I've been doing some javascript reading, and I've gathered that a closure has access only to the closure "wrapping" it, or, you might say it's immediate parent. Now I've been playing a bit, and I see in this jsfiddle that even deep nested functions have access to to vars defined way up.
Can anyone please explain that? Or explain what have I got completely wrong?
http://jsfiddle.net/tPQ4s/
function runNums() {
this.topVar = 'blah';
return function(){
(function() {
(function() {
console.log(topVar);
})();
})();
}
}
var someFunc = runNums();
someFunc();
Without going too deep into the details, a closure technically describes a array like variable within the such called Activation Object that is handled from the javascript engine. An ActivationObject contains Variables declared by var, function declarations and formal parameters.
That means, anytime a new function (-context) is invoked, internally a new Activation Object is created. That object is part of the new Execution Context, a typicall EC looks like:
this context variable
Activation Object
[[Scope]]
The interesting part here is [[Scope]]. That variable contains all Activation Objects of all parent context and is filled when the EC is called. So now, when a function wants to access a variable, the name resolution process first looks into its own Activation Object, if nothing is found the search continues in the "Scope chain", which is just an Indexed search through our [[Scope]] variable (which again, is an array of parent contexts). Thats why we also speak a lot about "lexical scope" in ECMA-/Javascript.
Note: The above behavior is not described entirely, that would need several pages of text. Also it describes the ECMAscript3 262 specification. Things work a little different in ES5, but its still around the same thing
That is because the chain runs further up to the top context.
In the example, that would be:
window < runNums < anonymous < anonymous < anonymous
Variables living in any of these will be available in the last anonymous function. In runNums, only variables living in runNums or window will be available. In the first anonymous function, only its variables and those living in runNums or window will be available, etc.
this is nothing but the Window object here.
Here runNums is a global function and runNums() is equal to window.runNums(). So this is window and this.topVar is window.topVar. Obviously it will be accessible from anywhere.
Try this and see the difference
var someFunc = new runNums();
someFunc();
The deep nested functions have not been executed. You did not return them for executing.
Related
I have two questions about hoisting:
The way function declarations are hoisted is that they go to the very top, even above variable declarations, to my understanding.
If we have this code:
function fn() {
callback();
}
function callback() {
console.log('why does this show');
}
I don't understand how this works, since both functions are hoisted to the top (which effectively produces the same code that already exists). But callback is still created below fn, and I don't see why we can access it in fn. My guess is it has something to do with the top level objects being able to access each other regardless of lexical position.
Similarly:
var a = 10;
function fn() {
console.log(a);
}
fn();
How does this work? Because the way I understand hoisting makes it seem like the function should be hoisted even above var a, which makes it seem like variables should always be inaccessible in functions.
We can go down the rabbit hole with this but I want to try to give a brief explanation on how your examples work.
Whenever the JavaScript engine creates an execution context (also known as calling stack) whether through functions or code in the global scope it creates an lexical environment. Which is a data structure that holds a collection of name/value pairs about variables and functions in it's own scope or from it's parent scope by using a reference.
About your first example. Both functions get added to the global execution context. If you call fn() in your first example initially. It will then add callback() to the call stack of fn() and execute it accordingly. So, the order of your functions don't really matter in this case.
You're second example is a different case. The execution context knowns you are referring to the global variable and therefor adding a reference to the lexical environment and that makes it able to use the variable inside fn().
This can be quite hard to get a grasp of. There are a ton of resources related to hoisting, scopes, lexical environments and execution contexts so be sure to check those out. :)
This is because how our Javascript Engine parse and compile the code.
I'm not an expert but V8 (The Chorme Engine for JS) in the first file get all the variables and functions names and store all the function references. So, you can "use" a function before "declare" because JS know where the function is.
Some languages, even C++ you could do that and is a nice feature :)
When I call a function, a local scope is erected for that call. Is there any way to directly reference that scope as an object? Just like window is a reference for the global scope object.
Example:
function test(foo){
var bar=1
//Now, can I access the object containing foo, bar, arguments and anything
//else within the local scope like this:
magicIdentifier.bar
}
Alternately, does anyone have a complete list of what is in the local scope on top of custom variables?
Background: I'm trying to get down to a way of completely shifting to global scope from within a function call, the with statement is a joke, call works a little better, but it still breaks for anything declared in function scope but not in global scope, therefore I would declare these few cases in global scope, but that requires me to know what they are. The IE function execScript makes a complete shift, but that only solves the problem for IE.
Note: To anyone loading JavaScript dynamically, setTimeout(code,1) is a simple effective hack to achieve global scope, but it will not execute immediately.
No, there's no way to reference the variable object of the execution context of a function binding object of the variable environment of the execution context (that's what that thing is called [now; hence the strikethrough]; details in §10.3 of the specification). You can only access the limited view to it you get with arguments (which is very limited indeed).
Usually when I've wanted to do this, I've just put everything I wanted on an object and then used that (e.g., passed it into a function). Of course, any functions created within the context have access to everything in scope where they're created, as they "close over" the context; more: Closures are not complicated.
I know this is hugely late, and you're probably not even slightly interested any more, but I was interested in the feasibility of this too and you should be able to make a work around of some sort using:
(function(global) {
var testVar = 1;
global.scope = function(s) {
return eval(s);
}
})(this);
then running:
scope('testVar'); // 1
returns the variable from within the closure. Not particularly nice, but theoretically possible to wrap that in an object, perhaps using some validation and getters and setters if you needed?
Edit: Having re-read the question, I assume you'd want to access it without having to specify a function in the scope itself, so this probably isn't applicable. I'll leave this here anyway.
Certain versions of Netscape had a magic property in the arguments object that did what you're looking for. (I can't remember what it was called)
What about something like this?
<script type="text/javascript">
var test = {
bar : 1,
foo : function () {
alert(this.bar);
}
}
test.foo();
</script>
You don't need a keyword to reference a variable in the local scope, because it's the scope you're in.
I was looking at adding comments to JSON and found this script that strips them out before processing making the JSON valid. I am just trying to understand how it works to make the JSON.minify() function available?
It starts with
(function(global){ ...
totally which is weird to me. I found that "global is a property of a RegExp instance, not the RegExp object" on MDN but I don't understand how it is works in this script if at all.
This snippet:
(function(global){
// your code here
// referring to the variable named "global" in this scope
// will be a reference to the default javascript global object
})(this);
is a construct for assigning the global object (whatever it might be) to an argument labeled global for all code that is inside this self-executing function.
The self executing function is used to define a separate execution scope so that any functions or variables you define inside this other scope will not interfere with or be directly accessible from outside this scope (insulating your scope from other code scopes).
In a browser, the global object is the window object, but if you intended to have code that might work in other javascript environments (like no node.js on a server) where the global object might not be window, this is a way of extracting the global value from the default this value, putting it into another variable which you can then refer to anywhere inside your code block.
For code mean to only run in a browser, there really is no point to this. You can just refer to window when you need the global object.
It's just a function parameter name. It might as well be froozboggles.
This code:
(function(foo) {
// In here, what's called "bar" in the outer scope is called "foo"
})(bar);
Defines an anonymous function taking one parameter bar and immediately calls it with the value of bar as the first parameter.
Apart from what jfriend00 mentions in his fine answer, it's also a good way of making sure that you don't leak variables and functions to the outer scope: If you declare, say, var baz = 17; in the top scope in javascript, it will be a property of window. If you wrap it in a function as in the pattern you mention, you can only export properties to window explicitly -- by assigning them to global, in the case of your example. Edit: As #josh3736 says in his comment, you can also leak to window by assigning without a previous declaration, e.g. quux = 4711;.
Just out of curiosity, do closures in JavaScript get a reference to the whole "outer environment", or is the returned function analyzed to see which variables in the outer scope it references and then only gets references to those?
Theoretically, a nested function in JavaScript has access to all variables in all containing scopes. When an identifier is encountered, it is resolved against the scope chain, which is a list that includes objects whose properties are variables and function parameters of each containing execution context (i.e. enclosing function), innermost first, plus the global object at the end. A function object drags its scope chain around with it wherever it goes.
However, these Variable objects and the scope chain are only specification constructs and are not accessible directly, so implementations are free to make whatever optimizations they like, including analyzing function code and only exposing variables that are accessed by a function and any functions nested within it, so long as the specification always appears to be satisfied. However, it's best to assume that if you have an enormous object that is accessible via a closure to a function, that enormous object is going to stick around at least until that function is garbage collected.
If you want further information about this, read the ECMAScript specification. A good starting point would be section 10.1.4: http://bclary.com/2004/11/07/#a-10.1.4. This is not the current version of the specification but is the baseline for what all current major browsers implement.
The answer is "yes and no". When a function "leaks" out of a function activation, the entire context is preserved*. However, as there's no way to refer to the context itself, the code of a function cannot "investigate" the context(s). Thus:
function leaker() {
var i = 100, j = "hello there";
return function() {
i = i - 1;
return i == 0;
}
}
The returned function can only ever refer to "i". The variable "j" may stick around, but there's no way for code in that returned function to "find" it.
* I write that the context is preserved, which I believe to be true, but technically that's the business of the interpreter/runtime.
When I call a function, a local scope is erected for that call. Is there any way to directly reference that scope as an object? Just like window is a reference for the global scope object.
Example:
function test(foo){
var bar=1
//Now, can I access the object containing foo, bar, arguments and anything
//else within the local scope like this:
magicIdentifier.bar
}
Alternately, does anyone have a complete list of what is in the local scope on top of custom variables?
Background: I'm trying to get down to a way of completely shifting to global scope from within a function call, the with statement is a joke, call works a little better, but it still breaks for anything declared in function scope but not in global scope, therefore I would declare these few cases in global scope, but that requires me to know what they are. The IE function execScript makes a complete shift, but that only solves the problem for IE.
Note: To anyone loading JavaScript dynamically, setTimeout(code,1) is a simple effective hack to achieve global scope, but it will not execute immediately.
No, there's no way to reference the variable object of the execution context of a function binding object of the variable environment of the execution context (that's what that thing is called [now; hence the strikethrough]; details in §10.3 of the specification). You can only access the limited view to it you get with arguments (which is very limited indeed).
Usually when I've wanted to do this, I've just put everything I wanted on an object and then used that (e.g., passed it into a function). Of course, any functions created within the context have access to everything in scope where they're created, as they "close over" the context; more: Closures are not complicated.
I know this is hugely late, and you're probably not even slightly interested any more, but I was interested in the feasibility of this too and you should be able to make a work around of some sort using:
(function(global) {
var testVar = 1;
global.scope = function(s) {
return eval(s);
}
})(this);
then running:
scope('testVar'); // 1
returns the variable from within the closure. Not particularly nice, but theoretically possible to wrap that in an object, perhaps using some validation and getters and setters if you needed?
Edit: Having re-read the question, I assume you'd want to access it without having to specify a function in the scope itself, so this probably isn't applicable. I'll leave this here anyway.
Certain versions of Netscape had a magic property in the arguments object that did what you're looking for. (I can't remember what it was called)
What about something like this?
<script type="text/javascript">
var test = {
bar : 1,
foo : function () {
alert(this.bar);
}
}
test.foo();
</script>
You don't need a keyword to reference a variable in the local scope, because it's the scope you're in.