Bluestream is a collection of NodeJS Streams and stream utilities that work well with promises and async functions. Think through2-concurrent with promise support. The goal is to reduce the edge cases when mixing streams and promises. In general, Promises are slower than callbacks, but these streams are a lot more forgiving than node core.

If you don't need streams but want to work with data over time, check out sister project streaming-iterables 🏄‍♂️!

Written in typescript, designed in NYC.

There are no dependencies.

npm install bluestream

import { read, transform, write, pipe } from 'bluestream'
import got from 'got'

// paginate an API
const pokeStream = read(async function () {
  this.offset = this.offset || 0
  const { body: { results } } = await got(`https://pokeapi.co/api/v2/pokemon/?offset=${this.offset}`, { json: true })
  if (results.length > 0) {
    this.offset += results.length
    for (const monster of results) {
      this.push(monster)
    }
  } else {
    return null
  }
})

const fetchMonsterInfo = transform({ concurrent: 2 }, async ({ url }) => {
  const { body } = await got(url, { json: true })
  return body
})

const logStream = write(pokemon => {
  console.log(`<h1>${pokemon.name}</h1><img src="${pokemon.sprites.front_default}">`)
})

await pipe(
  pokeStream,
  fetchMonsterInfo,
  logStream
)
console.log('caught them all')

• read()
• ReadStream()
• transform() (alias map)
• TransformStream()
• write()
• WriteStream()
• filter()
• reduce()
• tap()
• batch()
• wait()
• pipe()
• collect()
• readAsync()
• iterate()
• promise()

([opts:Options,] fn:(bytesWanted) => Promise)) => ReadStream

Creates a read-promise stream which accepts a function that takes the number of bytes or objects of wanted data as arguments and uses this.push or return to push values or promises. This function should return a promise that indicates when the object/chunk is fully processed. Return or push null to end the stream.

Options:

• read - An optional way to pass the read function

• objectMode - true or false

• all other Readable stream options

The other options are also passed to node's Read stream constructor.

A ReadStream works like a normal ReadableStream but the _read and push() methods have some noteworthy differences. (The _read method can be provided as the only argument, in a read key on the options, or as the _read method if you extend ReadStream.) Any returned, non-undefined value will automatically be pushed. Object mode is the default.

_read(bytesWanted)

• Is async function friendly, handles throws/rejects as error events
• Is called again only after it returns or resolves regardless of how many times you call .push
• Is called again if you don't push (to aid in control flow)
• Pushes any non undefined return values

this.push()

• Can be called in a promise, which will be resolved and then pushed as normal
• Returns true or false like a normal stream's push

This allows you to use it in some friendly ways:

// readable stream from an array
const list = [1, 2, 3]
const listStream = bstream.read(() => list.shift() || null)

// readable stream from redis scans
import Redis from 'io-redis'
const redis = new Redis()
let cursor = 0

const hscanStream = bstream.read(async () => {
  const [newCursor, keys] = await redis.scan('cursor', cursor)
  keys.map(key => this.push(key))
  if (newCursor === '0') {
    this.push(null)
  }
  cursor = newCursor
})

transform([opts:Options,] fn:(data[, enc]) => Promise)): TransformStream

map([opts:Options,] fn:(data[, enc]) => Promise)): TransformStream

Creates a transform-promise stream which accepts a function that takes data and encoding as arguments and uses this.push to push values or promises. Any returned, non-undefined value will automatically be pushed. This function should return a promise that indicates when the object/chunk is fully processed.

Options:

• transform - An optional way to pass the transform function

• concurrent - The maximum number of concurrent promises that are allowed. When this limit is reached, the stream will stop processing data and will start buffering incoming objects. Defaults to 1

• highWatermark - the size (in objects) of the buffer mentioned above. When this buffer fills up, the backpressure mechanism will activate. It's passed to node's transform stream.

The other options are also passed to node's Transform stream constructor.

write([opts:Options,] fn:(data[, enc]) => Promise)): WriteStream

new WriteStream(inputOpts: IWritableStreamOptions | writeFunction, fn?: writeFunction): WriteStream

Creates a write-promise stream which accepts a function that takes data and encoding as arguments and returns a promise that indicates when the object/chunk is fully processed.

Options:

• write - An optional way to pass the write function

• writev - Not supported, and passed directly to the underlying Writable stream

• concurrent - The maximum number of concurrent promises that are allowed. When this limit is reached, the stream will stop processing data and will start buffering incoming objects. Defaults to 1

• highWatermark - the size (in objects) of the buffer mentioned above. When this buffer fills up, the backpressure mechanism will activate. It's passed to node's write stream.

The other options are also passed to node's Write stream constructor.

filter([opts:Options,] fn: async (data[, enc]) => boolean): FilterStream

Creates a new FilterStream which accepts a function that takes data and encoding as arguments and returns a boolean to indicate whether the data value should pass to the next stream

Options: Same as transform

reduce([opts:Options,] fn: (acc, data[, enc]) => Promise): ReduceStream

Creates a new ReduceStream which accepts a function that takes the resolved current accumulator, data object, and encoding as arguments and returns the next accumulator or a promise for the next accumulator.

The ReduceStream has a promise() method which returns the final accumulator value

process.stdin.pipe(split()).pipe(es.reduce(function(acc, el) {
    return acc + el;
})).promise().then(function(sum) {

});

tap(opts?: ITransformStreamOptions | ITapFunction, fn?: ITapFunction) => TapStream
new TapStream(opts?: ITransformStreamOptions | ITapFunction, tapFunction?: ITapFunction)

A passthrough stream that intercepts data and lets you process it. Supports async tap functions which will delay processing. Supports concurrent if you need it.

import { pipe, tap, write } from 'bluestream'
import { ghoulGenerator, saveGhoul } from './util'

await pipe(
  ghoulGenerator(),
  tap(console.log),
  write(ghoul => saveGhoul(ghoul))
)
// Ghoul(1)
// Ghoul(2)
// Ghoul(3)
// ... 👻

batch(batchSize: number) => BatchStream
new BatchStream(batchSize: number)

A stream that collects a given number of objects and emits them in an array.

import { batch, pipe, write } from 'bluestream'
import { turkeyGenerator } from './util'

await pipe(
  turkeyGenerator(),
  batch(2),
  write(console.log)
)
// [turkey, turkey]
// [turkey, turkey]
// [turkey, turkey]
// [turkey, turkey]
// ... 🐧🐧

wait(stream: Stream): Promise<any>

Waits for the stream to end. Rejects on errors. If the stream has a .promise() method, it will resolve that value, e.g., from reduce.

pipe(readable: Readable, ...writableStreams: Writable[]): Promise<any>;

Pipes readable to writableStreams and forwards all errors to the resulting promise. The promise resolves when the destination stream ends. If the last writableStream has a .promise() method, it is resolved. If the last stream is a reduce stream the final value is resolved.

Generic Pipe example

import { pipe, read, write } from 'bluestream'
const values = [1, 2, 3, null]
await pipe(
  read(() => values.shift()),
  write(number => console.log(number))
)

Pipe example with reduce

import { pipe, read, reduce } from 'bluestream'
const values = [1, 2, 3, null]
const sum = await pipe(
  read(() => values.shift()),
  reduce((total, value) => total + value, 0)
)
console.log(sum)
// 6

collect(stream: Readable): Promise<null | string | any[] | Buffer>

Returns a Buffer, string or array of all the data events concatenated together. If there are no events, null is returned.

import { collect, read } from 'bluestream'
await collect(fs.readStream('file'))
// <Buffer 59 6f 75 20 61 72 65 20 63 6f 6f 6c 21>
await collect(fs.readStream('file', 'utf8'))
// 'You are cool!'
const values = [1, 2, 3, null]
await collect(read(() => values.shift()))
// [1, 2, 3]
await collect(read(() => null))
// null

readAsync(stream: Readable, count?: number): Promise<any>

Returns a count of bytes in a Buffer, characters in a string, or objects in an array. If no data arrives before the stream ends, null is returned.

iterate(stream: Readable): Readable | AsyncIterableIterator<any>

Returns an async iterator for any stream on node 8+

promise(stream: Readable) => Promise(any)

All bluestream streams implement a promise method that returns a promise that is fulfilled at the end of the stream or rejected if any errors are emitted by the stream.

For ReduceStreams, the promise is for the final reduction result. Any stream errors or exceptions encountered while reducing will result in a rejection of the promise.

const { pipe, map, tap, reduce } = require('bluestream')
const { Nodes } = require('./util')

let count = 0
const stats = await pipe(
  Nodes.scan({ fields: true }),
  map(generateStats),
  tap(() => count++),
  reduce(mergeGraphStats, {})
)
console.log({ count, stats })

Made by the loving team at @bustle and maybe you?