Ben Halpern

Posted on Aug 29, 2019

What is the oddest JavaScript behavior?

#discuss #javascript #languages

JavaScript is notorious for its inconsistencies in a lot of areas. It's also powerful and popular and has a lot going for it.

But can we poke fun at weird things? (whether or not the behavior is a good thing in general)

Let's start with an example:

Aug 12 '19

625

+'a' resolves to NaN ("Not a Number") because it coerces a string to a number, while the character a cannot be parsed as a number

document.write(+'a');

To lowercase it becomes banana.

Adding NaN to "ba" turns NaN into the string "NaN" due to type…

Open Full Answer

Top comments (85)

Paul • Aug 29 '19

'11' + 1 // 111
'11' - 1 // 10

XANOZOID • Aug 30 '19 • Edited

I love this. It just made me think about something pretty interesting... If you think of this in terms of what it actually could mean for a mathematical function, then the string appendage (regarding integers) could be thought of as its own math operation too - in a very round about way.

Max Cerrina • Sep 3 '19

Absolutely, and not even in a very roundabout way, you're pretty spot on :)

leob • Aug 30 '19 • Edited

Question: should we (as JS coders) assume that people know this, and not hesitate to write this sort of code? or should we refrain from this and use explicit type conversion ...

(well the answer is probably that if it is 100% clear that the data type of the first operand is always string and the second operand is always number then this would be okay - if in obscure corner cases it can sometimes suddenly be different then good luck debugging/troubleshooting ...)

Victor Aremu • Aug 29 '19

😂😂

Jacob Evans • Aug 30 '19

JS coercion at its best and worst in two lines lmao!

Brad • Aug 29 '19 • Edited

This took an hour of my life:

typeof null === 'object' //true

Anthony Marquez • Aug 30 '19

omg thank you for your service as I really didn’t know about this.

willsmart • Aug 30 '19 • Edited

yep, a good check for an actual object is

thing && typeof(thing)==='object'

Kind of insane they did that that.

Derk-Jan Karrenbeld • Aug 30 '19 • Edited

It's not insane. null, like nil in ruby, a singleton instance of NilClass is technically an object.

Technically an array is also an object, which is why there is Array.isArray (which has its own caveats).

willsmart • Aug 30 '19 • Edited

But at least ruby is rigorously consistent OO like that.

irb(main):004:0> nil.methods
=> [:&, :^, :|, :===, :inspect, :to_a, :to_s, :to_i, :to_f, :nil?, :to_h, :to_r, :rationalize, :to_c, :tap, :public_send, :instance_variables, :instance_variable_set, :instance_variable_defined?, :remove_instance_variable, :private_methods, :kind_of?, :is_a?, :instance_variable_get, :method, :public_method, :singleton_method, :instance_of?, :extend, :define_singleton_method, :to_enum, :enum_for, :<=>, :=~, :!~, :eql?, :respond_to?, :freeze, :object_id, :display, :send, :hash, :class, :singleton_class, :clone, :dup, :itself, :taint, :tainted?, :untaint, :untrust, :trust, :untrusted?, :methods, :protected_methods, :frozen?, :public_methods, :singleton_methods, :!, :==, :!=, :__send__, :equal?, :instance_eval, :instance_exec, :__id__]

In JS there are very few object-type calls that work in any way with nulls, so it's neat that they aimed to make this aspect of the language more OO, but they should have followed through better.
typeof(null) being 'object' is seldom anything but annoying for JS devs.

eg.

> Object.getOwnPropertyNames(null)
< Uncaught TypeError: Cannot convert undefined or null to object
    at Function.getOwnPropertyNames (<anonymous>)
    at <anonymous>:1:8
> null.foo
< Uncaught TypeError: Cannot read property 'foo' of null
    at <anonymous>:1:6
(anonymous) @ VM176:1
> for (const k of null);
< Uncaught TypeError: null is not iterable
    at <anonymous>:1:17

Basically there are very few times you might write typeof(thing)==='object' where you shouldn't do the null check too

Derk-Jan Karrenbeld • Aug 30 '19

Absolutely true. I was only referring to the use of the word "insane" which just sounds very unknowing/ignorant to me.

FWIW, now that we use TS for most things, this is not problem for us often, except for when something is nullable ánd multiple types (object/number/string/undefined).

My recommendation is to never use null and only use undefined. Replace your "null"s with an "EMPTY/UN_SET" data type.

willsmart • Sep 1 '19

Fair enough. I didn't mean insane in the sense of completely without reason, just in the sense that it was badly thought out.

The thing is that nulls are now baked into JS at a few levels, so in our own code we can avoid them, but they pop up throughout the various JS apis, eg

> /a/.exec('b')
< null
> document.body.getAttribute('a')
< null

That's what I mean by the loose "insane" comment. It was a mistake that was bad enough that as JS devs we're now best off to generally avoid the fundamental null value in JS.

I totally agree with you about TS, which corrects many of these issues, and how in JS we're best to use undefined whereever possible.

Brian Emilius • Aug 30 '19

Or
Object.prototype.toString.call(thing)

Rémy 🤖 • Aug 29 '19

Without a doubt I'll say this: jsfuck.com/

Ryan • Aug 30 '19

My work firewall thinks this is porn.

Yoandy Rodriguez Martinez • Aug 30 '19

Ryan, HR needs to now why you're searching for programmers porn during work hours

Ryan • Aug 31 '19

Clearly :D

Ben Lovy • Aug 29 '19

check out the source for a full list of the nonsense they leverage. I feel like "explain each line of this file to yourself" is probably a pretty good JS litmus test.

Kasey Speakman • Aug 29 '19

Wat is a funny talk about odd behaviors. JS starts at 1:22.

Alexander Holman • Aug 29 '19 • Edited

A colleague the other day noticed that the setTimeout/setInterval takes a 32 bit signed int as an argument, which is fine except that (as far as I know) JS doesn't natively support integers... but but instead Numbers are specifically 8 byte signed doubles. If you happen to have a interval or timeout every 2^31 miliseconds (ish) or more, then the value is interpreted as a negative number and therefore executes instantly and in terms of intervals repeatedly. So timeouts and intervals are limited to just under 25 days.

Shayne Darren • Aug 30 '19

I'm guessing a web app won't need a 25-day timeout or interval.

I tried a very basic test on node and it seems to be waiting on it to execute... any idea why?

Alexander Holman • Aug 30 '19 • Edited

not sure... i just ran it in node and experienced the same issue as the browser. The following line should make node print without pausing:

setInterval(() => {console.log('log')}, 2**31)

setInterval(() => {console.log('log')}, -1)

where the following will take 596.5 hours (roughly):

setInterval(() => {console.log('log')}, 2**31-1)

I am currently on a Windows machine running v8.9.1, perhaps its solved in later versions?

Seb Scholl • Aug 29 '19

My favorites...

alert(typeof NaN); //alerts 'Number'
alert(NaN === NaN); //evaluates false

Austin S. Hemmelgarn • Aug 30 '19

Most languages do this, but it could be a whole lot worse (fun fact, IEEE 754 actually allows for (2 ^ N) - 1 distinct NaN values not counting signed NaN values, where N is the bit width of the significand field in the binary representation (so for the 8-byte floats JS uses, there are 18014398509481982 possible NaN values including both signed values)).

The reason is simple, NaN is a numeric representation of a non-numeric value, so by definition it has to be the same type as numbers for the language (or, at minimum, it has to be at least one of the numeric types in the language), but you can't be certain what value it is, so you can't treat two NaN values as being identical.

Alex K. • Aug 30 '19 • Edited

I think most of the people get confused by the language's type coercion rules and operations related to those, which are quite often not intuitive:

0 == []            // true
[] == false        // true
[] + 1             // "1"
[] + "1"           // "1"
[] - 1             // -1
[] - "1"           // -1
{} + 1             // 1 
{} + "1"           // 1
{} - 1             // -1
{} - "1"           // -1
{} + []            // 0
{} == []           // Uncaught SyntaxError: Unexpected token ==
[] == {}           // false

Ben Calder • Aug 30 '19 • Edited

The only thing that surprises me here are the loose equality comparisons between object and array; but it's not as though any sane programmer would be doing this. Otherwise almost all the others look fairly intuitive: coercion is applied to satisfy the operator being applied.

It's perhaps more useful to understand all the values that can be coerced to false; and why sometimes it's a really bad idea to use a shortcut if(isSomeValueTruthy) style condition...

edit - actually [] + 1 is weird :D

Austin S. Hemmelgarn • Aug 30 '19

Yeah, this is a good one. Most languages try to coerce only one side of an expression at a time, and won't ever coerce collection types to non-collection types.

Jordan Kicklighter • Aug 30 '19

Wow, this is a new one for me. The array coercion actually caught me off guard.

Dave Cridland • Aug 30 '19 • Edited

Scoping rules with (or without) var are the weirdest:


// Prints the value of a.
function printA() {
  console.log(a);
}

// Prints '1'
function print1() {
  var a = 1;
  printA();
}

// Prints '2'
function print2() {
  var a = 2;
  printA();
}

// Prints whatever the value of b in the caller's scope is.
function printB() {
  a = b;
  printA();
}

// Prints 3.
function print3() {
  var b = 3;
  printB();
}

// Print 4, then 5.
function print45() {
  var b = 4;
  var a = 5;
  printB();
  printA();
  // Ha, no it doesn't, it prints 4 twice, and now:
  assert(a === 4);
}

João Ferreira • Aug 30 '19

loved

Matt Ellen • Aug 30 '19

you have just blown my tiny mind

Dave Cridland • Aug 30 '19

Of course, I made a mistake on it (now corrected), but still. If you want to see something batshit insane with scoping and hoisting, you can try and get your head around this:

function fib(n) {
  if (n) {
    a = 1;
    b = 0;
    c = n + 1;
  }
  if (c === 1) {
    return a + b;
  }
  c = c - 1;
  let d = a + b;
  b = a;
  a = d;
  return fib();
}

Matt Ellen • Aug 30 '19

😭

Antonin J. (they/them) • Aug 29 '19

Encountered this one yesterday and spent a day on it.

A File object cannot be serialized into JSON. There's a StackOverflow question about it.

But essentially if you do

console.log(someFile);

The browser will happily return something like:

File {
  name: 'filename',
  size: 1898921,
  // other properties
}

But doing

console.log(JSON.stringify(someFile))

will result in:

{}

With no properties. I've noticed that it does this for object spread as well so if you wanted to copy File info, you can't just do: { ...someFile }, you'll need to call and pick out the individual properties manually.

Victor Aremu • Aug 29 '19

This made my day 😂

Luke Westby • Aug 29 '19 • Edited

typeof document.all === 'undefined' // true
document.all instanceof HTMLAllCollection // true

Ken Bellows • Sep 1 '19

Imo the implicit type conversion is a little weird, but even ignoring that, I think the point was the difference in behavior between '11' + 1, which converts the number to a string, and '11' - 1, which converts the string to a number. This discrepancy makes implicit type casting feel unpredictable and weird, at noon least to me

Fernando Doglio • Aug 30 '19

Wouldn't you say that most of the "strange" behavior is actually associated with people not caring enough to read the actual specs? I mean, the example you listed is perfectly explained in the answers. It's not strange, it behaves as expected if you read the specs.
Same goes for a lot of the comments in this thread, especially those related to data types. If I write a spec for a new language saying that:

typeof 1 == number
typeof 2 == string

No matter how crazy that looks from the outside, if my language's inner logic makes sense, then it's completely valid and not crazy.

I guess what I'm trying to say here is that people need to start seeing JavaScript for the very special boy it is and stop trying to compare it with the rest of the class. It's different, but we love it anyway (at least I know I do :P)

CamDHall • Aug 30 '19

I don't really agree with this at all. Most of my experience is in JavaScript and I've read most of the specs if not all, at one time or another. It doesn't mean these things make "sense". Just because I know WHY it does dumb stuff, doesn't mean it's not dumb. Like, I know why people would try to rob a bank. Doesn't mean it's a perfectly rationale thing to do.

Fernando Doglio • Aug 30 '19

Well, the concept of "dumb" requires a context and to be compared against similar actions with different results. Thus you are taking into account expected behavior from others, even if you don't do it consciously. It's the same with your example: robbing a bank is bad, even if you know why they do it, because, in your context, you follow the law but then again, your context isn't the same as that of the robbers.
That is all I'm trying to say, the context here is everything. If we say JS's behavior is "dumb" because in other languages typeof NaN (or its equivalent) should never be "number", then we're taking things out of context. The ECMAScript standard states that Numbers should be IEEE-754 floating point data. This includes Infinity, -Infinity, and also NaN.

I don't mean to start a flame war here, so I hope I'm not offending anyone, all I'm just saying is that before claiming JS is stupid, dumb or behaving strangly, we should consider what we're comparing it against. That's all.

CamDHall • Aug 30 '19 • Edited

I started programming in JS. I spent probably 5 or 6 years working only with JS. But even before I started working with C# and C++, I thought the way JS handles things was clearly not the best way of handling them.

I can go further on the banking example. Even if I didn't follow the law, robbing a bank is not a smart move. There's easier targets. Less risky targets. Probably even targets that pay all most as well, with far less chance of being caught. So it's still pretty stupid, even in the context of being a criminal. I don't know a lot of thiefs, but I suspect if you ask a bunch of thiefs if robbing a bank is a good idea, they'd tell you no. Rob some poor guy leaving the atm or something.

JavaScript could definitely handle types better. Among other things. Even in the context of what it's used for. Typescript is maybe a good example of this.

I'm also not being malicious or anything. I Like JS, it's still what most of my side projects are in, even though I could choose to use something else. But it has problems. Some of which are self imposed.

lepinekong • Aug 31 '19 • Edited

From Deming / System Thinking: youtube.com/watch?v=OqEeIG8aPPk

A system must have an aim. What's the aim ;) Making the right things (good spec relative to that aim) is far more important than doing the things right (respect a spec) and that's all the problem in software industry in general: the problem is more about how to do the right product (so spec) than to do a product that respects a spec but people won't buy ;)

Doing a spec for a programming language often targets the machine first instead of human first, that should be reverse ;)

Jordan Kicklighter • Aug 30 '19

I mostly agree, however I would add that there were some less-than-ideal design decisions made earlier in the language's life that most people agree should be modified, but must remain so as to not break comparability. Which is a remarkable challenge.

Anton T • Aug 30 '19

The "everything is an object" approach is takes is such a double edged sword. On one hand it's nice to use index access for objects but typeof [] being object is just silly if you're coming from another language.

I had a stubborn bug where an array was refusing to be identified by Array.isArray in my code yet if I tried it in the console it would work. Even recreating it by inserting it into a blank array using spread syntax like [...myArray] still failed the isArray test :/

Adam Crockett 🌀 • Aug 29 '19

There is one that I can't remember precisely, but JavaScript cannot add 2.637 or something like that the result is wrong. I will have to look this up.

mlaj • Aug 29 '19

Adding floats. It won't give the right answer. It's a common problem in programming languages.

Adam Crockett 🌀 • Aug 30 '19

Yeah I had a suspension that was the case. Still it's odd and it doesn't not affect JavaScript so I'm technically correct. 🙃

Kasey Speakman • Aug 29 '19 • Edited

console.log(.3);
// 0.3

console.log(.1 + .2);
// 0.30000000000000004

Adam Crockett 🌀 • Aug 30 '19

That's the one! Thanks!

View full discussion (94 comments)

DEV Community

What is the oddest JavaScript behavior?

Top comments (85)

Read next

Middlewares = Magic? 3 Reasons Every Dev Should Adopt This Magic! ✨🔥

Creating a Multicolored Star Rating Card Component

Mastering JavaScript: Multiple Ways to Generate a Two-Dimensional Array

Someone finally fixed Javascript