shrinkRealFrac & shrinkDecimal problems

[peterwicksstringfield]

It looks like shrinkRealFrac can't shrink negative numbers and goes into an infinite loop on positive infinity. shrinkDecimal handles negatives okay but has the same problem on positive infinity.

Prelude Test.QuickCheck> shrinkRealFrac (-17)
[0.0]
Prelude Test.QuickCheck> shrinkRealFrac 17
[0.0,9.0,13.0,15.0,16.0,0.0,9.0,13.0,15.0,16.0]
Prelude Test.QuickCheck> 1/0 :: Float
Infinity
Prelude Test.QuickCheck> shrinkRealFrac (1/0 :: Float)
[0.0,0.0,0.0,0.0,0.0,^C,0.0,0.0,0Interrupted.
Prelude Test.QuickCheck> shrinkDecimal (1/0 :: Float)
[0.0,0.0,0.0,0.0,0.0,0.0,^C0.0,0.Interrupted.

This is with

$ nix-shell -p "haskell.packages.ghc8107.ghcWithPackages (pkgs: with pkgs; [ QuickCheck ] )"
$ ghci --version
The Glorious Glasgow Haskell Compilation System, version 8.10.7
$ cabal info QuickCheck
[...]
    Versions installed: 2.14.2
[...]

Looks to me like the problem is:

1. That the infinity-branch for shrinkRealFrac is overlaying the negative-number-branch; since (2*x + 1 > x) is false for negative numbers.
2. The "iterate (*2) 0" in both functions is a mistake?

Test/QuickCheck/Arbitrary.hs line 1221

shrinkRealFrac :: RealFrac a => a -> [a]
shrinkRealFrac x
  | not (x == x)  = 0 : take 10 (iterate (*2) 0) -- NaN
  | not (2*x+1>x) = 0 : takeWhile (<x) (iterate (*2) 0) -- infinity
  | x < 0 = negate x:map negate (shrinkRealFrac (negate x))
  | otherwise = [...]
shrinkDecimal :: RealFrac a => a -> [a]
shrinkDecimal x
  | not (x == x)  = 0 : take 10 (iterate (*2) 0)        -- NaN
  | not (2*abs x+1>abs x) = 0 : takeWhile (<x) (iterate (*2) 0) -- infinity
  | otherwise = [...]

[nick8325]

Thanks for the bug report! You're right: that code is obviously nonsense. I can only guess I meant to do iterate (*2) 1 in the first two cases, and the infinity test should look at the absolute value of x.