Provides Promise.all and Promise.allSettled variants that limit the number of simultaneously pending promises. This is useful e.g. when you need to make thousands of requests to a single server but do not want to hit it with all requests at once. Instead you might want to have at most 10 requests pending simultaneously and whenever one settles, the next one is initiated automatically.

This is a ES2019 CommonJS module, that works exactly the same way in CommonJS and ES modules (because there's only an unnamed default export). TypeScript typings are provided.

Installation

npm install capped-promise

Example

// Only necessary in node, fetch is natively available in a browser.
import fetch from "node-fetch";
import CappedPromise from "capped-promise";

const getText = async (url: string) => {
    const response = await fetch(url);

    if (!response.ok) {
        throw new Error("Unexpected response");
    }

    return await response.text();
};

const cssUrls = [
    "https://fonts.googleapis.com/css2?family=Roboto",
    "https://fonts.googleapis.com/css2?family=Open+Sans",
    "https://fonts.googleapis.com/css2?family=Poppins",
];

// Promise.all
// We pass already pending promises.
const cssPromises = cssUrls.map((url) => getText(url));
// All requests are made at the same time.
const promiseResults = await Promise.all(cssPromises);

// CappedPromise.all
// We pass promise-creating functions instead of promises.
const createCssPromises = cssUrls.map((url) => () => getText(url));
// We allow at most 2 simultaneous requests. So, in the beginning the first
// two promises are created right away. As soon as one of those is fulfilled,
// the third is automatically created. When the responses for all requests are
// in, the returned promise fulfills.
const cappedResults = await CappedPromise.all(2, createCssPromises);

Interface

Promise.all and Promise.allSettled accept iterables of awaitables (usually promises). By its very nature, once a promise has been created, it is in the state pending. That is, the underlying operation is already running. So, if you create e.g. 1000 Promise objects by calling fetch() repeatedly (without awaiting them), all those requests will be made quasi simultaneously, if you pass them to Promise.all and then await the result. This is why CappedPromise.all and CappedPromise.allSettled differ from their Promise counterparts as follows:

• Instead of awaitables, they accept parameterless functions that are expected to create and return awaitables. The CappedPromise implementation will then call these functions to create new awaitables as necessary.
• The first parameter is maxPending, which specifies how many promises can at most be pending simultaneously.

Otherwise, CappedPromise.all and CappedPromise.allSettled follow their Promise counterparts.

Behavior

Again, the CappedPromise behavior follows the Promise behavior as much as possible. However, due to the different interface there is the following additional concern: If a function passed to a CappedPromise method throws right away, e.g. () => { throw new Error("Oops!"); } then this is treated as an unintended catastrophic failure and is propagated as a rejection immediately, even if the function was passed to allSettled. If this is not what you expect, you should return a rejecting Promise, as follows: () => Promise.reject(new Error("Oops!")).