javascript toFixed() rounding bug?

[ersa]

im not sure if this is new but here is the problem i encountered with javascript:

var num = new Number(0.595);
alert(num.toFixed(2)); // 0.59

where is IE 7, it correctly alerts 0.60.

even funnier:

var num = new Number(2.595);
alert(num.toFixed(2)); // 2.60

[schapel]

It's due to the way floating point numbers are represented as binary numbers by computers. The numbers 0.595 and 2.595 cannot be represented exactly in binary, so they are represented as a number either slightly higher or slightly lower than that number. If it is higher, the number will round up. If it is lower, it will round down. The bottom line is that it is not a bug. You are expecting floating point numbers to be precise when they are not.

[j.j.]

where is IE 7, it correctly alerts 0.60.

IE has real bugs with toFixed(). Try alert((0.9).toFixed(0));

j.j.

[j.j.]

The bottom line is that it is not a bug

... in the Browser.

But it's a in the ECMAscript standard. They plan a workaround in ES4:

http://www.builderau.com.au/program/javascript/soa/Why-I-Love-ECMAScript-4-Real-Decimals/0,339028434,339289678,00.htm

http://wiki.ecmascript.org/doku.php?id=proposals:decimal&s=use+decimal

j.j.

[ersa]

It's due to the way floating point numbers are represented as binary numbers by computers. The numbers 0.595 and 2.595 cannot be represented exactly in binary, so they are represented as a number either slightly higher or slightly lower than that number. If it is higher, the number will round up. If it is lower, it will round down. The bottom line is that it is not a bug. You are expecting floating point numbers to be precise when they are not.

thanks. it explains.
to be holy honest, i was not "expecting floating point numbers to be precise", i was simply expecting it to do as explained in "https://developer.mozilla.org/en/Core_JavaScript_1.5_Reference/Global_Objects/Number/toFixed". and well, unexpectedly received a 'surprise'.

[schapel]

Yes, you were expecting floating point numbers to be precise. That's why you got the surprise, because the numbers you were using could not be represented precisely exactly in binary floating point.