Does BigNumber only support Integers? (answer: yes)

[thanhvk]

Hi everyone,

I have code as below:

export const computeChange = (open: string, close: string) => {
    const bigOpen = utils.bigNumberify(open)
    const bigClose = utils.bigNumberify(close)
    const percentMultiplier = utils.bigNumberify(100)

    if (bigOpen.eq(bigClose)) return 0

    const change = ((bigClose.sub(bigOpen)).mul(percentMultiplier)).div(bigOpen)
    return change
}

in case open = 252350000000000000000000000000000000000, close = 252320000000000000000000000000000000000

const change = ((bigClose.sub(bigOpen)).mul(percentMultiplier)).div(bigOpen) return 0 instead 0.011889663918834626.

So what I do wrong in here? or this is bug? and Do you have any experience or solution for this issue?

Thank you!

[ricmoo]

BigNumber is only for integers, it does not support floating-point or fixed-point math. In v5 (public beta soon) there is a FixedPoint object to help with these, but in the meantime, you can use the formatUnits to do this.

Keep in mind that floating-point numbers in JavaScript are NOT precise. They cannot be, it is not possible to represent arbitrary precision with a fixed number of bits (i.e. 64 bits)

You will need to choose a resolution you wish the result to be accurate to, but you can for example use:

// Multiply by 1000 to keep 3 decimal places
const change = ((bigClose.sub(bigOpen)).mul(percentMultiplier)).mul("1000").div(bigOpen);
// We use 2 here for decimals, since we multiplied by 1000 above.
const changeString = ethers.utils.formatUnits(change, 3);

In general though, if you plan to use decimal numbers there will be a lot of consequences you should familiarize yourself with regarding rounding, underflow (completely unrelated to overflow), parsing/formatting and representation.

[ricmoo]

Closing this now, but if you have more questions, please feel free to re-open or continue discussion (I monitor closed issues).

In v5, there is a FixedNumber type which can be used to perform fixed-point math operations too.