leventerkok / floatinghex Goto Github PK

It would be nice to support negative literals as well, like this:

[hf|-0x1p1]

Unfortunately, this is currently not possible due to a template-haskell shortcoming: There's no way to represent negative 0! (Or infinity for that matter.) See here: http://ghc.haskell.org/trac/ghc/ticket/13124

If TH is ever remedied to support proper floats, we can consider supporting negative literals as well.

Fyi, FloatingHex needs an instance that was only added with TH 2.10 and later

Configuring component lib from FloatingHex-0.2...
Preprocessing library FloatingHex-0.2...
[1 of 1] Compiling Data.Numbers.FloatingHex ( Data/Numbers/FloatingHex.hs, /tmp/matrix-worker/1484424259/dist-newstyle/build/x86_64-linux/ghc-7.8.4/FloatingHex-0.2/build/Data/Numbers/FloatingHex.o )

Data/Numbers/FloatingHex.hs:92:30:
    No instance for (TH.Lift Double) arising from a use of ‘TH.lift’
    In the expression: TH.lift d
    In a case alternative: Just d -> TH.lift d
    In the expression:
      case parseHexFloat s of {
        Just d -> TH.lift d
        Nothing
          -> fail
             $ "Invalid hexadecimal floating point number: |" ++ s ++ "|" }

I've already fixed up the meta-data on Hackage, e.g.

• https://hackage.haskell.org/package/FloatingHex-0.2/revisions/

Currently showHFloat is not compatible with printf's %a specifier. Using Haskell's floatToDigit it should be possible to remedy this situation.

Would be nice if the parser read nan/infinity etc. as well; as special cases. Note that we have to use :% in these cases to make sure reduction isn't done right away.

Due to rational requirement, the quasiquoter doesn't support negatives. But we should support it in the parse function, which should also be exported.