Asynchronous iterators for JavaScript


AsyncIterator is a lightweight JavaScript implementation of demand-driven object streams,
and an alternative to the two-way flow controlled Node.js Stream.
As opposed to Stream, you cannot _push_ anything into an AsyncIterator;
instead, an iterator _pulls_ things from another iterator.
This eliminates the need for expensive, complex flow control.

Read the full API documentation.

Data streams that only generate what you need

AsyncIterator allows functions to
return multiple _asynchronously_ and _lazily_ created values.
This adds a missing piece to JavaScript,
which natively supports returning a single value synchronously
and asynchronously (through Promise),
but multiple values only synchronously (through Iterable):

	single value	multiple values
synchronous	T getValue()	Iterable<T> getValues()
asynchronous	Promise<T> getValue()	AsyncIterator<T> getValues()

Like Iterable, an AsyncIterator only generates items when you ask it to.
This contrast with patterns such as Observable,
which are data-driven and don't wait for consumers to process items.

$3

An asynchronous iterator is an object that exposes a series of data items by:
- implementing EventEmitter
- returning an item when you call iterator.read (yielding null when none is available at the moment)
- informing when new items might be available through iterator.on('readable', callback)
- informing when no more items will become available through iterator.on('end', callback)
- streaming all of its items when you register through iterator.on('data', callback)

Any object that conforms to the above conditions can be used with the AsyncIterator library
(this includes Node.js Streams).
The AsyncIterator interface additionally exposes
several other methods and properties.

Example: fetching Wikipedia links related to natural numbers

In the example below, we create an iterator of links found on Wikipedia pages for natural numbers.
``JavaScript
import https from 'https';
import { resolve } from 'url';
import { IntegerIterator } from 'asynciterator';

// Iterate over the natural numbers
const numbers = new IntegerIterator({ start: 0, end: Infinity });
// Transform these numbers into Wikipedia URLs
const urls = numbers.map(n => https://en.wikipedia.org/wiki/${n});
// Fetch each corresponding Wikipedia page
const pages = urls.transform((url, done, push) => {
https.get(url, response => {
let page = '';
response.on('data', data => { page += data; });
response.on('end', () => { push(page); done(); });
});
});
// Extract the links from each page
const links = pages.transform((page, done, push) => {
let search = /href="([^"]+)"/g, match;
while (match = search.exec(page))
push(resolve('https://en.wikipedia.org/', match[1]));
done();
});
`

We could display a link every 0.1 seconds:
`JavaScript
setInterval(() => {
const link = links.read();
if (link)
console.log(link);
}, 100);
`

Or we can get the first 30 links and display them:
`JavaScript
links.take(30).on('data', console.log);
`

In both cases, pages from Wikipedia will only be fetched when needed—the data consumer is in control.
This is what makes AsyncIterator lazy.

If we had implemented this using the Observable pattern,
an entire flow of unnecessary pages would be fetched,
because it is controlled by the data publisher instead.

Usage


AsyncIterator implements the EventEmitter interface
and a superset of the Stream interface.

$3

By default, an AsyncIterator is in _on-demand_ mode,
meaning it only generates items when asked to.

The read method returns the next item,
or null when no item is available.

`JavaScript
const numbers = new IntegerIterator({ start: 1, end: 2 });
console.log(numbers.read()); // 1
console.log(numbers.read()); // 2
console.log(numbers.read()); // null
`

If you receive null,
you should wait until the next readable event before reading again.
This event is not a guarantee that an item _will_ be available.

`JavaScript
links.on('readable', () => {
let link;
while (link = links.read())
console.log(link);
});
`

The end event is emitted after you have read the last item from the iterator.

$3


An AsyncIterator can be switched to _flow_ mode by listening to the data event.
In flow mode, iterators generate items as fast as possible.

`JavaScript
const numbers = new IntegerIterator({ start: 1, end: 100 });
numbers.on('data', number => console.log('number', number));
numbers.on('end', () => console.log('all done!'));
`

To switch back to on-demand mode, simply remove all data listeners.

$3

An AsyncIterator can have custom properties assigned to it,
which are preserved when the iterator is cloned.
This is useful to pass around metadata about the iterator.

`JavaScript
const numbers = new IntegerIterator();
numbers.setProperty('rate', 1234);
console.log(numbers.getProperty('rate')); // 1234

const clone = numbers.clone();
console.log(clone.getProperty('rate')); // 1234

numbers.setProperty('rate', 4567);
console.log(clone.getProperty('rate')); // 4567
`

You can also attach a callback
that will be called as soon as the property is set:

`JavaScript
const numbers = new IntegerIterator();
numbers.getProperty('later', console.log);
numbers.setProperty('later', 'value');
// 'value'
`

$3

Due to the syntactical sugar EcmaScript's AsyncIterator
provides, our iterators can also be consumed as such.
If high performance over large iterators is required, this method of consumption not recommended.

`JavaScript
const numbers = new IntegerIterator({ start: 1, end: 100 });

for await (const number of numbers)
console.log('number', number);
console.log('all done!');
``

Error events emitted within the iterator can be caught by wrapping the for-await-block in a try-catch.

In cases where the returned EcmaScript AsyncIterator will not be fully consumed,
it is recommended to manually listen for error events on the main AsyncIterator
to avoid uncaught error messages.

License

The asynciterator library is copyrighted by Ruben Verborgh
and released under the MIT License.