Closed. This question does not meet Stack Overflow guidelines. It is not currently accepting answers.
This question does not appear to be about programming within the scope defined in the help center.
Closed 8 years ago.
Improve this question
In javascript, there are two patterns and I would like to weight the benefits of using one vs the other. What is the difference between returning an object vs returning a function, for example:
var returnFunction = function(name,age){
var n = name;
var a = name;
return function(){
anotherFunc : function(){},
oneMoreFunc : function(){}
}
}
I returned a function containing two more functions, and access to private variables name and age. I understand that I can invoke returnfunction, and I know that I can use it like a constructor. I want to know, what are the benefits of this style vs:
var returnObject = function(name,age){
var n = name;
var a = age;
return {
anotherFunc:function(){},
oneMoreFunc:function(){},
};
}
Is there a performance penalty?
Is it just a matter of style?
Are there any benefits to one or the other, or am I just overthinking this?
edit:
With regards to option A, I was referencing this particular syntax from Javascript: The Good Parts
Function.prototype.method = function(name, func) {
this.prototype[name] = func;
return this;
};
String.method('deentityify', function() {
// The entity table. It maps entity names to
// characters.
var entity = {
quot: '"',
lt: '<',
gt: '>'
};
// Return the deentityify method.
return function() {
// This is the deentityify method. It calls the string
// replace method, looking for substrings that start
// with '&' and end with ';'. If the characters in
// between are in the entity table, then replace the
// entity with the character from the table. It uses
// a regular expression (Chapter 7).
return this.replace(/&([^&;]+);/g,
function(a, b) {
var r = entity[b];
return typeof r === 'string' ? r : a;
}
);
};
}());
Option A was a contrived example meant to replicate this syntax.
Variant A doesn't work. It's a syntax error.
So what you're really comparing:
var returnFunction = function(name,age){
var n = name;
var a = name;
// return a function that returns
return function(){
// logic to construct the object
var obj = {
anotherFunc : function(){},
oneMoreFunc : function(){}
}
// return the object
return obj;
}
}
// vs.
var returnObject = function(name,age){
var n = name;
var a = age;
// return the object directly
return {
anotherFunc:function(){},
oneMoreFunc:function(){},
};
}
It depends on what does the object look like.
In most cases, you'd go with option B. Just return a simple object.
I don't know anything other than V8, but in V8 it looks like this:
-> new scope
-> assign some vars
-> create a function
-> compile the code in that function
-> return the function, close the scope
-> run the function
-> new scope
-> create the object
-> return the object, close the scope
vs.
-> new scope
-> assign some vars
-> create the object
-> return the object, close the scope
Obviously the first has more steps, but the speed difference is insignificant.
However, there are cases in which it would simply be impractical to return a complex object with multiple nested properties and native functions that have to be initialized. Case in which it is much more useful to generate the object and return it on-a-need basis. Which is option A.
But, better than having option A, and if you intend to make interventions on that returned object, it's just nicer to make it a class:
var returnObjectClass = function(name,age){
this.name = name;
this.class = class;
this.anotherFunc = function(){};
this.oneMoreFunc = function(){};
return this;
}
You can read more here: http://www.phpied.com/3-ways-to-define-a-javascript-class/
Related
I am working with Javascript and Appdescriptors in JSON format.
What I want to do is creating a instance of a Class, where the classname is saved as string in oModelConf[sModelName].type. If that is not the case I want to take "sap.ui.model.odata.ODataModel"
Related Question offers this solution:
function instantiate(className, args) {
var o, f, c;
c = window[className]; // get reference to class constructor function
f = function(){}; // dummy function
f.prototype = c.prototype; // reference same prototype
o = new f(); // instantiate dummy function to copy prototype properties
c.apply(o, args); // call class constructor, supplying new object as context
o.constructor = c; // assign correct constructor (not f)
return o;
}
This is not a very good solution I think.
EDIT It does not work for me because my class is not defined on window, so window[className] is undefined. I do not know where my function is defined in SAPUI5
A second Solution:
eval(`a = new ${oModelConf[sModelName].type || "sap.ui.model.odata.ODataModel"}(sServiceUrl, true);`);
This is not a better solution because we should not use eval().
Are there any better solutions?
EDIT2
Because of the url in pimskies answer I found an other solution:
Since window.sap.ui.model.odata.ODataModel is the same as sap.ui.model.odata.ODataModel and window.sap is the same as window[sap]
I could take my string, and replace all . with ][, put the right brackets to front and end.
I will not code that because it is not a going solution.(I should not have coded the evalthing too...)
You could use jQuery.sap.getObject to access the class:
var ModelClass = jQuery.sap.getObject(oModelConf[sModelName].type || "sap.ui.model.odata.ODataModel");
var model = new ModelClass();
Edit: An other way (which i would recommend if you use AMD)
If you are using the modern AMD modules and you don't know if the module containing your class has already been loaded, you should use sap.ui.require() to load the module asynchronously. It requires the module to be specified via its unified resource name (the conversion is probably the most ugly part):
var className = oModelConf[sModelName].type || "sap.ui.model.odata.ODataModel";
var urn = className.replace(".", "/"); //Convert to unified resource name
sap.ui.require([urn],function(ModelClass){
//This function is called when the module is available
var model = new ModelClass();
...
});
Maybe map the string to a class?
function Foo() {
console.log('new foo');
}
function Bar() {
console.log('new bar');
}
var objects = {
'foo': Foo,
'bar': Bar
};
var cls = objects.foo || Bar;
new cls();
https://jsfiddle.net/ckd56d9v/1/
Or take a look at this answer: https://stackoverflow.com/a/9804142/5930258
What not combine the two? Since window[className] is failing, replace it with
oModelConf[className].type || sap.ui.model.odata.ODataModel...
function instantiate(className, args) {
var o, f, c;
c = oModelConf[className] || sap.ui.model.odata.ODataModel;
f = function(){}; // dummy function
f.prototype = c.prototype; // reference same prototype
o = new f(); // instantiate dummy function to copy prototype properties
c.apply(o, args); // call class constructor, supplying new object as context
o.constructor = c; // assign correct constructor (not f)
return o;
}
Since I can determine the number of arguments a function expects to have by calling its Function.length property, is there any way for me to programmatically create the right number of parameters to insert into that function at runtime? Example:
var xyz = function(a,b) {};
var bcd = function(a,b,c,d,e,f) { }; // vararg example
var doc = document, func_length = xyz.length;
doc.xyz = (function() {
return function(a,b,c,d,e) { /* works for one but not the other */ } }).call(doc);
/* would prefer to `return function(a,b)` with only 2 parameters, if it is
used for function `xyz` (though returning 5 works fine in this case), and to
`return function(a,b,c,d,e,f)` with 6 if used for function `bcd` (which does
not work with only five params). */
// thinking about xyz.apply(null,arguments)...but which arguments..? :(
// Returning function(a,b,c,d,e) does not support functions with more than five
// parameters...which would mostly be varargs - hence my question
// I am also well aware that I can use an object or an array instead of
// using many params.
/* This is for incorporating a user-defined function for use in my code, while
* allowing for my function to do 'other stuff' afterward. (And allowing for
* varargs, of course).
* Because coding something like: doc.xyz = xyz is inflexible */
As you can see, I don't know how to do this, or if it is even possible. The search bar hasn't given me any other questions like this one, otherwise I would not have asked...
NOTE: This answer is a product of misunderstanding but
may help the future visitors of this site.
Another way:
Do you really need to add parameters? Writing the function this way would be enough:
function foo(){ //No arguments predefined
a = arguments[0] || ""; //first argument or (if not defined) empty string
b = arguments[1] || ""; //second argument etc.
c = arguments[2] || ""; //third argument etc.
alert(a+b+c);
}
foo("Hello ", "world!");
This alerts "Hello world".
The solution you want:
The simplest way:
This is what you've asked for but it's not as simple as the previous solution.
You can define a meta function with all the parameters and a handler function that changes over the time.
(function(){ //Wrapper
var foo_meta = function(a,b,c,d){ //Local meta of foo
alert(a+b+c+d); //Do the code
};
window.foo = function(a,b){ //Global foo
return foo_meta(a,b,"","");
};
window.redefine_foo = function(){ //Global foo-changer
//Rewrites foo
window.foo = function(a,b,c){
return foo_meta(a,b,c,"");
};
};
})(); //Wrapper
//Do some code
foo("a","b");
redefine_foo(); //Rewrite foo
foo("a","b","c");
//Note that foo_meta is not defined here
foo_meta == undefined; //It's safe in the wrapper :)
This will alert "ab" and then "abc". For the meaning of wrapper function, see the references.
Reference:
Arguments array: http://goo.gl/FaLM1H
Wrapping code: http://goo.gl/uQ5sd0
If you send two parameters 6 and 7 to a function doWork(a,b,c,d,e),a=7 and b=6 will be automatically set and rest of the parameters will be ignored.
Why not just pass one object into the function and use JQuery extend.
e.g.
var parm =
{ x: 1, y : 2};
f(p) {
p = $_.extend({...defaults here}, p);
...
}
This is an example for joining the arguments, regardless of the number of arguments, to show how function arguments can be turned into an array and then processed like any other array.
function foo(){ //No arguments predefined
// convert to real array
var args = Array.prototype.slice.call(arguments);
// or if Array generics are available
var args = Array.slice(arguments);
console.log(args.join(' '));
}
foo('Hello', 'world!');
foo('Hello', 'wonderful', 'world!');
Here is the fiddle
Ref: arguments MDN
Well, I think I've figured it out at last. I've realized that there may be no way to 'truly' add a parameter to a function the way that I was asking, but there is a way to emulate the same result:
var doc = document;
var xyz = function(a,b) {};
var bcd = function(a,b,c,d,e,f) {};
var obj = {};
// Now, here it is (mostly (sort of)):
obj.userFunc = function(args) {
var args_Array = [];
for (var i=0;i < arguments.length; i++ ) {
args_Array.push(arguments[i])
}
xyz.apply(null,args_Array); // or 'this'. or 'undefined'. Whatever you want.
// do other stuff
return this; // we know what to do to make 'this' scope explicit
} // end
var thisFunc = 'xyz'
doc[thisFunc] = obj.userFunc;
doc.xyz('h','i');
doc.xyz('h','i','j');
doc.xyz('h','i','j','k');
doc.xyz('h','i').xyz('j','l').xyz('j','q'); // etc.
The trick was to use the arguments object, which conveniently assimilated all the parameters into a readily available object, push each value into an array then apply the function.
In case you're wondering what the fuss was all about, I wanted to completely incorporate a user-defined function into another function, while still being able to do 'other stuff' afterward. My previous example worked to an extent, but did not have support for varargs. This does.
This approach is greatly more flexible than: doc[thisFunc] = userDefinedFunction
:) 4/26/2014
I would like to do the following.I have a code like this:
var obj = {
method : function() {}
};
var func = function() {
return method(); //method is undefined here
};
func(); // What to do here?
Is it possible to call func in a way that it will see the method inside from obj as it was given for example as a parameter. I want to use obj.method inside func, without writing 'obj.' before and without modifying func itself. Is there any hack possible to achieve this?
In other words, is it possible to force obj as a closure into func?
I tried:
with(obj) {
func();
}
But it doesn't work. Anyone, any ideas? Or is it the only option to get the body of the function as string, put 'with(obj)' inside it and then create a new function out of it?
Clarification:
Because this code will be in a helper class 'eval' is OK. Which I don't want is the modification of the function through .toString(), because browsers implement it differently.
This is a solution, using eval (MDN):
var obj = {
method : function() {console.log("it workes!");}
};
var func = function() {
return method(); //method is undefined here
};
var newfunc = (function (obj, func) {
var method = obj.method;
eval("var f = " + func.toString());
return f;
}(obj, func));
newfunc(); //it workes
Basically you're just creating a new scope with a local variable called method and re-evaluating the function body in this scope. So you're basically creating a new function with the same body. I don't really like this approach and I wouldn't recommend it, but considering your constraints, it might be the only solution.
And yes, it still requires you to write obj.method, but not inside of func. So I figured, it should be ok.
EDIT
So here is a version, in which you don't have to specify the property name manually:
var newfunc = (function (__obj__, __func__) {
for (var __key__ in __obj__) {
if (__obj__.hasOwnProperty(__key__)) {
eval("var " + __key__ + " = " + __obj__[__key__]);
}
}
eval("var __f__ = " + func.toString());
return __f__;
}(obj, func));
This also done by using eval().
Note that I changed all remaining local variables to a names containing underscores, to minimize the probability of name collisions with properties inside obj.
Note also that not all valid property names are valid variable names. You could have an object like this:
var obj = {
"my func": function () {}
}
But if you would use this object you would generate a syntax error with the above method, because it would try to evaluate:
var my func = ...
As apsillers said in the comment section, it gets even worse if you don't have control over the properties of obj. In this case you shouldn't use eval at all, because you would make cross-site scripting attacks very easy (example from apsillers):
var obj = {
"a; alert('xss'); var b": function () {}
}
would evaluate to 3 different statements:
var a;
alert('xss');
var b = function () {};
This is not possible unless you define method separately:
var obj = {
method : function() {}
},
method = obj.method;
// rest of code
This is because the method reference inside func() assumes the window. namespace; thus, without modifying func() itself, it can't be done sanely.
More clarified version based on basilikum's answer, and I've found a simplification with 'with':
var obj = {
method : function() { return "it workes!"; }
};
var func = function() {
return method(); //method is undefined here
};
(function (obj, func) {
with(obj) {
eval("var __res__ = (" + func.toString() + ")()");
}
return __res__;
}(obj, func));
>> "It workes!"
In some Javascript code (node.js specifically), I need to call a function with an unknown set of arguments without changing the context. For example:
function fn() {
var args = Array.prototype.slice.call(arguments);
otherFn.apply(this, args);
}
The problem in the above is that when I call apply, I'm change the context by passing this as the first argument. I'd like to pass args to the function being called without changing the context of the function being called. I essentially want to do this:
function fn() {
var args = Array.prototype.slice.call(arguments);
otherFn.apply(<otherFn's original context>, args);
}
Edit: Adding more detail regarding my specific question. I am creating a Client class that contains a socket (socket.io) object among other info pertaining to a connection. I am exposing the socket's event listeners via the client object itself.
class Client
constructor: (socket) ->
#socket = socket
#avatar = socket.handshake.avatar
#listeners = {}
addListener: (name, handler) ->
#listeners[name] ||= {}
#listeners[name][handler.clientListenerId] = wrapper = =>
# append client object as the first argument before passing to handler
args = Array.prototype.slice.call(arguments)
args.unshift(this)
handler.apply(this, args) # <---- HANDLER'S CONTEXT IS CHANGING HERE :(
#socket.addListener(name, wrapper)
removeListener: (name, handler) ->
try
obj = #listeners[name]
#socket.removeListener(obj[handler.clientListenerId])
delete obj[handler.clientListenerId]
Note that clientListenerId is a custom unique identifier property that is essentially the same as the answer found here.
If I understand you correctly:
changes context
| n | y |
accepts array n | func() | func.call() |
of arguments y | ???????? | func.apply() |
PHP has a function for this, call_user_func_array. Unfortunately, JavaScript is lacking in this regard. It looks like you simulate this behavior using eval().
Function.prototype.invoke = function(args) {
var i, code = 'this(';
for (i=0; i<args.length; i++) {
if (i) { code += ',' }
code += 'args[' + i + ']';
}
eval(code + ');');
}
Yes, I know. Nobody likes eval(). It's slow and dangerous. However, in this situation you probably don't have to worry about cross-site scripting, at least, as all variables are contained within the function. Really, it's too bad that JavaScript doesn't have a native function for this, but I suppose that it's for situations like this that we have eval.
Proof that it works:
function showArgs() {
for (x in arguments) {console.log(arguments[x]);}
}
showArgs.invoke(['foo',/bar/g]);
showArgs.invoke([window,[1,2,3]]);
Firefox console output:
--
[12:31:05.778] "foo"
[12:31:05.778] [object RegExp]
[12:31:05.778] [object Window]
[12:31:05.778] [object Array]
Simply put, just assign the this to what you want it to be, which is otherFn:
function fn() {
var args = Array.prototype.slice.call(arguments);
otherFn.apply(otherFn, args);
}
'this' is a reference to your function's context. That's really the point.
If you mean to call it in the context of a different object like this:
otherObj.otherFn(args)
then simply substitute that object in for the context:
otherObj.otherFn.apply(otherObj, args);
That should be it.
If you bind the function to an object and you use everywhere the bound function, you can call apply with null, but still get the correct context
var Person = function(name){
this.name = name;
}
Person.prototype.printName = function(){
console.log("Name: " + this.name);
}
var bob = new Person("Bob");
bob.printName.apply(null); //window.name
bob.printName.bind(bob).apply(null); //"Bob"
One way that you can work around the change of context that can happen in JavaScript when functions are called, is to use methods that are part of the object's constructor if you need them to be able to operate in a context where this is not going to mean the parent object, by effectively creating a local private variable to store the original this identifier.
I concede that - like most discussions of scope in JavaScript - this is not entirely clear, so here is an example of how I have done this:
function CounterType()
{
var counter=1;
var self=this; // 'self' will now be visible to all
var incrementCount = function()
{
// it doesn't matter that 'this' has changed because 'self' now points to CounterType()
self.counter++;
};
}
function SecondaryType()
{
var myCounter = new CounterType();
console.log("First Counter : "+myCounter.counter); // 0
myCounter.incrementCount.apply(this);
console.log("Second Counter: "+myCounter.counter); // 1
}
These days you can use rest parameters:
function fn(...args) {
otherFn(...args);
}
The only downside is, if you want to use some specific params in fn, you have to extract it from args:
function fn(...args) {
let importantParam = args[2]; //third param
// ...
otherFn(...args);
}
Here's an example to try (ES next version to keep it short):
// a one-line "sum any number of arguments" function
const sum = (...args) => args.reduce((sum, value) => sum + value);
// a "proxy" function to test:
var pass = (...args) => sum(...args);
console.log(pass(1, 2, 15));
I'm not going to accept this as an answer, as I'm still hoping for something more suitable. But here's the approach I'm using right now based upon the feedback on this question so far.
For any class that will be calling Client.prototype.addListener or Client.prototype.removeListener, I did added the following code to their constructor:
class ExampleClass
constructor: ->
# ...
for name, fn of this
this[name] = fn.bind(this) if typeof(fn) == 'function'
message: (recipient, body) ->
# ...
broadcast: (body) ->
# ...
In the above example, message and broadcast will always be bound to the new ExampleClass prototype object when it's instantiated, allowing the addListener code in my original question to work.
I'm sure some of you are wondering why I didn't just do something like the following:
example = new ExampleClass
client.addListener('message', example.bind(example))
# ...
client.removeListener('message', example.bind(example))
The problem is that every time .bind( ) is called, it's a new object. So that means that the following is true:
example.bind(example) != example.bind(example)
As such, the removeListener would never work successfully, thus my binding the method once when the object is instantiated.
Since you seem to want to be using the bind function as it is defined in Javascript 1.8.5, and be able to retrieve the original this object you pass the bind function, I recommend redefining the Function.prototype.bind function:
Function.prototype.bind = function (oThis) {
if (typeof this !== "function") {
throw new TypeError("Function.prototype.bind - what is trying to be bound is not callable");
}
var aArgs = Array.prototype.slice.call(arguments, 1),
fToBind = this,
fNOP = function () {},
fBound = function () {
return fToBind.apply(this instanceof fNOP && oThis
? this
: oThis,
aArgs.concat(Array.prototype.slice.call(arguments)));
};
fNOP.prototype = this.prototype;
fBound.prototype = new fNOP();
/** here's the additional code **/
fBound.getContext = function() {
return oThis;
};
/**/
return fBound;
};
Now you can retrieve the original context that you called the bind function with:
function A() {
return this.foo+' '+this.bar;
}
var HelloWorld = A.bind({
foo: 'hello',
bar: 'world',
});
HelloWorld(); // returns "hello world";
HelloWorld.getContext(); // returns {foo:"hello", bar:"world"};
I was just reminded of this question after a long time. Looking back now, I think what I was really trying to accomplish here was something similar to how the React library works with its automatic binding.
Essentially, each function is a wrapped bound function being called:
function SomeClass() {
};
SomeClass.prototype.whoami = function () {
return this;
};
SomeClass.createInstance = function () {
var obj = new SomeClass();
for (var fn in obj) {
if (typeof obj[fn] == 'function') {
var original = obj[fn];
obj[fn] = function () {
return original.apply(obj, arguments);
};
}
}
return obj;
};
var instance = SomeClass.createInstance();
instance.whoami() == instance; // true
instance.whoami.apply(null) == instance; // true
Just push properties directly to the function's object and call it with it's own "context".
function otherFn() {
console.log(this.foo+' '+this.bar); // prints: "hello world" when called from rootFn()
}
otherFn.foo = 'hello';
otherFn.bar = 'world';
function rootFn() {
// by the way, unless you are removing or adding elements to 'arguments',
// just pass the arguments object directly instead of casting it to Array
otherFn.apply(otherFn, arguments);
}
I am mlearning javascript and have some trouble creating an onject via prototype.
I have this:
<script type="text/javascript">
function myclass(a, b, c) {
if (arguments.length) { this.Init(a, b, c); }
}
myclass.prototype.Init = function(a, b, c) {
this.param1 = a;
this.param2 = b;
this.param3 = c;
};
myclass.prototype.Print = function() {
alert(this.param1 + '-' + this.param2 + '-' + this.param3);
};
var myObject = myclass(3, 5, 6);
myObject.Print();
</script>
but I get an error on line with this.Init(a, b, c);
Error: Object doesn't support this property or method
You forgot the new keyword when you declare myObject:
var myObject = new myclass(3, 5, 6);
Just out of curiosity is there a particular reason you have a separate "init" method?
The function that defines your "class" is called the "constructor" and you can just perform the setup there. If you wanted to "re-initialize" the object, then it may be helpful but it doesn't seem to serve a point here.
For instance:
// You might as well start wrapping your code now:
var myExample = (function myExample () {
// A common convention is to start the name of constructors with a
// capital letter, one reason is it help makes it more obvious
// when you forget the new keyword...Whether you use it or not
// is up to you. Also note, calling it "MyClass" is a little
// misleading because it's not a "class" really. You might
// confuse yourself if you think of it as a class too much.
// If you're wondering why I put the name twice, it's because
// otherwise it would be an anonymous function which can be
// annoying when debugging. You can just use var MyClass = function () {}
// if you want
var MyClass = function MyClass(a, b, c) {
// This will set each parameter to whatever was provided
// or if nothing is provided: null. If you leave out
// the || "" part then any
// time a value is not provided the parameter will
// return "undefined". This may be what you want in some cases.
this.param1 = a || "";
this.param2 = b || "";
this.param3 = c || "";
};
// likewise it's convention to start most variables/functions lowercase
// I think it's easier to type/looks better, but do as you please.
MyClass.prototype.print = function print() {
alert(this.param1 + '-' + this.param2 + '-' + this.param3);
};
var myObject = new MyClass();
myObject.print();
}());
The "wrapping" is
(function () {
//your code here
}());
It's mostly pointless here, but it's something you'll have to start doing eventually so might as well start now. That's just one way to "wrap" there are others as well.
Basically, the way your script was written, if the user ran another script that had a function called MyClass, it could overwrite yours or vice versa, causing problems.
The "wrapping" keeps it all within that function. If you need to make something available to outside stuff, you can expose it.
per comment:
You can access functions and variables from inside the wrapper by exposing them to the outside like so:
var myApp = (function myApp(){
// The constructor for our "class", this will be available from outside because
// we will expose it later
var myClass = function(){
//code to set up "class" etc
// See how we can use private function within myApp
privateFunction();
};
// Here we set up the private function, it will not be available outside myApp
// because will will not expose it
var privateFunction = function(){ };
// Another public function that we will expose later
var otherPublic = function(){};
//now we expose the stuff we want public by returning an object containing
// whatever it is we want public, in this case it's just myClass and otherPublic
return { myClass: myClass, otherPublic: otherPublic };
}());
Note in that example we are just exposing the constructor, if you wanted instance of the object
you'd have to collect them in a variable and expose that variable like:
var theInstance = new myClass();
return { theInstance : theInstance };
It would now be available outside myApp as myApp.theInstance
You can also use a more basic wrapping scheme:
var myApp = {
myClass: function(){
//if we want to call another function in myApp we have to do it like so:
myApp.publicFunction();
},
publicFunction: function(){},
someString: "this is a string"
};
There myApp is just an object literal containing your functions etc. The main difference is that EVERYTHING in myApp can be accessed from outside via myApp.name or myApp[name];