I have a function below, that checks if myValue exists before calling doSomething(), if it does not exist, it will throw. Typescript accepts this...
function myFunction(){
if (!myValue) {
const error = Error(errorName);
error.name = errorName.replace(/\s+/g, "-").toLowerCase();
throw error;
}
const something = doSomething(myValue);
}
The below implementation makes a Helper Function for throwing a named Error. Typescript doesn't like this, and errors that doSomething(myValue) is being passed myValue which is potentially undefined. How can I type the namedError so that it has the inferred typing from above, that it is Truthy?
function myFunction(){
if (!myValue) {
namedError("my value doesn't exist");
}
const something = doSomething(myValue);
}
function namedError(errorName: string){
const error = Error(errorName);
error.name = errorName.replace(/\s+/g, "-").toLowerCase();
throw error;
};
Since typescript 3.7 functions that return never have the same impact in control flow analysis as a throw expression. So this will work:
let myValue: string | undefined;
function myFunction() {
if (!myValue) {
namedError("my value doesn't exist");
}
const something = doSomething(myValue);
}
function namedError(errorName: string): never {
const error = Error(errorName);
error.name = errorName.replace(/\s+/g, "-").toLowerCase();
throw error;
};
function doSomething(value: string) {
}
Playground Link
Typescript null elimination doesn't work with nest function calls (unless it's an immediately invoked function expression) hence the error.
You can either move the doSomething invocation into an else block, or use the type assertion operator when you are sure a value will not be null or defined.
So either:
if (!myValue) {
namedError(...)
} else {
const something = doSomething(myValue);
}
or
if (!myValue) {
namedError(...)
}
// use type assertion (! postfix) to tell compiler that the value will never be null or undefined
const something = doSomething(myValue!);
I noticed not all the Javascript functions are constructors.
var obj = Function.prototype;
console.log(typeof obj === 'function'); //true
obj(); //OK
new obj(); //TypeError: obj is not a constructor
Question 1: How do I check if a function is a constructor so that it can be called with new keyword?
Question 2: When I create a function, is it possible to make it NOT a constructor?
A little bit of background:
ECMAScript 6+ distinguishes between callable (can be called without new) and constructable (can be called with new) functions:
Functions created via the arrow functions syntax or via a method definition in classes or object literals are not constructable.
Functions created via the class syntax are not callable.
Functions created in any other way (function expression/declaration, Function constructor) are callable and constructable.
Built-in functions are not constructrable unless explicitly stated otherwise.
About Function.prototype
Function.prototype is a so called built-in function that is not constructable. From the spec:
Built-in function objects that are not identified as constructors do not implement the [[Construct]] internal method unless otherwise specified in the description of a particular function.
The value of Function.prototype is create at the very beginning of the runtime initialization. It is basically an empty function and it is not explicitly stated that it is constructable.
How do I check if an function is a constructor so that it can be called with a new?
There isn't a built-in way to do that. You can try to call the function with new, and either inspect the error or return true:
function isConstructor(f) {
try {
new f();
} catch (err) {
// verify err is the expected error and then
return false;
}
return true;
}
However, that approach is not failsafe since functions can have side effects, so after calling f, you don't know which state the environment is in.
Also, this will only tell you whether a function can be called as a constructor, not if it is intended to be called as constructor. For that you have to look at the documentation or the implementation of the function.
Note: There should never be a reason to use a test like this one in a production environment. Whether or not a function is supposed to be called with new should be discernable from its documentation.
When I create a function, how do I make it NOT a constructor?
To create a function is truly not constructable, you can use an arrow function:
var f = () => console.log('no constructable');
Arrow functions are by definition not constructable. Alternatively you could define a function as a method of an object or a class.
Otherwise you could check whether a function is called with new (or something similar) by checking it's this value and throw an error if it is:
function foo() {
if (this instanceof foo) {
throw new Error("Don't call 'foo' with new");
}
}
Of course, since there are other ways to set the value of this, there can be false positives.
Examples
function isConstructor(f) {
try {
new f();
} catch (err) {
if (err.message.indexOf('is not a constructor') >= 0) {
return false;
}
}
return true;
}
function test(f, name) {
console.log(`${name} is constructable: ${isConstructor(f)}`);
}
function foo(){}
test(foo, 'function declaration');
test(function(){}, 'function expression');
test(()=>{}, 'arrow function');
class Foo {}
test(Foo, 'class declaration');
test(class {}, 'class expression');
test({foo(){}}.foo, 'object method');
class Foo2 {
static bar() {}
bar() {}
}
test(Foo2.bar, 'static class method');
test(new Foo2().bar, 'class method');
test(new Function(), 'new Function()');
You are looking for if a function has a [[Construct]] internal method. The internal method IsConstructor details the steps:
IsConstructor(argument)
ReturnIfAbrupt(argument). // (Check if an exception has been thrown; Not important.)
If Type(argument) is not Object, return false. // argument === Object(argument), or (typeof argument === 'Object' || typeof argument === 'function')
If argument has a [[Construct]] internal method, return true.
Return false.
Now we need to find places where IsConstructor is used, but [[Construct]] isn't called (usually by the Construct internal method.)
I found that it is used in the String function's newTarget (new.target in js), which can be used with Reflect.construct:
function is_constructor(f) {
try {
Reflect.construct(String, [], f);
} catch (e) {
return false;
}
return true;
}
(I could have used anything really, like Reflect.construct(Array, [], f);, but String was first)
Which yields the following results:
// true
is_constructor(function(){});
is_constructor(class A {});
is_constructor(Array);
is_constructor(Function);
is_constructor(new Function);
// false
is_constructor();
is_constructor(undefined);
is_constructor(null);
is_constructor(1);
is_constructor(new Number(1));
is_constructor(Array.prototype);
is_constructor(Function.prototype);
is_constructor(() => {})
is_constructor({method() {}}.method)
<note>
The only value that I found it didn't work for is Symbol, which, although new Symbol throws a TypeError: Symbol is not a constructor in Firefox, is_constructor(Symbol) === true. This is technically the correct answer, as Symbol does have a [[Construct]] internal method (Which means it can also be subclassed), but using new or super is special cased for Symbol to throw an error (So, Symbol is a constructor, the error message is wrong, it just can't be used as one.) You can just add if (f === Symbol) return false; to the top though.
The same for something like this:
function not_a_constructor() {
if (new.target) throw new TypeError('not_a_constructor is not a constructor.');
return stuff(arguments);
}
is_constructor(not_a_constructor); // true
new not_a_constructor; // TypeError: not_a_constructor is not a constructor.
So the intentions of the function of being a constructor can't be gotton like this (Until somthing like Symbol.is_constructor or some other flag is added).
</note>
There is a quick and easy way of determining if function can be instantiated, without having to resort to try-catch statements (which can not be optimized by v8)
function isConstructor(obj) {
return !!obj.prototype && !!obj.prototype.constructor.name;
}
First we check if object is part of a prototype chain.
Then we exclude anonymous functions
There is a caveat, which is: functions named inside a definition will still incur a name property and thus pass this check, so caution is required when relying on tests for function constructors.
In the following example, the function is not anonymous but in fact is called 'myFunc'. It's prototype can be extended as any JS class.
let myFunc = function () {};
With ES6+ Proxies, one can test for [[Construct]] without actually invoking the constructor. Here's a snippet:
const handler={construct(){return handler}} //Must return ANY object, so reuse one
const isConstructor=x=>{
try{
return !!(new (new Proxy(x,handler))())
}catch(e){
return false
}
}
If the passed item isn't an object, the Proxy constructor throws an error. If it's not a constructable object, then new throws an error. But if it's a constructable object, then it returns the handler object without invoking its constructor, which is then not-notted into true.
As you might expect, Symbol is still considered a constructor. That's because it is, and the implementation merely throws an error when [[Construct]] is invoked. This could be the case on ANY user-defined function that throws an error when new.target exists, so it doesn't seem right to specifically weed it out as an additional check, but feel free to do so if you find that to be helpful.
If the function is a constructor then it will have a "prototype" member which in turn has a "constructor" member that is equal to the function itself.
function isConstructor(func) {
return typeof func === 'function' && !!func.prototype && func.prototype.constructor === func;
}
There is a quick and easy way of determining if function can be instantiated, without having to resort to try-catch statements (which can not be optimized by v8)
function isConstructor(value) {
return !!value && !!value.prototype && !!value.prototype.constructor;
}
First, we check value is truthy.
Then checks if value is part of a prototype chain.
At last, simply checks if constructor is set ;-)
Note that above is named isConstructor not is isConstructable, I mean, this will return false for Bound-constructors as said in comments, because "bound" means to redirect to something, instead of being a real constructor directly.
So this answers title's "check ... is constructor" question, but not later "check ... can be called with new" question.
Example:
const myClazz = {method() {}};
myClazz.method = myClazz.method.bind(myClazz);
// We can call above with new keyword.
new (myClazz.method);
// But it's just a callback.
if (isConstructor(myClass))
throw new Error('expected to return false for arrow-functions and similar.');
Unit-testing
Below is based on Jasmine.
// Change import to wherever your common functions are
// (for me they're in src directory, outside of tests directory).
import * as common from '../common-tools';
let isMyClassCalled = false;
class MyClass {
constructor() {
isMyClassCalled = true;
}
}
describe('App isConstructor tool', () => {
it('should detect constructor', function () {
detect(class A {});
detect(Array);
detect(Function);
detect(new Function);
detect({method() {}}.method);
});
it('should NOT detect as constructor', function () {
noDetect();
noDetect(undefined);
noDetect(null);
noDetect(1);
noDetect(new Number(1));
noDetect(new (function(){}));
noDetect(Array.prototype);
noDetect(Function.prototype);
// Commented because optimizations convert below into function.
//noDetect((() => {}));
});
it('should NOT detect bound constructors', function () {
const clazz = {method() {}};
clazz.method = clazz.method.bind(clazz);
noDetect(clazz.method);
});
it('should never call constructor', function () {
common.isConstructor(MyClass);
expect(isMyClassCalled).toBe(false);
});
function detect(value, expecting = true) {
expect(common.isConstructor(value))
.withContext('For "' + value + '" value')
.toBe(expecting);
}
function noDetect(value) {
detect(value, false);
}
});
Alternative
All above tests pass with below as well.
function isConstructor(value) {
return typeof value === 'function' && !!value.prototype && value.prototype.constructor === value;
}
For question 1, what about this helper?
Function.isConstructor = ({ prototype }) => Boolean(prototype) && Boolean(prototype.constructor)
Function.isConstructor(class {}); // true
Function.isConstructor(function() {}); // true
Function.isConstructor(() => {}); // false
Function.isConstructor("a string"); // false
For question 2, the arrow function is the solution. It cannot be used as a constructor since it does not rely on the same scope as a regular function and does not have a prototype (definition of instances, similar to class definition for real OOP)
const constructable = function() { console.log(this); };
const callable = () => { console.log(this); };
constructable(); // Window {}
callable(); // Window {}
new constructable(); // aConstructableFunction {}
new callable(); // Uncaught TypeError: callable is not a constructor
I tried many workarounds but it didn't satisfy my needs, So i made my own workaround by using reflection metadata.
GOAL: Check if the current Function has it's own __class__ metadata which represent if this function is a constructor or not.
npm install reflect-metadata
Create class decorator factory #Class()
Create a helper function to check __class__ metadata attached to the current function or not.
NOTE: The only way in this workaround to distinguish between Constructor Function and Normal Function or Class is by using Class Decorator #Class()
import 'reflect-metadata';
type Constructor<T = any> = new (...args: any[]) => T;
function Class() {
return function (target: Constructor) {
if (!!Reflect.getOwnMetadata('__class__', target)) {
throw new Error(`Cannot apply #Class decorator on ${target.name} multiple times.`);
}
Reflect.defineMetadata('__class__', target, target);
};
}
function isConstructor<T>(type: Constructor<T>): boolean {
if (typeof type !== 'function') return false;
return !!Reflect.getOwnMetadata('__class__', type);
}
/*
* ------------------
* Example
* ------------------
*/
#Class()
class ServiceClass1 {}
class ServiceClass2 {}
function Test() {}
console.log(isConstructor(ServiceClass1)) // true
console.log(isConstructor(ServiceClass2)) // false
console.log(isConstructor(Test)) // false
As an addition to Felix Kling's answer, even if a function is not constructable, we can still use it like a constructor if it has a prototype property. We can do this with the help of Object.create(). Example:
// The built-in object Symbol is not constructable, even though it has a "prototype" property:
new Symbol
// TypeError: Symbol is not a constructor.
Object.create(Symbol.prototype);
// Symbol {}
// description: (...)
// __proto__: Symbol
Given a function:
function x(arg) { return 30; }
You can call it two ways:
result = x(4);
result = new x(4);
The first returns 30, the second returns an object.
How can you detect which way the function was called inside the function itself?
Whatever your solution is, it must work with the following invocation as well:
var Z = new x();
Z.lolol = x;
Z.lolol();
All the solutions currently think the Z.lolol() is calling it as a constructor.
As of ECMAScript 6, this is possible with new.target
new.target will be set as true if the function is called with new (or with Reflect.construct, which acts like new), otherwise it's undefined.
function Foo() {
if (new.target) {
console.log('called with new');
} else {
console.log('not called with new');
}
}
new Foo(); // "called with new"
Foo(); // "not called with new"
Foo.call({}); // "not called with new"
NOTE: This is now possible in ES2015 and later. See Daniel Weiner's answer.
I don't think what you want is possible [prior to ES2015]. There simply isn't enough information available within the function to make a reliable inference.
Looking at the ECMAScript 3rd edition spec, the steps taken when new x() is called are essentially:
Create a new object
Assign its internal [[Prototype]] property to the prototype property of x
Call x as normal, passing it the new object as this
If the call to x returned an object, return it, otherwise return the new object
Nothing useful about how the function was called is made available to the executing code, so the only thing it's possible to test inside x is the this value, which is what all the answers here are doing. As you've observed, a new instance of* x when calling x as a constructor is indistinguishable from a pre-existing instance of x passed as this when calling x as a function, unless you assign a property to every new object created by x as it is constructed:
function x(y) {
var isConstructor = false;
if (this instanceof x // <- You could use arguments.callee instead of x here,
// except in in EcmaScript 5 strict mode.
&& !this.__previouslyConstructedByX) {
isConstructor = true;
this.__previouslyConstructedByX = true;
}
alert(isConstructor);
}
Obviously this is not ideal, since you now have an extra useless property on every object constructed by x that could be overwritten, but I think it's the best you can do.
(*) "instance of" is an inaccurate term but is close enough, and more concise than "object that has been created by calling x as a constructor"
1) You can check this.constructor:
function x(y)
{
if (this.constructor == x)
alert('called with new');
else
alert('called as function');
}
2) Yes, the return value is just discarded when used in the new context
NOTE: This answer was written in 2008, when javascript was still in ES3 from 1999. A lot of new functionality has been added since then, so now better solutions exists. This answer is kept for historical reasons.
The benefit of the code below is that you don't need to specify the name of the function twice and it works for anonymous functions too.
function x() {
if ( (this instanceof arguments.callee) ) {
alert("called as constructor");
} else {
alert("called as function");
}
}
Update
As claudiu have pointed out in a comment below, the above code doesn't work if you assign the constructor to the same object it has created. I have never written code that does that and have newer seen anyone else do that eighter.
Claudius example:
var Z = new x();
Z.lolol = x;
Z.lolol();
By adding a property to the object, it's possible to detect if the object has been initialized.
function x() {
if ( (this instanceof arguments.callee && !this.hasOwnProperty("__ClaudiusCornerCase")) ) {
this.__ClaudiusCornerCase=1;
alert("called as constructor");
} else {
alert("called as function");
}
}
Even the code above will break if you delete the added property. You can however overwrite it with any value you like, including undefined, and it still works. But if you delete it, it will break.
There is at this time no native support in ecmascript for detecting if a function was called as a constructor. This is the closest thing I have come up with so far, and it should work unless you delete the property.
Two ways, essentially the same under the hood. You can test what the scope of this is or you can test what this.constructor is.
If you called a method as a constructor this will be a new instance of the class, if you call the method as a method this will be the methods' context object. Similarly the constructor of an object will be the method itself if called as new, and the system Object constructor otherwise. That's clear as mud, but this should help:
var a = {};
a.foo = function ()
{
if(this==a) //'a' because the context of foo is the parent 'a'
{
//method call
}
else
{
//constructor call
}
}
var bar = function ()
{
if(this==window) //and 'window' is the default context here
{
//method call
}
else
{
//constructor call
}
}
a.baz = function ()
{
if(this.constructor==a.baz); //or whatever chain you need to reference this method
{
//constructor call
}
else
{
//method call
}
}
Checking for the instance type of the [this] within the constructor is the way to go. The problem is that without any further ado this approach is error prone. There is a solution however.
Lets say that we are dealing with function ClassA(). The rudimentary approach is:
function ClassA() {
if (this instanceof arguments.callee) {
console.log("called as a constructor");
} else {
console.log("called as a function");
}
}
There are several means that the above mentioned solution will not work as expected. Consider just these two:
var instance = new ClassA;
instance.classAFunction = ClassA;
instance.classAFunction(); // <-- this will appear as constructor call
ClassA.apply(instance); //<-- this too
To overcome these, some suggest that either a) place some information in a field on the instance, like "ConstructorFinished" and check back on it or b) keep a track of your constructed objects in a list. I am uncomfortable with both, as altering every instance of ClassA is way too invasive and expensive for a type related feature to work. Collecting all objects in a list could provide garbage collection and resource issues if ClassA will have many instances.
The way to go is to be able to control the execution of your ClassA function. The simple approach is:
function createConstructor(typeFunction) {
return typeFunction.bind({});
}
var ClassA = createConstructor(
function ClassA() {
if (this instanceof arguments.callee) {
console.log("called as a function");
return;
}
console.log("called as a constructor");
});
var instance = new ClassA();
This will effectively prevent all attempts to trick with the [this] value. A bound function will always keep its original [this] context unless you call it with the new operator.
The advanced version gives back the ability to apply the constructor on arbitrary objects. Some uses could be using the constructor as a typeconverter or providing an callable chain of base class constructors in inheritance scenarios.
function createConstructor(typeFunction) {
var result = typeFunction.bind({});
result.apply = function (ths, args) {
try {
typeFunction.inApplyMode = true;
typeFunction.apply(ths, args);
} finally {
delete typeFunction.inApplyMode;
}
};
return result;
}
var ClassA = createConstructor(
function ClassA() {
if (this instanceof arguments.callee && !arguments.callee.inApplyMode) {
console.log("called as a constructor");
} else {
console.log("called as a function");
}
});
actually the solution is very possible and simple... don't understand why so many words been written for such a tiny thing
UPDATE: thanks to TwilightSun the solution is now completed, even for the test Claudiu suggested! thank you guys!!!
function Something()
{
this.constructed;
if (Something.prototype.isPrototypeOf(this) && !this.constructed)
{
console.log("called as a c'tor"); this.constructed = true;
}
else
{
console.log("called as a function");
}
}
Something(); //"called as a function"
new Something(); //"called as a c'tor"
demonstrated here: https://jsfiddle.net/9cqtppuf/
Extending Gregs solution, this one works perfectly with the test cases you provided:
function x(y) {
if( this.constructor == arguments.callee && !this._constructed ) {
this._constructed = true;
alert('called with new');
} else {
alert('called as function');
}
}
EDIT: adding some test cases
x(4); // OK, function
var X = new x(4); // OK, new
var Z = new x(); // OK, new
Z.lolol = x;
Z.lolol(); // OK, function
var Y = x;
Y(); // OK, function
var y = new Y(); // OK, new
y.lolol = Y;
y.lolol(); // OK, function
There is no reliable way to distinguish how a function is called in JavaScript code.1
However, a function call will have this assigned to the global object, while a constructor will have this assigned to a new object. This new object cannot ever be the global object, because even if an implementation allows you to set the global object, you still haven't had the chance to do it.
You can get the global object by having a function called as a function (heh) returning this.
My intuition is that in the specification of ECMAScript 1.3, constructors that have a defined behavior for when called as a function are supposed to distinguish how they were called using this comparison:
function MyClass () {
if ( this === (function () { return this; })() ) {
// called as a function
}
else {
// called as a constructor
}
}
Anyway, anyone can just use a function's or constructor's call or apply and set this to anything. But this way, you can avoid "initializing" the global object:
function MyClass () {
if ( this === (function () { return this; })() ) {
// Maybe the caller forgot the "new" keyword
return new MyClass();
}
else {
// initialize
}
}
1. The host (aka implementation) may be able to tell the difference, if it implements the equivalent to the internal properties [[Call]] and [[Construct]]. The former is invoked for function or method expressions, while the latter is invoked for new expressions.
In my testing for http://packagesinjavascript.wordpress.com/ I found the test if (this == window) to be working cross-browser in all cases, so that's the one I ended up using.
-Stijn
Until I saw this thread I never considered that the constructor might be a property of an instance, but I think the following code covers that rare scenario.
// Store instances in a variable to compare against the current this
// Based on Tim Down's solution where instances are tracked
var Klass = (function () {
// Store references to each instance in a "class"-level closure
var instances = [];
// The actual constructor function
return function () {
if (this instanceof Klass && instances.indexOf(this) === -1) {
instances.push(this);
console.log("constructor");
} else {
console.log("not constructor");
}
};
}());
var instance = new Klass(); // "constructor"
instance.klass = Klass;
instance.klass(); // "not constructor"
For most cases I'll probably just check instanceof.
From John Resig:
function makecls() {
return function(args) {
if( this instanceof arguments.callee) {
if ( typeof this.init == "function")
this.init.apply(this, args.callee ? args : arguments)
}else{
return new arguments.callee(args);
}
};
}
var User = makecls();
User.prototype.init = function(first, last){
this.name = first + last;
};
var user = User("John", "Resig");
user.name
If you're going hackish, then instanceof is the minimum solution after new.target as by other answers. But using the instanceof solution it would fail with this example:
let inst = new x;
x.call(inst);
Combining with #TimDown solution, you can use ES6's WeakSet if you want compatibility with older ECMAScript versions to prevent putting properties inside instances. Well, WeakSet will be used in order to allow unused objects be garbage collected. new.target won't be compatible in the same source code, as it is a ES6's syntax feature. ECMAScript specifies identifiers cannot be one of the reserved words, and new is not an object, anyways.
(function factory()
{
'use strict';
var log = console.log;
function x()
{
log(isConstructing(this) ?
'Constructing' :
'Not constructing'
);
}
var isConstructing, tracks;
var hasOwnProperty = {}.hasOwnProperty;
if (typeof WeakMap === 'function')
{
tracks = new WeakSet;
isConstructing = function(inst)
{
if (inst instanceof x)
{
return tracks.has(inst) ?
false : !!tracks.add(inst);
}
return false;
}
} else {
isConstructing = function(inst)
{
return inst._constructed ?
false : inst._constructed = true;
};
}
var z = new x; // Constructing
x.call(z) // Not constructing
})();
ECMAScript 3's instanceof operator of is specified as:
11.8.6 The instanceof operator
--- The production RelationalExpression: RelationalExpression instanceof ShiftExpression is evaluated
as follows:
--- 1. Evaluate RelationalExpression.
--- 2. Call GetValue(Result(1)).
--- 3. Evaluate ShiftExpression.
--- 4. Call GetValue(Result(3)).
--- 5. If Result(4) is not an object, throw a TypeError exception.
--- 6. If Result(4) does not have a [[HasInstance]] method, throw a TypeError exception.
--- 7. Call the [[HasInstance]] method of Result(4) with parameter Result(2).
--- 8. Return Result(7).
15.3.5.3 [[HasInstance]] (V)
--- Assume F is a Function object.
--- When the [[HasInstance]] method of F is called with value V, the following steps are taken:
--- 1. If V is not an object, return false.
--- 2. Call the [[Get]] method of F with property name "prototype".
--- 3. Let O be Result(2).
--- 4. If O is not an object, throw a TypeError exception.
--- 5. Let V be the value of the [[Prototype]] property of V.
--- 6. If V is **null**, return false.
--- 7. If O and V refer to the same object or if they refer to objects joined to each other (13.1.2), return true.
--- 8. Go to step 5.
And that means it'll be recursing the left hand side value after going to its prototype until it is not an object or until it is equal to the prototype of the right hand side object with the specified [[HasInstance]] method. What means it'll check if left hand side is an instance of the right hand side, consuming all internal prototypes of the left hand side though.
function x() {
if (this instanceof x) {
/* Probably invoked as constructor */
} else return 30;
}
maybe I`m wrong but (at the cost of a parasite) the following code seems like a solution:
function x(arg) {
//console.debug('_' in this ? 'function' : 'constructor'); //WRONG!!!
//
// RIGHT(as accepted)
console.debug((this instanceof x && !('_' in this)) ? 'function' : 'constructor');
this._ = 1;
return 30;
}
var result1 = x(4), // function
result2 = new x(4), // constructor
Z = new x(); // constructor
Z.lolol = x;
Z.lolol(); // function
Although this thread is ancient, I'm surprised that nobody has mentioned that under strict mode ('use strict') a function's default this value is undefined, instead of set to global/window as before, so to check if new is not used simply test for falsey value of !this
- EG:
function ctor() { 'use strict';
if (typeof this === 'undefined')
console.log('Function called under strict mode (this == undefined)');
else if (this == (window || global))
console.log('Function called normally (this == window)');
else if (this instanceof ctor)
console.log('Function called with new (this == instance)');
return this;
}
If you test that function as-is, you will get undefined as this value, due to the 'use strict' directive at the start of the function. Of course, if already has strict mode on then it won't change if you remove the 'use strict' directive, but otherwise if you remove it the this value will be set to window or global.
If you use new to call the function then the this value will match the instanceof check (although if you checked the other things, then instance is last option so this check is not needed, and to be avoided if you want to inherit instances anyway)
function ctor() { 'use strict';
if (!this) return ctor.apply(Object.create(ctor.prototype), arguments);
console.log([this].concat([].slice.call(arguments)));
return this;
}
This will log the this value and any arguments you pass to the function to console, and return the this value. If the this value is falsey then it creates a new instance using Object.create(ctor.prototype) and uses Function.apply() to re-call the constructor with the same params but with correct instance as this. If the this value is anything other than falsey then it is assumed to be a valid instance and returned.
I believe the solution is to turn your Constructor function into a wrapper of the real Constructor function and its prototype Constructor if required. This method will work in ES5 from 2009 and also work in strict mode. In the code window below I have an example using the module pattern, to hold the real constructor and its prototype's constructor, in a closure, which is accessible through scope within the constructor(wrapper). This works because no property is added to the "this" keyword within the Constructor(wrapper) and the Constructor(wrapper).prototype is not set, so is Object by default; thus the array returned from Object.getpropertyNames will have a length equal to 0, if the new keyword has been used with the Constructor(wrapper). If true then return new Vector.
var Vector = (function() {
var Vector__proto__ = function Vector() {
// Vector methods go here
}
var vector__proto__ = new Vector__proto__();;
var Vector = function(size) {
// vector properties and values go here
this.x = 0;
this.y = 0;
this.x = 0;
this.maxLen = size === undefined? -1 : size;
};
Vector.prototype = vector__proto__;
return function(size){
if ( Object.getOwnPropertyNames(this).length === 0 ) {
// the new keyword WAS USED with the wrapper constructor
return new Vector(size);
} else {
// the new keyword was NOT USED with the wrapper constructor
return;
};
};
})();
Tim Down I think is correct. I think that once you get to the point where you think you need to be able to distinguish between the two calling modes, then you should not use the "this" keyword. this is unreliable, and it could be the global object, or it could be some completely different object. the fact is, that having a function with these different modes of activation, some of which work as you intended, others do something totally wild, is undesirable. I think maybe you're trying to figure this out because of that.
There is an idiomatic way to create a constructor function that behaves the same no matter how it's called. whether it's like Thing(), new Thing(), or foo.Thing(). It goes like this:
function Thing () {
var that = Object.create(Thing.prototype);
that.foo="bar";
that.bar="baz";
return that;
}
where Object.create is a new ecmascript 5 standard method which can be implemented in regular javascript like this:
if(!Object.create) {
Object.create = function(Function){
// WebReflection Revision
return function(Object){
Function.prototype = Object;
return new Function;
}}(function(){});
}
Object.create will take an object as a parameter, and return a new object with that passed in object as its prototype.
If however, you really are trying to make a function behave differently depending on how it's called, then you are a bad person and you shouldn't write javascript code.
If you don't want to put a __previouslyConstructedByX property in the object - because it pollutes the object's public interface and could easily be overwritten - just don't return an instance of x:
function x() {
if(this instanceof x) {
console.log("You invoked the new keyword!");
return that;
}
else {
console.log("No new keyword");
return undefined;
}
}
x();
var Z = new x();
Z.lolol = x;
Z.lolol();
new Z.lolol();
Now the x function never returns an object of type x, so (I think) this instanceof x only evaluates to true when the function is invoked with the new keyword.
The downside is this effectively screws up the behaviour of instanceof - but depending on how much you use it (I don't tend to) that may not be a problem.
If you're goal is for both cases to return 30, you could return an instance of Number instead of an instance of x:
function x() {
if(this instanceof x) {
console.log("You invoked the new keyword!");
var that = {};
return new Number(30);
}
else {
console.log("No new");
return 30;
}
}
console.log(x());
var Z = new x();
console.log(Z);
Z.lolol = x;
console.log(Z.lolol());
console.log(new Z.lolol());
I had this same problem when I tried to implement a function that returns a string instead of an object.
It seems to be enough to check for the existence of "this" in the beginning of your function:
function RGB(red, green, blue) {
if (this) {
throw new Error("RGB can't be instantiated");
}
var result = "#";
result += toHex(red);
result += toHex(green);
result += toHex(blue);
function toHex(dec) {
var result = dec.toString(16);
if (result.length < 2) {
result = "0" + result;
}
return result;
}
return result;
}
Anyway, in the end I just decided to turn my RGB() pseudoclass into an rgb() function, so I just won't try to instantiate it, thus needing no safety check at all. But that would depend on what you're trying to do.
function createConstructor(func) {
return func.bind(Object.create(null));
}
var myClass = createConstructor(function myClass() {
if (this instanceof myClass) {
console.log('You used the "new" keyword');
} else {
console.log('You did NOT use the "new" keyword');
return;
}
// constructor logic here
// ...
});
On the top of the question, below code will auto-fix the issue in case function is called without new.
function Car() {
if (!(this instanceof Car)) return new Car();
this.a = 1;
console.log("Called as Constructor");
}
let c1 = new Car();
console.log(c1);
This can achieved without using ES6 new.target. You can run your code in strict mode and in this case value of this will be undefined if called without new otherwise it will be empty object.
Example::
"use strict"
function Name(){
console.log(this)
if(this){
alert("called by new")
}
else
alert("did not called using new")
}
new Name()
Use this instanceof arguments.callee (optionally replacing arguments.callee with the function it's in, which improves performance) to check if something is called as a constructor. Do not use this.constructor as that can be easily changed.