Testing haskell equality on Doubles with HUnit?

Question

Does HUnit have some way of doing approximate equalities? Clearly, this fails:

test1 = TestCase (assertEqual "Should equal" (get_scores) [(0.3, 0.3), (0.6, 0.36), (1.0, 0.3399999)])

Answer 1

For floating point approximate comparison, there's a library with some nice utils:

http://hackage.haskell.org/package/ieee754

Answer 2

Since this question was first asked, the HUnit-approx package has been released.

It uses an implicit variable, epsilon , and an approximate equality operator, ~?~ , instead of ~=? .

A short example:

tweak :: Float -> Float
tweak x = x * 1.0001

egApproxTest  = let ?epsilon = 0.01 in
    [ "strictEqual" ~: 0 ~=? tweak 0,
      "approxEqual" ~: 2 ~?~ tweak 2 ]

Answer 3

Note: I don't know if there's a correct/official/accepted way to do this.

---

Here's the source code for assertEqual :

assertEqual :: (Eq a, Show a) => String -- ^ The message prefix 
                              -> a      -- ^ The expected value 
                              -> a      -- ^ The actual value
                              -> Assertion
assertEqual preface expected actual =
  unless (actual == expected) (assertFailure msg)
 where msg = (if null preface then "" else preface ++ "\n") ++
             "expected: " ++ show expected ++ "\n but got: " ++ show actual

Based on that and on JUnit's function for testing double equality , we could create our own in the same style:

import Control.Monad (unless)

assertEquals ::   String  -- ^ The message prefix
               -> Double  -- ^ The maximum difference between expected and actual
               -> Double  -- ^ The expected value
               -> Double  -- ^ The actual value
               -> Assertion
assertEquals preface delta expected actual = 
  unless (abs (expected - actual) < delta) (assertFailure msg)
 where msg = ... same as above ...

Answer 4

For my purposes, this helper function worked well enough:

assertFloatEqual text a b = 
  assertEqual text (take 6 (show a)) (take 6 (show b))

Answer 5

I made my data type an instance of equal with code there to use 2 decimal places. Here is an example using a cartesian point:

data Point =  Point { x_axis :: Double, y_axis :: Double, z_axis :: Double } 
              deriving (Show)

Instance of equal

In order to avoid double rounding errors and trig errors which cause the same point, and thus CornerPoints, to be /= due to tiny differences, give it a range of .01 , and still allow the points to be equal.

axisEqual :: (Eq a, Num a, Ord a, Fractional a) => a -> a -> Bool

axisEqual  a b
  | (abs (a - b)) <= 0.011 = True
  | otherwise      = False

instance Eq Point where

  Point x y z == Point xa ya za
    | (axisEqual x xa) && (axisEqual y ya)  &&(axisEqual z za) = True 
    | otherwise = False