#13328: Foldable, Functor, and Traversable deriving handle phantom types badly -------------------------------------+------------------------------------- Reporter: dfeuer | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: 8.4.1 Component: Compiler | Version: 8.1 Resolution: | Keywords: deriving-perf Operating System: Unknown/Multiple | Architecture: Type of failure: Runtime | Unknown/Multiple performance bug | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by RyanGlScott):
The question about `fmap` strictness looks like a red herring to me. There's no real difference between the empty case definition and the `coerce` definition.
Sure, I just wanted to note the "conflict".
The illegitimate `fmap` is of very little interest to me. Someone who cares about preserving their map count or whatever just won't be able to use `Functor` deriving for types explicitly mentioning their bogus one.
That's also fine, just make sure we note this if this gets implemented. Luckily, `Functor`/`Foldable`/`Traversable` happen to be pretty rigid, law-abiding classes, so this restriction won't really hurt anyone in practice. What is your opinion on the `RoleAnnotations` example above? This is the only thing that genuinely concerns me, since I think having the strictness of a derived `Foldable` instance change depending on whether a type parameter is phantom or non-phantom is quite unintuitive. I'm not sure of a good workaround, however. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/13328#comment:7 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler