impl Div<Decimal256> for Uint256

[ewoolsey]

Currently this is not implemented, my guess as to why being that the Output type is not obvious.
The most obvious workaround being:

let res = my_uint * my_dec.inv();

This comes with the drawback that if my_dec is very large you begin to lose significant precision with .inv(), to the point that it could become 0. Another possible solution is perhaps:

let decimal_res = Decimal256::new(my_uint * DECIMAL_FRACTION) / my_dec;
// and if you want a Uint256
let uint_res = decimal_res.atomics() / DECIMAL_FRACTIONAL

My question is then, Do you plan to implement this? and if not, is this method the best way to go about it?

[ewoolsey]

This also raises the question about why we don't have

impl From<Decimal256> for Uint256 // should floor by default

and

impl TryFrom<Uint256> for Decimal256

I imagine these would be very useful to many.

[ewoolsey]

After playing around I suppose this is probably the cleanest solution:

pub fn div(numerator: Uint256, denominator: Decimal256) -> Uint256 {
    numerator * Decimal256::one().atomics() / denominator.atomics()
}

and perhaps to prevent overflow during the first mul then this would be the most robust version:

pub fn checked_div(numerator: Uint256, denominator: Decimal256) -> StdResult<Uint256> {
    Ok(numerator.full_mul(Decimal256::one().atomics()).checked_div(Uint512::from(denominator.atomics()))?.try_into()?)
}

[webmaster128]

Mixed type arithmetic is discouraged and should not be implemented. I was a mistake to implement it in the first place. As a consequence, let res = my_uint * my_dec.inv(); (uint times decimal) will not be possible from 2.0 onwards. The better way is to do a clean conversion and then operate on decimals or integers, depending on the situation. Some of those conversions between ints and decimals are probably missing and we should add them.

However, #1566 added things like checked_div_ceil which you can use on the integer.

[ewoolsey]

Mixed type arithmetic is discouraged and should not be implemented. I was a mistake to implement it in the first place. As a consequence, let res = my_uint * my_dec.inv(); (uint times decimal) will not be possible from 2.0 onwards. The better way is to do a clean conversion and then operate on decimals or integers, depending on the situation. Some of those conversions between ints and decimals are probably missing and we should add them.

However, #1566 added things like checked_div_ceil which you can use on the integer.

Ah yeah I suppose that makes sense. Adding those conversions would definitely be really helpful I think. I've personally found it a bit cumbersome! Thanks for the reply though. Feel free to close this issue, or change the title to implementing those conversions!

[webmaster128]

Here you find an overview of the conversions between unsigned math types. Hope this helps. Is there anything missing that you need?

[webmaster128]

With #1825 being shipped in 1.4, I think we are good here. Closing for now.