GitLab

Bodigrim pushed to branch wip/fix-toRational at Glasgow Haskell Compiler / GHC

Commits:

8aa42af4

by Andrew Lelechenko at 2025-06-15T17:25:10+01:00

Fix issues with toRational for types capable to represent infinite and not-a-number values

This commit fixes all of the following pitfalls:

> toRational (read "Infinity" :: Double)
179769313486231590772930519078902473361797697894230657273430081157732675805500963132708477322407536021120113879871393357658789768814416622492847430639474124377767893424865485276302219601246094119453082952085005768838150682342462881473913110540827237163350510684586298239947245938479716304835356329624224137216 % 1
> toRational (read "NaN" :: Double)
269653970229347386159395778618353710042696546841345985910145121736599013708251444699062715983611304031680170819807090036488184653221624933739271145959211186566651840137298227914453329401869141179179624428127508653257226023513694322210869665811240855745025766026879447359920868907719574457253034494436336205824 % 1

> realToFrac (read "NaN" :: Double) -- With -O0
Infinity
> realToFrac (read "NaN" :: Double) -- With -O1
NaN

> realToFrac (read "NaN" :: Double) :: CDouble
Infinity
> realToFrac (read "NaN" :: CDouble) :: Double
Infinity

Implements https://github.com/haskell/core-libraries-committee/issues/338

4 changed files:

libraries/base/changelog.md
libraries/ghc-internal/src/GHC/Internal/Float.hs
libraries/ghc-internal/src/GHC/Internal/Real.hs
testsuite/tests/numeric/should_run/T9810.stdout

Changes:

libraries/base/changelog.md

@@ -11,6 +11,7 @@
    * `instance Functor NonEmpty` is now specified using `map` (rather than duplicating code). ([CLC proposal #300](https://github.com/haskell/core-libraries-committee/issues/300))
    * `fail` from `MonadFail` now carries `HasCallStack` constraint. ([CLC proposal #327](https://github.com/haskell/core-libraries-committee/issues/327))
    * The `Data.Enum.enumerate` function was introduced ([CLC #306](https://github.com/haskell/core-libraries-committee/issues/306))
 +  * Fix issues with toRational for types capable to represent infinite and not-a-number values ([CLC proposal #338](https://github.com/haskell/core-libraries-committee/issues/338)
    * Worker threads used by various `base` facilities are now labelled with descriptive thread labels ([CLC proposal #305](https://github.com/haskell/core-libraries-committee/issues/305), [GHC #25452](https://gitlab.haskell.org/ghc/ghc/-/issues/25452)). Specifically, these include:
      * `Control.Concurrent.threadWaitRead`
      * `Control.Concurrent.threadWaitWrite`

libraries/ghc-internal/src/GHC/Internal/Float.hs

@@ -430,14 +430,10 @@ naturalToFloat# (NB b) = case integerToBinaryFloat' (IP b) of
  -- | @since base-2.01
  --
 --- Beware that 'toRational' generates garbage for non-finite arguments:
 ---
 --- >>> toRational (1/0 :: Float)
 --- 340282366920938463463374607431768211456 % 1
 --- >>> toRational (0/0 :: Float)
 --- 510423550381407695195061911147652317184 % 1
 ---
  instance  Real Float  where
 +    toRational x
 +      | isInfinite x = if x > 0 then infinity else -infinity
 +      | isNaN x = notANumber
      toRational (F# x#)  =
          case decodeFloat_Int# x# of
            (# m#, e# #)
@@ -686,14 +682,10 @@ naturalToDouble# (NB b) = case integerToBinaryFloat' (IP b) of
  -- | @since base-2.01
  --
 --- Beware that 'toRational' generates garbage for non-finite arguments:
 ---
 --- >>> toRational (1/0)
 --- 179769313 (and 300 more digits...) % 1
 --- >>> toRational (0/0)
 --- 269653970 (and 300 more digits...) % 1
 ---
  instance  Real Double  where
 +    toRational x
 +      | isInfinite x = if x > 0 then infinity else -infinity
 +      | isNaN x = notANumber
      toRational (D# x#)  =
          case integerDecodeDouble# x# of
            (# m, e# #)

libraries/ghc-internal/src/GHC/Internal/Real.hs

@@ -703,15 +703,6 @@ fromIntegral = fromInteger . toInteger
  -- | General coercion to 'Fractional' types.
  --
 --- WARNING: This function goes through the 'Rational' type, which does not have values for 'NaN' for example.
 --- This means it does not round-trip.
 ---
 --- For 'Double' it also behaves differently with or without -O0:
 ---
 --- > Prelude> realToFrac nan -- With -O0
 --- > -Infinity
 --- > Prelude> realToFrac nan
 --- > NaN
  realToFrac :: (Real a, Fractional b) => a -> b
  {-# NOINLINE [1] realToFrac #-}
  -- See Note [Allow time for type-specialisation rules to fire]

testsuite/tests/numeric/should_run/T9810.stdout

  ## Double ##
  Infinity
  -Infinity
 -Infinity
 +NaN
  Infinity
  -Infinity
  Infinity
  ## Float ##
  Infinity
  -Infinity
 -Infinity
 +NaN
  Infinity
  -Infinity
  Infinity