function is designed to adapt functions that invoke their callbacks multiple times and emit the results as a stream.

Let’s break down how to use it in the following example:

1
import { Stream, Effect, Chunk, Option, StreamEmit } from "effect"
2

3
const events = [1, 2, 3, 4]
4

5
const stream = Stream.async((emit: StreamEmit.Emit<never, never, number, void>) => {
6
  events.forEach((n) => {
7
    setTimeout(() => {
8
      if (n === 3) {
9
        // Terminate the stream
10
        emit(Effect.fail(Option.none()))
11
      } else {
12
        // Add the current item to the stream
13
        emit(Effect.succeed(Chunk.of(n)))
14
      }
15
    }, 100 * n)
16
  })
17
})
18

19
Effect.runPromise(Stream.runCollect(stream)).then(console.log)
20
// { _id: 'Chunk', values: [ 1, 2 ] }

The StreamEmit.Emit<R, E, A, void> type represents an asynchronous callback that can be called multiple times. This callback takes a value of type Effect<Chunk<A>, Option<E>, R>. Here’s what each of the possible outcomes means:

When the value provided to the callback results in a Chunk<A> upon success, it signifies that the specified elements should be emitted as part of the stream.
If the value passed to the callback results in a failure with Some<E>, it indicates the termination of the stream with the specified error.
When the value passed to the callback results in a failure with None, it serves as a signal for the end of the stream, essentially terminating it.

To put it simply, this type allows you to specify how your asynchronous callback interacts with the stream, determining when to emit elements, when to terminate with an error, or when to signal the end of the stream.

From Iterables

fromIterable

You can create a pure stream from an Iterable of values using the Stream.fromIterable constructor. It’s a straightforward way to convert a collection of values into a stream.

1
import { Stream, Effect } from "effect"
2

3
const numbers = [1, 2, 3]
4

5
const stream = Stream.fromIterable(numbers)
6

7
Effect.runPromise(Stream.runCollect(stream)).then(console.log)
8
// { _id: 'Chunk', values: [ 1, 2, 3 ] }

fromIterableEffect

When you have an effect that produces a value of type Iterable, you can employ the Stream.fromIterableEffect constructor to generate a stream from that effect.

For instance, let’s say you have a database operation that retrieves a list of users. Since this operation involves effects, you can utilize Stream.fromIterableEffect to convert the result into a Stream:

1
import { Stream, Effect, Context } from "effect"
2

3
class Database extends Context.Tag("Database")<
4
  Database,
5
  { readonly getUsers: Effect.Effect<Array<string>> }
6
>() {}
7

8
const getUsers = Database.pipe(Effect.andThen((_) => _.getUsers))
9

10
const stream = Stream.fromIterableEffect(getUsers)
11

12
Effect.runPromise(
13
  Stream.runCollect(
14
    stream.pipe(
15
      Stream.provideService(Database, {
16
        getUsers: Effect.succeed(["user1", "user2"]),
17
      }),
18
    ),
19
  ),
20
).then(console.log)
21
// { _id: 'Chunk', values: [ 'user1', 'user2' ] }

This enables you to work seamlessly with effects and convert their results into streams for further processing.

fromAsyncIterable

Async iterables are another type of data source that can be converted into a stream. With the Stream.fromAsyncIterable constructor, you can work with asynchronous data sources and handle potential errors gracefully.

1
import { Stream, Effect } from "effect"
2

3
const myAsyncIterable = async function* () {
4
  yield 1
5
  yield 2
6
}
7

8
const stream = Stream.fromAsyncIterable(
9
  myAsyncIterable(),
10
  (e) => new Error(String(e)), // Error Handling
11
)
12

13
Effect.runPromise(Stream.runCollect(stream)).then(console.log)
14
// { _id: 'Chunk', values: [ 1, 2 ] }

In this code, we define an async iterable and then create a stream named stream from it. Additionally, we provide an error handler function to manage any potential errors that may occur during the conversion.

From Repetition

Repeating a Single Value

You can create a stream that endlessly repeats a specific value using the Stream.repeatValue constructor:

1
import { Stream, Effect } from "effect"
2

3
const stream = Stream.repeatValue(0)
4

5
Effect.runPromise(Stream.runCollect(stream.pipe(Stream.take(5)))).then(console.log)
6
// { _id: 'Chunk', values: [ 0, 0, 0, 0, 0 ] }

Repeating a Stream’s Content

Stream.repeat allows you to create a stream that repeats a specified stream’s content according to a schedule. This can be useful for generating recurring events or values.

1
import { Stream, Effect, Schedule } from "effect"
2

3
// Creating a stream that repeats a value indefinitely
4
const stream = Stream.repeat(Stream.succeed(1), Schedule.forever)
5

6
Effect.runPromise(Stream.runCollect(stream.pipe(Stream.take(5)))).then(console.log)
7
// { _id: 'Chunk', values: [ 1, 1, 1, 1, 1 ] }

Repeating an Effect’s Result

Imagine you have an effectful API call, and you want to use the result of that call to create a stream. You can achieve this by creating a stream from the effect and repeating it indefinitely.

Here’s an example of generating a stream of random numbers:

1
import { Stream, Effect, Random } from "effect"
2

3
const stream = Stream.repeatEffect(Random.nextInt)
4

5
Effect.runPromise(Stream.runCollect(stream.pipe(Stream.take(5)))).then(console.log)
6
/*
7
Example Output:
8
{
9
  _id: 'Chunk',
10
  values: [ 1666935266, 604851965, 2194299958, 3393707011, 4090317618 ]
11
}
12
*/

Repeating an Effect with Termination

You can repeatedly evaluate a given effect and terminate the stream based on specific conditions.

In this example, we’re draining an Iterator to create a stream from it:

1
import { Stream, Effect, Option } from "effect"
2

3
const drainIterator = <A>(it: Iterator<A>): Stream.Stream<A> =>
4
  Stream.repeatEffectOption(
5
    Effect.sync(() => it.next()).pipe(
6
      Effect.andThen((res) => {
7
        if (res.done) {
8
          return Effect.fail(Option.none())
9
        }
10
        return Effect.succeed(res.value)
11
      }),
12
    ),
13
  )

Generating Ticks

You can create a stream that emits void values at specified intervals using the Stream.tick constructor. This is useful for creating periodic events.

1
import { Stream, Effect } from "effect"
2

3
const stream = Stream.tick("100 millis")
4

5
Effect.runPromise(Stream.runCollect(stream.pipe(Stream.take(5)))).then(console.log)
6
/*
7
Output:
8
{
9
  _id: 'Chunk',
10
  values: [ undefined, undefined, undefined, undefined, undefined ]
11
}
12
*/

From Unfolding/Pagination

In functional programming, the concept of unfold can be thought of as the counterpart to fold.

With fold, we process a data structure and produce a return value. For example, we can take an Array<number> and calculate the sum of its elements.

On the other hand, unfold represents an operation where we start with an initial value and generate a recursive data structure, adding one element at a time using a specified state function. For example, we can create a sequence of natural numbers starting from 1 and using the increment function as the state function.

Unfold

unfold

The Stream module includes an unfold function defined as follows:

declare const unfold: <S, A>(
  initialState: S,
  step: (s: S) => Option.Option<readonly [A, S]>,
) => Stream<A>

Here’s how it works:

initialState. This is the initial state value.
step. The state function step takes the current state s as input. If the result of this function is None, the stream ends. If it’s Some<[A, S]>, the next element in the stream is A, and the state S is updated for the next step process.

For example, let’s create a stream of natural numbers using Stream.unfold:

1
import { Stream, Effect, Option } from "effect"
2

3
const stream = Stream.unfold(1, (n) => Option.some([n, n + 1]))
4

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

unfoldEffect

Sometimes, we may need to perform effectful state transformations during the unfolding operation. This is where Stream.unfoldEffect comes in handy. It allows us to work with effects while generating streams.

Here’s an example of creating an infinite stream of random 1 and -1 values using Stream.unfoldEffect:

1
import { Stream, Effect, Option, Random } from "effect"
2

3
const stream = Stream.unfoldEffect(1, (n) =>
4
  Random.nextBoolean.pipe(Effect.map((b) => (b ? Option.some([n, -n]) : Option.some([n, n])))),
5
)
6

7
Effect.runPromise(Stream.runCollect(stream.pipe(Stream.take(5)))).then(console.log)
8
// Example Output: { _id: 'Chunk', values: [ 1, 1, 1, 1, -1 ] }

Additional Variants

There are also similar operations like Stream.unfoldChunk and Stream.unfoldChunkEffect tailored for working with Chunk data types.

Pagination

paginate

Stream.paginate is similar to Stream.unfold but allows emitting values one step further.

For example, the following stream emits 0, 1, 2, 3 elements:

1
import { Stream, Effect, Option } from "effect"
2

3
const stream = Stream.paginate(0, (n) => [n, n < 3 ? Option.some(n + 1) : Option.none()])
4

5
Effect.runPromise(Stream.runCollect(stream)).then(console.log)
6
// { _id: 'Chunk', values: [ 0, 1, 2, 3 ] }

Here’s how it works:

We start with an initial value of 0.
The provided function takes the current value n and returns a tuple. The first element of the tuple is the value to emit (n), and the second element determines whether to continue (Option.some(n + 1)) or stop (Option.none()).

Additional Variants

There are also similar operations like Stream.paginateChunk and Stream.paginateChunkEffect tailored for working with Chunk data types.

Unfolding vs. Pagination

You might wonder about the difference between the unfold and paginate combinators and when to use one over the other. Let’s explore this by diving into an example.

Imagine we have a paginated API that provides a substantial amount of data in a paginated manner. When we make a request to this API, it returns a ResultPage object containing the results for the current page and a flag indicating whether it’s the last page or if there’s more data to retrieve on the next page. Here’s a simplified representation of our API:

1
import { Chunk, Effect } from "effect"
2

3
type RawData = string
4

5
class PageResult {
6
  constructor(
7
    readonly results: Chunk.Chunk<RawData>,
8
    readonly isLast: boolean,
9
  ) {}
10
}
11

12
const pageSize = 2
13

14
const listPaginated = (pageNumber: number): Effect.Effect<PageResult, Error> => {
15
  return Effect.succeed(
16
    new PageResult(
17
      Chunk.map(Chunk.range(1, pageSize), (index) => `Result ${pageNumber}-${index}`),
18
      pageNumber === 2, // Return 3 pages
19
    ),
20
  )
21
}

Our goal is to convert this paginated API into a stream of RowData events. For our initial attempt, we might think that using the Stream.unfold operation is the way to go:

1
import { Chunk, Effect, Stream, Option } from "effect"
2

24 collapsed lines
3
type RawData = string
4

5
class PageResult {
6
  constructor(
7
    readonly results: Chunk.Chunk<RawData>,
8
    readonly isLast: boolean,
9
  ) {}
10
}
11

12
const pageSize = 2
13

14
const listPaginated = (pageNumber: number): Effect.Effect<PageResult, Error> => {
15
  return Effect.succeed(
16
    new PageResult(
17
      Chunk.map(Chunk.range(1, pageSize), (index) => `Result ${pageNumber}-${index}`),
18
      pageNumber === 2, // Return 3 pages
19
    ),
20
  )
21
}
22

23
const firstAttempt = Stream.unfoldChunkEffect(0, (pageNumber) =>
24
  listPaginated(pageNumber).pipe(
25
    Effect.map((page) => {
26
      if (page.isLast) {
27
        return Option.none()
28
      }
29
      return Option.some([page.results, pageNumber + 1] as const)
30
    }),
31
  ),
32
)
33

34
Effect.runPromise(Stream.runCollect(firstAttempt)).then(console.log)
35
/*
36
Output:
37
{
38
  _id: "Chunk",
39
  values: [ "Result 0-1", "Result 0-2", "Result 1-1", "Result 1-2" ]
40
}
41
*/

However, this approach has a drawback, it doesn’t include the results from the last page. To work around this, we perform an extra API call to include those missing results:

1
import { Chunk, Effect, Stream, Option } from "effect"
2

24 collapsed lines
3
type RawData = string
4

5
class PageResult {
6
  constructor(
7
    readonly results: Chunk.Chunk<RawData>,
8
    readonly isLast: boolean,
9
  ) {}
10
}
11

12
const pageSize = 2
13

14
const listPaginated = (pageNumber: number): Effect.Effect<PageResult, Error> => {
15
  return Effect.succeed(
16
    new PageResult(
17
      Chunk.map(Chunk.range(1, pageSize), (index) => `Result ${pageNumber}-${index}`),
18
      pageNumber === 2, // Return 3 pages
19
    ),
20
  )
21
}
22

23
const secondAttempt = Stream.unfoldChunkEffect(Option.some(0), (pageNumber) =>
24
  Option.match(pageNumber, {
25
    // We already hit the last page
26
    onNone: () => Effect.succeed(Option.none()),
27
    // We did not hit the last page yet
28
    onSome: (pageNumber) =>
29
      listPaginated(pageNumber).pipe(
30
        Effect.map((page) =>
31
          Option.some([page.results, page.isLast ? Option.none() : Option.some(pageNumber + 1)]),
32
        ),
33
      ),
34
  }),
35
)
36

37
Effect.runPromise(Stream.runCollect(secondAttempt)).then(console.log)
38
/*
39
Output:
40
{
41
  _id: 'Chunk',
42
  values: [
43
    'Result 0-1',
44
    'Result 0-2',
45
    'Result 1-1',
46
    'Result 1-2',
47
    'Result 2-1',
48
    'Result 2-2'
49
  ]
50
}
51
*/

While this approach works, it’s clear that Stream.unfold isn’t the most friendly option for retrieving data from paginated APIs. It requires additional workarounds to include the results from the last page.

This is where Stream.paginate comes to the rescue. It provides a more ergonomic way to convert a paginated API into an Effect stream. Let’s rewrite our solution using Stream.paginate:

1
import { Chunk, Effect, Stream, Option } from "effect"
2

24 collapsed lines
3
type RawData = string
4

5
class PageResult {
6
  constructor(
7
    readonly results: Chunk.Chunk<RawData>,
8
    readonly isLast: boolean,
9
  ) {}
10
}
11

12
const pageSize = 2
13

14
const listPaginated = (pageNumber: number): Effect.Effect<PageResult, Error> => {
15
  return Effect.succeed(
16
    new PageResult(
17
      Chunk.map(Chunk.range(1, pageSize), (index) => `Result ${pageNumber}-${index}`),
18
      pageNumber === 2, // Return 3 pages
19
    ),
20
  )
21
}
22

23
const finalAttempt = Stream.paginateChunkEffect(0, (pageNumber) =>
24
  listPaginated(pageNumber).pipe(
25
    Effect.andThen((page) => {
26
      return [page.results, page.isLast ? Option.none<number>() : Option.some(pageNumber + 1)]
27
    }),
28
  ),
29
)
30

31
Effect.runPromise(Stream.runCollect(finalAttempt)).then(console.log)
32
/*
33
Output:
34
{
35
  _id: 'Chunk',
36
  values: [
37
    'Result 0-1',
38
    'Result 0-2',
39
    'Result 1-1',
40
    'Result 1-2',
41
    'Result 2-1',
42
    'Result 2-2'
43
  ]
44
}
45
*/

From Queue and PubSub

In Effect, there are two essential asynchronous messaging data types: Queue and PubSub. You can easily transform these data types into Streams by utilizing Stream.fromQueue and Stream.fromPubSub, respectively.

From Schedule

We can create a stream from a Schedule that does not require any further input. The stream will emit an element for each value output from the schedule, continuing for as long as the schedule continues:

1
import { Effect, Stream, Schedule } from "effect"
2

3
// Emits values every 1 second for a total of 10 emissions
4
const schedule = Schedule.spaced("1 second").pipe(Schedule.compose(Schedule.recurs(10)))
5

6
const stream = Stream.fromSchedule(schedule)
7

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