Re: Decimal: Formal review begins.

From: Rainer Deyke (rdeyke_at_[hidden])
Date: 2025-01-19 12:28:50

On 18.01.25 17:44, Matt Borland via Boost wrote:
> Using the example from the webpage DECIMAL(5,2) gives us the range
> -999.99 to 999.99, but decimal will store it as -9.9999e+02 to
> 9.9999e+02. As long as you limit yourself to chars_format::fixed on
> any potential output to the user you should be fine. One of the
> original users wanted to add bitcoin to their trading platform but
> the smallest divisible unit of a bitcoin, a satoshi = 1/100 million
> of a bitcoin, exceeded the precision of their in-house fixed point
> system so they switched to using decimal64.

Using floating point when you mean fixed point is a really bad practice. 
  You are going to end up with fractional satoshis, and you won't even 
see them because you are using chars_format::fixed.  All you'll see is 
that 0.000000001 + 0.000000001 is sometimes mysteriously 0.000000003 
instead of 0.000000002.  You might as well use binary instead of decimal 
floating point at that point.

-- 
Rainer Deyke - rainerd_at_[hidden]