On 2019-12-16 1:42 a.m., Viktor Dukhovni wrote:
Not sure where to post about this, and it has likely been noted elsewhere already, but reading through the report I see in Section 3.17.2
I'm not sure either, but one place you could report it is at https://github.com/haskell/rfcs/issues No guarantees, but it would make the issue slightly less likely to be forgotten.
• Consider the following declarations:
newtype N = N Bool data D = D !Bool
These examples illustrate the difference in pattern matching between types defined by data and newtype:
(\ (N True) -> True) ⊥ ⇒ ⊥ (\ (D True) -> True) ⊥ ⇒ ⊥ (\ ~(D True) -> True) ⊥ ⇒ True
Additional examples may be found in Section 4.2.3.
But unlike the examples in 4.2.3, which do show differences between "newtype" and "data", the above does not seem to show any actual difference. For that the pattern after the constructor would I believe have to be irrefutable, which "True" is not.
Thus, e.g. the below (which duplicates 4.2.3) would do the job:
(\ (N _) -> True) ⊥ ⇒ True (\ (D _) -> True) ⊥ ⇒ ⊥
Perhaps I misunderstood something, but if not, and this has not been noted before, it may be worth fixing in the next revision.