Crate numeric_float

This crate defines variants of the f32 and f64 types which exclude Not-a-Number. These types are collectively called "numeric". As a consequence of never being NaN, these types implement Eq, Ord, and Hash.

They also attempt to define all relevant operations one would expect to be able to perform on floating-point numbers. The problem is what to do about operations that might result in NaN.

The n*f types handle operations that result in NaN by instead returning the underlying f* result directly. For example, adding two n32fs results in an f32.

The n*p types handle operations that result in NaN by panicking when NaN is produced. Some operations which panic also have try_* alternatives that return Option<Self> instead.

Supported Traits

Where possible, the types in this crate implement all traits which are implemented on the corresponding f* type. In addition to those in the standard library, the following crates are also optionally supported:

• conv
• num
• rustc-serialize
• serde

The following are explicitly not supported:

• DecodableFloat, and RawFloat are internal and not stable to implement.
• One, Zero, and *Assign are unstable.
• ieee754: the Ieee754 trait cannot be implemented for new types.
• num: Float, Num, One, Signed, and Zero cannot be implemented for n*f.

Structs

n32f	A 32-bit floating point type which excludes Not-a-Number.
n32p	A 32-bit floating point type which excludes Not-a-Number.
n64f	A 64-bit floating point type which excludes Not-a-Number.
n64p	A 64-bit floating point type which excludes Not-a-Number.

Enums

ParseNumericError

An error which can be returned when parsing a numeric float (e.g. n32f, n64p).

Traits

Numeric

This trait is implemented for all numeric float types in this crate.