non-IO test driver (for multiple tests)

[jwaldmann]

I want to write code like this:

append :: [a] -> [a] -> [a]
append a b = case a of
  [] -> b
  x : y -> x : append y b

assoc op = \ a b c ->
    op a (op b c) == op (op a b) c

comm op = \ a b -> op a b == op b a

left_neutral op a = \ b -> op a b == b 
right_neutral op b = \ a -> op a b == a 

tests :: Either String String
tests = do
    testE "assoc" 1000 
        $ assoc ( append ::[Bool] ->[Bool]->[Bool])
    testE "right_neutral" 1000 
        $ right_neutral ( append ::[Bool] ->[Bool]->[Bool]) []
    testE "left_neutral" 1000 
        $ left_neutral ( append ::[Bool] ->[Bool]->[Bool]) []

it should have

• a pure test driver (no IO) so I can process the result in a pure program and without re-parsing
• run several tests, give OK only if all pass
• give human-readable output

I can achieve this with

import Test.SmallCheck.Property
import Control.Monad.Error

testE :: Testable a 
      => String -> Int -> a
      -> Either String String
testE name num prop 
  = case 
        filter ( \ r -> not (resultIsOk $ result r))
      $ take num
      $ do depth <- [0..] ; test prop depth
    of [] -> Right $ unwords 
          [ "property", name, "passed", show num , "tests" ]
       r : _ -> Left $ unwords $
         [ "property", name, "fails for inputs"
         ] ++ arguments r

so I guess I'd like to have that (or something similar to the same effect) in the library somewhere.

[UnkindPartition]

I like the idea. Will add something like this.

[jwaldmann]

Here is some background on why I want such a feature:
http://dfa.imn.htwk-leipzig.de/bugzilla/show_bug.cgi?id=304

[UnkindPartition]

Sorry for the delay.

This is added in cd42a8d. It's a bit different from what you proposed:

• no pretty-printing is done, instead the raw data is returned
• it doesn't accept an additional bound parameter ('num' in your code)

If you are happy with this, I'll upload an updated version to hackage.