fluture-retry

Toolset for retrying potentially failing computations represented by [Fluture][] Futures.

Usage no npm install needed!

<script type="module">
  import flutureRetry from 'https://cdn.skypack.dev/fluture-retry';
</script>

README

Fluture retry

Toolset for retrying potentially failing computations represented by Fluture Futures.

Usage

Node

$ npm install --save fluture fluture-retry

On Node 12 and up, this module can be loaded directly with import or require. On Node versions below 12, require or the esm-loader can be used.

Deno and Modern Browsers

You can load the EcmaScript module from various content delivery networks:

Old Browsers and Code Pens

There's a UMD file included in the NPM package, also available via jsDelivr: https://cdn.jsdelivr.net/npm/fluture-retry@3.1.0/dist/umd.js

This file adds flutureRetry to the global scope, or use CommonJS/AMD when available.

Usage Example

Let's say we have the following Future Error String that may fail occasionally:

import {Future} from 'fluture';
const task = Future ((rej, res) => {
  const fail = Math.random () > 0.8;
  setTimeout (fail ? rej : res, 100, fail ? new Error ('rej') : 'res');
});

We might simply want to try again when it does fail, a certain amount of times, waiting a certain length of time in between tries.

Basic usage

The retryLinearly export will take a Future and produce a Future which retries the computation five times, at linearly increasing intervals (1 second, 2 seconds, 3 seconds, etc). If all tries fail, the Future rejects with the last encountered rejection reason. So we can simply wrap our task from before, and we get back a Future Error String with increased odds of success.

import {fork} from 'fluture';
import {retryLinearly} from 'fluture-retry';
const retriedTask = retryLinearly (task);
fork (retriedTask) (console.error) (console.log);

Advanced usage

The pre-baked retry strategies may not include exactly what you need. The retry function puts you in control of the following:

  • How much time is in between every try, and how the amount of failures affect the waiting time.
  • How many times a Future is retried.
  • What to do with the accumulated errors.

In the following example, we retry our task 32 times, with an exponentially increasing interval starting at 64ms. It will have retried 32 times after about two and a half minutes, waiting just over a minute at most. At the end we list all unique error messages.

import {fork} from 'fluture';
import {retry, exponentially} from 'fluture-retry';

//    retriedTask :: Future (Array Error) String
const retriedTask = retry (exponentially (64)) (32) (task);

fork (retriedTask) (
  errors => console.error (
    `All tries failed. The following errors were encountered: \n  ${
      Array.from (
        new Set (errors.map (({message}) => message))
      ).join ('\n  ')
    }.`
  )
) (console.log);

API

retry :: (Number -⁠> Number) -⁠> Number -⁠> Future a b -⁠> Future (Array a) b

Create a retrying Future using the given parameters:

  1. A function over the amount of failures to determine waiting time. See exponentially, linearly and statically for pre-baked functions of this sort.
  2. The maximum number of retries before failing.
  3. A Future representing the computation to retry.

See Advanced usage for an example.

exponentially :: Number -⁠> Number -⁠> Number

Takes two numbers and returns the result of multiplying the first by the second raised to the power of two. To be partially applied and used as a first argument to retry.

linearly :: Number -⁠> Number -⁠> Number

Takes two numbers and returns the result of multiplying them. To be partially applied and used as a first argument to retry.

statically :: a -⁠> b -⁠> a

Takes two values and returns the first. To be partially applied and used as a first argument to retry.

linearSeconds :: Number -⁠> Number

Takes a number and multiplies it by 1000.

retryLinearly :: Future a b -⁠> Future a b

A pre-baked retry strategy. See Basic usage.