Join 10350+ others. No spamming.
I promise!

Follow us at github.



itchyny/sjsp

222

itchyny / sjsp

Haskell

Simple JavaScript Profiler


READ ME

sjsp - Simple JavaScript Profiler

Why?

This is a JavaScript profiler, injecting profiling codes into your JavaScript files.

Applications written in JavaScript are getting larger these days and more complicated. There are many JavaScript Frameworks and they sometimes make us difficult to profile our applications. The default profilers of the Web Browsers get useless as our applications become huge and the frameworks do tricky things. The profiler sometimes lists many many functions of the frameworks.

Let's get back to what we really need.

We want to profile the code we write.

In a really simple way.

function test() {
  var start_time = Date.now(); // grab the current time at the top

  // our code

  log_profile("test", Date.now() - start_time); // grab the current time again and log the time the function consumed.
}

Do we have to write the profiling codes in all the functions by hand?

Of course not.

So, here comes sjsp, a tool for injecting profiling codes into JavaScript files.

Installation

There are two ways to build sjsp; cabal and stack. If you do not have Haskell environment yet, stack is a simpler way.

cabal

 $ git clone https://github.com/itchyny/sjsp
 $ cd sjsp
 $ sudo cabal install

When you failed to install with the error The program happy version >=x.xx is required but it could not be found., add the path to the environment variable.

export PATH=$PATH:$HOME/.cabal/bin

stack

Prepare stack command to download it from here and then run the following script:

 $ git clone https://github.com/itchyny/sjsp
 $ cd sjsp
 $ stack install
export PATH=$PATH:$HOME/.local/bin

Usage

 $ sjsp test.js            # generates test.sjsp.js
  1. Use sjsp command on the JavaScript files you want to profile. The command sjsp does not break your test.js, but creates a new file.
  2. Use the generated test.sjsp.js instead of test.js
  3. Open the page with your favorite browser and look into the JavaScript console. The profiling result will be reported every 10 seconds.

The profiling result will look like the following.

========== SORT BY TIME ==========
time: 30.202sec   count:  71      test6  test.js  (line: 31, col: 18)  function test6() {
time: 16.474sec   count:  41      test7  test.js  (line: 37, col: 18)  function test7() {
time: 15.490sec   count: 133      test4  test.js  (line: 19, col: 18)  function test4() {
time:  5.981sec   count: 216      test1  test.js  (line:  1, col: 18)  function test1() {
time:  4.375sec   count:  18      test5  test.js  (line: 25, col: 18)  function test5() {
time:  3.241sec   count: 512      test3  test.js  (line: 13, col: 18)  function test3() {
time:  0.874sec   count:  67  anonymous  test.js  (line: 49, col: 24)  setInterval(function() {
time:  0.808sec   count:   2      test2  test.js  (line:  7, col: 18)  function test2() {
time:  0.445sec   count:   2  anonymous  test.js  (line: 43, col: 23)  setTimeout(function() {
========== SORT BY COUNT ==========
time:  3.241sec   count: 512      test3  test.js  (line: 13, col: 18)  function test3() {
time:  5.981sec   count: 216      test1  test.js  (line:  1, col: 18)  function test1() {
time: 15.490sec   count: 133      test4  test.js  (line: 19, col: 18)  function test4() {
time: 30.202sec   count:  71      test6  test.js  (line: 31, col: 18)  function test6() {
time:  0.874sec   count:  67  anonymous  test.js  (line: 49, col: 24)  setInterval(function() {
time: 16.474sec   count:  41      test7  test.js  (line: 37, col: 18)  function test7() {
time:  4.375sec   count:  18      test5  test.js  (line: 25, col: 18)  function test5() {
time:  0.808sec   count:   2      test2  test.js  (line:  7, col: 18)  function test2() {
time:  0.445sec   count:   2  anonymous  test.js  (line: 43, col: 23)  setTimeout(function() {

The result is easy to read and shows the functions you have to improve the performance of.

How it works

Suppose test.js looks like the following.

function test() {
  console.log('test');
}

The sjsp command generates test.sjsp.js.

/* some dirty codes of sjsp */ function test() { var sjsp__state = sjsp__start("test.js",1,1,"test","function test() {");
  console.log('test');; sjsp__end(sjsp__state);
}

It simply inserts sjsp__start and sjsp__end function calls at the top and the end of the functions. The local variable sjsp__state holds the starting time. It also saves the name, line number and column number of the function and the whole line. When the sjsp__end function is called, the profiling result is stored.

It just inserts the two statements for each functions. However, remember that functions can be aborted with return statements. How does it handle return statements?

Suppose the expression which is returned by the function is heavy.

function test() {  
  return someHeavyExpression;
}

Firstly consider the following code.

function test() { var sjsp__state = sjsp__start("test.js",1,1,"test","function test() {  ");  
  return someHeavyExpression; sjsp__end(sjsp__state);
}

Unfortunately, the sjsp__end function will never be called. Then what about placing the function before the return statement?

function test() { var sjsp__state = sjsp__start("test.js",1,1,"test","function test() {  ");  
  sjsp__end(sjsp__state); return someHeavyExpression;
}

The function will surely be called but the profiling result is not correct. Now, let's see how sjsp handles return statements.

function test() { var sjsp__state = sjsp__start("test.js",1,1,"test","function test() {  ");  
  return (function(arguments){ var sjsp__return = someHeavyExpression; sjsp__end(sjsp__state); return sjsp__return; } ).call(this,arguments);; sjsp__end(sjsp__state);
}

It creates an anonymous function, captures the result and calls the function instantly. This way does not break the logic and the profiling result is correct.

Author

itchyny (https://github.com/itchyny)

License

This software is released under the MIT License, see LICENSE.