123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242
/* eslint no-console: 0, global-require: 0 */'use strict';var Benchmark = require('benchmark');var markdownTable = require('markdown-table');global.ss = require('simple-statistics');global.ssMaster = require('../');global.science = require('science');global.jStat = require('jstat').jStat;global.mathjs = require('mathjs');global.d3 = require('d3-array');
// We have a very simple way to benchmark single-array-input// methods, but jStat is a little more complicated - you specify// if you want _sample_ variance by adding a boolean argument to the// method. So I add it here.global.jStat.sampleVariance = function(input) {    return global.jStat.variance(input, true);};
function setupInput() {    var i;    var input10 = [];    for (i = 0; i < 10; i++) {        input10.push(Math.random());    }    var input100 = [];    for (i = 0; i < 100; i++) {        input100.push(Math.random());    }    var input1000 = [];    for (i = 0; i < 1000; i++) {        input1000.push(Math.random());    }
    var input10a = [];    for (i = 0; i < 10; i++) {        input10a.push(Math.random());    }    var input100a = [];    for (i = 0; i < 100; i++) {        input100a.push(Math.random());    }    var input1000a = [];    for (i = 0; i < 1000; i++) {        input1000a.push(Math.random());    }}
// This is the start of what will eventually be a markdown table// included in the readme.var results = [[    '', 'ss', 'science', 'jStat', 'mathjs', 'd3', 'ssMaster']];
// This is a generic 'suite maker' method that we use to test different// libraries doing the same work - it gets a name as a first argument, like// 'median', and the second argument is an array of strings that are the names// of the implementations of that function in different libraries.function suiteForMethods(name, methods, type) {
    var suite = new Benchmark.Suite({        name: name,        onError: function(error) {            console.log(error);        }    });
    methods.forEach(function (method) {        if (type === 'two-arrays') {            suite.add(method, {                setup: setupInput,                fn: method + '(input10, input10a);' + method + '(input100, input100a);' +                    method + '(input1000, input1000a);'            });        } else {            suite.add(method, {                setup: setupInput,                fn: method + '(input10);' + method + '(input100);' + method + '(input1000);'            });        }    });
    suite        .on('cycle', function(event) {            console.error(String(event.target));        })        .on('complete', function() {        var row = [name];        for (var i = 1; i < results[0].length; i++) {            row[i] = '';        }        var fastest = this.filter('fastest')[0];        this.forEach(function (event) {            // This gross code figures out which column to put            // each result in, because we don't have implementations            // of every method in every library.            var idx = results[0].indexOf(event.name.split('.')[1]);            row[idx] = Math.round(event.hz).toLocaleString();            if (event === fastest) {                row[idx] = '**' + row[idx] + '**';            }        });        results.push(row);    })    .run();}
// Here we declare all of the basic operations that we'll test, and the// functions in each library that do them.var tests = [    [        'variance',        [            'global.ss.variance',            'global.ssMaster.variance',            'global.jStat.variance'        ]    ],    [        'sampleVariance',        [            'global.ss.sampleVariance',            'global.ssMaster.sampleVariance',            'global.science.stats.variance',            'global.jStat.sampleVariance',            'global.d3.variance'        ]    ],    [        'covariance',        [            'global.ss.sampleCovariance',            'global.ssMaster.sampleCovariance',            'global.jStat.covariance'        ],        'two-arrays'    ],    [        'skewness',        [            'global.ss.sampleSkewness',            'global.ssMaster.sampleSkewness',            'global.jStat.skewness'        ]    ],    [        'geometricMean',        [            'global.ss.geometricMean',            'global.ssMaster.geometricMean',            'global.jStat.geomean'        ]    ],    [        'medianAbsoluteDeviation',        [            'global.ss.medianAbsoluteDeviation',            'global.ssMaster.medianAbsoluteDeviation',            'global.mathjs.mad',            'global.jStat.meddev'        ]    ],    [        'median',        [            'global.ss.median',            'global.ssMaster.median',            'global.mathjs.median',            'global.jStat.median',            'global.science.stats.median',            'global.d3.median'        ]    ],    [        'mode',        [            'global.ss.mode',            'global.ssMaster.mode',            'global.mathjs.mode',            'global.jStat.mode',            'global.science.stats.mode'        ]    ],    [        'mean',        [            'global.ss.mean',            'global.ssMaster.mean',            'global.mathjs.mean',            'global.jStat.mean',            'global.science.stats.mode'        ]    ],    [        'min',        [            'global.ss.min',            'global.ssMaster.min',            'global.mathjs.min',            'global.jStat.min',            'global.d3.min'        ]    ],    [        'max',        [            'global.ss.max',            'global.ssMaster.max',            'global.mathjs.max',            'global.jStat.max',            'global.d3.max'        ]    ],    [        'sum',        [            'global.ss.sum',            'global.ssMaster.sum',            'global.mathjs.sum',            'global.jStat.sum',            'global.d3.sum'        ]    ]];
tests.forEach(function (test) {    suiteForMethods(test[0], test[1], test[2]);});
// Finally, console.log spits out the content of readme.mdconsole.log('# Benchmarks');console.log('\n\nThis document is generated by `npm start`. It compares simple-statistics performance against similar libraries in JavaScript.');console.log('\n\n' + markdownTable(results));console.log('\n\nFastest libraries are in bold. Numbers are in operations per second.');console.log('\n\nssMaster is simple-statistics master branch, used for testing and proofing - it is not guaranteed to be any released version.');
console.log('\n\n### Versions\n');
['simple-statistics', 'science', 'jstat', 'mathjs', 'd3-array'].forEach(function (p) {    console.log('* ' + p + ': ' + require(p + '/package').version);});