BloombergLP::bdldfp::DecimalConvertUtil::decimal128FromFloat

static
Decimal128
decimal128FromFloat(
    float binary,
    int digits = 0);

Description

If binary is singular (+/-NaN, +/-Inf, or +/-0) or is not within the representable range of the return type, return a corresponding decimal singular value.

Optionally specify digits to indicate the number of significant digits to produce in the returned value. The digits parameter is treated as follows:

If digits is larger than the number of digits in the destination type, it will be reduced to that number of digits.

If digits is positive, the result is binary rounded to that many significant digits.

If digits is negative, the decimal value with the fewest significant digits that converts back to binary is returned if possible, and otherwise the value closest to binary is returned. Note that this provides the most visually appealing result but is the most expensive to compute.

If digits is not specified or 0, a default value will be used (possibly depending on the value of binary) based on the premise that binary is a converted decimal value of no more significant digits than is guaranteed to have a uniquely converted binary value (15 for double, 6 for float in general, and 7 for float in the range [ .0009999995 .. 8589972000 ]). Note that this is likely to have the best performance for "business" numbers (i.e., numbers that originate as decimal values in external market quote feeds).

Note that the purpose of these functions is to restore a decimal value that has been converted to a binary floating-point type. It is more efficient to use conversion constructors when all that is needed is the nearest decimal to the binary value.

Note that if binary is a float value that was converted from an IBM/Perkin-Elmer/Interdata binary float value itself converted from a decimal value of no more than 6 significant digits, specifying 6 for digits will recover the original decimal value. Not specifying digits may result in a value having a spurious seventh digit.