Since in JavaScript, every uninitiated property returns undefined.
var a = {b:1};
a.x; // Undefined
Then to check if value exists or not, it's common to do this:
if (a.x !=== undefined) {..
I know hasOwnProperty is the better choice. However, you do find code that does the above, or simply !!a.x.
This is a problem. Now you don't know whether the value didn't get assigned or the value was assigned, but what was being assigned was empty.
Given this fact, should it be considered a bad practice to do the following anytime...
a.x = undefined;
Or are there situations where this is required or even preferred?
To remove a property of an Object
delete a.x;
I wouldn't say setting a.x = undefined is bad practice, if the goal is to set it to undefined.
But if used to simulate removing a property, it is.
One difference is that values set with undefined will still show up when using a for-in loop, or when using Object.keys() while unset or missing values will obviously not show up.
Personally I would use null for values that are purposefully supposed to be empty or uninitialized since null is valid in JSON while undefined is not, but if you're not worried about that then you can use undefined to.
As a somewhat contrived example off the top of my head, this could be helpful to explicitly unset an object's variables through some kind of extend function that uses a for-in loop. Or alternatively you could use it to prevent overwriting certain variables.
Check out the following example:
class MyClass {
static greet(){
if(!this.say){
return console.log("I don't know how to greet")
}
if(!this.to){
return console.log("I don't know who to greet")
}
console.log(this.say, this.to)
}
}
MyClass.say = "Hello"
MyClass.to = "World"
function extend(klass, obj, canOverwriteWithUndefined){
for(let key in obj){
if(typeof obj[key] === 'undefined' && canOverwriteWithUndefined === false){
throw "Cannot overwrite value with undefined key: " + key
}
klass[key] = obj[key]
}
}
const a = { say : "Hello", to : "World" }
const b = { say : undefined, to : "StackOverflow" }
const c = { say : "Hello" }
const d = { to : undefined }
extend(MyClass, a)
MyClass.greet() // Hello World
extend(MyClass, b)
MyClass.greet() // I don't know how to greet
extend(MyClass, c)
MyClass.greet() // Hello StackOverflow
extend(MyClass, d, false)
MyClass.greet() // Error
There is a difference between an unassigned value and an empty value. By default the value is always undefined and your code should always check for undefined to avoid unwanted output.
If for some reason the value of the object's property has to be removed, you should set it to empty (like an empty string or numeric 0 or an empty object/array). So the code written using that property never fails.
If at all you need to set it undefined, you may delete the property itself from the object but that would mean, your code would need to check if the property exists using the hasOwnProperty method.
I've been writing JavaScript for quite a long time now, and I have never had a reason to use null. It seems that undefined is always preferable and serves the same purpose programmatically. What are some practical reasons to use null instead of undefined?
I don't really have an answer, but according to Nicholas C. Zakas, page 30 of his book "Professional JavaScript for Web Developers":
When defining a variable that is meant
to later hold an object, it is
advisable to initialize the variable
to null as opposed to anything else.
That way, you can explicitly check for the value null to determine if
the variable has been filled with an object reference at a later time
At the end of the day, because both null and undefined coerce to the same value (Boolean(undefined) === false && Boolean(null) === false), you can technically use either to get the job done. However, there is right way, IMO.
Leave the usage of undefined to the JavaScript compiler.
undefined is used to describe variables that do not point to a reference. It is something that the JS compiler will take care for you. At compile time the JS engine will set the value of all hoisted variables to undefined. As the engine steps through the code and values becomes available the engine will assign respective values to respective variables. For those variables for whom it did not find values, the variables would continue to maintain a reference to the primitive undefined.
Only use null if you explicitly want to denote the value of a variable as having "no value".
As #com2gz states: null is used to define something programmatically empty. undefined is meant to say that the reference is not existing. A null value has a defined reference to "nothing". If you are calling a non-existing property of an object, then you will get undefined. If I would make that property intentionally empty, then it must be null so you know that it's on purpose.
TLDR; Don't use the undefined primitive. It's a value that the JS compiler will automatically set for you when you declare variables without assignment or if you try to access properties of objects for which there is no reference. On the other hand, use null if and only if you intentionally want a variable to have "no value".
Sidebar: I, personally, avoid explicitly setting anything to undefined (and I haven't come across such a pattern in the many codebases/third party libs I've interacted with). Also, I rarely use null. The only times I use null is when I want to denote the value of an argument to a function as having no value, i.e.,:
function printArguments(a,b) {
console.log(a,b);
}
printArguments(null, " hello") // logs: null hello
null and undefined are essentially two different values that mean the same thing. The only difference is in the conventions of how you use them in your system. As some have mentioned, some people use null for meaning "no object" where you might sometimes get an object while undefined means that no object was expected (or that there was an error). My problem with that is its completely arbitrary, and totally unnecessary.
That said, there is one major difference - variables that aren't initialized (including function parameters where no argument was passed, among other things) are always undefined.
Which is why in my code I never use null unless something I don't control returns null (regex matching for example). The beauty of this is it simplifies things a lot. I never have to check if x === undefined || x === null, I can just check x === undefined. And if you're in the habit of using == or simply stuff like if(x) ... , stop it.
!x will evaluate to true for an empty string, 0, null, NaN - i.e. things you probably don't want. If you want to write javascript that isn't awful, always use triple equals === and never use null (use undefined instead). It'll make your life way easier.
undefined is where no notion of the thing exists; it has no type, and it's never been referenced before in that scope; null is where the thing is known to exist, but it has no value.
Everyone has their own way of coding and their own internal semantics, but over the years I have found this to be the most intuitive advice that I give people who ask this question: when in doubt, do what JavaScript does.
Let's say you are working with object properties like options for a jQuery plugin...ask yourself what value JavaScript gives a property that has yet to be defined -- the answer is undefined. So in this context, I would initialize these types of things with 'undefined' to be consistent with JavaScript (for variables, you can do var myVar; instead of var myVar = undefined;).
Now let's say you are doing DOM manipulation...what value does JavaScript assign to non-existent elements? The answer is null. This is the value I would initialize with if you are creating a placeholder variable that will later hold a reference to an element, document fragment, or similar that relates to the DOM.
If you're working with JSON, then a special case needs to be made: for undefined property values, you should either set them to "" or null because a value of undefined is not considered proper JSON format.
With this said, as a previous poster has expressed, if you find that you're initializing stuff with null or undefined more than once in a blue moon, then maybe you should reconsider how you go about coding your app.
You might adopt the convention suggested here, but there really is no good reason to. It is not used consistently enough to be meaningful.
In order to make the convention useful, you first must know that the called function follows the convention. Then you have to explicitly test the returned value and decide what to do. If you get undefined, you can assume that some kind of error occurred that the called function knew about. But if an error happened, and the function knew about it, and it is useful to send that out into the wider environment, why not use an error object? i.e. throw an error?
So at the end of the day, the convention is practically useless in anything other than very small programs in simple environments.
A few have said that it is ok to initialise objects to null. I just wanted to point out that destructuring argument defaults don't work with null. For example:
const test = ({ name } = {}) => {
console.log(name)
}
test() // logs undefined
test(null) // throws error
This requires performing null checks prior to calling the function which may happen often.
A useful property in null that undefined does not qualifies:
> null + 3
3
> undefined + 3
NaN
I use null when I want to 'turn off' a numeric value,
or to initialize some. My last use was manipulating css transform:
const transforms = { perspective : null, rotateX : null };
// if already set, increase, if not, set to x
runTimeFunction((x) => { trasforms.perspective += x; });
// still useful, as setting perspective to 0 is different than turning it off
runTimeFunction2((x) => { transforms.perspective = null; });
// toCss will check for 'null' values and not set then at all
runTimeFunction3(() => { el.style.transform = toCss(transforms); });
Not sure if I should use this property thought...
DOM nodes and elements are not undefined, but may be null.
The nextSibling of the last child of an element is null.
The previousSibling of the first child is null.
A document.getElementById reference is null if the element does not exist in the document.
But in none of these cases is the value undefined; there just is no node there.
Unknown variable: undefined.
Known variable yet no value: null.
You receive an object from a server, server_object.
You reference server_object.errj. It tells you it’s undefined. That means it doesn’t know what that is.
Now you reference server_object.err. It tells you it’s null. That means you’re referencing a correct variable but it’s empty; therefore no error.
The problem is when you declare a variable name without a value (var hello) js declares that as undefined: this variable doesn’t exist; whereas programmers mostly mean: “I’ve not given it a value yet”, the definition of null.
So the default behavior of a programmer—declaring a variable without a value as nothing—is at odds with js—declaring it as not existing. And besides, !undefined and !null are both true so most programmers treat them as equivalent.
You could of course ensure you always do var hello = null but most won’t litter their code as such to ensure type sanity in a deliberately loosely-typed language, when they and the ! operator treat both undefined and null as equivalent.
In JavaScript, the value null represents the intentional absence of any object value. null expresses a lack of identification, indicating that a variable points to no object.
The global undefined property represents the primitive value undefined.
undefined is a primitive value automatically assigned to variables.
undefined is meant to say that the reference is not existing.
I completely disagree that usage null or undefined is unnecessary.
undefined is thing which keeping alive whole prototype chaining process.
So compiler only with null can't check if this property just equal to null, or its not defined in endpoint prototype. In other dynamic typed languages(f.e. Python) it throws exception if you want access to not defined property, but for prototype-based languages compiler should also check parent prototypes and here are the place when undefined need most.
Whole meaning of using null is just bind variable or property with object which is singleton and have meaning of emptiness,and also null usage have performance purposes. This 2 code have difference execution time.
var p1 = function(){this.value = 1};
var big_array = new Array(100000000).fill(1).map((x, index)=>{
p = new p1();
if(index > 50000000){
p.x = "some_string";
}
return p;
});
big_array.reduce((sum, p)=> sum + p.value, 0)
var p2 = function(){this.value = 1, p.x = null};
var big_array = new Array(100000000).fill(1).map((x, index)=>{
p = new p2();
if(index > 50000000){
p.x = "some_string";
}
return p;
});
big_array.reduce((sum, p)=> sum + p.value, 0)
I'm working through this exact question right now, and looking at the following philosophy:
Any function that is intended to return a result should return null if it fails to find a result
Any function that is NOT intended to return a result implicitly returns undefined.
For me, this question is significant because anyone calling a function that returns a result should have no question as to whether to test for undefined vs null.
This answer does not attempt to address:
Property values of null vs undefined
Variables within your functions being null vs undefined
In my opinion, variables are your own business and not a part of your API, and properties in any OO system are defined and therefore should be defined with value different from what they would be if not defined (null for defined, undefined is what you get when accessing something that is not in your object).
Here's a reason: var undefined = 1 is legal javascript, but var null = 1 is a syntax error. The difference is that null is a language keyword, while undefined is, for some reason, not.
If your code relies on comparisons to undefined as if it's a keyword (if (foo == undefined) -- a very easy mistake to make) that only works because nobody has defined a variable with that name. All that code is vulnerable to someone accidentally or maliciously defining a global variable with that name. Of course, we all know that accidentally defining a global variable is totally impossible in javascript...
Just wanna add that with usage of certain javascript libraries, null and undefined can have unintended consequences.
For example, lodash's get function, which accepts a default value as a 3rd argument:
const user = {
address: {
block: null,
unit: undefined,
}
}
console.log(_.get(user, 'address.block', 'Default Value')) // prints null
console.log(_.get(user, 'address.unit', 'Default Value')) // prints 'Default Value'
console.log(_.get(user, 'address.postalCode', 'Default Value')) // prints 'Default Value'
Another example: If you use defaultProps in React, if a property is passed null, default props are not used because null is interpreted as a defined value.
e.g.
class MyComponent extends React.Component {
static defaultProps = {
callback: () => {console.log('COMPONENT MOUNTED')},
}
componentDidMount() {
this.props.callback();
}
}
//in some other component
<MyComponent /> // Console WILL print "COMPONENT MOUNTED"
<MyComponent callback={null}/> // Console will NOT print "COMPONENT MOUNTED"
<MyComponent callback={undefined}/> // Console WILL print "COMPONENT MOUNTED"
There are already some good answers here but not the one that I was looking for. null and undefined both "technically" do the same thing in terms of both being falsy, but when I read through code and I see a "null" then I'm expecting that it's a user defined null, something was explicitly set to contain no value, if I read through code and see "undefined" then I assume that it's code that was never initialized or assigned by anything. In this way code can communicate to you whether something was caused by uninitialized stuff or null values. Because of that you really shouldn't assign "undefined" manually to something otherwise it messes with the way you (or another developer) can read code. If another developer sees "undefined" they're not going to intuitively assume it's you who made it undefined, they're going to assume it's not been initialized when in fact it was. For me this is the biggest deal, when I read code I want to see what it's telling me, I don't want to guess and figure out if stuff has "actually" been initialized.
Not even to mention that using them in typescript means two different things. Using:
interface Example {
name?: string
}
Means that name can be undefined or a string, but it can't be null. If you want it null you have to explicitly use:
interface Example {
name: string | null
}
And even then you'll be forced to initialize it at least with "null".
That's of course only true if you're using "strictNullChecks": true in tsconfig.json.
Based on a recent breakage we ran into, the example below shows why I prefer to use undefined over null, unless there is a specific reason to do otherwise:
function myfunc (myArg) {
if (typeof myArg === 'string') {
console.log('a', myArg);
} else if (typeof abc === 'object') {
console.log('b', myArg);
if (myArg.id) {
console.log('myArg has an id');
} else {
console.log('myArg has an id');
}
} else {
console.log('no value');
}
}
The following values will play nicely:
'abc'
{}
undefined
{ id: 'xyz' }
On the other hand the assumption of null and undefined being equivalent here breaks the code. The reason being is that null is of type of object, where as undefined is of type undefined. So here the code breaks because you can't test for a member on null.
I have seen a large number of cases with code of similar appearance, where null is just asking for problems:
if (typeof myvar === 'string') {
console.log(myvar);
} else if (typeof myvar === 'object') {
console.log(myvar.id);
}
The fix here would be to explicitly test for null:
if (typeof myvar === 'string') {
console.log(myvar);
} else if (myvar !== null && typeof myvar === 'object') {
console.log(myvar.id);
}
My attitude is to code for the weaknesses of a language and the typical behaviours of programmers of that language, hence the philosophy here of going with 'undefined' bey default.
To write simple code you need to keep complexity and variation down. When a variable or a property on an object does not have a value it is undefined , and for a value to be null you need to assign it a null value.
Undeclared vs Null
null is both an Object "type" and one of the 7 unique primitive value types called null
undefined is both a global scope property and type called undefined and one of the 7 unique primitive value types called undefined (window.undefined) .
It is the primitive types we use as values we are interested in.
In the case of null, as a value type it means an empty value has been assigned to a variable, but the variable type (Number, String, etc) is still defined. It just has no value. That is what null means. It means a variable has an empty value but it is still a value. It also reinitializes the variable with some kind of value, but is not undefined as a type.
undefined is a special case. When you declare a variable (or use a missing value not yet declared) it is of type undefined, as the browser does not know what type of data has been assigned to it yet. If the variable is declared but not assigned a value is is assigned the primitive calue undefined by default prior to assigning a value, and implies the variable does not exist or exists but has no value assigned.
Like null, undefined is also a primitive value type. But unlike null it means the variable does not exist, where null means the value does not exist. That is why its always better to check if the variable exists and has been assigned a variable using undefined before checking if the value is null or empty. undefined implies no variable or object exists in the compilation at all. The variable has either not been declared or declared with a missing value so not initialized. So checking for undefined is a very good way to avoid many types of errors in JavaScript and supersedes null.
That is why I would not rely on "truthy" checks for true/false with null and undefined, even though they will both return a false response, as undefined implies an additional step for missing feature, object, or variable, not just a true/false check. It implies something more. If you have a missing undeclared variable, truthy statements will trigger an ERROR!
Let's look at undefined first:
//var check1;// variable doesnt even exist so not assigned to "undefined"
var check2;// variable declared but not initialized so assigned "undefined"
var check3 = 'hello world';// variable has a value so not undefined
console.log('What is undefined?');
//console.log(check1 === undefined);// ERROR! check1 does not exist yet so not assigned undefined!
console.log(check2 === undefined);// True
console.log(check3 === undefined);// False
console.log(typeof check1 === 'undefined');// True - stops the ERROR!
console.log(typeof check2 === 'undefined');// True
console.log(typeof check3 === 'undefined');// False
As you can see undeclared variables, or declared but not initialized, both are assigned a type of undefined. Notice declared variables that are not initialized are assigned a value of undefined, the primitive value type but variables that do not exist are undefined types.
null has nothing to do with missing variables or variables not yet assigned values, as null is still a value. So anything with a null is already declared and initialized. Also notice a variable assigned a null value is actually an object type unlike undefined types. For example...
var check4 = null;
var check5 = 'hello world';
console.log('What is null?');
console.log(check4 === undefined);// False
console.log(check5 === undefined);// False
console.log(typeof check4 === 'undefined');// False
console.log(typeof check5 === 'undefined');// False
console.log(typeof check4);// return 'object'
console.log(typeof check5);// return 'string'
As you can see each act differently and yet both are primitive values you can assign any variable. Just understand they represent different states of variables and objects.
Does this code would work as expected in all browsers? Is there any notes in specification about it?
var attributes = this._attributes ? this._attributes : (this._attributes = []);
I.e. if *this._attributes* not initialized, then new array will be created and that array will be assigned to *this._attributes* and to attributes.
There's nothing special about that expression, and you'll have no problems in any major browser. You could shorten it by using the || operator:
var attributes = this._attributes || (this._attributes = []);
That will work in all browsers.
It could be actually made terser with...
var attributes = this._attributes || (this._attributes = []);
No, I think unfortunately you may not access _attributes when it's undefined. So you have to check typeof attributes != "undefined".
I don't see any reason why not. I don't think I'd write it that way, I'm not a fan of assignments with side-effects, but syntactically it's fine.
This works just fine, when accessing an undefined property of any object, that access will return undefined. The single thing you have to watch out for is that you don't extend the Object.prototype to have a _attributes attribute because this will screw you up, but then again, never extend native prototypes.
From the spec :
8.12.2 [[GetProperty]] (P)
Let prop be the result of calling the [[GetOwnProperty]] internal method of O with property name P.
If prop is not undefined, return prop.
Let proto be the value of the [[Prototype]] internal property of O.
If proto is null, return undefined.
Return the result of calling the [[GetProperty]] internal method of proto with argument P.
So it checks whether the object has the property, if so it returns it, if not it searches up the prototype chain, if it finds something there it returns it, otherwise it returns undefined.
In the book Javascript the good parts, on the opening page of Ch3 on objects, it states:
An object is a container of properties, where a property has a name and a value. A property name can be any string, including the empty string. A property value can be any Javascript value except for undefined.
Note: undefined is highlighted in the book to denote that is is a literal.
In practice, however, I am able to do it.
var a = { "name": undefined };
What is wrong with my understanding ?
I believe the answer is that he's wrong.
As you observe, you can set var a = { "name": undefined };.
a.name === undefined
a.name === a.someFakeProperty
Here's where they're different, though:
'someFakeProperty' in a === false
'name' in a === true
Or, to do it a different way,
a.hasOwnProperty('someFakeProperty') === false
a.hasOwnProperty('name') === true
Using the somewhat infamous for..in loop,
for (var i in a) {
alert(i);
}
... will give you name.
So, by value you may not be able to distinguish undefined and undefined, but they are quite different internally.
Addition: he's wrong about the property names, too - a[window] = 43; a[window] == 43; is just fine. Sure, you can't then do a.window, but a.___ is just syntactic sugar for a['___']. As noted in the comments, property names are cast to string so I was wrong about this bit.
I don't like the terminology that Crockford uses, he seem to mix the concept of undefined and undeclared.
The statement:
A property value can be any Javascript value except for undefined.
Is completely wrong IMO, because undefined is a primitive value of the language.
See also:
Difference between undefined and not being defined in Javascript
I think what he's trying to say is that a property's value cannot be undefined because undefined is exactly how JavaScript denotes properties that don't exist. In other words, if you have the following object
var a = { "name": undefined };
Then a.name is undefined, but so is a.someFakeProperty. However, as Ben Lee points out in his comment, name will still show up when you iterate the properties of a using a for loop, while someFakeProperty will not. Therefore, it seems that Crockford was a bit imprecise in conveying the idea in question here.
Perhaps this:
var a = {"name": x}; //x is undefined, so it will cause an error.
Just my understanding.
Say I have the following code:
function One() {}
One.prototype.x = undefined;
function Two() {}
var o = new One();
var t = new Two();
o.x and t.x will both evaluate to undefined. o.hasOwnProperty('x') and t.hasOwnProperty('x') will both return false; the same goes for propertyIsEnumerable. Two questions:
Is there any way to tell that o.x is defined and set to undefined?
Is there ever any reason to? (should the two be semantically equivalent?)
A small caveat: doing (for propName in o) loop will yield 'x' as one of the strings, while doing it in t will not - so there IS a difference in how they're represented internally (at least in Chrome).
A slightly simpler way than your method is to use the Javascript in operator
alert('x' in o); // true
alert('x' in t); // false
object.hasOwnProperty(name) only returns true for objects that are in the same object, and false for everything else, including properties in the prototype.
function x() {
this.another=undefined;
};
x.prototype.something=1;
x.prototype.nothing=undefined;
y = new x;
y.hasOwnProperty("something"); //false
y.hasOwnProperty("nothing"); //false
y.hasOwnProperty("another"); //true
"someting" in y; //true
"another" in y; //true
Additionally the only way do delete a property is to use delete. Setting it to undefined do NOT delete it.
The proper way to do it is to use in like roborg said.
Update:
undefined is a primitive value, see ECMAScript Language Specification section 4.3.2 and 4.3.9.
Ach, one for the anti-patterns thread there.
undefined is not a keyword.
When you assign write var foo = undefined; you are assigning it the value of the symbol undefined which you haven't defined and hence the value you'll get is "undefined". You would produce exactly the same result if you had assigned it UnDeFiNeD or NotFineAtAll or _qwertyuiop
Why is this so terrible? Well apart from the fact that it leads to a false understanding of what's going on, what happens if you happen to load a library or hire a dev who writes var undefined = true;
Something like this?
function isDefined(var){
return (typeof(window[var]) == 'undefined') ? false : true;
}
or..
function isDefined(var){
return (window.var === undefined) ? false : true;
}
No. I don't think a variable equal to undefined should be recognized as "defined".
Setting it equal to undefined directly is just a lazy way of deleting it -- as apposed to using the delete keyword. I believe this just means that garbage collection hasn't taken affect on the variable/property.
[EDIT]
Regarding your comment about hasOwnProperty and propertyIsEnumerable, prototype methods/properties are neither the object's own nor enumerable.
Object.prototype.foo = 'something';
var o = {bar: 'anything'};
o.hasOwnProperty('foo'); //=> false
o.hasOwnProperty('bar'); //=> true
Object.prototype.hasOwnProperty('foo'); //=> true
One way to do it is this:
var definedProperties = [];
for (var prop in o) definedProperties.push(prop);
if (o.x === undefined && contains(prop, 'x')) {
//x is defined, but set to undefined
}
where contains just checks whether any element in the array is 'x'. Is this the only way?
You could try putting a watch on One.x but that would only tell you at the time it gets undefined.