Re

Do it again, if it doesn't work the first time. Supports various configurable retry strategies, including: constant, exponential backoff and linear backoff.

Functions are styled to match the simplicity and ease of use found in the async library.

Install

npm install re

Quick Example

1var Re = require('re'),
2    re = new Re();
3
4re.try(repeatMe, doMeAtTheEnd);
5
6var repeatMe = function(retryCount, done){
7    if(retryCount < 2) done(new Error("Not there yet!"));
8    else done(null, retryCount);
9};
10
11var doMeAtTheEnd = function(err, retryCount){
12    console.log("It took this many tries: " + retryCount);
13};

In the Browser

Tested in recent versions of Internet Explorer, Firefox and Chrome. Usage:

1<script type="text/javascript" src="re.js"></script>
2<script type="text/javascript">
3  var re = new Re();
4
5  re.try(repeatMe, doMeAtTheEnd);
6      
7  // repeatMe and doMeAtTheEnd are exactly as above
8
9</script>

Try it in your browser with this test: test/test.html or play with the test in this fiddle: re-fiddle (these pages don't work in IE, because it's recently gone from lax to pedantic).

Usage

If you like the defaults, call it like this:

1var Re = require('re'),
2    re = new Re();
3
4re.try(function(retryCount, done){
5    if(retryCount < 2) done(new Error("Not there yet!"));
6    else done(null, retryCount);
7  },
8  function(err, retryCount){
9    console.log("It took this many retries: " + retryCount);
10});

The re.try function takes two arguments, a function to call until it works (or we run out of retries) and a function to call when it finally succeeds (or we fail too many times). As the name suggests we automatically wrap your function in a standard try block and, if an exception occurs, call it again according to the retry schedule.

This first function passed to re.try should take 2 arguments like this:

1function operation(retryCount, done)

The retryCount argument is the number of the current retry. It'll be zero the first time and get bigger every time.

The done argument is a function to call when you've completed your operation. If you encounter an error condition, pass in the err object as the first argument. If you don't encounter an error, pass in a falsy first argument (null works). If you give us a falsy error and no exception happens, we call your callback with all the arguments passed into this function.

The second function passed to re.try can take as many arguments as you like but should always start with an error parameter. This will be falsy, if no error happens.

The re.do function is like re.try except it doesn't wrap your operation in a try...catch.

Options

The default options look like this:

1var options = {
2    retries : 10,
3    strategy : {
4      "type": Re.STRATEGIES.EXPONENTIAL,
5      "initial":100,
6      "base":2
7    }
8}

You pass this options object into the Re constructor.

1var Re = require('re'),
2    re = new Re(options);

This gives you 10 retries and an exponential backoff strategy with the following progression (in milliseconds): 100, 200, 400, 800, 1600, 3200, 6400, 12800, 25600, 51200

Retry Strategy Examples

The following will retry every 400 milliseconds:

1{"type": Re.STRATEGIES.CONSTANT, "initial": 400}

The following will give a linear backoff strategy that has the following progression (when paired with retries: 10) : 200, 400, 600, 800, 1000, 1200, 1400, 1600, 1800, 1800

1{"type": Re.STRATEGIES.LINEAR, "initial": 200, "max":1800}

Both progressive strategies accept the max option. All strategies also accept a rand option. This is a Boolean that adds a random multiplier between 1 and 2. This makes them act like the tradition backoff function. This option is set to false by default.

Stability

Test coverage is good and expanding. We use mocha.

Technical Details

The traditional exponential backoff function is described here: Exponential Backoff in Distributed Systems. This is equivalent to our exponential backoff function with the rand option set to true.

Our formula for exponential backoff looks something like this, when using all the options:

1return Math.min(random * initial * Math.pow(base, retry), max);

Where random is a random number in the half-open interval [1, 2). When randomness is turned off, the value of this variable is always 1.

If you don't specify the max option, the formula looks like this:

1return random * initial * Math.pow(base, retry);

I'm shamelessly stealing the following link from node-retry just because it's fun for nerdy math people to play with. You can use it to calculate the exact value you need for the base option so that all retry intervals sum to a desired amount: Wolfram Alpha.