如何在运行时获取 Javascript Function 调用/跟踪

Question

As I interact with my AJAX based application at RUNTIME I'd like the console to spit out all the functions it's calling. (so no stack trace, or breakpoints, or profiling or anything)

So for example, let's say I pressed a button on the page. I'd like for it to return all the functions it went through when that happened:

So I'd see in the console something like (when I pressed a button):

1. button1Clicked();
2.     calculating();
3.          printingResults();

Which basically means that button1Clicked() called calculating() which called printingResults()

Is there a utility, or plugin, browser, or maybe some way in the language to do this? I'm using google chrome, btw.

ps and NO I do not want to go through each function and add a "console.log("inside function X")" b/c that's too much work

pps as an added bonus I'd like to see the arguments passed into the functions too, but maybe that's pushing it.:>

Answer 1

I can't think of a great way to intercept all function calls globally to insert logging (though there is a decent workaround in the update section below).

Instead, how about only adding logging to functions in a certain namespace that you care about? You can do this with the following setup code:

var functionLogger = {};

functionLogger.log = true;//Set this to false to disable logging 

/**
 * Gets a function that when called will log information about itself if logging is turned on.
 *
 * @param func The function to add logging to.
 * @param name The name of the function.
 *
 * @return A function that will perform logging and then call the function. 
 */
functionLogger.getLoggableFunction = function(func, name) {
    return function() {
        if (functionLogger.log) {
            var logText = name + '(';

            for (var i = 0; i < arguments.length; i++) {
                if (i > 0) {
                    logText += ', ';
                }
                logText += arguments[i];
            }
            logText += ');';

            console.log(logText);
        }

        return func.apply(this, arguments);
    }
};

/**
 * After this is called, all direct children of the provided namespace object that are 
 * functions will log their name as well as the values of the parameters passed in.
 *
 * @param namespaceObject The object whose child functions you'd like to add logging to.
 */
functionLogger.addLoggingToNamespace = function(namespaceObject){
    for(var name in namespaceObject){
        var potentialFunction = namespaceObject[name];

        if(Object.prototype.toString.call(potentialFunction) === '[object Function]'){
            namespaceObject[name] = functionLogger.getLoggableFunction(potentialFunction, name);
        }
    }
};

Then, for whatever namespaceObject you want to add logging to, you just call:

functionLogger.addLoggingToNamespace(yourNamespaceObject);

Here's a fiddle to see it in action.

UPDATE
Note that you can call functionLogger.addLoggingToNamespace(window); to add logging to all global functions at the time of the call. Also, if you really want, you can traverse the tree to find any functions and update them accordingly. The one downfall of this method is that it only works on functions that exist at the time. Thus, it's still not the greatest solution, but it's a LOT less work than adding logging statements by hand:)

Answer 2

This is called profiling and Chrome and Firebug have it built in. In Chrome developer Tools , go to the profiles tab and click the record (circle) button. Perform your ajax and after your response, click the record button again to stop. The results of the profiling will appear in the right pane.

Note, this is going to give you everything so if you are using a library like jQuery, the vast majority of the function calls are going to be garbage to you. I've tried this a few times and I find it is much more helpful to do the console.log('inside <method>') thing.

Answer 3

I just found out that you could do that with a console.trace()

Answer 4

Maybe you can have JavaScript do some of the work of adding console.log for you:

Adding console.log to every function automatically

Also this blog by Paul Irish might help:

http://paulirish.com/2009/log-a-lightweight-wrapper-for-consolelog/

It includes a link to some JavaScript specifically targeted at logging arguments:

http://pastie.org/1033665

Answer 5

Give a try to diyism_trace_for_javascript.htm:

https://code.google.com/p/diyism-trace/downloads/list

eval('window.c=function(){3+5;}');
declare_ticks_for(window);

function a(k, c) {
  return k + 2;
}

function b() {
  4 + 3;
  a(3, {'a':'c','b':'d'});
  c();
  return 5 + 4;
}

b();

View logs in console tab of chrome or firefox

Answer 6

A variation on Briguy37's solution, I wrote one that accepts a function to call before each method. It also works with ECMAScript 6 classes, where methods are not enumerated by for...in. I'm using it to modify Object prototypes, to add logging to all new instances of my object.

function inject(obj, beforeFn) {
    for (let propName of Object.getOwnPropertyNames(obj)) {
        let prop = obj[propName];
        if (Object.prototype.toString.call(prop) === '[object Function]') {
            obj[propName] = (function(fnName) {
                return function() {
                    beforeFn.call(this, fnName, arguments);
                    return prop.apply(this, arguments);
                }
            })(propName);
        }
    }
}

function logFnCall(name, args) {
    let s = name + '(';
    for (let i = 0; i < args.length; i++) {
        if (i > 0)
            s += ', ';
        s += String(args[i]);
    }
    s += ')';
    console.log(s);
}

inject(Foo.prototype, logFnCall);

Answer 7

Let me throw a third solution into the ring: an omniscient debugger .

Note that all other answers offer two kinds of solutions:

1. Manually patch your JS functions in run-time, and log them to console
   • Yes it can get the job done, but will be useless once your project grows to a certain size. It does not give you sufficient controlability, unless you keep spending time on keeping on deleloping this feature.
2. Jeff proposes using a profiler for debugging purposes
   • Not very helpful, as the profiler views (at least for now) are designed to help you analyze performance, not the call graph; does not work well, unless you spend a lot of time training yourself to getting used to the counter-productive user interface.

That is why I wrote Dbux - a VSCode extension that provides an omniscient debugger with dynamic execution analysis tools, code annotations and a full-blown dynamic call graph visualization tool, aimed at helping developers with Program Comprehension and Debugging.

Some examples -

Call Graph of a fibonacci(6) :

Call Graph with code and other Dbux Tools in one screen:

Links:

• Overview + Documentation for the Dbux Call Graph
• Dbux VSCode Extension (with a lot of documentation and some pictures)
• Dbux Github Repo
• Video: Introduction to Dbux (with timestamps )
• Video: a Perspective on what interactive debugging might look like in the near future

Answer 8

You can trace function calls with help of putout code transformer . Plugin will look this way:

const {template, types, operator} = require('putout');
const {replaceWith} = operator;
const {BlockStatement} = types;

// create nodes
const buildLog = template(`console.log('TYPE' + ' ' + 'NAME')`);
const buildLogEnter = template(`console.log('enter' + ' ' + 'NAME' + '(' + JSON.stringify(Array.from(arguments)) + ')')`);
const buildLogException = template(`console.log('TYPE' + ' ' + 'NAME' + ': ' + trace$error.message); throw trace$error`);
const buildTryCatch = template(`try {
        BLOCK;
    } catch(trace$error) {
        CATCH;
    } finally {
        FINALLY;
    }
`);

const JSON = 'JSON';

// nodes we are searching for
module.exports.include = () => [
    'Function',
];

module.exports.fix = (path) => {
    const name = getName(path);
    
    // create 3 types of events
    const enterLog = buildLogEnter({
        NAME: name,
        JSON,
    });
    const exitLog = buildLogEvent(name, 'exit');
    const errorLog = buildLogExceptionEvent(name);
    
    // move function body into try-catch
    const bodyPath = path.get('body');
    replaceWith(bodyPath, BlockStatement([buildTryCatch({
        BLOCK: path.node.body.body,
        CATCH: errorLog,
        FINALLY: exitLog,
    })]));
    
    // add into the beginning of function "console.log" with "enter" event
    bodyPath.node.body.unshift(enterLog);
};


// get name of a function
function getName(path) {
    if (path.isClassMethod())
        return path.node.key.name;
    
    if (path.isFunctionDeclaration())
        return path.node.id.name;
    
    const {line} = path.node.loc.start;
    return `<anonymous:${line}>`;
}

// build logger
function buildLogEvent(name, type) {    
    return buildLog({
        NAME: name,
        TYPE: type,
    });
}

// build logger that throws
function buildLogExceptionEvent(name) {    
    return buildLogException({
        NAME: name,
        TYPE: 'error',
    });
}

Let's suppose that this is the code you want to trace:

const processFile = (a) => a;
process([]);

function process(runners) {
    const files = getFiles(runners);
    const linted = lintFiles(files);
    
    return linted;
}

function getFiles(runners) {
    const files = [];
    
    for (const run of runners) {
        files.push(...run());
    }
    
    return files;
}

function lintFiles(files) {
    const linted = [];
    
    for (const file of files) {
        linted.push(processFile(file));
    }
   
    return linted;
}

Here is a full picture:

If you save processed source as trace.js and run it with node, you will have:

> node trace.js
enter process([[]])
enter getFiles([[]])
exit getFiles
enter lintFiles([[]])
exit lintFiles
exit process

There is putout issue related to tracing functions .

Answer 9

I've used @Briguy37's solution with an improvement. In my case, I did not want to trace functions from some libraries, so I added some code to exclude them. Here is how it is used:

• First, include the definition of the functions you don't want to trace;
• excludeLoggingToNamespace to list the functions defined up to now and exclude them;
• Include the definition of the functions you want to trace;
• Call addLoggingToNamespace to add the logging capability to the functions defined in the above step.

Example:

<script src="js/someLibrary.js"></script>
<script>
    functionLogger.excludeLoggingToNamespace(window);
</script>
<script src="js/codeIWantToTraceHere.js"></script>
<script>
    functionLogger.addLoggingToNamespace(window);
</script>

Here is the code I added to @Briguy37's solution:

var excludedFunctions = {};

        functionLogger.excludeLoggingToNamespace = function(namespaceObject){
            for(var name in namespaceObject){
                var potentialFunction = namespaceObject[name];

                if(Object.prototype.toString.call(potentialFunction) === '[object Function]') {
                    excludedFunctions[name] = name;
                }
            }
        }; 

And I had to modify @Briguy37's addLoggingToNamespace method to take into accound the excludedFunctions hash:

functionLogger.addLoggingToNamespace = function(namespaceObject){
    for(var name in namespaceObject){
        var potentialFunction = namespaceObject[name];

        if(Object.prototype.toString.call(potentialFunction) === '[object Function]' && 
           !excludedFunctions[name]) {
            namespaceObject[name] = functionLogger.getLoggableFunction(potentialFunction, name);
        }
    }
};