mamrhein/fpdec.rs

Fast implementation of Decimal fixed-point arithmetics

fpdec.rs's Issues

Cast to primitive types

Does fpdec lib has any conversion to another type(isize, u8, i16)?

Bug; multiplication overflow causes crashing for very small values during float conversion

Issue occurs here

-1e-22 is just small enough to trigger the crash

'subtract with overflow ' panic in fn into_float::n_signif_bits

Converting a Decimal with absolute value less than 1 and significant bits larger than the number of mantissa bits of the target type causes the function 'n_signif_bits' to be called with parameter 0, which leads to the 'subtract with overflow ' panic.

Wrong rounding in conversion of Decimal to f64

f64::from(Decimal::new_raw(9802266554071127_i128, 16)) gives 0.9802266554071128 instead of 0.9802266554071127.

lshift overflow in from_float.rs:217

Converting from certain f64 values results in a left-shift overflow during conversion
The value 113078212145816600000000000000000000000000000000000000000000000000000000000.0 is one such example

Debug output gives crazy result

The debug output implementation seems broken. Here's a minimal reproducible example

This small test program:

fn main() {
    let decimal = fpdec::Decimal::new_raw(-42472286074876, 18);
    println!("{}", decimal);
    println!("{:?}", decimal);
}

Gives the output:

-0.000042472286074876
Dec!(-1.999957527713925124)

fpdec 0.8.0
rustc 1.72.0-nightly (04075b320 2023-06-22)

Requesting implementations for Hash, and TryInto<i/u/f/*>

The inability to convert between rust numeric primitives and Decimal is a limitation I am running into

I am currently converting the decimal to a string and re-parsing it, which is far from ideal

An implementation of Hash would also simplify things on my end

increasing precision

I'd like to use this library with the 'quantities' crate to do some quantum mechanics work.
Planck's constant is 6.62607015 × 10−34 , so I'd like about 42 digits of precision.

Is forking this library the only way to increase the precision?
(Fork and adjust the MAX_N_FRAC_DIGITS constant defined here: fpdec-core/src/lib.rs)

Thanks!

Feature request: checked_pow

To exponentiate fixed-p values I currently have to drop both operands to float first, checked_pow would round out the available set of checked operators