streaming-iterables

streaming-iterables 🏄‍♂️

A Swiss army knife for async iterables. Designed to help replace your streams. These utilities have a comparable speed, friendlier error handling, and are easier to understand than most stream based workloads.

Streams were our last best hope for processing unbounded amounts of data. Since Node 10 they have become something greater, they've become async iterable. With async iterators you can have less code, do more work, faster.

If you still need streams with async functions, check out sister project bluestream🏄‍♀️!

We support and test against LTS node releases, but may work with older versions of node.

Install

There are no dependencies.

npm install streaming-iterables

We ship esm, umd and types.

Overview

Every function is curryable, you can call it with any number of arguments. For example:

import { map } from 'streaming-iterables'

for await (const str of map(String, [1,2,3])) {
  console.log(str)
}
// "1", "2", "3"

const stringable = map(String)
for await (const str of stringable([1,2,3])) {
  console.log(str)
}
// "1", "2", "3"

Since this works with async iterators it requires node 10 or higher.

API

• batch()
• batchWithTimeout()
• buffer()
• collect()
• concat()
• consume()
• flatMap()
• flatten()
• flatTransform()
• fromStream()
• filter()
• getIterator()
• map()
• merge()
• parallelMap()
• parallelMerge()
• pipeline()
• reduce()
• take()
• tap()
• throttle()
• time()
• transform()
• writeToStream()

batch

function batch<T>(size: number, iterable: AsyncIterable<T>): AsyncGenerator<T[]>
function batch<T>(size: number, iterable: Iterable<T>): Generator<T[]>

Batch objects from iterable into arrays of size length. The final array may be shorter than size if there is not enough items. Returns a sync iterator if the iterable is sync, otherwise an async iterator. Errors from the source iterable are immediately raised.

size can be between 1 and Infinity.

import { batch } from 'streaming-iterables'
import { getPokemon } from 'iterable-pokedex'

// batch 10 pokemon while we process them
for await (const pokemons of batch(10, getPokemon())) {
  console.log(pokemons) // 10 pokemon at a time!
}

batchWithTimeout

function batchWithTimeout<T>(size: number, timeout: number, iterable: AsyncIterable<T>): AsyncGenerator<T[]>
function batchWithTimeout<T>(size: number, timeout: number, iterable: Iterable<T>): Generator<T[]>

Like batch but flushes early if the timeout is reached. The batches may be shorter than size if there are not enough items. Returns a sync iterator if the iterable is sync, otherwise an async iterator. Errors from the source iterable are immediately raised.

size can be between 1 and Infinity. timeout can be between 0 and Infinity.

import { batchWithTimeout } from 'streaming-iterables'
import { getPokemon } from 'iterable-pokedex'

// batch 10 pokemon while we process them
for await (const pokemons of batchWithTimeout(10, 100, getPokemon())) {
  console.log(pokemons) // Up to 10 pokemon at a time!
}

buffer

function buffer<T>(size: number, iterable: AsyncIterable<T>): AsyncIterable<T>
function buffer<T>(size: number, iterable: Iterable<T>): AsyncIterable<T>

Buffer keeps a number of objects in reserve available for immediate reading. This is helpful with async iterators as it will pre-fetch results so you don't have to wait for them to load. For sync iterables it will pre-compute up to size values and keep them in reserve. The internal buffer will start to be filled once .next() is called for the first time and will continue to fill until the source iterable is exhausted or the buffer is full. Errors from the source iterable will be raised after all buffered values are yielded.

size can be between 0 and Infinity.

import { buffer } from 'streaming-iterables'
import { getPokemon, trainMonster } from 'iterable-pokedex'

// load 10 monsters in the background while we process them one by one
for await (const monster of buffer(10, getPokemon())) {
  await trainMonster(monster) // got to do some pokéwork
}

collect

function collect<T>(iterable: Iterable<T>): T[]
function collect<T>(iterable: AsyncIterable<T>): Promise<T[]>

Collect all the values from an iterable into an array. Returns an array if you pass it an iterable and a promise for an array if you pass it an async iterable. Errors from the source iterable are raised immediately.

import { collect } from 'streaming-iterables'
import { getPokemon } from 'iterable-pokedex'

console.log(await collect(getPokemon()))
// [bulbasaur, ivysaur, venusaur, charmander, ...]

concat

function concat(...iterables: Array<Iterable<any>>): Iterable<any>
function concat(...iterables: Array<AnyIterable<any>>): AsyncIterable<any>

Combine multiple iterators into a single iterable. Reads each iterable completely one at a time. Returns a sync iterator if all iterables are sync, otherwise it returns an async iterable. Errors from the source iterable are raised immediately.

import { concat } from 'streaming-iterables'
import { getPokemon } from 'iterable-pokedex'
import { getTransformers } from './util'

for await (const hero of concat(getPokemon(2), getTransformers(2))) {
  console.log(hero)
}
// charmander
// bulbasaur <- end of pokemon
// megatron
// bumblebee <- end of transformers

consume

export function consume<T>(iterable: Iterable<T>): void
export function consume<T>(iterable: AsyncIterable<T>): Promise<void>

A promise that resolves after the function drains the iterable of all data. Useful for processing a pipeline of data. Errors from the source iterable are raised immediately.

import { consume, map } from 'streaming-iterables'
import { getPokemon, trainMonster } from 'iterable-pokedex'

const train = map(trainMonster)
await consume(train(getPokemon())) // load all the pokemon and train them!

flatMap

function flatMap<T, B>(func: (data: T) => FlatMapValue<B>, iterable: AnyIterable<T>): AsyncGenerator<B>

Map func over the iterable, flatten the result and then ignore all null or undefined values. It's the transform function we've always needed. It's equivalent to;

(func, iterable) => filter(i => i !== undefined && i !== null, flatten(map(func, iterable)))

note: The return value for func is FlatMapValue<B>. Typescript doesn't have recursive types but you can nest iterables as deep as you like.

The ordering of the results is guaranteed. Errors from the source iterable are raised after all mapped values are yielded. Errors from func are raised after all previously mapped values are yielded.

import { flatMap } from 'streaming-iterables'
import { getPokemon, lookupStats } from 'iterable-pokedex'

async function getDefeatedGyms(pokemon) {
  if (pokemon.gymBattlesWon > 0) {
    const stats = await lookupStats(pokemon)
    return stats.gyms
  }
}

for await (const gym of flatMap(getDefeatedGyms, getPokemon())) {
  console.log(gym.name)
}
// "Pewter Gym"
// "Cerulean Gym"
// "Vermilion Gym"

flatten

function flatten<B>(iterable: AnyIterable<B | AnyIterable<B>>): AsyncIterableIterator<B>

Returns a new iterator by pulling every item out of iterable (and all its sub iterables) and yielding them depth-first. Checks for the iterable interfaces and iterates it if it exists. If the value is a string it is not iterated as that ends up in an infinite loop. Errors from the source iterable are raised immediately.

note: Typescript doesn't have recursive types but you can nest iterables as deep as you like.

import { flatten } from 'streaming-iterables'

for await (const item of flatten([1, 2, [3, [4, 5], 6])) {
  console.log(item)
}
// 1
// 2
// 3
// 4
// 5
// 6

flatTransform

function flatTransform<T, R>(concurrency: number, func: (data: T) => FlatMapValue<R>, iterable: AnyIterable<T>): AsyncIterableIterator<R>

Map func over the iterable, flatten the result and then ignore all null or undefined values. Returned async iterables are flattened concurrently too. It's the transform function we've always wanted.

It's similar to;

const filterEmpty = filter(i => i !== undefined && i !== null)
(concurrency, func, iterable) => filterEmpty(flatten(transform(concurrency, func, iterable)))

note: The return value for func is FlatMapValue<B>. Typescript doesn't have recursive types but you can nest iterables as deep as you like. However only directly returned async iterables are processed concurrently. (Eg, if you use an async generator function as func it's output will be processed concurrently, but if it's nested inside other iterables it will be processed sequentially.)

Order is determined by when async operations resolve. And it will run up to concurrency async operations at once. This includes promises and async iterables returned from func. Errors from the source iterable are raised after all transformed values are yielded. Errors from func are raised after all previously transformed values are yielded.

concurrency can be between 1 and Infinity.

Promise Example;

import { flatTransform } from 'streaming-iterables'
import { getPokemon, lookupStats } from 'iterable-pokedex'

async function getDefeatedGyms(pokemon) {
  if (pokemon.gymBattlesWon > 0) {
    const stats = await lookupStats(pokemon)
    return stats.gyms
  }
}

// lookup 10 stats at a time
for await (const gym of flatTransform(10, getDefeatedGyms, getPokemon())) {
  console.log(gym.name)
}
// "Pewter Gym"
// "Cerulean Gym"
// "Vermilion Gym"

Async Generator Example

import { flatTransform } from 'streaming-iterables'
import { getPokemon } from 'iterable-pokedex'
import { findFriendsFB, findFriendsMySpace } from './util'


async function* findFriends (pokemon) {
  yield await findFriendsFB(pokemon.name)
  yield await findFriendsMySpace(pokemon.name)
}

for await (const pokemon of flatTransform(10, findFriends, getPokemon())) {
  console.log(pokemon.name)
}
// Pikachu
// Meowth
// Ash - FB
// Jessie - FB
// Misty - MySpace
// James - MySpace

fromStream

function fromStream<T>(stream: Readable): AsyncIterable<T>

Wraps the stream in an async iterator or returns the stream if it already is an async iterator.

note: Since Node 10, streams already async iterators. This function may be used to ensure compatibility with older versions of Node.

import { fromStream } from 'streaming-iterables'
import { createReadStream } from 'fs'

const pokeLog = fromStream(createReadStream('./pokedex-operating-system.log'))

for await (const pokeData of pokeLog) {
  console.log(pokeData) // Buffer(...)
}

filter

function filter<T>(filterFunc: (data: T) => boolean | Promise<boolean>, iterable: AnyIterable<T>): AsyncIterableIterator<T>

Takes a filterFunc and a iterable, and returns a new async iterator of the same type containing the members of the given iterable which cause the filterFunc to return true.

import { filter } from 'streaming-iterables'
import { getPokemon } from 'iterable-pokedex'

const filterWater = filter(pokemon => pokemon.types.include('Water'))

for await (const pokemon of filterWater(getPokemon())) {
  console.log(pokemon)
}
// squirtle
// vaporeon
// magikarp

getIterator

function getIterator<T>(values: Iterableish<T>): Iterator<T> | AsyncIterator<T>

Get the iterator from any iterable or just return an iterator itself.

map

function map<T, B>(func: (data: T) => B | Promise<B>, iterable: AnyIterable<T>): AsyncIterableIterator<B>

Map a function or async function over all the values of an iterable. Errors from the source iterable and func are raised immediately.

import { consume, map } from 'streaming-iterables'
import got from 'got'

const urls = ['https://http.cat/200', 'https://http.cat/201', 'https://http.cat/202']
const download = map(got)

// download one at a time
for await (page of download(urls)) {
  console.log(page)
}

merge

function merge(...iterables: Array<AnyIterable<any>>): AsyncIterableIterator<any>

Combine multiple iterators into a single iterable. Reads one item off each iterable in order repeatedly until they are all exhausted. If you care less about order and want them faster see parallelMerge().

parallelMap

function parallelMap<T, R>(concurrency: number, func: (data: T) => R | Promise<R>, iterable: AnyIterable<T>): AsyncIterableIterator<R>

Map a function or async function over all the values of an iterable and do them concurrently. Errors from the source iterable are raised after all mapped values are yielded. Errors from func are raised after all previously mapped values are yielded. Just like map().

concurrency can be between 1 and Infinity.

If you don't care about order, see the faster transform() function.

import { consume, parallelMap } from 'streaming-iterables'
import got from 'got'

const urls = ['https://http.cat/200', 'https://http.cat/201', 'https://http.cat/202']
const download = parallelMap(2, got)

// download two at a time
for await (page of download(urls)) {
  console.log(page)
}

parallelMerge

function parallelMerge<T>(...iterables: Array<AnyIterable<T>>): AsyncIterableIterator<T>

Combine multiple iterators into a single iterable. Reads one item off of every iterable and yields them as they resolve. This is useful for pulling items out of a collection of iterables as soon as they're available. Errors iterables are raised immediately.

import { parallelMerge } from 'streaming-iterables'
import { getPokemon, getTransformer } from 'iterable-pokedex'

// pokemon are much faster to load btw
const heros = parallelMerge(getPokemon(), getTransformer())
for await (const hero of heros) {
  console.log(hero)
}
// charmander
// bulbasaur
// megatron
// pikachu
// eevee
// bumblebee
// jazz

pipeline

function pipeline(firstFn: Function, ...fns: Function[]): any;

Calls firstFn and then every function in fns with the result of the previous function. The final return is the result of the last function in fns.

import { pipeline, map, collect } from 'streaming-iterables'
import { getPokemon } from 'iterable-pokedex'
const getName = map(pokemon => pokemon.name)

// equivalent to `await collect(getName(getPokemon()))`
await pipeline(getPokemon, getName, collect)
// charmander
// bulbasaur
// MissingNo.

reduce

function reduce<T, B>(func: (acc: B, value: T) => B, start: B, iterable: AnyIterable<T>): Promise<B>

An async function that takes a reducer function, an initial value and an iterable.

Reduces an iterable to a value which is the accumulated result of running each value from the iterable thru func, where each successive invocation is supplied the return value of the previous. Errors are immediate raised.

take

function take<T>(count: number, iterable: AsyncIterable<T>): AsyncIterableIterator<T>
function take<T>(count: number, iterable: Iterable<T>): IterableIterator<T>

Returns a new iterator that reads a specific number of items from iterable. When used with generators it advances the generator, when used with arrays it gets a new iterator and starts from the beginning.

tap

function tap<T>(func: (data: T) => any, iterable: AnyIterable<T>): AsyncIterableIterator<T>

Returns a new iterator that yields the data it consumes, passing the data through to a function. If you provide an async function, the iterator will wait for the promise to resolve before yielding the value. This is useful for logging, or processing information and passing it along.

throttle

function throttle<T>(limit: number, interval: number, iterable: AnyIterable<T>): AsyncGenerator<T>

Throttles iterable at a rate of limit per interval without discarding data. Useful for throttling rate limited APIs.

limit can be greater than 0 but less than Infinity. interval can be greater than or equal to 0 but less than Infinity.

import { throttle } from 'streaming-iterables'
import { getPokemon, trainMonster } from 'iterable-pokedex'

// load monsters at a maximum rate of 1 per second
for await (const monster of throttle(1, 1000, getPokemon())) {
  await trainMonster(monster)
}

time

function time<T>(config?: ITimeConfig, iterable: AsyncIterable<R>): AsyncIterableIterator<R>
function time<T>(config?: ITimeConfig, iterable: Iterable<R>): IterableIterator<R>

interface ITimeConfig {
    progress?: (delta: [number, number], total: [number, number]) => any;
    total?: (time: [number, number]) => any;
}

Returns a new iterator that yields the data it consumes and calls the progress and total callbacks with the hrtime it took for iterable to provide a value when .next() was called on it. That is to say, the time returned is the time this iterator spent waiting for data, not the time it took to finish being read. The hrtime tuple looks like [seconds, nanoseconds].

import { consume, transform, time } from 'streaming-iterables'
import got from 'got'

const urls = ['https://http.cat/200', 'https://http.cat/201', 'https://http.cat/202']
const download = transform(1000, got)
const timer = time({
  total: total => console.log(`Spent ${total[0]} seconds and ${total[1]}ns downloading cats`),
})
// download all of these at the same time
for await (page of timer(download(urls))) {
  console.log(page)
}

transform

function transform<T, R>(concurrency: number, func: (data: T) => R | Promise<R>, iterable: AnyIterable<T>): AsyncIterableIterator<R>

Map a function or async function over all the values of an iterable. Order is determined by when func resolves. And it will run up to concurrency async func operations at once. If you care about order see parallelMap(). Errors from the source iterable are raised after all transformed values are yielded. Errors from func are raised after all previously transformed values are yielded.

concurrency can be between 1 and Infinity.

import { consume, transform } from 'streaming-iterables'
import got from 'got'

const urls = ['https://http.cat/200', 'https://http.cat/201', 'https://http.cat/202']
const download = transform(1000, got)

// download all of these at the same time
for await (page of download(urls)) {
  console.log(page)
}

writeToStream

function writeToStream(stream: Writable, iterable: AnyIterable<any>): Promise<void>

Writes the iterable to the stream respecting the stream back pressure. Resolves when the iterable is exhausted, rejects if the stream errors during calls to write() or if there are error events during the write.

As it is when working with streams there are a few caveats;

• It is possible for the stream to error after writeToStream() has finished writing due to internal buffering and other concerns, so always handle errors on the stream as well.
• writeToStream() doesn't close the stream like stream.pipe() might. This is done so you can write to the stream multiple times. You can call stream.write(null) or any stream specific end function if you are done with the stream.

import { pipeline, map, writeToStream } from 'streaming-iterables'
import { getPokemon } from 'iterable-pokedex'
import { createWriteStream } from 'fs'

const file = createWriteStream('pokemon.ndjson')
const serialize = map(pokemon => `${JSON.stringify(pokemon)}\n`)
await pipeline(getPokemon, serialize, writeToStream(file))
file.end() // close the stream
// now all the pokemon are written to the file!

Types

Iterableish

type Iterableish<T> = Iterable<T> | Iterator<T> | AsyncIterable<T> | AsyncIterator<T>

Any iterable or iterator.

AnyIterable

type AnyIterable<T> = Iterable<T> | AsyncIterable<T>

Literally any Iterable (async or regular).

FlatMapValue

type FlatMapValue<B> = B | AnyIterable<B> | undefined | null | Promise<B | AnyIterable<B> | undefined | null>

A value, an array of that value, undefined, null or promises for any of them. Used in the flatMap and flatTransform functions as possible return values of the mapping function.

Contributors wanted!

Writing docs and code is a lot of work! Thank you in advance for helping out.