attempt

Automatically retry functions that fail, in crazily customizable ways.

npm install attempt
11 downloads in the last day
559 downloads in the last week
6 030 downloads in the last month

Attempt

Retries functions that throw or call back with an error, in crazily customizable ways.

Build Status

Installation

In your project folder, type:

npm install attempt

Usage

Whenever you have code that looks like this:

function flakyApiCall(function(err, result) {
    if (err)
        console.log('Flaky API died AGAIN!', err);
    else
        doSomething(result);
});

Just drop that code in an attempt!

var attempt = require('attempt');
attempt(
    function() {
        flakyApiCall(this);
    },
    function(err, result) {
        if (err)
            console.log('Flaky API failed 3 times.', err);
        else
            doSomething(result);
    });

Details

Attempt will re-run your attempted function if it throws an error or if it calls back with a non-empty first argument (following the first-arg-is-an-error standard Node.js convention). The function call looks like this:

attempt([options], tryFunc, [callback])

tryFunc is called with one argument: attempts. It is the number of times the tryFunc has been run before.

attempt(function(attempts) {
    if (attempts)
        console.log('This is retry #' + attempts);
    else
        console.log('This is the first attempt!');
});

Options

The following options are set per request:

retries

Default: 2. The number of times to retry the tryFunc before giving up and sending the error to the callback.

attempt(
    { retries: 15 },
    function() {
        flakyApiCall(this);
    },
    function(err, result) {
        if (err)
            console.log('Failed 16 times.', err);
        else
            doSomething(result);
    });

interval

Default: 0. The number of milliseconds to wait between attempts.

attempt(
    { interval: 5000 },
    function() {
        flakyApiCall(this);
    },
    function(err, result) {
        if (err)
            console.log('5 retries * 5 seconds = 25 seconds of failure.', err);
        else
            doSomething(result);
    });

factor

Default: 1. The factor by which the interval should be multiplied per attempt. If set to 2 with an interval of 5, the first retry will execute after 5 seconds, the second after 10, the third after 20, and so on.

This allows an exponential retry scheme. For a smaller gap between retries, floats like 1.2 can be used to grow the interval at a slower rate.

onError

Default: null. Function to call when the tryFunc fails with an error. The first argument is the error.

attempt(
    {
        onError: function(err) {
            console.log(err);
        }
    },
    function() { flakyApiCall(this); },
    function(err, result) { /* ... */ });

The second argument, if it exists, is a callback function that will need to be called in order for the next attempt to continue. This is useful if you have to do something asynchronous to fix the error.

attempt(
    {
        onError: function(err, done) {
            log.write(err, done);
        }
    },
    function() { flakyApiCall(this); },
    function(err, result) { /* ... */ });

By default, calling done() will ignore the retry interval. If you still want it to be observed, call done(true).

max

Default: Infinity. The maximum number of milliseconds to wait before retrying. If the interval or factor causes a wait time larger than 'max', 'max' will be used.

random

Default: 0. Increase the wait interval by a random factor. Generally, this should be a number between 0 (no randomness) and 1 (wait time could double). For example, if .5 is used, the interval could be anywhere between 100% and 150% of its original calculated time.

attempts

Default: 0. The number of attempts to fake Attempt into believing were already completed. This is mostly used by Attempt internally, but can be useful for hacking interval times.

License

Attempt is distributed under the MIT license.

Credits

Attempt was created by Tom Frost in 2012. Because webservice APIs be flaky, man.

npm loves you