Perf Metrics

Get performance metrics of js classes/functions with simple metrics reports.

How to use

First you will need to install the lib to get some metrics:

  npm install @perfmetrics/lib

Next just call the lib from your code:

const perfMetrics = require("@perfmetrics/lib");

The perfMetrics is a function that takes two arguments:

• target: class, function, object, anything that can be used with proxy.
• log, a callback (by default = console.log) with the arguments: (name, start, end, duration) => ...

  • name: the name of function, or ClassName.NameFunc

  • start: start time (higth resolution time in milliseconds)

  • end: end time (higth resolution time in milliseconds)

  • duraction: time that takes to execute the function (higth resolution time in milliseconds)

Now lets track a function:

const perfMetrics = require("@perfmetrics/lib");

function quadraticTime(n) { // n^2
    let r = [];
    for (let i = 0; i < n; i++) {
        for (let j = 0; j < n; j++) {
            r.push([i, j]);
        }
    }
}

const quadraticTimePerf = perfMetrics(quadraticTime);

// Now when you call quadraticTimePerf will track all the execution time
// In this case it will be logged to console.log

for (let i = 0; i < 100; i++) {
  quadraticTimePerf(1000 + i);
}

Its easy to implement a new logger, but we can use the filestats logger, just install

npm install @perfmetrics/filestats

Then we can change the code to use the logger like this:

const perfMetrics = require("@perfmetrics/lib");

function quadraticTime(n) { // n^2
    let r = [];
    for (let i = 0; i < n; i++) {
        for (let j = 0; j < n; j++) {
            r.push([i, j]);
        }
    }
}

// Output, save file every 1 second
const fsLogger = fileStats("./stats.log", 1000);
const quadraticTimePerf = perfMetrics(quadraticTime, fsLogger);

// Now when you call quadraticTimePerf will track all the execution time
// In this case it will be logged to console.log

for (let i = 0; i < 100; i++) {
  quadraticTimePerf(1000 + i);
}

Afther running the code we will get a stats.log file. To generate a report we first install package:

npm install @perfmetrics/cli --global

And finally just run:

perfmetrics -s stats.log --html stats

This will generate a stats.html report like this one: https://fsvieira.github.io/perfmetrics/

To get more information on how to use the cli, just run

perfmetrics -h

For more information check out the examples on the examples folder of this repo.

How it works ?

Using a proxy perfMetrics will try to keep track of the execution time of all functions generated from the proxy funcion/class, the time statistics and function information is logged using a callback function on perfMetrics, currently there in this package there is only a file logger.

Using reporter like node reports/html.js stats.log stats will create a stats.html report.

Complexity calculation

The complexity is calculated using regressions functions (https://github.com/Tom-Alexander/regression-js) on the collected data. It goes like this:

• Order the duraction data ascending,
• Run multiple methods and choose the 2 (if they exist) with the best r-square score.

This means that more data will give better results. It also means that a function may not correspond to code classification but that can be a good thing, why spending time otimizing functions that never reach the worst case in real cenario usage.

Notes

This is still a work in progress, but alredy a working beta.

What I would like to do in the future:

• Add and improve loggers: using webworker compatible api, rest-api.
• Add more reports, including a live webapp using a rest-api server.
• Compare two reports of same funcitions so that we can understand if there is improvements.