perfect

Run simple performance tests.

npm install perfect --save

With perfect you can run simple performance tests to see which variations of a piece of code (f.e. different algos to achieve the same thing) are the fastest.

Requirements

• The source is not transpiled for older browsers (yet) so you may have to transpile this in your own build setup.

Basic Usage

See the examples.

The API looks like this:

import { performanceTest } from 'perfect'
 
performanceTest({
 
    // describe the test (similar to describe() in other testing frameworks)
    description: 'compare sorting algorithms',
 
    iterations: 100000, // optional, default is 500000
 
    testCases: {
 
        // define a test case. The property key is the name of the test
        // (similar to it() in other test frameworks)
        'merge sort': {
 
            // setup anything you need for the test on `this`, which persists for
            // every iteration of the test. Called only once, before all iterations
            setup() {},
 
            // called before each iteration
            before() {},
 
            // The code to test. Called for each iteration. Receives start and end
            // markers. Wrap the specific part of the functionality you are testing
            // with start() and end() calls.
            test(start, end) {},
 
            // called after each iteration
            after() {},
 
            // called only once, after all iterations
            cleanup() {},
        },
 
        // define another test case
        'quick sort': {
            // ...
        },
 
        // ...define any number of test cases to compare performance of...
    },
)

The performanceTest function will start a performance test right away (synchronously). Async tests are not supported yet (they will be good for testing async functionality like worker responses or network conditions).

Each test in testCases will be executed as many times as specified by iterations, which defaults to 500000.

The test function of each test case is what the performance test is interested in timing. Inside a test function, call start() and end() to specify exactly which part of the test function we care about timing. If start() and end() are not called, then the performance test will time the entire execution of the test function. Calling start() or end() are completely optional, but you may like to use them to get best results. If you omit start(), then the timer will start from the beginning of the execution of the test function. Likewise, if you omit end(), then timing ends after the test has returned.

The setup() function is called once, at the beginning of a testCase, before any iterations call the test function. In all functions, this is a context that you can use throughout all invocations of any functions. Use this in setup() to set up anything that you may across all iterations of the test function.

before and after are called during each iteration, before and after the test function is called. This useful for resetting state, incrementing variables, etc.

Finally, cleanup is called once after interation has finished. Use this to cleanup anything you no longer need, for example in case your tests are part of a long running application and you don't want things to be left over in memory. Plus, it can be nice to reset the playing field for each test case, if you need to.