Math.round() doesn't follow specification for exact 0.5

Sun 2019.08.11

http://www.espruino.com/Reference#l_Math­_round

https://developer.mozilla.org/en-US/docs­/Web/JavaScript/Reference/Global_Objects­/Math/round
"If the fractional portion of the argument is greater than 0.5, the argument is rounded to the integer with the next higher absolute value. If it is less than 0.5, the argument is rounded to the integer with the lower absolute value. If the fractional portion is exactly 0.5, the argument is rounded to the next integer in the direction of +8. Note that this differs from many languages' round() functions, which often round this case to the next integer away from zero, instead giving a different result in the case of negative numbers with a fractional part of exactly 0.5."

While rounding towards positive infinity does work for negative numbers, it is not rounding
correctly for exact 0.5 values.

>console.log(Math.round(-100.6), Math.round(-100.4), Math.round(-101.5), Math.round(-100.5));
-101 -100 -101 -100
=undefined

>console.log(Math.round(100.6), Math.round(100.4), Math.round(101.5), Math.round(100.5));
101 100 101 100
=undefined
>

Correct values
> -101 -100 -101 -100
> 101 100 102 101

>process.env
={
  VERSION: "2v04",
  GIT_COMMIT: "3956264e",
  BOARD: "ESPRUINOWIFI",

EDIT:

Had a thought, could this anomaly be related to the floating point number of digits to right of zero not matching different browsers and node.js output? hmmmm. . . .

http://forum.espruino.com/conversations/­336605/
test files #1

EDIT: Tue 2019.08.20

see snippet that may be helpful in resolving this
http://forum.espruino.com/comments/14867­790/

@Robin

If you like take a look at the source code to find and fix the error ;-)

Or it it might be related to this issue

https://github.com/espruino/Espruino/iss­ues/1573

Just to add:

$ ./espruino
>(new Float64Array([101.5])).buffer
=new Uint8Array([0, 0, 0, 0, 0, 96, 89, 64]).buffer

$ node
> (new Float64Array([101.5])).buffer
ArrayBuffer { byteLength: 8 }
> new Uint8Array((new Float64Array([101.5])).buffer)
Uint8Array [ 0, 0, 0, 0, 0, 96, 89, 64 ]

So this would appear not to be a parsing issue (like 1573) but is to do with the rounding code itself

It seems like Math.round() is a better solution, but it is not! In some cases it will NOT round correctly. Also, toFixed() will NOT round correctly in some cases.

To correct the rounding problem with the previous Math.round() and toFixed(), you can define a custom JavaScript round function that performs a "nearly equal" test to determine whether a fractional value is sufficiently close to a midpoint value to be subject to midpoint rounding. The following function return the value of the given number rounded to the nearest integer accurately.

Number.prototype.roundTo = function(decimal) {
return +(Math.round(this + "e+" + decimal) + "e-" + decimal);
}

var num = 9.7654;
console.log( num.roundTo(2)); //output 9.77

Do you think this is an issue specific to Espruino?

If so, maybe you could look at helping to contribute a fix for it?

I also think Math.round() does not behave as expected

Math.round(1.3)
=1
>Math.round(1.4)
=1
>Math.round(1.5)
=1
>Math.round(1.6)
=2
>Math.round(1.0)
=1

Math.round(1.5) should return 2

Thanks - just created a new issue for this: https://github.com/espruino/Espruino/iss­ues/2258

Not sure how that happened as that's pretty much the first thing you'd test when implementing Math.round!

That reminds me: I was planning to report

Math.ceil(Math.sqrt(9))
=4

but totally forgot about it (whoops).

Edit: reported now