Two fixed/decimal numeric types:

A fast fixed-scale decimal backed by a signed 128-bit integer.

• P — total precision (digits), 1..38
• S — fixed scale (fraction digits), 0..P
• Stored as raw = value * 10^S
• Supports parsing, formatting, + - * /, comparisons, and conversions to long double
• Errors are tracked inside the value (Err::None/Invalid/Overflow/DivByZero)

Example:

using N = unumber::numeric::Numeric128<38, 4>;

N a("12.34005");
N b("2.5000");
auto c = a * b;          // rounded half-up
std::cout << c << "\n";  // 30.8501

A PostgreSQL-like arbitrary-precision decimal with a dynamic scale.

• Magnitude is stored in base 1e9 limbs (std::vector<uint32_t>)

• scale is the number of fractional digits

• Limits match PostgreSQL defaults:

  • up to 131072 digits before the decimal point
  • up to 16383 digits after the decimal point

• Supports parsing, formatting, + - * /, rescaling with rounding (Trunc / HalfUp)

Example:

using unumber::numeric::Numeric;

Numeric a("1.23");
Numeric b("2.5");

auto s = a + b;                              // 3.73
auto m = Numeric::mul(a, b, 2, Rounding::HalfUp); // 3.08
auto d = Numeric::div(a, b, 4, Rounding::Trunc);  // 0.4920

• Rounding::Trunc — toward zero
• Rounding::HalfUp — rounds away from zero on a 0.5 tie

Unit tests use GoogleTest and include randomized cross-checks against builtin __int128 reference math when available.