We Keep Coding

Why doesn't JavaScript ES6 support multi-constructor classes?

I want to write my Javascript class like below.
class Option {
constructor() {
this.autoLoad = false;
}
constructor(key, value) {
this[key] = value;
}
constructor(key, value, autoLoad) {
this[key] = value;
this.autoLoad = autoLoad || false;
}
}
I think it would be nice if we can write out class in this way.
Expect to happen:
var option1 = new Option(); // option1 = {autoLoad: false}
var option2 = new Option('foo', 'bar',); // option2 = {foo: 'bar'}
var option3 = new Option('foo', 'bar', false); // option3 = {foo: 'bar', autoLoad: false}

I want to write my Javascript class like below
You can't, in the same way you can't overload standard functions like that. What you can do is use the arguments object to query the number of arguments passed:
class Option {
constructor(key, value, autoLoad) {
// new Option()
if(!arguments.length) {
this.autoLoad = false;
}
// new Option(a, [b, [c]])
else {
this[key] = value;
this.autoLoad = autoLoad || false;
}
}
}
Babel REPL Example
Of course (with your updated example), you could take the approach that you don't care about the number of arguments, rather whether each individual value was passed, in which case you could so something like:
class Option {
constructor(key, value, autoLoad) {
if(!key) { // Could change this to a strict undefined check
this.autoLoad = false;
return;
}
this[key] = value;
this.autoLoad = autoLoad || false;
}
}

What you want is called constructor overloading. This, and the more general case of function overloading, is not supported in ECMAScript.
ECMAScript does not handle missing arguments in the same way as more strict languages. The value of missing arguments is left as undefined instead of raising a error. In this paradigm, it is difficult/impossible to detect which overloaded function you are aiming for.
The idiomatic solution is to have one function and have it handle all the combinations of arguments that you need. For the original example, you can just test for the presence of key and value like this:
class Option {
constructor(key, value, autoLoad = false) {
if (typeof key !== 'undefined') {
this[key] = value;
}
this.autoLoad = autoLoad;
}
}

Another option would be to allow your constructor to take an object that is bound to your class properties:
class Option {
// Assign default values in the constructor object
constructor({key = 'foo', value, autoLoad = true} = {}) {
this.key = key;
// Or on the property with default (not recommended)
this.value = value || 'bar';
this.autoLoad = autoLoad;
console.log('Result:', this);
}
}
var option1 = new Option();
// Logs: {key: "foo", value: "bar", autoLoad: true}
var option2 = new Option({value: 'hello'});
// Logs: {key: "foo", value: "hello", autoLoad: true}
This is even more useful with Typescript as you can ensure type safety with the values passed in (i.e. key could only be a string, autoLoad a boolean etc).

Guessing from your sample code, all you need is to use default values for your parameters:
class Option {
constructor(key = 'foo', value = 'bar', autoLoad = false) {
this[key] = value;
this.autoLoad = autoLoad;
}
}
Having said that, another alternative to constructor overloading is to use static factories. Suppose you would like to be able to instantiate an object from plain parameters, from a hash containing those same parameters or even from a JSON string:
class Thing {
constructor(a, b) {
this.a = a;
this.b = b;
}
static fromHash(hash) {
return new this(hash.a, hash.b);
}
static fromJson(string) {
return this.fromHash(JSON.parse(string));
}
}
let thing = new Thing(1, 2);
// ...
thing = Thing.fromHash({a: 1, b: 2});
// ...
thing = Thing.fromJson('{"a": 1, "b": 2}');

Here's a hack for overloading based on arity (number of arguments). The idea is to create a function from a number of functions with different arities (determined by looking at fn.length).
function overloaded(...inputs) {
var fns = [];
inputs.forEach(f => fns[f.length] = f);
return function() {
return fns[arguments.length].apply(this, arguments);
};
}
var F = overloaded(
function(a) { console.log("function with one argument"); },
function(a, b) { console.log("function with two arguments"); }
);
F(1);
F(2, 3);
Of course this needs a lot of bullet-proofing and cleaning up, but you get the idea. However, I don't think you'll have much luck applying this to ES6 class constructors, because they are a horse of a different color.

you can use static methods,look at my answer to same question
class MyClass {
constructor(a,b,c,d){
this.a = a
this.b = b
this.c = c
this.d = d
}
static BAndCInstance(b,c){
return new MyClass(null,b,c)
}
}
//a Instance that has b and c params
MyClass.BAndCInstance(b,c)

Use object.assigne with arguments with this
This={...this,...arguments}

Its not the overload I wanted, but this is a basic version of how I faked my way through creating an obj1 with some different initialization behavior. I realize I could have expanded the arguments as stated above, but I already had a nasty set of arguments and relatively different data sources to deconstruct that would have really distorted my objectives; this just made it cleaner for my situation...
class obj1{
constructor(v1, v2){
this.a = v1;
this.b = v2;
}
}
class obj1Alt{
constructor(v1, v2){
return new obj1(v1*2,v2*2);
}
}
new obj1(2,4) // returns an obj1
new obj1Alt(2,4) // also returns an obj1
Disclaimer: I've been programming for a long time, but I am fairly new to JS; probably not a best practice.

Javascript simulate real private variable and manage it with getter/setter

Do I use correctly simulation of private variable with this way?
var C = function(a) {
var _private = a + 1;
// more code...
Object.defineProperties(this, {
'privateProp': {
get: function() {
return _private;
},
set: function(b) {
_private = b + 2;
}
}
});
}
So with the getter/setter I can manage that private variable. And value of it can be get/set with custom methods.
Let's say this example
var c = new C(5);
console.log(c.privateProp); // 6
c.privateProp = 2;
console.log(c.privateProp); // 4

There is no correct way to simulate private variables, because they are not part of the Javascript specification.
However, there are still a number of ways to simulate them:
With Getters and Setters
This is the same as your answer but with a slightly different syntax:
function Rocket() {
var _fuel = 100;
return {
get fuel() {
return _fuel;
}
set fuel(value) {
_fuel = value;
}
};
}
This syntax allows you to define the getter and setter functions in a more natural way, but comes with the restriction that they must be inside an object literal.
With a Closure
function Rocket() {
var fuel = 100;
return {
getFuel: function() { return fuel; },
setFuel: function(value) { fuel = value; }
};
}
This is arguably the most elegant way, because this is how Javascript has worked ever since anonymous functions were added to the specification.
By Convention
function Rocket() {
this._fuel = 100;
this.getFuel = function() {
return this._fuel;
};
this.setFuel = function(value) {
this._fuel = value;
};
}
The final way is to expose the variable as public, but mark it as private by using an underscore at the beginning of the name.

Method chaining and/or value caching

Say i have a constructor, and some instance methods, like
function MyClass(name) {
this.name = name || '';
}
MyClass.prototype = {
constructor: MyClass,
isEmptyName: function() {
return this.name === '';
}
}
Now i can write
var myClass = new MyClass('Ben');
myClass.isEmptyName();
which would return false. Now if i make another method what would also return a Boolean
MyClass.prototype = {
constructor: MyClass,
isEmptyName: function() {
return this.name === '';
}
longerThan: function(len) {
return this.name.length > len;
}
}
i would like to chain these methods like this (somehow, thats my question :) )
myClass.isEmptyName().and.longerThan(2);
Just omit now the '.and.' part. I want the upper statement to finally return a value
false && true -> false
Or a more realistic sample:
myClass.notEmptyName().and.longerThan(4);
To summarize my problem i would say, i want my methods return a boolean value if they are called 'directly' myClass.notEmptyName() should return true, but work like i wrote in the samples, otherwise.
Other libraries do this somehow, but i can't guess how, npm's should is a good example:
user.should.have.property('pets').with.lengthOf(4);
user.pets.should.be.instanceof(Array).and.have.lengthOf(4);
Thanks

That's not possible. A method can't return either a boolean or be chainable depending on how it's used later on, because it doesn't know how it will be used later on.
You can chain methods that validate the object in different ways, but you need to get the result at the end if you want to use it in an expression, either by reading a property or calling a method:
function MyClass(name) {
this.name = name;
this.and = this;
}
MyClass.prototype = {
value: true,
isNotEmpty: function() {
this.value = this.value && this.name.length > 0; return this;
},
isLongerThan: function(len) {
this.value = this.value && this.name.length > len; return this;
},
evaluate: function() {
return this.value;
}
};
console.log(new MyClass('Adam').isLongerThan(2).evaluate());
console.log(new MyClass('Bob').isNotEmpty().and.isLongerThan(3).evaluate());
Demo: http://jsfiddle.net/Guffa/62e8dLwL/
Edit:
To allow evaluation more than once, you would reset the value in the evaluate method:
evaluate: function() {
var v = this.value;
this.value = true;
return v;
}

Sure, you can do that. We will define a new intermediate status object, called ChainResult, which remembers the underlying object, the current value, and a pending operation (a function to use to combine the next test). We give this object a valueOf method, so that when JS tries to evaluate it as a primitive, it "looks" like it has a value. To make this work, it turns out that ChainResult actually needs to be a function, and so we hang the necessary properties off the function.
function ChainResult(obj, val) {
function x() { }
x.obj = obj;
x.val = val;
x.op = null;
// the `valueOf` function spits out the current value when the object is evaluated
x.valueOf = function() { return this.val; };
// the test functions combine the results with the current value
// using the current operation as set by a preceding `and` or `or`
x.isEmptyName = function() {
x.val = x.op(x.val, x.obj._isEmptyName());
return this;
};
x.isLongerThan = function(len) {
x.val = x.op(x.val, x.obj._isLongerThan(len));
return this;
};
// we implement `and` and `or` via getters which set the operation
// on the ChainResult object, and return `this` so we can keep chaining
Object.defineProperties(x, {
and: {
get: function() { x.op = function(a,b) { return a && b; }; return x; }
},
or: {
get: function() { x.op = function(a,b) { return a || b; }; return x; }
}
});
return x;
}
The MyClass definition needs a bit of tweaking:
function MyClass(name) {
this.name = name || '';
}
MyClass.prototype = {
constructor: MyClass,
// we implement the testers as pseudo-private functions
_isEmptyName: function() { return this.name === ''; },
_isLongerThan: function(len) { return this.name.length > len; },
// when the public tester functions are invoked directly on the object
// (when they are the first link in the chain), we construct and return a
// ChainResult object with the initial value set correctly
isEmptyName: function() { return ChainResult(this, this._isEmptyName()); },
isLongerThan: function(len) { return ChainResult(this, this._isLongerThan(len)) }
};
Flow:
new MyClass('Bob') // create MyClass object
.isEmptyName() // create ChainResult object with value `false`
.or // remember `or` operation in ChainResult object
.isLongerThan(2) // update value of ChainResult object
; // JS tries to convert to scalar, calls valueOf
// true
This needs to be bullet-proofed and tightened up, but you get the idea.

i want my methods return a boolean value if they are called 'directly' myClass.notEmptyName() should return true
Your methods are always called directly on the instance, and would always need to return a primitive boolean value. By that, the context (myClass) is lost and you cannot have an and method (or property getter) on the result.
I would recommend you to have a look at functional programming, partial application and currying, which helps a lot with fluent interfaces like this. Given
function and(fn1, fn2) {
return function(val) {
return fn1(val) && fn2(val);
};
}
function or(fn1, fn2) {
return function(val) {
return fn1(val) || fn2(val);
};
}
function hasEmptyName: function(val) {
return val.name === '';
}
function hasNameLongerThan: function(len) {
return function(val) {
return val.name.length > len;
};
}
you could write
and(hasEmptyName, hasNameLongerThan(2))(myClass);
Making these functions methods of anything is complicated however. Maybe something like this:
function method(name) {
var args = Array.prototype.slice.call(arguments, 1);
return function(instance) {
return instance[name].apply(instance, args);
};
}
Function.prototype.and = function (fn2) {
var fn1 = this;
return function(val) {
return fn1(val) && fn2(val);
};
}
Function.prototype.or = function (fn2) {
var fn1 = this;
return function(val) {
return fn1(val) || fn2(val);
};
}
Object.defineProperty(Object.prototype, "test", function(pred) {
return pred(this);
});
Now you could write
myClass.test(method("notEmptyName").and(method("longerThan", 4)));

This is an answer (better call it to outcome) for my own question:
Finally i came out with another solution based on the responses in this thread (Thanks guys!) because the original problem can not be solved, since the javascript runtime can't find out wether to return a value, or return itself (the object) when chained. Explanation is messy, sorry :(
Check out my lib, where i have methods like:
check(value).isString() etc..
Originally i wanted to chain these like check(value).isString().and.not.empty() but this, in this way, can not be done. (Challenge me)
Finally i created tokens for chaining, so instead of
check(value).isString().and.not.empty()
I can write
check(value).isstring.and.not.empty.test()
Not nice, but still something.
check.js
For review, visit checkjs on my github repo.
Note: README is outdated.

If you use promises, you could write functions that return values and can be chained. Mind you, these are all asynchronous.
Find this JS plugin: Kris Owal's Q, or if you like to use a JS library, they usually contain deferred objects and promises.

Using "dot" inside a prototype name in JavaScript

Lets say I have this class:
function classA(n){
this.name = n
}
classA.prototype.getName = function(){
return this.name
}
var x = new classA('john')
console.log(x.getName())
My question is: can I group multiple methods inside a namespace? So I would like to do that:
var x = new classA('john')
console.log(x.CONSTANT.getName())
So I would like to call some methods as x.someMethod() and others as x.CONSTANT.otherMethod()
PS: I'm looking for a cross-browser method. Bind is not working in Safari and IE9.

You can do it, for example, via bind. Google es5 shim for implementation of bind in browsers, which don't support it natively.
function MyClass(name) {
this.name = name;
this.CONSTANT.otherMethod = this.CONSTANT.otherMethod.bind(this);
}
MyClass.prototype.CONSTANT = {
otherMethod: function() {
alert(this.name);
}
};

As far as I know a constant is just a property and it can't contain methods, you need to separate your objects and use methods to have the same effect:
function A (id) {
this.id = id;
this.showId = function () { return this.id; }
};
function B (a) {
this.a = a;
this.getA = function () { return this.a; }
}
var a = new A(12);
var b = new B(a);
b.getA().showId();
edit:
You can use a literal object as follow
function B (id) {
this.id = id;
this.CONSTANT = { otherMethod: function () { alert("..."); } };
someMethod = function () { return this.id; }
}
but the literal CONSTANT object can't access B-object methods,
Consider the #kirilloid post to round this.

You can, but you have to be careful because it won't act like you think it will. The this for the method will be the namespace, not the root object.
For example, in x.CONSTANT.getName(), the this object will be x.CONSTANT, and not x.
Here's some sample code which kinda does what you ask (or in jsfiddle):
function MyClass() {}
MyClass.prototype.CONSTANT = {
getName: function() {
alert('Foo');
}
};
var c = new MyClass();
c.CONSTANT.getName();
To make sure the this is right, you need to do much more.

You can use getters/setters (read this article) to achieve this. For example you may define it like this:
classA.prototype.__defineGetter__('CONSTANT', function() {
var that = this;
return {
getName: function() {
return that.name;
}
};
});
Note that holding reference to the object. It will work now
x = new classA('test');
x.CONSTANT.getName();
// result - test

Javascript Reflection

Is there a way to get all methods (private, privileged, or public) of a javascript object from within? Here's the sample object:
var Test = function() {
// private methods
function testOne() {}
function testTwo() {}
function testThree() {}
// public methods
function getMethods() {
for (i in this) {
alert(i); // shows getMethods, but not private methods
}
}
return { getMethods : getMethods }
}();
// should return ['testOne', 'testTwo', 'testThree', 'getMethods']
Test.getMethods();
The current issue is the code in getMethods(), the simplified example will return just the public methods, but not to private ones.
edit: my test code may (or may not) be overcomplicating what i'm hoping to get at. given the following:
function myFunction() {
var test1 = 1;
var test2 = 2;
var test3 = 3;
}
is there a way to find out what variables exist in myFunction() from within myFunction(). the pseudo-code would look like this:
function myFunction() {
var test1 = 1;
var test2 = 2;
var test3 = 3;
alert(current.properties); // would be nice to get ['test1', 'test2', 'test3']
}

The technical reason why those methods are hidden is twofold.
First, when you execute a method on the Test object, "this" will be the untyped object returned at the end of the anonymous function that contains the public methods per the Module Pattern.
Second, the methods testOne, testTwo, and testThree aren't attached to a specific object, and exist only in the context of the anonymous function. You could attach the methods to an internal object and then expose them through a public method, but it wouldn't be quite as clean as the original pattern and it won't help if you're getting this code from a third party.
The result would look something like this:
var Test = function() {
var private = {
testOne : function () {},
testTwo : function () {},
testThree : function () {}
};
function getMethods() {
for (i in this) {
alert(i); // shows getMethods, but not private methods
}
for (i in private) {
alert(i); // private methods
}
}
return { getMethods : getMethods }
}();
// will return ['getMethods', 'testOne', 'testTwo', 'testThree']
Test.getMethods();
edit:
Unfortunately, no. The set of local variables aren't accessible via a single, automatic keyword.
If you remove the "var" keyword they would be attached to the global context (usually the window object), but that's the only behavior that I know of that is similar to what you're describing. There would be a lot of other properties and methods on that object if you did that, though.

From http://netjs.codeplex.com/SourceControl/changeset/view/91169#1773642
//Reflection
~function (extern) {
var Reflection = this.Reflection = (function () { return Reflection; });
Reflection.prototype = Reflection;
Reflection.constructor = Reflection;
Reflection.getArguments = function (func) {
var symbols = func.toString(),
start, end, register;
start = symbols.indexOf('function');
if (start !== 0 && start !== 1) return undefined;
start = symbols.indexOf('(', start);
end = symbols.indexOf(')', start);
var args = [];
symbols.substr(start + 1, end - start - 1).split(',').forEach(function (argument) {
args.push(argument);
});
return args;
};
extern.Reflection = extern.reflection = Reflection;
Function.prototype.getArguments = function () { return Reflection.getArguments(this); }
Function.prototype.getExpectedReturnType = function () { /*ToDo*/ }
} (this);

Javascript doesn't really have the notion of private anything. Because of that, javascript doesn't have a reflection API as such. The technique you're using doesn't so much make them private as render them inaccessible; they're hidden, not private. I think you could manage something by putting those methods somewhere manually.

Part of the issue with your test code is that Test is the object created by your return statement: "{ getMethods : getMethods }" It has no testOne, testTwo, or testThree methods; instead, those are only available within the same namespace as the original getMethods function.

With a little change to the way the function is defined you can achieve what you want. Wrap the actual implementation of the function in an object literal it would then look like this:
(function() {
var obj = {
// private methods
testOne: function () {},
testTwo : function () {},
testThree: function () {},
// public methods
getMethods : function () {
for (i in this) {
alert(i); // shows getMethods, but not private methods
}
}
};
return { getMethods : function(){return obj.getMethods();} }
})();

you can use var that = this; trick:
var Test = function() {
var that = this;
function testOne() {}
function testTwo() {}
function testThree() {}
function getMethods() {
for (i in that) {
alert(i);
}
}
return { getMethods : getMethods }
}();

If you call getMethods() like that, isn't it static? Surely you'd need to properly init the class for this to work as expected?
var t = new Test();
t.getMethods();
If that doesn't work, please take a look at the JS Serializer. I used it a while back for some debug and I think it worked for private vars.