I think I may have been confused for 2 different reasons.
First, I forgot that we are coercing in your example from Dimensional Int Double -> Double, which should work according to rule 2 in figure 2. I had it mixed up with converting from Dimensional Int Double -> Dimensional Int x (where there is a coercion from Double to x).
Second, it doesn't appear to actually be possible to export the Quantity' constructor from the defining module at all. (If I put it in the export list, GHCi tells me that it isn't in scope even though it is and I checked the spelling three times.) So the scenario I was thinking of isn't
#9580: Possible excessive leniency in interaction between coerce and data families? -------------------------------------+------------------------------------- Reporter: dmcclean | Owner: simonpj Type: bug | Status: closed Priority: low | Milestone: Component: Compiler | Version: 7.8.3 (Type checker) | Keywords: Resolution: fixed | Architecture: Unknown/Multiple Operating System: | Difficulty: Unknown Unknown/Multiple | Blocked By: Type of failure: | Related Tickets: None/Unknown | Test Case: indexed- | types/should_fail/T9580 | Blocking: | Differential Revisions: | -------------------------------------+------------------------------------- Comment (by dfeuer): Replying to [comment:8 dmcclean]: possible for that reason too; I hadn't realized that this was a rule.
I now think that with the change you made everything is correct, sorry
for the confusion. Did you use the proper export syntax, `Dimensional (Quantity')`? That's supposed to work. If you didn't get an unknown whatever error when exporting `Quantity'` from the defining module, I'm ''guessing'' that it exported the `Quantity'` ''type'' (produced by `DataKinds`) rather than the data constructor. I don't really know how that extension is supposed to work with module exports though. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/9580#comment:9 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler