wait-in-parallel

For Javascript developers who have accepted async/await and TypeScript into our hearts, our lives have gotten so much better but there are still some edge cases where you need to be more careful with asynchronous behavior or just where you'd like to have a concise way of expressing parallel execution that's easy to grok.

This is an attempt at that.

Maybe you think you don't need this help. Maybe you don't. But let's be honest, the async/await semantics are great when we're doing one thing after another ... not quite as good we're doing things in parallel. Yes you can do:

await Promise.all([thing1, thing2, thing3]);
await Promise.all([anotherThing1, anotherThing2]);

What about the looping trap? Eh, what trap? I gave it a name to make it sound more formal and scary. Basically what I mean is when you are looping around anything and one of the steps within the loop is asynchronous:

things.maps(async thing => await doSomething(thing));
doSomethingElse();

Without really intending to i've fired off a set of parallel executions and then immediately executed doSomethingElse. Probably not what you wanted. What if I could instead:

things.maps(thing => inParallel.add(doSomething(thing)));
await inParallel.isDone();
doSomethingElse();

Yeah well that's what I've been saying all along. You need this. Treat yourself. Show yourself that you really do care.

Installation

# yarn
yarn add wait-in-parallel
# npm
yarn install --save wait-in-parallel

Basic Usage

This package exposes a Parallel Object which exposed as the default export of wait-in-parallel:

import Parallel from "wait-in-parallel";
const inParallel = new Parallel();

Adding Parallel executions

inParallel.add("red", addThing("red"));
inParallel.add("blue", addThing("blue"));

Waiting for completion

Waiting is very basic ... use the await keyword off of the isDone property ... and if you care about the returned promises then they will be returned with the resolution the await:

const results = await inParallel.isDone();

So in this case imagine the addThing function resolves eventually with a random number, you'd get a results variable that would look like:

{
  red: 42,
  blue: 99
}

Imagine, however, that the "blue" promise failed. Then instead of completing the await block would error. In this example we didn't have it surrounded by a try/catch block but had we then what we could have caught would look like:

{
  name: "ParallelError",
  message: " 1 of 2 promises failed; failing messages were: [ blue ] ",
  failed: [ "blue" ],
  successful: [ "red" ],
  incomplete: [],
  errors: { blue: "TypeError: the blue type could't be found" },
  stacks: { blue: [ ... ] },
  // from promises which DID succeed
  results: {
    red: 42
  }
}

Advanced Usage

Similar "things" in Parallel

The initial example above was a bit non-descript but imagine that blue and red represent two different data structures or "things". It is quite common for you to want to evalute unlike things in parallel but it is equally common to want to run a bunch of like things together in parallel. In this latter scenario there is nothing preventing us from using exactly the same approach as above but in many cases there are some optimizations you will want to take advantage of.

First off is stronger typing of your results. The Parallel object is a "generic" and can be typecast to indicate what your parallel tasks will return. So for instance, imagine we have a Person interface we are expecting to come back from all our individual tasks we would then initialize the object instance like so:

const p = new Parallel<Person>();
// where people is an array of "person IDs" 
// and getPerson async gets them from DB
Object
  .keys(people)
  .map(person => p.add(person, this.getPerson(person)));
const results = p.isDone();

Great, so now all of our person objects have come back and they are strongly typed to being Person objects. And I can reference them by the TaskId I used when adding:

const name = results["id1234"].name;

Great, but often we want to do some array processing on our results and that we received a dictionary (which is great for quick lookups) is less convenient than just an array. No problem, you can have an array if that's what you want with an alternative to isDone called isDoneAsArray:

const results = p.isDoneAsArray();
const names = results.map(p => p.name);

Pretty nice, right? Ok now you may have noticed that this approach just returns an array of Person objects but the TaskId we passed in with add() is not there. True. So should we just NOT have to have a TaskId? Well typically the answer is "no" because that TaskId is super useful when you have errors. Also, not having it in this example is not "lossy" because the Person object almost surely has a person.id attribute which serves as an identity property. That said, there may be situations where you do want to preserve the TaskId so in these situations you can do the following:

type PersonWithTaskId = Person & { taskId: string };
const p = new Parallel<PersonWithTaskId>();
Object
  .keys(people)
  .map(person => p.add(person, this.getPerson(person)));
const results = p.isDoneAsArray('taskId');

Whatever string property you put in as a parameter to isDoneAsArray(prop: string) will be added to the resulting array of hashes.

Fail Fast

As you probably noticed in the "basic" use case the strategy assumed in running in parallel is to run each promise to completion regardless if there is a failure. That may be what be what you want but there are certainly cases where you'd prefer to "fail fast" at the point where the first promise to fail represents failure.

This is possible, and is achieved by calling failFast() the fluent interface:

try {
  const result = await inParallel
    .create()
    .failFast()
    .add("red", addThing("red"))
    .add("blue", addThing("blue"))
    .isDone();
  // success
} catch (e) {
  // failure
}

Now the await will fail as soon as our call to addThing("blue") fails regardless of whether "red" has completed or not. If it hasn't completed then you'll see it in the incomplete array from above.

Fail Slow ... but let me know

It may be that you do really want to wait for all the promises to complete but you'd like to be alerted as the failures come in so you can handle that in an appropriate manner. No problem, this is what the notifyOnFailure(fn) feature is for:

try {
  const fn: IParallelFailureNotification = (which: string, error: Error) => {
    console.log(`The ${which} promise failed with error: ${error.message}`);
  };
  const result = await Parallel.create()
    .add("red", addThing("red"))
    .add("blue", addThing("blue"))
    .notifyOnFailure(fn)
    .isDone();
  // success
} catch (e) {
  // failure
}

Slow ... but I don't have ALL DAY

There are often cases where you need/want to wait for completion of all promises to complete but only to a point. In essence you want a "timeout" to fire on any of the promises if they're go beyond a certain timeframe. Well clearly you can just add this timeout functionality to your promises but to make this functionality easier to implement there is an optional third parameter to the .add() method:

try {
  const result = await Parallel.create()
    .add("red", addThing("red"), 5000))   // timeout of 5 seconds
    .add("blue", addThing("blue"), 2000)) // timeout of 2 seconds
    .isDone();
  // success
} catch (e) {
  // failure
}

Delayed Start

In the cases demonstrated so far every addition (aka, call to add) has passed in a promise which is executing. In this always-hurring world, "right away" makes sense a lot of the time but occationally it might make sense to delay execution of the promises. This can be done by passing in a function which returns a promise:

// using an async function
const aFn = async () => "an async function";
// an old fashioned promise
const p = () => Promise.resolve("I'm a promise");

This allows you to allows you to load up a number of parallel execution groups but they don't start executing until the call to .isDone() is called:

import Parallel from "wait-in-parallel";

const group1 = Parallel.create()
  .add(() => job1)
  .add(() => job2);
const group2 = Parallel.create()
  .add(() => job3)
  .add(() => job4);

setTimeout(console.log(await group2.isDone()), 500);

In the above example, nothing is executed for 500ms, then group2 is kicked off and when it completes it logs to the console. job1 and job2 are never executed as group1 is still delayed ... awaiting the start signal.

Other Notes

• this library is available in commonjs, es2015, and umd module definitions
• you don't need to use TypeScript, I just think you should

License

Copyright (c) 2018 LifeGadget Ltd

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.