So, after all it turned out, rounding bugs will always haunt you, no matter how hard you try to compensate them. Hence the problem should be attacked by representing numbers exactly in decimal notation.
Number.prototype.toFixedDown = function(digits) {
var re = new RegExp("(\\d+\\.\\d{" + digits + "})(\\d)"),
m = this.toString().match(re);
return m ? parseFloat(m[1]) : this.valueOf();
};
[ 5.467.toFixedDown(2),
985.943.toFixedDown(2),
17.56.toFixedDown(2),
(0).toFixedDown(1),
1.11.toFixedDown(1) + 22];
// [5.46, 985.94, 17.56, 0, 23.1]
Old error-prone solution based on compilation of others':
Number.prototype.toFixedDown = function(digits) {
var n = this - Math.pow(10, -digits)/2;
n += n / Math.pow(2, 53); // added 1360765523: 17.56.toFixedDown(2) === "17.56"
return n.toFixed(digits);
}
Consider taking advantage of the double tilde: ~~.
Take in the number. Multiply by significant digits after the decimal so that you can truncate to zero places with ~~. Divide that multiplier back out. Profit.
function truncator(numToTruncate, intDecimalPlaces) {
var numPower = Math.pow(10, intDecimalPlaces); // "numPowerConverter" might be better
return ~~(numToTruncate * numPower)/numPower;
}
I'm trying to resist wrapping the ~~ call in parentheses; order of operations should make that work correctly, I believe.
alert(truncator(5.1231231, 1)); // is 5.1
alert(truncator(-5.73, 1)); // is -5.7
alert(truncator(-5.73, 0)); // is -5
I've unintentionally also handled cases to round off left of the decimal as well.
alert(truncator(4343.123, -2)); // gives 4300.
The logic's a little wacky looking for that usage, and may benefit from a quick refactor. But it still works. Better lucky than good.
Here is my take on the subject:
convert.truncate = function(value, decimals) {
decimals = (decimals === undefined ? 0 : decimals);
return parseFloat((value-(0.5/Math.pow(10, decimals))).toFixed(decimals),10);
};
It's just a slightly more elaborate version of
(f - 0.005).toFixed(2)