loading...
Cover image for JavaScript quirks in one image from the Internet

JavaScript quirks in one image from the Internet

mkrl profile image Mikhail Korolev ・5 min read

Recently I stumbled upon this picture on Google Images:

Thanks for inventing JavaScript

The man on the picture is Brendan Eich by the way, the creator of JavaScript and a co-founder of the Mozilla project.

Even with some of the examples are not really related to the language itself, I still find it interesting enough to give it a short breakdown, considering it doesn't make much sense for some people with the classic "strict language"-only programming background.

The breakdown

Starting with the "boring" parts:

Floating-point arithmetic

> 9999999999999999
< 10000000000000000

> 0.5+0.1==0.6
< true

> 0.1+0.2==0.3
< false

Nothing really surprising here, it's an old concept that has been around for quite a while. And it is, of course, not related to JavaScript "features". Instead of trying to explain it here, I'll just leave a link the to this great "explainlikeimfive" website dedicated exclusively to explaining floating-point math.

Not A Number is a Number

> typeof NaN
< "number"

What is "NaN" after all? It is, in fact, a representation of a specific value that can't be presented within the limitations of the numeric type (the only JS numeric primitive is, in fact float). NaN was introduced in the IEEE 754 floating-point standard.

So, it's just a number that a computer can't calculate in this particular environment.

Type conversion

JavaScript is a dynamic type language, which leads to the most hated "why it this like that" debugging sessions for those who is not familiar with the silent (implicit) type coercion.

The simple part: strict equality with ===

> true === 1
< false

Strict equality compares two values. Neither value is implicitly converted to some other value before being compared. If the values have different types, the values are considered unequal. Boolean variable is not equal to 1, which is a number.

On the other hand, there is this:

> true == 1
< true

This is an example of implicit type coercion. Implicit type coercion is being triggered when you apply operators to values of different types: 2+'2', 'true'+false, 35.5+new RegExp('de*v\.to') or put a value into a certain context which expects it to be of a certain type, like if (value) { (coerced to boolean).

JavaScript type conversion is not the most trivial part, so I would suggest further reading like this great article by Alexey Samoshkin and this little MDN doc on equality comparisons. There is also this equality comparison cheatsheet that may come in handy.

Anyway, let's get back to our picture.

> [] + []
< ""

There are 2 types of variables in JS: objects and primitives, with primitives being number, string, boolean, undefined, null and symbol. Everything else is an object, including functions and arrays.

When an expression with operators that call implicit conversion is being executed, the entire expression is being converted to one of three primitive types:

  • string
  • number
  • boolean

Primitive conversions follow certain rules that are pretty straightforward.

As for the objects: In case of boolean, any non-primitive value is always coerced to true. For string and number, the following internal operation ToPrimitive(input, PreferredType) is being run, where optional PreferredType is either number or string. This executes the following algorithm:

  1. If input is already a primitive, return it as it is
  2. Otherwise, input is treated like an object. Call input.valueOf(). Return if the result is a primitive.
  3. Otherwise, call input.toString(). If the result is a primitive, return it.
  4. Otherwise, throw a TypeError.

Swap 2 and 3 if PreferredType is string.

Take a look at this pseudo-implementation of the above in actual JavaScript, plus the boolean conversion (the original is a courtesy of Alexey Samoshkin via the article mentioned previously).

function ToPrimitive(input, preferredType){

  switch (preferredType){
    case Boolean:
      return true;
      break;
    case Number:
      return toNumber(input);
      break;
    case String:
      return toString(input);
      break
    default:
      return toNumber(input);  
  }

  function isPrimitive(value){
    return value !== Object(value);
  }

  function toString(){
    if (isPrimitive(input.toString())) return input.toString();
    if (isPrimitive(input.valueOf())) return input.valueOf();
    throw new TypeError();
  }

  function toNumber(){
    if (isPrimitive(input.valueOf())) return input.valueOf();
    if (isPrimitive(input.toString())) return input.toString();
    throw new TypeError();
  }
}

So, at the end of the day, the original [] + [] == "" is being interpreted as:

ToPrimitive([]) + ToPrimitive([])

Both arrays return an empty string as a result of toString([]). The final result is a concatenation of two empty strings.

Now, onto the:

> [] + {}
< "[object Object]"

Because of the String({}) resulting in a [object Object] the result is a simple concatenation of "" and "[object Object]". Simple enough. Now what the hell is going on here then?

> {} + []
< 0

Turns out, JavaScript interprets the first {} as a code block! When the input is being parsed from start to end, it treats { as the beginning of the block, following by closing } immediately. Hence, using our pseudo-implementation the previous example will be evaluated into the following:

ToPrimitive(+[])

..which is 0. The + is an unary prefix operator that converts the operand into a number.

Loose equality == and binary + operators always trigger default preferredType, which assumes numeric conversion (except Date that returns string). That explains true+true+true===3 and true==1. Thus, as expected true===1 returns false because there are no operators on the left side of the expression and === does not trigger implicit type coercion. Same with []==0 which is roughly equivalent to Number([]) == 0.

Everything brings up interesting examples like the one we have here:

> (!+[]+[]+![]).length
< 9

Breaking it down,

  • (!+[]) + [] + (![])
  • !0 + [] + false
  • true + [] + false
  • true + '' + false
  • 'truefalse'

'truefalse'.length === 9

Very simple.

And last (and the least, to be honest):

Math.max() < Math.min() ?

> Math.max()
< -Infinity
> Math.min()
< Infinity

This may be considered as a minor language flaw, in terms of returning a kind of an unexpected result from a function that wants certain arguments.

But there actually is a little bit of some actual math behind that.

Let's make a guess on how (probably) Math.max() works and write down another imaginary transcript into actual JavaScript:

Math.max = function () {
  let temp = -Infinity;
  for ( let i = 0; i < arguments.length; i++ ) {
    let num = Number(arguments[i]);
    if ( num > temp ) {
      temp = num;
    }
  }
  return Number(temp);
}

Now it kind of makes sense to return -Infinity when no arguments are passed.

-Infinity is an identity element of Math.max(). Identity element for a binary operation is an element that leaves any other element unchanged after applying said operation to both elements.

So, 0 is the identity of addition, 1 is the identity of multiplication. x+0and x*1 is always x. Out of -Infinity and x, x will always be the maximum number.

There is an absolutely gorgeous article by Charlie Harvey that deeply dives into this topic.


Summing up, implicit type coercion is a very important concept you should always keep in mind. Avoid loose equality. Think about what are you comparing, use explicit conversion whenever possible. Consider switching to TypeScript if the above scares you :)

And if you want to see more "wtf" JavaScript, there is a great resource called, literally, wtfjs, that is also available as a handbook via npm.

Posted on Mar 1 '19 by:

Discussion

markdown guide
 

being a python dev who writes desktop applications & games, I'm still mystified by js. its interesting though, to actually see WHY it does what it does. this is the first time ive seen javascript's flaws discussed in a way that actually makes sense, even if those flaws still make my python-centric developer brain hurt. :)

 

It's ironic how JavaScript continues to dominate only because we can't get rid of it in the browser. In a lot ways I wish the world ran on Python instead. 😁

 

because of my experience working on SBTCVM, a base 3 virtual machine inspired by 50s Russian computers, specifically its 2 custom languages, I know language design isn't necessarily easy...

Whether its a scripting language or a compiled one, its not something that goes well if you rush, nor is it easy in any sense.

as far as scripting languages, they often depend heavily on what environment they are built for, even if such dependence isn't always obvious.

On another note, compiled languages come with their own set of challenges, challenges I'm sure some braver web developers might dive into once web assembly becomes more common.

Then again, writing programs for a 6.5Khz ternary CPU emulated in python, is a tad different than scripting hypertext :p

Sounds absolutely wild, I'd love to hear more about it. I'm looking forward to web assembly. I could see it potentially opening the flood gates for new web technologies. Best of luck with your ternary emulations. Edit: Looks like you've written quite a bit on the subject. Small typo in your bio 😁

 

Loved your article Mikhail. While I understood most examples and even knew how they work on the surface, your deep explanation and linked articles made me learn something new. Thank you.