Why `-0.0` is parsed to zero without sign?

[eirnym]

as subject says, -0.0 is parsed to zero without sign, while it clearly has and in calculations such sign is important. More so, I can't add it later on

[akubera]

It's due to the fact that the internal representation uses num_bigint::BigInt which specifically has a NoSign variant on the Sign enum for representing zero. I've been working and brainstorming on new implementations of the core, and any extra thoughts/input you have is valuable. How are you using it and what would you like the behavior to be? Is it generated particularly when an operation is rounded to zero from the negative side?

I've been thinking about another type in the library that wraps BigDecimal with extra metadata, allowing for NaN and Infinities and the like; signed zero sounds more basic than that, shouldn't need a wrapper type.

Python Decimal does allow negative-zero with the following properties (which is what I'd expect):

from decimal import *

zero = Decimal("0")

zero.is_zero() # True
zero.is_signed() # False

negzero = Decimal('-0.0')

negzero.is_zero() # True
negzero.is_signed() # True
negzero < 0  # False
negzero == 0 # True

zero == negzero  # True

[eirnym]

One of the option is to use BigUInt and carry the sign and special cases inside the BigDecimal as flags.

The Python's Decimal implementation for zero is quite reasonable and matches the IEEE754 standard. Basically the zero is a "sign without a value" and usually comes from rounding.

[eirnym]

@akubera What do you think on the proposition?