[tambourine_man]

Agreed, although some things like:

  NaN === NaN // -> false 

Are pretty f*&^%$ up, no matter how you look at them

Edit: well, apparently there is a non f&^%$ up way to look at it: https://news.ycombinator.com/item?id=14891810

[xyclos]

NaN is meant to represent a non-sensical mathematical operation (like divide by 0). One non-sensical mathematical operation is not the same as some other non-sensical mathematical operation (1/0 !== Infinity/Infinity)

[Pigo]

That's a good point, a lot of this seems like it can be explained with some context or just asking why you'd ever want to do that.

[gmiller123456]

>NaN is meant to represent a non-sensical mathematical operation

This is not a very good explanation. NaN is used to represent a nonsensical value. By your explanation (1/0 === 1/0) should be true since they represent the same invalid opperation. The fact that NaN != NaN is not meant to mean anything, it is merely defined that way to prevent a class of bugs from occurring. Anyone not familiar with the definition is correct to be confused by it. I'd think if the spec was designed today, this special case (hack) wouldn't be there and we'd have exceptions in its place.

[justinpombrio]

Nor is it the same as the same non-sensical mathematical operator :-)

1/0 !== 1/0

[tambourine_man]

That's mad. In JS, 1/0 is Infinity, not Undefined. And Infinity is not equal to itself.

[tambourine_man]

That's a good explanation, thanks.

[roywiggins]

That's just normal floating point behavior. NaN isn't equal to anything, not even itself. It's in the standard.

[tambourine_man]

Floating point can be a bit mind bending. Still, something not being equal to itself almost pushes the barrier into philosophy.

[roywiggins]

As long as you can define what's going on rigorously, it's just mathematics :) The equality operator means just what I choose it to mean- neither more nor less, as Humpty Dumpty would say.

Also fun: (NaN < x) should be false for every x, and (NaN > x) should be false for every x. It's not a number, so it's not less than or greater than any number. But then by process of elimination, NaN == x for every x. Which is obviously nonsense. So equality and comparison is completely "broken" anyway when you start comparing NaN to things.

[fmihaila]

NaN is not a name for one thing. You could think of not a number as not any particular thing. There is no comparison with itself, because there is no it.

[goatlover]

How is isNaN() implemented?

[Gaelan]

    function isNaN(x) { return x !== x }

It’s a native browser function so it may not actually be done that way, but that’s one way you could do it.

[tambourine_man]

We really are deep in the realm of Philosophy.

[fny]

There's a really great SO on why this is an IEEE standard: https://stackoverflow.com/questions/1565164/what-is-the-rati...

[tambourine_man]

Great discussion indeed. I was thinking along the same lines, the comparison may be meaningless, but “false” is not a good answer either, even though a Boolean is either true or false by definition and NotABool would be even more crazy. What a rabbit hole.

[aisofteng]

That's part of IEEE 754.

[tambourine_man]

I didn't know it was part of IEEE 754, thanks. So JS is not alone in this.