float-cmp

Documentation is available at https://mikedilger.github.io/float-cmp

float-cmp defines traits for approximate comparison of floating point types which have fallen away from exact equality due to the limited precision available within floating point representations. Implementations of these traits are provided for f32 and f64 types.

The recommended go-to solution (although it may not be appropriate in all cases) is the approx_eq() function in the ApproxEq trait. An epsilon test is performed first, which handles very small numbers, zeroes, and differing signs of very small numbers, considering them equal if the difference is less than the given epsilon (e.g. f32::EPSILON). For larger numbers, floating point representations are spaced further apart, and in these cases the ulps test comes to the rescue. Numbers are considered equal if the number of floating point representations between them is below a specified bound (Ulps are a cardinal count of floating point representations that separate two floating point numbers).

Several other traits are provided including Ulps, ApproxEqUlps, ApproxOrdUlps, and ApproxEqRatio.

The problem

Floating point operations must round answers to the nearest representable number. Multiple operations may result in an answer different from what you expect. In the following example, the assert will fail, even though the printed output says "0.45 == 0.45":

  let a = 0.15_f32 + 0.15_f32 + 0.15_f32;
  let b = 0.1_f32 + 0.1_f32 + 0.25_f32;
  println!("{} == {}", a, b);
  assert!(a==b)  // Fails, because they are not exactly equal

This fails because the correct answer to most operations isn't exactly representable, and so your computer's processor chooses to represent the answer with the closest value it has available. This introduces error, and this error can accumulate as multiple operations are performed.

The solution

With ApproxEq, we can get the answer we intend:

  let a = 0.15_f32 + 0.15_f32 + 0.15_f32;
  let b = 0.1_f32 + 0.1_f32 + 0.25_f32;
  println!("{} == {}", a, b);
  assert!(a.approx_eq(&b, 2.0 * ::std::f32::EPSILON, 2)) // They are equal, within 2 ulps

I recommend you use a small integer for the ulps parameter (2 to 10 or so), and a similar small multiple of the floating point's EPSILON constant (2.0 to 10.0 or so), but please send feedback if you think you have better advice.

Some explanation

We use the term ULP (units of least precision, or units in the last place) to mean the difference between two adjacent floating point representations (adjacent meaning that there is no floating point number between them). This term is borrowed from prior work (personally I would have chosen "quanta"). The size of an ULP (measured as a float) varies depending on the exponents of the floating point numbers in question. That is a good thing, because as numbers fall away from equality due to the imprecise nature of their representation, they fall away in ULPs terms, not in absolute terms. Pure epsilon-based comparisons are absolute and thus don't map well to the nature of the additive error issue. They work fine for many ranges of numbers, but not for others (consider comparing -0.0000000028 to +0.00000097).

Implementing these traits

You can implement ApproxEq for your own complex types. The trait and type parameter notation can be a bit tricky, especially if your type is type parameterized around floating point types. So here is an example (you'll probably not specify the Copy trait directly, but use some other NumTraits type floating point trait):

use float_cmp::{Ulps, ApproxEq};

pub struct Vec2<F> {
  pub x: F,
  pub y: F,
}

impl<F: Ulps + ApproxEq<Flt=F> + Copy> ApproxEq for Vec2<F> {
  type Flt = F;

  fn approx_eq(&self, other: &Self,
               epsilon: <F as ApproxEq>::Flt,
               ulps: <<F as ApproxEq>::Flt as Ulps>::U) -> bool
  {
    self.x.approx_eq(&other.x, epsilon, ulps)
      && self.y.approx_eq(&other.y, epsilon, ulps)
  }
}

Inspiration

This crate was inspired by this Random ASCII blog post: