I am trying to create function findStrings which should share it's variables with the function init, stringSplit and stringSearch.
String.prototype.findStrings = (function () {
var action = 'split';
var params = ',';
var init = function (action = 'split', params = ',')
{ this.stringSplit(); }
var stringSplit = function (action)
{ return this.init(); }
var stringSearch = function (action)
{return this.init(); }
return init;
})();
What I need.
When I call test.findStrings('split'); it should 1) initiate variables by calling init 2) then call stringSplit so this should be string test.
When I call var result = test.findStrings( 'search', { params ... } ); 1) initiate variables, process arguments from object params 2) then call stringsSearch and this should be test.
var test = '#a > :matches(.b, #c)';
test.findStrings('split', "," );
var result = test.findStrings( 'search', { params ... } );
Error which I got is saying that this does not have stringSplit (because this is Window) and I need to this to be test.
How to access the test string within the functions and make it possible to access the variables and arguments within them?
Actually, I could remove the function searchStrings to simplify it.
Solution:
String.prototype.findStrings = (function () {
var text = this;
...
String.prototype.init = function (action = 'split', params = ',') {
text.stringSplit();
}
}
There were two errors.
I need to this to be test.
When init() is first called, this does equal test. The problem is that stringSplit and stringSearch are just variables, and have nothing to do with the current context.
Either way, the code is pretty difficult to understand. If you really want to override String, then why don't you do something like:
String.prototype.findStrings = function(action = 'split', params = ',') {
// do logic here (not entirely sure what you want to accomplish)
return (action === 'split') ? this.stringSplit() : this.stringSearch();
}
// Add other functions also to string prototype
String.prototype.stringSplit = function() { ... }
String.prototype.stringSearch = function() { ... }
Personally though, I don't like adding methods to types like String for various reasons. Why not adopt a more functional approach, and create functions that you simply pass strings to, and get an appropriate result?
For example:
function findStrings(str, params){
// do logic and return result
}
First of all, you have a circular reference in your code: the init function calls stringSplit which calls init.
But to reach to that problem you should remove the this keyword from your prototype function. this will get the value of the object that is calling your function, in your case the window object, but your functions are created in a closure so you can't access them using the window object.
We have some js code splitted in many files. We have a core file that defines code used by many other js files.
Currently we have something like this:
core.js:
window.mycore = function() {
var myfunction1 = function() {
};
var myfunction2 = function() {
};
var myfunction3 = function() {
//..
var a = myfunction1(b);
//..
};
//...
// many "myfunction"
//...
var myfunctionN = function() {
};
var publish = function() {
for(var i = 0; i < arguments.length; i++) {
try {
window.mycore[arguments[i]] = eval('(' + arguments[i] + ')');
}
catch(e) {
Log.err(600, arguments[i], e);
}
}
};
publish("myfunction1", "myfunction7", "myfunction8",/*...*/"myfunctionM")
}
app.js:
// ...
// ...
var result = window.core.myfunction1("myparam");
// ...
// ...
Note that none core methods are declared as members of the window.core object. Instead they are attached to the core object with the publish function.
This has some pros:
The core code can reference any core function without the need of writing "window.core."
We avoid writing "var myfunction = window.mycore.myfunction = function() ..." in every public function declaration
The exposed methods can be seen centraliced.
But, the use of eval in the publish function is bringing us problems when using code analysis tools since they don't tend to understand eval declarations.
So, here is my question.
Which is the better way to improve this code, so we can keep the advantages mentioned but eradicating the eval declaration.
I am aware of the solution of sending to the publish function some name/value pairs like publish({'myfunction1': myfunction1}, ... ), but I also want to avoid function name repetitions.
Consider that I am not looking for radical changes since there is a lot of code written already.
Thanks!
I'm not sure I understand completely your reasons for using the "publish" method, but is there any reason your not just returning an object with the correct functions from your constructor?
ie:
window.mycore = (function() {
var myFunc1 = function(a) {
alert(a);
};
var myFunc2 = function(b) {
// call to other function in the same scope
myFunc1(b);
}
...
// at the end just expose the public members you want
return {
myFunc1: myFunc1,
myFunc2: myFunc2
};
})();
or
window.mycore = (function() {
return {
myFunc1: function(a) {
alert(a);
},
myFunc2: function(b) {
this.myFunc1(b);
}
};
})();
or, yet another way to end up with the same object :) ... as always there are different ways to get there
(function(){
var o = {};
o.func1 = function(a) {
alert(a);
}
o.func2 = function(b) {
this.func1(b);
}
window.mycore = o;
})();
So, at a fundamental level, I think it would have benefitted you to have written those name spaces as objects. But thats a whole different subject entirely. (and it disqualifies based on the fact that you dont want to do a lot of refactoring).
With that said, my first idea was that you could probably sidestep the need for eval by using the .call() or .apply() method. What they allow you to do is to chain a function call out of your function name. but that doesn't apply to a "string" which is what you're giving your publish function.
so after googling, this is how you execute a function from a string:
var fn = window[settings.functionName];
if(typeof fn === 'function') {
fn(t.parentNode.id);
}
https://stackoverflow.com/a/912642/680578
Personally I prefer the #Jaime approach, but maybe you may do something like
window.mycore = function() {
function myfunction1() {
};
function myfunction2() {
};
function myfunction3() {
//..
var a = myfunction1(b);
//..
};
//...
// many "myfunction"
//...
function myfunctionN() {
};
var publish = function() {
for(var i = 0; i < arguments.length; i++) {
try {
window.mycore[arguments[i].name] = arguments[i];
}
catch(e) {
Log.err(600, arguments[i].name, e);
}
}
};
publish(myfunction1, myfunction7, myfunction8,/*...*/myfunctionM);
}
Suppose I have a variables in the global scope.
Suppose I wish to define a function which I can guarantee will not have access to this variable, is there a way to wrap the function, or call the function, that will ensure this?
In fact, I need any prescribed function to have well defined access to variables, and that access to be defined prior to, and separate from that function definition.
Motivation:
I'm considering the possibility of user submitted functions. I should be able to trust that the function is some variety of "safe" and therefore be happy publishing them on my own site.
Run the code in an iframe hosted on a different Origin. This is the only way to guarantee that untrusted code is sandboxed and prevented from accessing globals or your page's DOM.
Using embedded Web Workers could allow to run safe functions. Something like this allows a user to enter javascript, run it and get the result without having access to your global context.
globalVariable = "I'm global";
document.getElementById('submit').onclick = function() {
createWorker();
}
function createWorker() {
// The text in the textarea is the function you want to run
var fnText = document.getElementById('fnText').value;
// You wrap the function to add a postMessage
// with the function result
var workerTemplate = "\
function userDefined(){" + fnText +
"}\
postMessage(userDefined());\
onmessage = function(e){console.log(e);\
}"
// web workers are normally js files, but using blobs
// you can create them with strings.
var blob = new Blob([workerTemplate], {
type: "text/javascript"
});
var wk = new Worker(window.URL.createObjectURL(blob));
wk.onmessage = function(e) {
// you listen for the return.
console.log('Function result:', e.data);
}
}
<div>Enter a javascript function and click submit</div>
<textarea id="fnText"></textarea>
<button id="submit">
Run the function
</button>
You can try these for example by pasting it in the textarea:
return "I'm a safe function";
You can see that it's safe:
return globalVariable;
You can even have more complex scripts, something like this:
var a = 4, b = 5;
function insideFn(){
// here c is global, but only in the worker context
c = a + b;
}
insideFn();
return c;
See info about webworkers here, especially embedded web workers:
https://developer.mozilla.org/en-US/docs/Web/API/Web_Workers_API/Using_web_workers#Embedded_workers
A little late, but maybe it will help you a bit
function RestrictFunction(params) {
params = ( params == undefined ? {} : params );
var scope = ( params.scope == undefined ? window : params.scope );
var data = ( params.data == undefined ? {} : params.data );
var script = ( params.script == undefined ? '' : params.script );
if (typeof params.script == 'function') {
script = params.script.toString();
script = script.substring(script.indexOf("{") + 1, script.lastIndexOf("}"));
}
// example: override native functions that on the white list
var setTimeout = function(_function,_interval) {
// this is important to prevent the user using `this` in the function and access the DOM
var interval = scope.setTimeout( function() {
RestrictFunction({
scope:scope,
data:data,
script:_function
});
} , _interval );
// Auto clear long user intervals
scope.setTimeout( function() {
scope.clearTimeout(interval);
} , 60*1000 );
return interval;
}
// example: create custom functions
var trace = function(str) {
scope.console.log(str);
}
return (function() {
// remove functions, objects and variables from scope
var queue = [];
var WhiteList = [
"Blob","Boolean","Date","String","Number","Object","Array","Text","Function",
"unescape","escape","encodeURI","encodeURIComponent","parseFloat","parseInt",
"isNaN","isFinite","undefined","NaN",
"JSON","Math","RegExp",
"clearTimeout","setTimeout"
];
var properties = Object.getOwnPropertyNames(scope);
for (var k = 0; k<properties.length; k++ ) {
if (WhiteList.indexOf(properties[k])!=-1) continue;
queue.push("var "+properties[k]+" = undefined;");
}
for (var k in scope) {
if (WhiteList.indexOf(k)!=-1) continue;
queue.push("var "+k+" = undefined;");
}
queue.push("var WhiteList = undefined;");
queue.push("var params = undefined;") ;
queue.push("var scope = undefined;") ;
queue.push("var data = undefined;") ;
queue.push("var k = undefined;");
queue.push("var properties = undefined;");
queue.push("var queue = undefined;");
queue.push("var script = undefined;");
queue.push(script);
try {
return eval( '(function(){'+ queue.join("\n") +'}).apply(data);' );
} catch(err) { }
}).apply(data);
}
Example of use
// dummy to test if we can access the DOM
var dummy = function() {
this.notify = function(msg) {
console.log( msg );
};
}
var result = RestrictFunction({
// Custom data to pass to the user script , Accessible via `this`
data:{
prop1: 'hello world',
prop2: ["hello","world"],
prop3: new dummy()
},
// User custom script as string or function
script:function() {
trace( this );
this.msg = "hello world";
this.prop3.notify(this.msg);
setTimeout( function() {
trace(this);
} , 10 );
trace( data );
trace( params );
trace( scope );
trace( window );
trace( XMLHttpRequest );
trace( eval );
return "done!"; // not required to return value...
},
});
console.log( "result:" , result );
You can't restrict the scope of a Function using the "call" or "apply" methods, but you can use a simple trick using "eval" and scoping to essentially hide any specific global variables from the function to be called.
The reason for this is because the function has access to the "global" variables that are declared at the scope that the function itself what declared. So, by copying the code for the method and injecting it in eval, you can essentially change the global scope of the function you are looking to call. The end result is essentially being able to somewhat sandbox a piece of javascript code.
Here's a full code example:
<html>
<head>
<title>This is the page title.</title>
<script>
function displayTitle()
{
alert(document.title);
}
function callMethod(method)
{
var code = "" +
// replace global "window" in the scope of the eval
"var window = {};" +
// replace global "document" in the scope of the eval
"var document = {}; " +
"(" +
// inject the Function you want to call into the eval
method.toString() +
// call the injected method
")();" +
"";
eval(code);
}
callMethod(displayTitle);
</script>
</head>
<body></body>
</html>
The code that gets eval'd looks like this:
var window = {};
var document = {};
(function displayTitle()
{
alert(document.title);
})();
You can use WebWorkers to isolate your code:
Create a completely separate and parallel execution environment (i.e. a separate thread or process or equivalent construct), and run the rest of these steps asynchronously in that context.
Here is a simple example:
someGlobal = 5;
//As a worker normally take another JavaScript file to execute we convert the function in an URL: http://stackoverflow.com/a/16799132/2576706
function getScriptPath(foo) {
return window.URL.createObjectURL(new Blob([foo], {
type: 'text/javascript'
}));
}
function protectCode(code) {
var worker = new Worker(getScriptPath(code));
}
protectCode('console.log(someGlobal)'); // prints 10
protectCode('console.log(this.someGlobal)');
protectCode('console.log(eval("someGlobal"))');
protectCode('console.log(window.someGlobal)');
This code will return:
Uncaught ReferenceError: someGlobal is not defined
undefined
Uncaught ReferenceError: someGlobal is not defined and
Uncaught ReferenceError: window is not defined
so you code is now safe.
EDIT: This answer does not hide the window.something variables. But it has a clean way to run user-defined code. I am trying to find a way to mask the window variables
You can use the javascript function Function.prototype.bind() to bind the user submitted function to a custom scope variable of your choosing, in this custom scope you can choose which variables to share with the user defined function, and which to hide. For the user defined functions, the code will be able to access the variables you shared using this.variableName. Here is an example to elaborate on the idea:
// A couple of global variable that we will use to test the idea
var sharedGlobal = "I am shared";
var notSharedGlobal = "But I will not be shared";
function submit() {
// Another two function scoped variables that we will also use to test
var sharedFuncScope = "I am in function scope and shared";
var notSharedFuncScope = "I am in function scope but I am not shared";
// The custom scope object, in here you can choose which variables to share with the custom function
var funcScope = {
sharedGlobal: sharedGlobal,
sharedFuncScope: sharedFuncScope
};
// Read the custom function body
var customFnText = document.getElementById("customfn").value;
// create a new function object using the Function constructor, and bind it to our custom-made scope object
var func = new Function(customFnText).bind(funcScope);
// execute the function, and print the output to the page.
document.getElementById("output").innerHTML = JSON.stringify(func());
}
// sample test function body, this will test which of the shared variables does the custom function has access to.
/*
return {
sharedGlobal : this.sharedGlobal || null,
sharedFuncScope : this.sharedFuncScope || null,
notSharedGlobal : this.notSharedGlobal || null,
notSharedFuncScope : this.notSharedFuncScope || null
};
*/
<script type="text/javascript" src="app.js"></script>
<h1>Add your custom body here</h1>
<textarea id="customfn"></textarea>
<br>
<button onclick="submit()">Submit</button>
<br>
<div id="output"></div>
The example does the following:
Accept a function body from the user
When the user clicks submit, the example creates a new function object from the custom body using the Function constructor. In the example we create a custom function with no parameters, but params can be added easily as the first input of the Function constructor
The function is executed, and its output is printed on the screen.
A sample function body is included in comments, that tests which of the variables does the custom function has access to.
Create a local variable with the same name.
If you have a global variable like this:
var globalvar;
In your function:
function noGlobal(); {
var globalvar;
}
If the function refers to globalvar, it will refers to the local one.
In my knowledge, in Javascript, any variable declared outside of a function belongs to the global scope, and is therefore accessible from anywhere in your code.
Each function has its own scope, and any variable declared within that function is only accessible from that function and any nested functions. Local scope in JavaScript is only created by functions, which is also called function scope.
Putting a function inside another function could be one possibility where you could achieve reduced scope ( ie nested scope)
if you are talking about a function that is exposed to you by loading a third party script, you are pretty much out of luck. that's because the scope for the function is defined in the source file it's defined in. sure, you can bind it to something else, but in most cases, that's going to make the function useless if it needs to call other functions or touch any data inside it's own source file - changing it's scope is only really feasible if you can predict what it needs to be able to access, and have access to that yourself - in the case of a third party script that touches data defined inside a closure, object or function that's not in your scope, you can't emulate what might need.
if you have access to the source file then it's pretty simple - look at the source file, see if it attempts to access the variable, and edit the code so it can't.
but assuming you have the function loaded, and it doesn't need to interact with anything other than "window", and for academic reasons you want to do this, here is one way - make a sandbox for it to play in. here's a simple shim wrapper that excludes certain items by name
function suspectCode() {
console.log (window.document.querySelector("#myspan").innerText);
console.log('verbotten_data:',typeof verbotten_data==='string'?verbotten_data:'<BANNED>');
console.log('secret_data:',typeof secret_data==='string'?secret_data:'<BANNED>'); // undefined === we can't
console.log('window.secret_data:',typeof window.secret_data==='string'?window.secret_data:'<BANNED>');
secret_data = 'i changed the secret data !';
console.log('secret_data:',typeof secret_data==='string'?secret_data:'<BANNED>'); // undefined === we can't
console.log('window.secret_data:',typeof window.secret_data==='string'?window.secret_data:'<BANNED>');
}
var verbotten_data = 'a special secret';
window.secret_data = 'special secret.data';
console.log("first call the function directly");
suspectCode() ;
console.log("now run it in a sandbox, which banns 'verbotten_data' and 'secret_data'");
runFunctionInSandbox (suspectCode,[
'verbotten_data','secret_data',
// we can't touch items tied to stack overflows' domain anyway so don't clone it
'sessionStorage','localStorage','caches',
// we don't want the suspect code to beable to run any other suspect code using this method.
'runFunctionInSandbox','runSanitizedFunctionInSandbox','executeSandboxedScript','shim',
]);
function shim(obj,forbidden) {
const valid=Object.keys(obj).filter(function(key){return forbidden.indexOf(key)<0;});
var shimmed = {};
valid.forEach(function(key){
try {
shimmed[key]=obj[key];
} catch(e){
console.log("skipping:",key);
}
});
return shimmed;
}
function fnSrc (fn){
const src = fn.toString();
return src.substring(src.indexOf('{')+1,src.lastIndexOf('}')-1);
}
function fnArgs (fn){
let src = fn.toString();
src = src.substr(src.indexOf('('));
src = src.substr(0,src.indexOf(')')-1);
src = src.substr(1,src.length-2);
return src.split(',');
}
function runSanitizedFunctionInSandbox(fn,forbidden) {
const playground = shim(window,forbidden);
playground.window = playground;
let sandboxed_code = fn.bind(playground,playground.window);
sandboxed_code();
}
function runFunctionInSandbox(fn,forbidden) {
const src = fnSrc(fn);
const args = fnArgs(fn);
executeSandboxedScript(src,args,forbidden);
}
function executeSandboxedScript(sourceCode,arg_names,forbidden) {
var script = document.createElement("script");
script.onload = script.onerror = function(){ this.remove(); };
script.src = "data:text/plain;base64," + btoa(
[
'runSanitizedFunctionInSandbox(function(',
arg_names,
['window'].concat(forbidden),
'){ ',
sourceCode,
'},'+JSON.stringify(forbidden)+')'
].join('\n')
);
document.body.appendChild(script);
}
<span id="myspan">Page Access IS OK<span>
or a slightly more involved version that allows arguments to be passed to the function
function suspectCode(argument1,argument2) {
console.log (window.document.querySelector("#myspan").innerText);
console.log(argument1,argument2);
console.log('verbotten_data:',typeof verbotten_data==='string'?verbotten_data:'<BANNED>');
console.log('secret_data:',typeof secret_data==='string'?secret_data:'<BANNED>'); // undefined === we can't
console.log('window.secret_data:',typeof window.secret_data==='string'?window.secret_data:'<BANNED>');
secret_data = 'i changed the secret data !';
console.log('secret_data:',typeof secret_data==='string'?secret_data:'<BANNED>'); // undefined === we can't
console.log('window.secret_data:',typeof window.secret_data==='string'?window.secret_data:'<BANNED>');
}
var verbotten_data = 'a special secret';
window.secret_data = 'special secret.data';
console.log("first call the function directly");
suspectCode('hello','world') ;
console.log("now run it in a sandbox, which banns 'verbotten_data' and 'secret_data'");
runFunctionInSandbox (suspectCode,['hello','sandboxed-world'],
[
'verbotten_data','secret_data',
// we can't touch items tied to stack overflows' domain anyway so don't clone it
'sessionStorage','localStorage','caches',
// we don't want the suspect code to beable to run any other suspect code using this method.
'runFunctionInSandbox','runSanitizedFunctionInSandbox','executeSandboxedScript','shim',
]);
function shim(obj,forbidden) {
const valid=Object.keys(obj).filter(function(key){return forbidden.indexOf(key)<0;});
var shimmed = {};
valid.forEach(function(key){
try {
shimmed[key]=obj[key];
} catch(e){
console.log("skipping:",key);
}
});
return shimmed;
}
function fnSrc (fn){
const src = fn.toString();
return src.substring(src.indexOf('{')+1,src.lastIndexOf('}')-1);
}
function fnArgs (fn){
let src = fn.toString();
src = src.substr(src.indexOf('('));
src = src.substr(0,src.indexOf(')'));
src = src.substr(1,src.length);
return src.split(',');
}
function runSanitizedFunctionInSandbox(fn,args,forbidden) {
const playground = shim(window,forbidden);
playground.window = playground;
let sandboxed_code = fn.bind(playground,playground.window);
sandboxed_code.apply(this,new Array(forbidden.length).concat(args));
}
function runFunctionInSandbox(fn,args,forbidden) {
const src = fnSrc(fn);
const arg_names = fnArgs(fn);
executeSandboxedScript(src,args,arg_names,forbidden);
}
function executeSandboxedScript(sourceCode,args,arg_names,forbidden) {
var script = document.createElement("script");
script.onload = script.onerror = function(){ this.remove(); };
let id = "exec"+Math.floor(Math.random()*Number.MAX_SAFE_INTEGER).toString();
window.execArgs=window.execArgs||{};
window.execArgs[id]=args;
let script_src = [
'runSanitizedFunctionInSandbox(function(',
['window'].concat(forbidden),
(arg_names.length===0?'':','+arg_names.join(","))+'){',
sourceCode,
'},',
'window.execArgs["'+id+'"],',
JSON.stringify(forbidden)+');',
'delete window.execArgs["'+id+'"];'
].join('\n');
let script_b64 = btoa(script_src);
script.src = "data:text/plain;base64," +script_b64;
document.body.appendChild(script);
}
<span id="myspan">hello computer...</span>
I verified #josh3736's answer but he didn't leave an example
Here's one to verify it works
parent.html
<h1>parent</h1>
<script>
abc = 'parent';
function foo() {
console.log('parent foo: abc = ', abc);
}
</script>
<iframe></iframe>
<script>
const iframe = document.querySelector('iframe');
iframe.addEventListener('load', function() {
console.log('-calling from parent-');
iframe.contentWindow.foo();
});
iframe.src = 'child.html';
</script>
child.html
<h1>
child
</h1>
<script>
abc = 'child';
function foo() {
console.log('child foo: abc = ', abc);
}
console.log('-calling from child-');
parent.foo();
</script>
When run it prints
-calling from child-
parent foo: abc = parent
-calling from parent-
child foo: abc = child
Both child and parent have a variable abc and a function foo.
When the child calls into the parent's foo that function in the parent sees the parent's global variables and when the parent calls the child's foo that function sees the child's global variables.
This also works for eval.
parent.html
<h1>parent</h1>
<iframe></iframe>
<script>
const iframe = document.querySelector('iframe');
iframe.addEventListener('load', function() {
console.log('-call from parent-');
const fn = iframe.contentWindow.makeFn(`(
function() {
return abc;
}
)`);
console.log('from fn:', fn());
});
iframe.src = 'child.html';
</script>
child.html
<h1>
child
</h1>
<script>
abc = 'child';
function makeFn(s) {
return eval(s);
}
</script>
When run it prints
-call from parent-
from fn: child
showing that it saw the child's abc variable not the parent's
note: if you create iframes programmatically they seem to have to be added to the DOM or else they won't load. So for example
function loadIFrame(src) {
return new Promise((resolve) => {
const iframe = document.createElement('iframe');
iframe.addEventListener('load', resolve);
iframe.src = src;
iframe.style.display = 'none';
document.body.appendChild(iframe); // iframes don't load if not in the document?!?!
});
}
Of course in the child above we saw that the child can reach into the parent so this code is NOT SANDBOXED. You'd probably have to add some stuff to hide the various ways to access the parent if you want make sure the child can't get back but at least as a start you can apparently use this technique to give code a different global scope.
Also note that of course the iframes must be in the same domain as the parent.
Here's another answer. This one's based on how chained scopes work in javascript. It also uses Function(), which produces faster code than eval.
/** This takes a string 'expr', e.g. 'Math.max(x,1)' and returns a function (x,y)=>Math.max(x,1).
* It protects against malicious input strings by making it so that, for the function,
* (1) no names are in scope other than the parameters 'x' and 'y' and a whitelist
* of other names like 'Math' and 'Array'; (2) also 'this' binds to the empty object {}.
* I don't think there's any way for malicious strings to have any effect.
* Warning: if you add something into the global scope after calling make_fn but
* before executing the function you get back, it won't protect that new thing.
*/
function make_fn(expr) {
const whitelist = ['Math', 'Array']; // Warning: not 'Function'
let scope = {};
for (let obj = this; obj; obj = Object.getPrototypeOf(obj)) {
Object.getOwnPropertyNames(obj).forEach(name => scope[name] = undefined);
}
whitelist.forEach(name => scope[name] = this[name]);
const fn = Function("scope","x","y","with (scope) return "+expr).bind({});
return (x,y) => fn(scope,x,y);
}
This is how it behaves: https://jsfiddle.net/rkq5otme/
make_fn("x+y")(3,5) ==> 8
make_fn("Math.max(x,y)")(3,5) ==> 5
make_fn("this")(3,5) ==> {}
make_fn("alert('oops')") ==> TypeError: alert is not a function
make_fn("trace((function(){return this}()))")(3,5) ==> ReferenceError: trace is not defined
Explanation. Consider a simpler version
Function("x", "return Math.max(x, window, this);")
This creates a function with the specified body which has two chained scopes: (1) the function scope which binds x, (2) the global scope. Let's spell out how the symbols Math, x, window and this are all resolved.
x is bound to the property in the function scope
window is bound to the property in the next chained scope, global, i.e. window['window'].
We don't want this! To prevent it, we create our own scope object scope = {window:undefined,...} and write with (scope) return Math.max(x,window,this). The with statement is only allowed in non-strict code. It adds an additional scope, so the chain is now: (0) the specified scope object we created, (1) the function scope which binds x, (2) the global scope. Because name 'window' is found in our scope object, it can never bind to the one in global scope.
Math will suffer the same fate.
We want to whitelist it! So we make our scope object {Math:global.Math, window:undefined, ...}
this is bound upon invocation of the function to the global object, window.
We don't want this! To prevent it, we call .bind({}) on the function, which wraps the function in a wrapper which sets this={}. Alas bind(null) didn't seem to bind.
Note: there's another possible construction of the scope object, using Proxy. It doesn't seem particularly better.
const scope = new Proxy({}, {
has: (obj, key) => !['x','y'].includes(key),
get: (obj, key) => (whitelist.includes(key)) ? this[key] : undefined,
set: (obj, key, value) => {},
deleteProperty: (obj, key) => {},
enumerate: (obj, key) => [],
ownKeys: (obj, key) => [],
defineProperty: (obj, key, desc) => {},
getOwnPropertyDescriptor: (obj, key) => undefined,
});
I am having some trouble getting a callback function to work. Here is my code:
SomeObject.prototype.refreshData = function()
{
var read_obj = new SomeAjaxCall("read_some_data", { }, this.readSuccess, this.readFail);
}
SomeObject.prototype.readSuccess = function(response)
{
this.data = response;
this.someList = [];
for (var i = 0; i < this.data.length; i++)
{
var systemData = this.data[i];
var system = new SomeSystem(systemData);
this.someList.push(system);
}
this.refreshList();
}
Basically SomeAjaxCall is making an ajax request for data. If it works we use the callback 'this.readSuccess' and if it fails 'this.readFail'.
I have figured out that 'this' in the SomeObject.readSuccess is the global this (aka the window object) because my callbacks are being called as functions and not member methods. My understanding is that I need to use closures to keep the 'this' around, however, I have not been able to get this to work.
If someone is able show me what I should be doing I would appreciate it greatly. I am still wrapping my head around how closures work and specifically how they would work in this situation.
Thanks!
Well the most straightforward thing to do is to just wrap "this.readSuccess" in another function:
SomeObject.prototype.refreshData = function()
{
var obj = this;
var read_obj = new SomeAjaxCall("read_some_data", { },
function() { obj.readSuccess(); }, function() { obj.readFail(); });
}
Some Javascript frameworks provide a utility to "bind" a function to an object, which simply means that it creates one of those little functions for you. Note that the variable "obj" will be "remembered" by those little functions, so when your handlers are called the "this" reference will be to the object that was used to call "refreshData".
Your problem here is not exactly a closure or scoping problem. The problem is that when you assign this.readSuccess to a variable, you assign the function itself without any notion of the object it originaly belongs to.
In the same way, you can take a regular, "stand-alone" function and use it as method of an object:
function hello() {
alert("Hello "+this.planet);
}
var planet = "Earth";
hello(); // -> 'Hello Earth'
var Venus = {
planet: "Venus"
};
hello.apply(Venus); // -> 'Hello Venus'
Venus.hello = hello;
Venus.hello(); // -> 'Hello Venus'
And your problem can be replicated in this example
var helloVenus = Venus.hello;
helloVenus(); // -> 'Hello Earth'
So your problem is to assign this.readSuccess to some variable and having it called as a method of this. Which can be done with a closure as demonstrated by Pointy. Since I don't know what "SomeAjaxCall" actually does, it's hard to know if the value of this is actually lost and if var obj = this is actually needed. Chances are that it's not, so you can be fine with this kind of code:
var helloVenus = function() { Venus.hello() }
helloVenus(); // -> 'Hello Venus'
In your case, that would be (edit: adding the arguments passed to the handler) :
SomeObject.prototype.refreshData = function()
{
var read_obj = new SomeAjaxCall(
"read_some_data",
{ },
function () { this.readSuccess.apply(this, arguments) },
function () { this.readFail.apply(this, arguments) }
);
}
As noted previously, several js frameworks offer a bind function to simplify this kind of issue. But you don't need a complete framework just for this : here is a perfectly fine Function#bind method that works an plain javascript:
Function.prototype.bind = function(obj) {
var __method = this;
var args = [];
for(var i=1; i<arguments.length; i++)
args.push(arguments[i]);
return function() {
var args2 = [];
for(var i=0; i<arguments.length; i++)
args2.push(arguments[i]);
return __method.apply(obj, args.concat(args2));
};
}
With the help of Function#bind, you can write:
SomeObject.prototype.refreshData = function()
{
var read_obj = new SomeAjaxCall(
"read_some_data",
{ },
this.readSuccess.bind(this),
this.readFail.bind(this)
);
}
The following site seems to suggest that the problem may be more in your "class" building than in the usage. If you read down to the "rewrite using prototype properties", they write that that particular method of "class" structuring will keep the methods global instead of instance-based. Perhaps another creation method?
http://devedge-temp.mozilla.org/viewsource/2001/oop-javascript/