This package has been extracted from the original fbjs library and it exposes single fetchWithRetries. This fetch solves two common problems:

fetch timeouts, and
retries

Installation

yarn add @mrtnzlml/fetch

Usage

import fetchWithRetries from '@mrtnzlml/fetch';

fetchWithRetries('//localhost', {
  // see: https://github.com/github/fetch/
  // ...

  fetchTimeout: 15000,
  retryDelays: [1000, 3000],
})
  .then(response => console.warn(response))
  .catch(error => console.error(error));

It does two things:

it will timeout the connection after 15 seconds, and
it will retry after 1 second and then after 3 seconds

Retries are performed in these situations:

original fetch request failed
fetch returned HTTP status <200 or >=300 (but not 401 or 403 since these errors are not transitive)
when the timeout (fetchTimeout) occurs

This package uses fetch ponyfill internally (cross-fetch) so it supports server JS as well as browsers.

Error handling

You have to catch errors while fetching the response. This fetch throws these exceptions:

standard fetch exception (Error) when request failed for some reason
TimeoutError when fetch fails because of defined timeout
ResponseError when final response returns HTTP status <200 or >=300

Example:

import fetchWithRetries, {TimeoutError, ResponseError} from '@mrtnzlml/fetch';

(async () => {
  try {
    const response = await fetchWithRetries('//localhost');
    // TODO: do something with the response
  } catch (error) {
    if (error instanceof TimeoutError) {
      console.error('request timeouted');
    } else if (error instanceof ResponseError) {
      console.error('unsuccessful response', error.response);
    } else {
      console.error('unknown error');
    }
  }
})();

FAQ

How does the timing work in this case?

const config = {
  fetchTimeout: 2000,
  retryDelays: [100, 3000],
};

There are many situations that may occur (skipping happy path):

request failed with HTTP status code 500
retries after 100ms
request successful (we are done), OR fail again
retries after 3000ms again for the last time

Example with timeouts:

first request takes too long and it's terminated after 2000ms
next retry was scheduled to be after 100ms but we already burned 2000ms so it's going to be executed immediately
second request takes to long as well and is terminated after 2000ms
last request will wait 3000ms minus the burned timeout => 1000ms
last attempt (it will timeout or resolve correctly)

Current Tags

2.0.0 ... latest (2 years ago)

Installation

Usage

Error handling

FAQ

How does the timing work in this case?

Current Tags

11 Versions