What does ~~ ("double tilde") do in Javascript?

The first ~ operator forces the operand to an integer (possibly after coercing the value to a string or a boolean), then inverts the lowest 31 bits. Officially ECMAScript numbers are all floating-point, but some numbers are implemented as 31-bit integers in the SpiderMonkey engine.

You can use it to turn a 1-element array into an integer. Floating-points are converted according to the C rule, ie. truncation of the fractional part.

The second ~ operator then inverts the bits back, so you know that you will have an integer. This is not the same as coercing a value to boolean in a condition statement, because an empty object {} evaluates to true, whereas ~~{} evaluates to false.

js>~~"yes"
0
js>~~3
3
js>~~"yes"
0
js>~~false
0
js>~~""
0
js>~~true
1
js>~~"3"
3
js>~~{}
0
js>~~{a:2}
0
js>~~[2]
2
js>~~[2,3]
0
js>~~{toString: function() {return 4}}
4
js>~~NaN
0
js>~~[4.5]
4
js>~~5.6
5
js>~~-5.6
-5

In ECMAScript 6, the equivalent of ~~ is Math.trunc:

Returns the integral part of a number by removing any fractional digits. It does not round any numbers.

Math.trunc(13.37)   // 13
Math.trunc(42.84)   // 42
Math.trunc(0.123)   //  0
Math.trunc(-0.123)  // -0
Math.trunc("-1.123")// -1
Math.trunc(NaN)     // NaN
Math.trunc("foo")   // NaN
Math.trunc()        // NaN

The polyfill:

function trunc(x) {
    return x < 0 ? Math.ceil(x) : Math.floor(x);
}

It removes everything after the decimal point because the bitwise operators implicitly convert their operands to signed 32-bit integers. This works whether the operands are (floating-point) numbers or strings, and the result is a number.

In other words, it yields:

function(x) {
  if(x < 0) return Math.ceil(x);
  else return Math.floor(x);
}

only if x is between -(2³¹) and 2³¹ - 1. Otherwise, overflow will occur and the number will "wrap around".

This may be considered useful to convert a function's string argument to a number, but both because of the possibility of overflow and that it is incorrect for use with non-integers, I would not use it that way except for "code golf" (i.e. pointlessly trimming bytes off the source code of your program at the expense of readability and robustness). I would use +x or Number(x) instead.

---

How this is the NOT of the NOT

The number -43.2, for example is:

-43.2₁₀ = 11111111111111111111111111010101₂

as a signed (two's complement) 32-bit binary number. (JavaScript ignores what is after the decimal point.) Inverting the bits gives:

NOT -43₁₀ = 00000000000000000000000000101010₂ = 42₁₀

Inverting again gives:

NOT 42₁₀ = 11111111111111111111111111010101₂ = -43₁₀

This differs from Math.floor(-43.2) in that negative numbers are rounded toward zero, not away from it. (The floor function, which would equal -44, always rounds down to the next lower integer, regardless of whether the number is positive or negative.)