Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 8 years ago.
Improve this question
I've been looking at the source code of my.class.js to find out what makes it so fast on Firefox. Here's the snippet of code used to create a class:
my.Class = function () {
var len = arguments.length;
var body = arguments[len - 1];
var SuperClass = len > 1 ? arguments[0] : null;
var hasImplementClasses = len > 2;
var Class, SuperClassEmpty;
if (body.constructor === Object) {
Class = function () {};
} else {
Class = body.constructor;
delete body.constructor;
}
if (SuperClass) {
SuperClassEmpty = function() {};
SuperClassEmpty.prototype = SuperClass.prototype;
Class.prototype = new SuperClassEmpty();
Class.prototype.constructor = Class;
Class.Super = SuperClass;
extend(Class, SuperClass, false);
}
if (hasImplementClasses)
for (var i = 1; i < len - 1; i++)
extend(Class.prototype, arguments[i].prototype, false);
extendClass(Class, body);
return Class;
};
The extend function is simply used to copy the properties of the second object onto the first (optionally overriding existing properties):
var extend = function (obj, extension, override) {
var prop;
if (override === false) {
for (prop in extension)
if (!(prop in obj))
obj[prop] = extension[prop];
} else {
for (prop in extension)
obj[prop] = extension[prop];
if (extension.toString !== Object.prototype.toString)
obj.toString = extension.toString;
}
};
The extendClass function copies all the static properties onto the class, as well as all the public properties onto the prototype of the class:
var extendClass = my.extendClass = function (Class, extension, override) {
if (extension.STATIC) {
extend(Class, extension.STATIC, override);
delete extension.STATIC;
}
extend(Class.prototype, extension, override);
};
This is all pretty straightforward. When you create a class, it simply returns the constructor function you provide it.
What beats my understanding however is how does creating an instance of this constructor execute faster than creating an instance of the same constructor written in Vapor.js.
This is what I'm trying to understand:
How do constructors of libraries like my.class.js create so many instances so quickly on Firefox? The constructors of the libraries are all very similar. Shouldn't the execution time also be similar?
Why does the way the class is created affect the execution speed of instantiation? Aren't definition and instantiation separate processes?
Where is my.class.js gaining this speed boost from? I don't see any part of the constructor code which should make it execute any faster. In fact traversing a long prototype chain like MyFrenchGuy.Super.prototype.setAddress.call should slow it down significantly.
Is the constructor function being JIT compiled? If so then why aren't the constructor functions of other libraries also being JIT compiled?
I don't mean to offend anyone, but this sort of thing really isn't worth the attention, IMHO. Almost any speed-difference between browsers is down to the JS engine. The V8 engine is very good at memory management, for example; especially when you compare it to IE's JScript engines of old.
Consider the following:
var closure = (function()
{
var closureVar = 'foo',
someVar = 'bar',
returnObject = {publicProp: 'foobar'};
returnObject.getClosureVar = function()
{
return closureVar;
};
return returnObject;
}());
Last time I checked, chrome actually GC'ed someVar, because it wasn't being referenced by the return value of the IIFE (referenced by closure), whereas both FF and Opera kept the entire function scope in memory.
In this snippet, it doesn't really matter, but for libs that are written using the module-pattern (AFAIK, that's pretty much all of them) that consist of thousands of lines of code, it can make a difference.
Anyway, modern JS-engines are more than just "dumb" parse-and-execute things. As you said: there's JIT compilation going on, but there's also a lot of trickery involved to optimize your code as much as possible. It could very well be that the snippet you posted is written in a way that FF's engine just loves.
It's also quite important to remember that there is some sort of speed-battle going on between Chrome and FF about who has the fastest engine. Last time I checked Mozilla's Rhino engine was said to outperform Google's V8, if that still holds true today, I can't say... Since then, both Google and Mozilla have been working on their engines...
Bottom line: speed differences between various browsers exist - nobody can deny that, but a single point of difference is insignificant: you'll never write a script that does just one thing over and over again. It's the overall performance that matters.
You have to keep in mind that JS is a tricky bugger to benchmark, too: just open your console, write some recursive function, and rung it 100 times, in FF and Chrome. compare the time it takes for each recursion, and the overall run. Then wait a couple of hours and try again... sometimes FF might come out on top, whereas other times Chrome might be faster, still. I've tried it with this function:
var bench = (function()
{
var mark = {start: [new Date()],
end: [undefined]},
i = 0,
rec = function(n)
{
return +(n === 1) || rec(n%2 ? n*3+1 : n/2);
//^^ Unmaintainable, but fun code ^^\\
};
while(i++ < 100)
{//new date at start, call recursive function, new date at end of recursion
mark.start[i] = new Date();
rec(1000);
mark.end[i] = new Date();
}
mark.end[0] = new Date();//after 100 rec calls, first element of start array vs first of end array
return mark;
}());
But now, to get back to your initial question(s):
First off: the snippet you provided doesn't quite compare to, say, jQuery's $.extend method: there's no real cloning going on, let alone deep-cloning. It doesn't check for circular references at all, which most other libs I've looked into do. checking for circular references does slow the entire process down, but it can come in handy from time to time (example 1 below). Part of the performance difference could be explained by the fact that this code simply does less, so it needs less time.
Secondly: Declaring a constructor (classes don't exist in JS) and creating an instance are, indeed, two different things (though declaring a constructor is in itself creating an instance of an object (a Function instance to be exact). The way you write your constructor can make a huge difference, as shown in example 2 below. Again, this is a generalization, and might not apply to certain use-cases on certain engines: V8, for example, tends to create a single function object for all instances, even if that function is part of the constructor - or so I'm told.
Thirdly: Traversing a long prototype-chain, as you mention is not as unusual as you might think, far from it, actually. You're constantly traversing chains of 2 or three prototypes, as shown in example 3. This shouldn't slow you down, as it's just inherent to the way JS resolves function calls or resolves expressions.
Lastly: It's probably being JIT-compiled, but saying that other libs aren't JIT-compiled just doesn't stack up. They might, then again, they might not. As I said before: different engines perform better at some tasks then other... it might be the case that FF JIT-compiles this code, and other engines don't.
The main reason I can see why other libs wouldn't be JIT-compiled are: checking for circular references, deep cloning capabilities, dependencies (ie extend method is used all over the place, for various reasons).
example 1:
var shallowCloneCircular = function(obj)
{//clone object, check for circular references
function F(){};
var clone, prop;
F.prototype = obj;
clone = new F();
for (prop in obj)
{//only copy properties, inherent to instance, rely on prototype-chain for all others
if (obj.hasOwnProperty(prop))
{//the ternary deals with circular references
clone[prop] = obj[prop] === obj ? clone : obj[prop];//if property is reference to self, make clone reference clone, not the original object!
}
}
return clone;
};
This function clones an object's first level, all objects that are being referenced by a property of the original object, will still be shared. A simple fix would be to simply call the function above recursively, but then you'll have to deal with the nasty business of circular references at all levels:
var circulars = {foo: bar};
circulars.circ1 = circulars;//simple circular reference, we can deal with this
circulars.mess = {gotcha: circulars};//circulars.mess.gotcha ==> circular reference, too
circulars.messier = {messiest: circulars.mess};//oh dear, this is hell
Of course, this isn't the most common of situations, but if you want to write your code defensively, you have to acknowledge the fact that many people write mad code all the time...
Example 2:
function CleanConstructor()
{};
CleanConstructor.prototype.method1 = function()
{
//do stuff...
};
var foo = new CleanConstructor(),
bar = new CleanConstructor);
console.log(foo === bar);//false, we have two separate instances
console.log(foo.method1 === bar.method1);//true: the function-object, referenced by method1 has only been created once.
//as opposed to:
function MessyConstructor()
{
this.method1 = function()
{//do stuff
};
}
var foo = new MessyConstructor(),
bar = new MessyConstructor();
console.log(foo === bar);//false, as before
console.log(foo.method1 === bar.method1);//false! for each instance, a new function object is constructed, too: bad performance!
In theory, declaring the first constructor is slower than the messy way: the function object, referenced by method1 is created before a single instance has been created. The second example doesn't create a method1, except for when the constructor is called. But the downsides are huge: forget the new keyword in the first example, and all you get is a return value of undefined. The second constructor creates a global function object when you omit the new keyword, and of course creates new function objects for each call. You have a constructor (and a prototype) that is, in fact, idling... Which brings us to example 3
example 3:
var foo = [];//create an array - empty
console.log(foo[123]);//logs undefined.
Ok, so what happens behind the scenes: foo references an object, instance of Array, which in turn inherits form the Object prototype (just try Object.getPrototypeOf(Array.prototype)). It stands to reason, therefore that an Array instance works in pretty much the same way as any object, so:
foo[123] ===> JS checks instance for property 123 (which is coerced to string BTW)
|| --> property not found #instance, check prototype (Array.prototype)
===========> Array.prototype.123 could not be found, check prototype
||
==========> Object.prototype.123: not found check prototype?
||
=======>prototype is null, return undefined
In other words, a chain like you describe isn't too far-fetched or uncommon. It's how JS works, so expecting that to slow things down is like expecting your brain to fry because your thinking: yes, you can get worn out by thinking too much, but just know when to take a break. Just like in the case of prototype-chains: their great, just know that they are a tad slower, yes...
I'm not entirely sure, but I do know that when programming, it is good practice to make the code as small as possible without sacrificing functionality. I like to call it minimalist code.
This can be a good reason to obfuscate code. Obfuscation shrinks the size of the file by using smaller method and variable names, making it harder to reverse-engineer, shrinking the file size, making it faster to download, as well as a potential performance boost. Google's javascript code is intensely obfuscated, and that contributes to their speed.
So in JavaScript, bigger isn't always better. When I find a way I can shrink my code, I implement it immediately, because I know it will benefit performance, even if by the smallest amount.
For example, using the var keyword in a function where the variable isn't needed outside the function helps garbage collection, which provides a very small speed boost versus keeping the variable in memory.
With a library like this this that produces "millions of operations per second" (Blaise's words), small performance boosts can add up to a noticeable/measurable difference.
So it is possible that my.class.js is "minimalist coded" or optimized in some manner. It could even be the var keywords.
I hope this helped somewhat. If it didn't help, then I wish you luck in getting a good answer.
Related
In Bluebird's util.js file, it has the following function:
function toFastProperties(obj) {
/*jshint -W027*/
function f() {}
f.prototype = obj;
ASSERT("%HasFastProperties", true, obj);
return f;
eval(obj);
}
For some reason, there's a statement after the return function, which I'm not sure why it's there.
As well, it seems that it is deliberate, as the author had silenced the JSHint warning about this:
Unreachable 'eval' after 'return'. (W027)
What exactly does this function do? Does util.toFastProperties really make an object's properties "faster"?
I've searched through Bluebird's GitHub repository for any comments in the source code or an explanation in their list of issues, but I couldn't find any.
2017 update: First, for readers coming today - here is a version that works with Node 7 (4+):
function enforceFastProperties(o) {
function Sub() {}
Sub.prototype = o;
var receiver = new Sub(); // create an instance
function ic() { return typeof receiver.foo; } // perform access
ic();
ic();
return o;
eval("o" + o); // ensure no dead code elimination
}
Sans one or two small optimizations - all the below is still valid.
Let's first discuss what it does and why that's faster and then why it works.
What it does
The V8 engine uses two object representations:
Dictionary mode - in which object are stored as key - value maps as a hash map.
Fast mode - in which objects are stored like structs, in which there is no computation involved in property access.
Here is a simple demo that demonstrates the speed difference. Here we use the delete statement to force the objects into slow dictionary mode.
The engine tries to use fast mode whenever possible and generally whenever a lot of property access is performed - however sometimes it gets thrown into dictionary mode. Being in dictionary mode has a big performance penalty so generally it is desirable to put objects in fast mode.
This hack is intended to force the object into fast mode from dictionary mode.
Bluebird's Petka himself talks about it here.
These slides (wayback machine) by Vyacheslav Egorov also mentions it.
The question "*https://stackoverflow.com/questions/23455678/pros-and-cons-of-dictionary-mode*" and its accepted answer are also related.
This slightly outdated article is still a fairly good read that can give you a good idea on how objects are stored in v8.
Why it's faster
In JavaScript prototypes typically store functions shared among many instances and rarely change a lot dynamically. For this reason it is very desirable to have them in fast mode to avoid the extra penalty every time a function is called.
For this - v8 will gladly put objects that are the .prototype property of functions in fast mode since they will be shared by every object created by invoking that function as a constructor. This is generally a clever and desirable optimization.
How it works
Let's first go through the code and figure what each line does:
function toFastProperties(obj) {
/*jshint -W027*/ // suppress the "unreachable code" error
function f() {} // declare a new function
f.prototype = obj; // assign obj as its prototype to trigger the optimization
// assert the optimization passes to prevent the code from breaking in the
// future in case this optimization breaks:
ASSERT("%HasFastProperties", true, obj); // requires the "native syntax" flag
return f; // return it
eval(obj); // prevent the function from being optimized through dead code
// elimination or further optimizations. This code is never
// reached but even using eval in unreachable code causes v8
// to not optimize functions.
}
We don't have to find the code ourselves to assert that v8 does this optimization, we can instead read the v8 unit tests:
// Adding this many properties makes it slow.
assertFalse(%HasFastProperties(proto));
DoProtoMagic(proto, set__proto__);
// Making it a prototype makes it fast again.
assertTrue(%HasFastProperties(proto));
Reading and running this test shows us that this optimization indeed works in v8. However - it would be nice to see how.
If we check objects.cc we can find the following function (L9925):
void JSObject::OptimizeAsPrototype(Handle<JSObject> object) {
if (object->IsGlobalObject()) return;
// Make sure prototypes are fast objects and their maps have the bit set
// so they remain fast.
if (!object->HasFastProperties()) {
MigrateSlowToFast(object, 0);
}
}
Now, JSObject::MigrateSlowToFast just explicitly takes the Dictionary and converts it into a fast V8 object. It's a worthwhile read and an interesting insight into v8 object internals - but it's not the subject here. I still warmly recommend that you read it here as it's a good way to learn about v8 objects.
If we check out SetPrototype in objects.cc, we can see that it is called in line 12231:
if (value->IsJSObject()) {
JSObject::OptimizeAsPrototype(Handle<JSObject>::cast(value));
}
Which in turn is called by FuntionSetPrototype which is what we get with .prototype =.
Doing __proto__ = or .setPrototypeOf would have also worked but these are ES6 functions and Bluebird runs on all browsers since Netscape 7 so that's out of the question to simplify code here. For example, if we check .setPrototypeOf we can see:
// ES6 section 19.1.2.19.
function ObjectSetPrototypeOf(obj, proto) {
CHECK_OBJECT_COERCIBLE(obj, "Object.setPrototypeOf");
if (proto !== null && !IS_SPEC_OBJECT(proto)) {
throw MakeTypeError("proto_object_or_null", [proto]);
}
if (IS_SPEC_OBJECT(obj)) {
%SetPrototype(obj, proto); // MAKE IT FAST
}
return obj;
}
Which directly is on Object:
InstallFunctions($Object, DONT_ENUM, $Array(
...
"setPrototypeOf", ObjectSetPrototypeOf,
...
));
So - we have walked the path from the code Petka wrote to the bare metal. This was nice.
Disclaimer:
Remember this is all implementation detail. People like Petka are optimization freaks. Always remember that premature optimization is the root of all evil 97% of the time. Bluebird does something very basic very often so it gains a lot from these performance hacks - being as fast as callbacks isn't easy. You rarely have to do something like this in code that doesn't power a library.
V8 developer here. The accepted answer is a great explanation, I just wanted to highlight one thing: the so-called "fast" and "slow" property modes are unfortunate misnomers, they each have their pros and cons. Here is a (slightly simplified) overview of the performance of various operations:
struct-like properties
dictionary properties
adding a property to an object
--
+
deleting a property
---
+
reading/writing a property, first time
-
+
reading/writing, cached, monomorphic
+++
+
reading/writing, cached, few shapes
++
+
reading/writing, cached, many shapes
--
+
colloquial name
"fast"
"slow"
So as you can see, dictionary properties are actually faster for most of the lines in this table, because they don't care what you do, they just handle everything with solid (though not record-breaking) performance. Struct-like properties are blazing fast for one particular situation (reading/writing the values of existing properties, where every individual place in the code only sees very few distinct object shapes), but the price they pay for that is that all other operations, in particular those that add or remove properties, become much slower.
It just so happens that the special case where struct-like properties have their big advantage (+++) is particularly frequent and really important for many apps' performance, which is why they acquired the "fast" moniker. But it's important to realize that when you delete properties and V8 switches the affected objects to dictionary mode, then it isn't being dumb or trying to be annoying: rather it attempts to give you the best possible performance for what you're doing. We have landed patches in the past that have achieved significant performance improvements by making more objects go to dictionary ("slow") mode sooner when appropriate.
Now, it can happen that your objects would generally benefit from struct-like properties, but something your code does causes V8 to transition them to dictionary properties, and you'd like to undo that; Bluebird had such a case. Still, the name toFastProperties is a bit misleading in its simplicity; a more accurate (though unwieldy) name would be spendTimeOptimizingThisObjectAssumingItsPropertiesWontChange, which would indicate that the operation itself is costly, and it only makes sense in certain limited cases. If someone took away the conclusion "oh, this is great, so I can happily delete properties now, and just call toFastProperties afterwards every time", then that would be a major misunderstanding and cause pretty bad performance degradation.
If you stick with a few simple rules of thumb, you'll never have a reason to even try to force any internal object representation changes:
Use constructors, and initialize all properties in the constructor. (This helps not only your engine, but also understandability and maintainability of your code. Consider that TypeScript doesn't quite force this but strongly encourages it, because it helps engineering productivity.)
Use classes or prototypes to install methods, don't just slap them onto each object instance. (Again, this is a common best practice for many reasons, one of them being that it's faster.)
Avoid delete. When properties come and go, prefer using a Map over the ES5-era "object-as-map" pattern. When an object can toggle into and out of a certain state, prefer boolean (or equivalent) properties (e.g. o.has_state = true; o.has_state = false;) over adding and deleting an indicator property.
When it comes to performance, measure, measure, measure. Before you start sinking time into performance improvements, profile your app to see where the hotspots are. When you implement a change that you hope will make things faster, verify with your real app (or something extremely close to it; not just a 10-line microbenchmark!) that it actually helps.
Lastly, if your team lead tells you "I've heard that there are 'fast' and 'slow' properties, please make sure that all of ours are 'fast'", then point them at this post :-)
Reality from 2021 (NodeJS version 12+).
Seems like a huge optimization is done, objects with deleted fields and sparse arrays don't become slow. Or I'm missing smth?
// run in Node with enabled flag
// node --allow-natives-syntax script.js
function Point(x, y) {
this.x = x;
this.y = y;
}
var obj1 = new Point(1, 2);
var obj2 = new Point(3, 4);
delete obj2.y;
var arr = [1,2,3]
arr[100] = 100
console.log('obj1 has fast properties:', %HasFastProperties(obj1));
console.log('obj2 has fast properties:', %HasFastProperties(obj2));
console.log('arr has fast properties:', %HasFastProperties(arr));
both show true
obj1 has fast properties: true
obj2 has fast properties: true
arr has fast properties: true
// run in Node with enabled flag
// node --allow-natives-syntax script.js
function Point(x, y) {
this.x = x;
this.y = y;
}
var obj2 = new Point(3, 4);
console.log('obj has fast properties:', %HasFastProperties(obj2)) // true
delete obj2.y;
console.log('obj2 has fast properties:', %HasFastProperties(obj2)); //true
var obj = {x : 1, y : 2};
console.log('obj has fast properties:', %HasFastProperties(obj)) //true
delete obj.x;
console.log('obj has fast properties:', %HasFastProperties(obj)); //fasle
Function and object look different
Is there any difference between the run speeds of a constructor function when compared to an equivalent object initializer?
For example
function blueprint(var1, var2){
this.property1 = var1;
this.property2 = var2;
}
var object1 = new blueprint(value1,value2);
vs
object1 = {property1:value1, property2:value2};
If there is, is it relevant enough to be of concern when optimizing code or would file size take priority?
If there is, is it relevant enough to be of concern when optimizing code or would file size take priority?
Neither.
It's extremely rare for decisions like this to have any (positive) effect on the system performance. Even if current browsers (or whatever your execution environment) show an observable advantage one way or another, that difference is not terribly likely to persist over new releases.
"It's much easier to optimize correct code than to correct optimized code."
Write readable, maintainable code and when it is all correct, check to see whether it is objectionably slow or the files are unreasonably large and make the optimizations.
Ran in console:
function blueprint(var1, var2){
this.property1 = var1;
this.property2 = var2;
}
var start = new Date();
var stop;
var object1;
for (var i = 0; i < 1000000; i++) {
object1 = new blueprint(1,1);
}
stop = new Date();
console.log(stop - start);
Results...
Google Chrome: 2832 milliseconds
Firefox 3.6.17: 3441 milliseconds
Ran in console:
var start = new Date();
var stop;
var object1;
for (var i = 0; i < 1000000; i++) {
object1 = {
'property1': 1,
'property2': 1
};
}
stop = new Date();
console.log(stop - start);
Results...
Google Chrome: 2302 milliseconds
Firefox 3.6.17: 2285 milliseconds
Offhand, it's pretty obvious which one is faster. However, unless you are going through a significant amount of operations I think you should use whatever is more readable and not worry about it.
I think object intializer will be faster than using constructor because constructor has a function call and it has to maintain its own instance too.
As a side note, use constructor if you want to create multiple instances of similar objects other wise go for object initializer if only single object is required.
Using a constructor to create a trivial object with just value properties is counter-productive. Just creating a simple object literal from scratch each time is faster. You can always define a function if it is to be called from lots of different places. Hey you just created a basic constructor function :lol:
If your object becomes non-trivial, for example including getters, setters, or full-blown methods, then a constructor (with the javascript in a prototype to be shared) is orders of magnitude faster than creating an object from scratch. Of course you are talking about a few micro-seconds (on a typical desktop) for creating an object with a small amount of embedded javascript vs less than a microsecond for calling a constructor, so in most cases it isn't important. Creating an object with only value properties is another order of magnitude faster.
Remember also that the initial creation of the constructor is an expensive operation, which may be more important if it is only to be used a few times. In some cases the constructor can be pre-compiled, for example if it is defined in a javascript code module in a Firefox addon, and then it is a win-win.
There are also more formal methods for creating objects such as the Object.create() function. However this is complicated and cumbersome to use and doesn't appear to be well optimised in any current browser. In all the tests I've run it is desperately slow compared to other methods, but might be useful when you need advanced capabilities and aren't going to be calling it hundreds of times.
The constructor function is used for multiple entries under the same "object".
The object initializer should only be used for a limited amount of entries, for example 3.
The constructor function is faster for multiple entries while the ...
object initializer is faster for few entries, at least in theory, I have not tested the speeds because I doubt the difference is catastrophic.
I wouldn't worry about it. The overhead of the constructor is an additional function call and a few extra properties to set (like the prototype). With modern JIT engines, it should hardly matter.
For a long time I have been throwing around the idea of making my JavaScript more object oriented. I have looked at a few different implementations of this as well but I just cannot decide if it is necessary or not.
What I am trying to answer are the following questions
Is John Resig's simple inheritance structure safe to use for production?
Is there any way to be able to tell how well it has been tested?
Besides Joose what other choices do I have for this purpose? I need one that is easy to use, fast, and robust. It also needs to be compatible with jQuery.
Huh. It looks much more complicated than it needs to be, to me.
Actually looking more closely I really take exception to what it is doing with providing this._super() whilst in a method, to call the superclass method.
The code introduces a reliance on typeof==='function' (unreliable for some objects), Function#toString (argh, function decomposition is also unreliable), and deciding whether to wrap based on whether you've used the sequence of bytes _super in the function body (even if you've only used it in a string. and if you try eg. this['_'+'super'] it'll fail).
And if you're storing properties on your function objects (eg MyClass.myFunction.SOME_PRIVATE_CONSTANT, which you might do to keep namespaces clean) the wrapping will stop you from getting at those properties. And if an exception is thrown in a method and caught in another method of the same object, _super will end up pointing at the wrong thing.
All this is just to make calling your superclass's method-of-the-same name easier. But I don't think that's especially hard to do in JS anyway. It's too clever for its own good, and in the process making the whole less reliable. (Oh, and arguments.callee isn't valid in Strict Mode, though that's not really his fault since that occurred after he posted it.)
Here's what I'm using for classes at the moment. I don't claim that this is the “best” JS class system, because there are loads of different ways of doing it and a bunch of different features you might want to add or not add. But it's very lightweight and aims at being ‘JavaScriptic’, if that's a word. (It isn't.)
Function.prototype.makeSubclass= function() {
function Class() {
if (!(this instanceof Class))
throw 'Constructor function requires new operator';
if ('_init' in this)
this._init.apply(this, arguments);
}
if (this!==Object) {
Function.prototype.makeSubclass.nonconstructor.prototype= this.prototype;
Class.prototype= new Function.prototype.makeSubclass.nonconstructor();
}
return Class;
};
Function.prototype.makeSubclass.nonconstructor= function() {};
It provides:
protection against accidental missing new. The alternative is to silently redirect X() to new X() so missing new works. It's a toss-up which is best; I went for explicit error so that one doesn't get used to writing without new and causing problems on other objects not defined like that. Either way is better than the unacceptable JS default of letting this. properties fall onto window and mysteriously going wrong later.
an inheritable _init method, so you don't have to write a constructor-function that does nothing but call the superclass constructor function.
and that's really all.
Here's how you might use it to implement Resig's example:
var Person= Object.makeSubclass();
Person.prototype._init= function(isDancing) {
this.dancing= isDancing;
};
Person.prototype.dance= function() {
return this.dancing;
};
var Ninja = Person.makeSubclass();
Ninja.prototype._init= function() {
Person.prototype._init.call(this, false);
};
Ninja.prototype.swingSword= function() {
return true;
};
var p= new Person(true);
p.dance(); // => true
var n = new Ninja();
n.dance(); // => false
n.swingSword(); // => true
// Should all be true
p instanceof Person &&
n instanceof Ninja && n instanceof Person
Superclass-calling is done by specifically naming the method you want and calling it, a bit like in Python. You could add a _super member to the constructor function if you wanted to avoid naming Person again (so you'd say Ninja._super.prototype._init.call, or perhaps Ninja._base._init.call).
JavaScript is prototype based and not class based. My recommendation is not to fight it and declare subtypes the JS way:
MyDerivedObj.prototype = new MySuperObj();
MyDerivedObj.prototype.constructor = MyDerivedObj;
See how far you can get without using inheritance at all. Treat it as a performance hack (to be applied reluctantly where genuinely necessary) rather than a design principle.
In an a highly dynamic language like JS, it is rarely necessary to know whether an object is a Person. You just need to know if it has a firstName property or an eatFood method. You don't often need to know if an object is an array; if it has a length property and some other properties named after integers, that's usually good enough (e.g. the Arguments object). "If it walks like a duck and quacks like a duck, it's a duck."
// give back a duck
return {
walk: function() { ... },
quack: function() { ... }
};
Yes, if you're making very large numbers of small objects, each with dozens of methods, then by all means assign those methods to the prototype to avoid the overhead of creating dozens of slots in every instance. But treat that as a way of reducing memory overhead - a mere optimisation. And do your users a favour by hiding your use of new behind some kind of factory function, so they don't even need to know how the object is created. They just need to know it has method foo or property bar.
(And note that you won't really be modelling classical inheritance in that scenario. It's merely the equivalent of defining a single class to get the efficiency of a shared vtable.)
Here's a ugly bit of Javascript it would be nice to find a workaround.
Javascript has no classes, and that is a good thing. But it implements fallback between objects in a rather ugly way. The foundational construct should be to have one object that, when a property fails to be found, it falls back to another object.
So if we want a to fall back to b we would want to do something like:
a = {sun:1};
b = {dock:2};
a.__fallback__ = b;
then
a.dock == 2;
But, Javascript instead provides a new operator and prototypes. So we do the far less elegant:
function A(sun) {
this.sun = sun;
};
A.prototype.dock = 2;
a = new A(1);
a.dock == 2;
But aside from elegance, this is also strictly less powerful, because it means that anything created with A gets the same fallback object.
What I would like to do is liberate Javascript from this artificial limitation and have the ability to give any individual object any other individual object as its fallback. That way I could keep the current behavior when it makes sense, but use object-level inheritance when that makes sense.
My initial approach is to create a dummy constructor function:
function setFallback(from_obj, to_obj) {
from_obj.constructor = function () {};
from_obj.constructor.prototype = to_obj;
}
a = {sun:1};
b = {dock:2};
setFallback(a, b);
But unfortunately:
a.dock == undefined;
Any ideas why this doesn't work, or any solutions for an implementation of setFallback?
(I'm running on V8, via node.js, in case this is platform dependent)
Edit:
I've posted a partial solution to this below, that works in the case of V8, but isn't general. I'd still appreciate a more general solution.
You could just use Object.create. It's part of ES5 so it's already available natively in some browsers. I believe it does exactly what you want.
Okay, some more research and cross-platform checking and there's some more information (though not a general solution).
Some implementations have basically what I did for my __fallback__. It is called __proto__ and is about perfect:
a = {sun:1};
b = {dock:2};
a.__proto__ = b;
a.dock == 2;
It seems that, what happens in when a new object is constructed is roughly this:
a = new Constructor(...args...);
produces behavior roughly equivalent to:
object.constructor = constructor;
object.__proto__ = constructor.prototype;
constructor.call(this, ...args...);
So it is no wonder that coming along later and adjusting an object's constructor or constructor.prototype has no effect, because the __proto__ setting is already set.
Now for my v8 application, I can just use __proto__, but I understand it that this isn't exposed on the IE VM (I don't run windows, so I can't tell). So it is not a general solution to the problem.
In another question, a user pointed out that the new keyword was dangerous to use and proposed a solution to object creation that did not use new. I didn't believe that was true, mostly because I've used Prototype, Script.aculo.us and other excellent JavaScript libraries, and everyone of them used the new keyword.
In spite of that, yesterday I was watching Douglas Crockford's talk at YUI theater and he said the exactly same thing, that he didn't use the new keyword anymore in his code (Crockford on JavaScript - Act III: Function the Ultimate - 50:23 minutes).
Is it 'bad' to use the new keyword? What are the advantages and disadvantages of using it?
Crockford has done a lot to popularize good JavaScript techniques. His opinionated stance on key elements of the language have sparked many useful discussions. That said, there are far too many people that take each proclamation of "bad" or "harmful" as gospel, refusing to look beyond one man's opinion. It can be a bit frustrating at times.
Use of the functionality provided by the new keyword has several advantages over building each object from scratch:
Prototype inheritance. While often looked at with a mix of suspicion and derision by those accustomed to class-based OO languages, JavaScript's native inheritance technique is a simple and surprisingly effective means of code re-use. And the new keyword is the canonical (and only available cross-platform) means of using it.
Performance. This is a side-effect of #1: if I want to add 10 methods to every object I create, I could just write a creation function that manually assigns each method to each new object... Or, I could assign them to the creation function's prototype and use new to stamp out new objects. Not only is this faster (no code needed for each and every method on the prototype), it avoids ballooning each object with separate properties for each method. On slower machines (or especially, slower JS interpreters) when many objects are being created this can mean a significant savings in time and memory.
And yes, new has one crucial disadvantage, ably described by other answers: if you forget to use it, your code will break without warning. Fortunately, that disadvantage is easily mitigated - simply add a bit of code to the function itself:
function foo()
{
// if user accidentally omits the new keyword, this will
// silently correct the problem...
if ( !(this instanceof foo) )
return new foo();
// constructor logic follows...
}
Now you can have the advantages of new without having to worry about problems caused by accidentally misuse.
John Resig goes into detail on this technique in his Simple "Class" Instantiation post, as well as including a means of building this behavior into your "classes" by default. Definitely worth a read... as is his upcoming book, Secrets of the JavaScript Ninja, which finds hidden gold in this and many other "harmful" features of the JavaScript language (the chapter on with is especially enlightening for those of us who initially dismissed this much-maligned feature as a gimmick).
A general-purpose sanity check
You could even add an assertion to the check if the thought of broken code silently working bothers you. Or, as some commented, use the check to introduce a runtime exception:
if ( !(this instanceof arguments.callee) )
throw new Error("Constructor called as a function");
Note that this snippet is able to avoid hard-coding the constructor function name, as unlike the previous example it has no need to actually instantiate the object - therefore, it can be copied into each target function without modification.
ES5 taketh away
As Sean McMillan, stephenbez and jrh noted, the use of arguments.callee is invalid in ES5's strict mode. So the above pattern will throw an error if you use it in that context.
ES6 and an entirely harmless new
ES6 introduces Classes to JavaScript - no, not in the weird Java-aping way that old-school Crockford did, but in spirit much more like the light-weight way he (and others) later adopted, taking the best parts of prototypal inheritance and baking common patterns into the language itself.
...and part of that includes a safe new:
class foo
{
constructor()
{
// constructor logic that will ONLY be hit
// if properly constructed via new
}
}
// bad invocation
foo(); // throws,
// Uncaught TypeError: class constructors must be invoked with 'new'
But what if you don't want to use the new sugar? What if you just want to update your perfectly fine old-style prototypal code with the sort of safety checks shown above such that they keep working in strict mode?
Well, as Nick Parsons notes, ES6 provides a handy check for that as well, in the form of new.target:
function foo()
{
if ( !(new.target) )
throw new Error("Constructor called as a function");
// constructor logic follows...
}
So whichever approach you choose, you can - with a bit of thought and good hygiene - use new without harm.
I have just read some parts of Crockford's book "JavaScript: The Good Parts". I get the feeling that he considers everything that ever has bitten him as harmful:
About switch fall through:
I never allow switch cases to fall
through to the next case. I once found
a bug in my code caused by an
unintended fall through immediately
after having made a vigorous speech
about why fall through was sometimes
useful. (page 97, ISBN
978-0-596-51774-8)
About ++ and --:
The ++ (increment) and -- (decrement)
operators have been known to
contribute to bad code by encouraging
excessive trickiness. They are second
only to faulty architecture in
enabling viruses and other security
menaces. (page 122)
About new:
If you forget to include the new
prefix when calling a constructor
function, then this will not be
bound to the new object. Sadly, this
will be bound to the global object, so
instead of augmenting your new object,
you will be clobbering global
variables. That is really bad. There
is no compile warning, and there is no
runtime warning. (page 49)
There are more, but I hope you get the picture.
My answer to your question: No, it's not harmful. but if you forget to use it when you should you could have some problems. If you are developing in a good environment you notice that.
In the 5th edition of ECMAScript there is support for strict mode. In strict mode, this is no longer bound to the global object, but to undefined.
JavaScript being a dynamic language, there are a zillion ways to mess up where another language would stop you.
Avoiding a fundamental language feature such as new on the basis that you might mess up is a bit like removing your shiny new shoes before walking through a minefield just in case you might get your shoes muddy.
I use a convention where function names begin with a lowercase letter and 'functions' that are actually class definitions begin with an uppercase letter. The result is a really quite compelling visual clue that the 'syntax' is wrong:
var o = MyClass(); // This is clearly wrong.
On top of this, good naming habits help. After all, functions do things and therefore there should be a verb in its name whereas classes represent objects and are nouns and adjectives without any verb.
var o = chair() // Executing chair is daft.
var o = createChair() // Makes sense.
It's interesting how Stack Overflow's syntax colouring has interpreted the code above.
I am newbie to JavaScript so maybe I am just not too experienced in providing a good view point to this. Yet I want to share my view on this "new" thing.
I have come from the C# world where using the keyword "new" is so natural that it is the factory design pattern that looks weird to me.
When I first code in JavaScript, I don't realize that there is the "new" keyword and code like the one in YUI pattern and it doesn't take me long to run into disaster. I lose track of what a particular line is supposed to be doing when looking back the code I've written. More chaotic is that my mind can't really transit between object instances boundaries when I am "dry-running" the code.
Then, I found the "new" keyword which, to me, "separates" things. With the new keyword, it creates things. Without the new keyword, I know I won't confuse it with creating things unless the function I am invoking gives me strong clues of that.
For instance, with var bar=foo(); I don’t have any clues as what bar could possibly be.... Is it a return value or is it a newly created object? But with var bar = new foo(); I know for sure bar is an object.
Another case for new is what I call Pooh Coding. Winnie-the-Pooh follows his tummy. I say go with the language you are using, not against it.
Chances are that the maintainers of the language will optimize the language for the idioms they try to encourage. If they put a new keyword into the language they probably think it makes sense to be clear when creating a new instance.
Code written following the language's intentions will increase in efficiency with each release. And code avoiding the key constructs of the language will suffer with time.
And this goes well beyond performance. I can't count the times I've heard (or said) "why the hell did they do that?" when finding strange looking code. It often turns out that at the time when the code was written there was some "good" reason for it. Following the Tao of the language is your best insurance for not having your code ridiculed some years from now.
I wrote a post on how to mitigate the problem of calling a constructor without the new keyword.
It's mostly didactic, but it shows how you can create constructors that work with or without new and doesn't require you to add boilerplate code to test this in every constructor.
Constructors without using "new"
Here's the gist of the technique:
/**
* Wraps the passed in constructor so it works with
* or without the new keyword
* #param {Function} realCtor The constructor function.
* Note that this is going to be wrapped
* and should not be used directly
*/
function ctor(realCtor) {
// This is going to be the actual constructor
return function wrapperCtor() {
var obj; // The object that will be created
if (this instanceof wrapperCtor) {
// Called with new
obj = this;
} else {
// Called without new. Create an empty object of the
// correct type without running that constructor
surrogateCtor.prototype = wrapperCtor.prototype;
obj = new surrogateCtor();
}
// Call the real constructor function
realCtor.apply(obj, arguments);
return obj;
}
function surrogateCtor() {}
}
Here's how to use it:
// Create our point constructor
Point = ctor(function(x, y) {
this.x = x;
this.y = y;
});
// This is good
var pt = new Point(20, 30);
// This is OK also
var pt2 = Point(20, 30);
The rationale behind not using the new keyword, is simple:
By not using it at all, you avoid the pitfall that comes with accidentally omitting it. The construction pattern that YUI uses, is an example of how you can avoid the new keyword altogether:
var foo = function () {
var pub = { };
return pub;
}
var bar = foo();
Alternatively, you could do this:
function foo() { }
var bar = new foo();
But by doing so you run risk of someone forgetting to use the new keyword, and the this operator being all FUBAR. As far as I know, there isn't any advantage to doing this (other than you being used to it).
At The End Of The Day: It's about being defensive. Can you use the new statement? Yes. Does it make your code more dangerous? Yes.
If you have ever written C++, it's akin to setting pointers to NULL after you delete them.
I think "new" adds clarity to the code. And clarity is worth everything. It is good to know there are pitfalls, but avoiding them by avoiding clarity doesn't seem like the way for me.
Case 1: new isn't required and should be avoided
var str = new String('asd'); // type: object
var str = String('asd'); // type: string
var num = new Number(12); // type: object
var num = Number(12); // type: number
Case 2: new is required, otherwise you'll get an error
new Date().getFullYear(); // correct, returns the current year, i.e. 2010
Date().getFullYear(); // invalid, returns an error
Here is the briefest summary I could make of the two strongest arguments for and against using the new operator:
Arguments against new
Functions designed to be
instantiated as objects using the
new operator can have disastrous
effects if they are incorrectly
invoked as normal functions. A
function's code in such a case will
be executed in the scope where the
function is called, instead of in
the scope of a local object as
intended. This can cause global
variables and properties to get
overwritten with disastrous
consequences.
Finally, writing function Func(),
and then calling Func.prototype
and adding stuff to it so that you
can call new Func() to construct
your object seems ugly to some
programmers, who would rather use
another style of object inheritance
for architectural and stylistic
reasons.
For more on this argument check out Douglas Crockford's great and concise book JavaScript: The Good Parts. In fact, check it out anyway.
Arguments in favor of new
Using the new operator along with
prototypal assignment is fast.
That stuff about accidentally
running a constructor function's
code in the global namespace can
easily be prevented if you always
include a bit of code in your
constructor functions to check to
see if they are being called
correctly, and, in the cases where
they aren't, handling the call
appropriately as desired.
See John Resig's post for a simple explanation of this technique, and for a generally deeper explanation of the inheritance model he advocates.
I agree with PEZ and some here.
It seems obvious to me that "new" is self descriptive object creation, where the YUI pattern Greg Dean describes is completely obscured.
The possibility someone could write var bar = foo; or var bar = baz(); where baz isn't an object creating method seems far more dangerous.
I think new is evil, not because if you forget to use it by mistake it might cause problems, but because it screws up the inheritance chain, making the language tougher to understand.
JavaScript is prototype-based object-oriented. Hence every object must be created from another object like so: var newObj=Object.create(oldObj). Here oldObj is called the prototype of newObj (hence "prototype-based"). This implies that if a property is not found in newObj then it will be searched in oldObj. newObj by default will thus be an empty object, but due to its prototype chain, it appears to have all the values of oldObj.
On the other hand, if you do var newObj=new oldObj(), the prototype of newObj is oldObj.prototype, which is unnecessarily difficult to understand.
The trick is to use
Object.create=function(proto){
var F = function(){};
F.prototype = proto;
var instance = new F();
return instance;
};
It is inside this function and it is only here that new should be used. After this, simply use the Object.create() method. The method resolves the prototype problem.
In my not-so-humble opinion, "new" is a flawed concept in 2021 JavaScript. It adds words where none are needed. It makes the return value of a function/constructor implicit and forces the use of this in the function/constructor. Adding noise to code is never a good thing.
// With new
function Point(x, y) {
this.x = x
this.y = y
}
let point = new Point(0, 0)
Vs.
// Without new
function Point(x, y) {
return { x, y }
}
let point = Point(0, 0)