Any plans for Fixed Point numbers?

[fulara]

Are there any plans to support Fixed Point numbers?
I did a search on crates and did not find any implementation.
BigRational is too slow for me.
and I could not get Ratio working for my case. (overflow issues).

[cuviper]

See issue #8 and PR #224 for BigDecimal, although I think there are still some unanswered questions.

[fulara]

Yep, saw that. But thats regarding the BigDecimal - which would have infinite precision.
I am looking form something having decimal representation but not being as slow as BigDecimal probably would be.

Any plans for d64 or d128 in this library?

[hauleth]

In theory it shouldn't be that hard. However truly fixed point arithmetic isn't possible until we get Π types so you could write:

let a: Fixed64<10> = 10.into(); // 64 bit digit with 10 digits after decimal point
let b: Fixed64<5> = 10.into();

// a + b would be impossible as their type is different

Currently only solution would be using typenum crate which is quite slow when compiling. So I hope that in future it will be possible (as I would ❤️ to have Π types as I could make good use of them in my Octavo library).

[icefoxen]

It should be easy to write a macro that constructs a fixed-point number type to the given specifications, so:

// Create a signed 64-bit number with 10 bits after the decimal point
fixed_point!(Fix64_10, i64, 10);

let a: Fix64_10 = 10.into();
let b: Fix64_10 = 20.into();
a + b

[scottjmaddox]

There's a typenum-based implementation here, that looks interesting. It might be better to limit the library to defining the macro's, and let the library users instantiate the types of interest (as suggested by icefoxen), though. Otherwise compilation time could quickly explode with the number of types, especially if you tried to provide all possible converting functions (mul, cast, etc.).

[cuviper]

I don't have any plans to implement this in the rust-num organization, but maybe fpa or some other crate could work for you.