7.2.1 Flonums🔗ℹ

A flonum is a number that is represented using an IEEE 64-bit floating-point representation. A flonum is a plain number, so number operations like + work on flonums and mixtures of flonums with other numbers. In some cases, using functions and operators from rhombus/flonum can reduce run time by allowing flonum operations to more efficiently communicate intermediate results.

The functions exported by rhombus/flonum explicitly require flonum arguments and produce flonum results. They can produce different results from functions like math.sqrt, typically returning #nan when the result would not be a flonum, such as when a negative number is passed to math.sqrt.

Some operations, like +, statically specialize for performance when all arguments have static information for the Flonum annotation. There is no performance difference between that inferred specialization and these operators, and a conversion is inferred only when the result value is the same for unspecialized and flonum-specific operations.

Some related operations, such as math.sqrt, do not always produce flonum results for flonum arguments. In those cases, the corresponding function like flonum.sqrt produces #nan, instead. For example, flonum.sqrt(-1.0) is #nan instead of the complex number #{0.0+1.0i}.

A flonum array potentially provides better performance than a normal array of flonums, because the elements are known to be flonums without a check on access. In particular, flonum calculations that read and write flonum arrays may perform better by reducing allocation compared to normal arrays.

Like an array, a flonum array is indexable using […], assignable via […] and :=, supports membership tests using the in operator, and can be used as sequence. Two flonum arrays are equal by is_now as long as they have the same length and their elements are pairwise equal by flonum.(==).

The flonum.Array.contains method accepts any argument value to find, but it will only succeed for flonums. The comparsion operation is always flonum.(==).