Crate real_float
source · [−]Expand description
Intro
This crate is for dealing checked floating point numbers.
It exports two types: Real and Finite. Real is checked at runtime
to ensure that it is never NaN, while Finite adds the additional constraint
that it can never be infinite.
For the sake of brevity, we will only discuss Real, but understand that
everything applies to Finite as well.
Checking behavior
A NaN-check is inserted in every single operation and method. If a NaN
ever surfaces, it will result in a runtime panic.
Note that these checks will normally only be present in debug builds.
This is consistent with Rust’s philosphy for integer overflowing: panic
in debug mode, and allow fast-but-likely-incorrect bevhavior in release mode.
If you want these checks to be present no matter what, enable the strict feature.
Fallible API
The types in this crate also support fallible APIs for any operation that would
otherwise panic. These are the try_* methods defined on Real and Finite,
and they will perform NaN checks whether or not debug mode is enabled.
Comparison with similar crates
TODO
Modules
Module containinig traits that define required operations for floating point numbers.
If the optional num-traits feature is enabled, these will automatically be implemented for
any type implementing num_traits::Float.
Macros
Constructor for Finite that never checks the value, and can be used in a const context.
Constructor for NonNegative that never checks the value, and can be used in a const context.
Constructor for Real that never checks the value, and can be used in a const context.
Structs
The error produced when infinity or NaN is encountered.
The error produced when NaN is encountered.
The error produced when a negative or NaN value is encountered.