basic-performance-testing.md

Basic Performance Testing

This guide will follow a worked example on how to go about comparatively performance testing functions. Performance testing is not a subject that should worry beginners. It is very interesting to look at and will become useful on complex projects.

The following is almost certainly not the only (or necessarily the best) way, however it is fairly easy to understand and use.

The Functions

We are tasked with performance testing two different ways of generating random strings 50 characters in length (see random test cases guide).

Our original function took the following form:

function makestring() {
    var text = "";
    var possible = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789";

    for( var i=0; i < 50; i++ ) {
        text += possible.charAt(Math.floor(Math.random() * possible.length));
    }
    return text;
}

This received feedback that it is extremely inefficient to sum n strings as it generates n-1 intermediate objects; it is instead always better to use an array and finally perform a .join("") over them. So we created the following:

function makestring2() {
   var array = [];
   var possible = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789";
   for( var i=0; i < 50; i++ ) {
       array.push(possible.charAt(Math.floor(Math.random() * possible.length)));
   }
   return array.join("");
}

Performance Test

Now, rather than just accepting the advice, how do we actually test these?

Our chosen method is to make use of JavaScript's new Date() constructor. This gives the time since 01 January 1970 00:00:00 UTC. The number is specified in milliseconds. Therefore, if we wrap a for loop repeating the function between two variables that are assigned new Date() and find the difference between these variables we should have the time taken to perform the function. To get a representative measure of performance, consider repeating the function many times over. Here we have used 20,000 times.

var start = new Date();  // log start timestamp
for (var i = 0; i < 20000; i++) {
    console.log(makestring());
}
var end = new Date();  // log end timestamp
var diff = end - start;
console.log(diff.toString());

and for the second function:

var start1 = new Date();  // log start timestamp
for (var i = 0; i < 20000; i++) {
    console.log(makestring2());
}
var end1 = new Date();  // log end timestamp
var diff1 = end1 - start1;
console.log(diff1.toString());

The Comparison

By comparing the time taken to perform each function we are able to get a rough idea of performance. In this example, our makestring() took approximately 2322 ms and our array function makestring1() took approximately 323 ms.

Note: Values may not be the same between seperate runs, but will be in a similar range. Likewise consider moving console.log(diff.toString()) to the end of the function for ease of reading.

Moving beyond this example, the same method should be applicable to any other set of functions. Wrap a loop repeating the test function(s) between date variables, find the difference, compare!

Appendix

Complete Script

function makestring() {
    var text = "";
    var possible = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789";

    for( var i = 0; i < 50; i++ ) {
        text += possible.charAt(Math.floor(Math.random() * possible.length));
    }
    return text;
}

function makestring2() {
   var array = [];
   var possible = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789";

   for( var i = 0; i < 50; i++ ) {
       array.push(possible.charAt(Math.floor(Math.random() * possible.length)));
   }
   return array.join("");
}

var start = new Date();  // log start timestamp
for (var i = 0; i < 20000; i++) {
    console.log(makestring());
}
var end =  new Date();  // log end timestamp
var diff = end - start;

var start = new Date();  // log start timestamp
for (var i = 0; i < 20000; i++) {
    console.log(makestring2());
}
var end = new Date();  // log end timestamp
var diff1 = end - start;

console.log(diff.toString());
console.log(diff1.toString());