Why no exposed `NullAddr#` literal?
Haskell Community [Unofficial]
May 27, 2026
tomjaguarpaw:
> I’m not sure it is fair, or at least there could in principle be an exception for values, and a newtype constructor applied to a value would be value. Maybe that adds too much extra complexity to be worth it.
I worked on this during my internship at Well-Typed. It is quite complicated. I managed to implement a basic form for unlifted top-level binding in Core (for optimizations), but that is stuck on supporting this in GHCi: !10841.
Discussion in the ATmosphere