CODEWITHSUNDEEP
genericsfloating-pointrust
user72961
user72961

Reputation: 143

Writing generic floating point code without a lot of boilerplate type constraints

Is there a way to specify that the generic type T can only be either f32 or f64? Otherwise I end up imposing a lot of boilerplate type constraints: Copy, Add<Output=T>, Sub<Output=T>, Mul..., Div..., Zero, One, etc., which gets tedious after a while. Also I am afraid this list will grow to include Sin, Cos, Tan, etc. in the future.

Upvotes: 4

Views: 554

Answers (1)

mdup
mdup

Reputation: 8529

You want num::traits::Float, available in crate num.

Every type satisfying Float is guaranteed to be Num + Copy + NumCast + PartialOrd + Neg<Output=Self>. In turn, PartialEq + Zero + One + Add + Sub + Mul + Div + Rem are required for Num. This answer your first requests for basic arithmetic. Other operations like sin(), cos() are available as member functions in Float.

Finally, f32 and f64 are supplied out-of-the-box as implementors of Float.

Upvotes: 6

Related Questions