26 Jul
2018
26 Jul
'18
6:31 p.m.
On Thurs, Jul 26, 2018 at 05:39:00, Emil Axelsson wrote: >> Den 2018-07-26 kl. 01:02, skrev Anthony Clayden: >> >>> * On Tue, Jul 24, 2018 at 5:52 AM,Ryan Reich wrote: *>*>> This seems like the canonical illustration of the dangers of * >>>* UndecidableInstances.* >> >>* No I don't buy that explanation. The worst UndecidableInstances can do * >>* is send the type checker into a loop.* > I used to have a counter-example to that claim: https://mail.haskell.org/pipermail/haskell-cafe/2010-January/072357.html > That example was causing looping dictionaries at run time on GHC 6.10. > But now, on GHC 8.2, the program terminates. 6.10 was rather a long time ago. The whole type inference/improvement engine has had several overhauls since 2010. To be clear: you can write looping programs without type classes at all -- such as mutually recursive functions. You can write looping programs with type classes -- by turning those functions into methods in classes. You can write looping programs with type classes that need `UndecidableInstances` -- by making those classes mutual superclasses or mutual constraints on instances. That doesn't mean the `UndecidableInstances` are to 'blame' for the looping. I would blame the functions/method calls for not reducing the problem. > But one wonders if there are other ways to create looping dictionaries? With `UndecidableInstances`, yes there's plenty. And I expect that'll make the typechecker loop at compile time. The O.P. here is not getting typechecker looping, so I think the explanation is to do with method calls -- for which we need to see what's going on in the imports. AntC