Floating-Point

Cloth provides two binary floating-point types, f32 and f64, conforming to the IEEE 754 single- and double-precision formats.

Types

Type	Format	Bits	Sign	Exponent	Mantissa
`f32`	IEEE 754 binary32	32	1	8	23
`f64`	IEEE 754 binary64	64	1	11	52

Approximate ranges

Type	Smallest positive normal	Largest finite	Decimal precision
`f32`	≈ 1.175 × 10⁻³⁸	≈ 3.403 × 10³⁸	≈ 7 digits
`f64`	≈ 2.225 × 10⁻³⁰⁸	≈ 1.798 × 10³⁰⁸	≈ 15 digits

Aliases

The following names are aliases for the IEEE 754 floating-point types.

Alias	Notes
`float`	Single-precision floating-point.
`double`	Double-precision floating-point.
`real`	A real-number floating-point type.

Literals

Floating-point literals contain a decimal point or exponent and have type f64 by default.


let pi = 3.14159;
let g  = 9.8;

The keyword NaN is the IEEE 754 not-a-number value. It is a legal expression of floating-point type and can appear in any position where a floating-point literal is permitted.


let invalid = NaN;

Special values

f32 and f64 follow the IEEE 754 model and include three categories of special values:

Signed zero — +0.0 and -0.0 are distinct bit patterns but compare equal under ==.
Infinities — produced by overflow and by exact arithmetic such as 1.0 / 0.0.
NaN — produced by undefined operations such as 0.0 / 0.0 or sqrt of a negative number. NaN is unequal to every value, including itself; comparisons against NaN always evaluate to false except for !=, which is true.

Arithmetic

The arithmetic operators (+, -, *, /, %) and comparison operators (==, !=, <, <=, >, >=) are defined on floating-point types. Bitwise operators are not defined on floating-point types.

I/O

The standard library provides a println overload for f64:


import cloth.io.Out::{ println };

println(3.14);