Floating-point types

Float32 and Float64 store IEEE 754 floating-point values. Float is an alias for Float64.

Floating-point values are useful for measurements and calculations in which fractional results and rounding are expected. They should not be used when an exact decimal amount—such as money—must remain exact.

Literals

A decimal point or base-ten exponent makes a floating-point literal. Without an expected type it is Float64; an annotation can select Float32:

local width: Float32 = 12.5
local distance = 1_000.25
local large = 6.02e23
local small = 2e-3

Arithmetic behavior

Floating-point values support +, -, *, and / when both operands have the same exact type. Results are rounded to that type’s precision.

Unlike integer division, floating-point division by zero follows IEEE 754 behavior and can produce infinity rather than a Pop division-by-zero trap.

Ordering supports <, <=, >, and >= for matching floating-point types. Every ordering comparison with NaN is false. Default == and ~= are not available for floats in this release.

There is no implicit conversion between Float32, Float64, and integer types. Use explicit target-type syntax:

local count: Int = 7
local measurement = Float64(count)
local whole = Int32(measurement) -- truncates toward zero; traps if out of range

Continue with Decimal floating-point literals and Numeric conversions for the detailed rules.