float_set

Summary. Construct an XFL from a separate exponent and mantissa.

Signature.

int64_t float_set(int32_t exponent, int64_t mantissa);

Parameters.

Name	Type	Description
exponent	int32_t	Base-10 exponent.
mantissa	int64_t	Signed mantissa; its sign becomes the XFL's sign.

Return value. The normalized XFL. A mantissa of 0 returns the XFL 0 (the exponent is ignored). Errors: INVALID_FLOAT (-10024) if the value overflows the XFL range or normalizes to a value that cannot be represented; otherwise the specific normalization error (MANTISSA_OVERSIZED -26, EXPONENT_OVERSIZED -28, etc.).

Common failure patterns.

• An exponent/mantissa combination outside the representable range → INVALID_FLOAT (on overflow) or a specific mantissa/exponent size error.

Caveats / notes.

• The mantissa need not be pre-normalized; float_set normalizes it to 16 significant digits and adjusts the exponent. float_set(-2, 1234) and float_set(0, 12) … produce the canonical XFL for the same numeric value regardless of input scaling.
• To build a whole number n, use float_set(0, n).

Minimal example.

int64_t half = float_set(-1, 5);   // 5 * 10^-1 = 0.5

Practical example.

// A mantissa of 0 is the canonical zero regardless of exponent.
ASSERT(float_set(0, 0) == 0);
// 1 * 10^0 == float_one()
ASSERT(float_set(0, 1) == float_one());
// negative mantissa yields a negative XFL
int64_t neg = float_set(-5, -9123456789LL);

Related APIs. float_one, float_mantissa, float_int.