#12369: data families shouldn't be required to have return kind *, data instances should -------------------------------------+------------------------------------- Reporter: ekmett | Owner: Type: feature request | Status: new Priority: normal | Milestone: 8.2.1 Component: Compiler (Type | Version: 8.0.1 checker) | Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by Iceland_jack): [http://cs.brynmawr.edu/~rae/papers/2016/thesis/eisenberg-thesis-draft.pdf Eisenberg]:
A ''data family'' defines a family of datatypes. An example shows best how this works:
{{{#!hs data family Array a -- compact storage of elements of type a data instance Array Bool = MkArrayBool ByteArray data instance Array Int = MkArrayInt (Vector Int) }}}
With such a definition, we can have a different runtime representation for `Array Bool` than we do for `Array Int`, something not possible with more traditional parameterized types.
The part about **runtime representation** made me wonder if you could define {{{#!hs -- × × × FAILS × × × data family Array a :: TYPE rep }}} I dunno about the instances, maybe using [https://ghc.haskell.org/trac/ghc/wiki/UnliftedDataTypes#Proposal3:Allownewty... Proposal 3: Allow newtypes over unlifted types] {{{#!hs newtype instance Array Int# :: TYPE PtrRepUnlifted where MkInt# :: ByteArray# -> Array Int# newtype instance Array (##) :: TYPE IntRep where MkInt# :: Int# -> Array (##) }}} I don't really understand representation polymorphism, may suffer from same problem as #11471. ---- {{{#!hs -- WORKS FINE data family Array (a :: TYPE rep) :: Type -- × × × FAILS × × × -- -- Unexpected type ‘Int#’ data Array Int# -- Should this fail? }}} -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/12369#comment:5 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler