[Haskell-cafe] Why does GHC not warn about unnamed uninitialized fields?

Eelis van der Weegen eelis at eelis.net
Wed Nov 23 06:33:15 CET 2011


Today I noticed that GHC is more concerned (when using -Wall) about 
uninitialized fields when those fields have names:

   data A = A {a::Integer}
   data B = B Integer

   x :: A
   x = A{} -- Gives a nice warning: "Fields of `A' not initialised: a"

   y :: B
   y = B{} -- No warning!

Is this on purpose? If so, what is the rationale?

The context in which I encountered this boils down to the following:

   {-# LANGUAGE RecordWildCards #-}

   data C = C {c::Integer}

   f :: C -> C
   f C{..} = C{..}

   g :: C -> Integer
   g (C i) = i

   main :: IO ()
   main =  print (g (f (C 3)))

This code worked fine (and printed "3"), until I made C's Integer 
nameless. This made it crash with following runtime error instead:

   T: T.hs:6:11-15: Missing field in record construction

The crash and error are perfectly understandable, but it would have been 
more helpful if GHC had warned about this at compile-time! The reason it 
didn't, though, is because of its aforementioned cavalier attitude 
towards uninitialized fields that don't have names... :-)

(I'm using GHC 7.0.4.)

Cheers,

Eelis




More information about the Haskell-Cafe mailing list