System.Future in contrib is strange, especially because it relies on two backend-specific functions, prim__makeFuture and prim__awaitFuture which are very different: the first one is a special primitive function that compiler is aware of, when the second one is a normal %foreign function. The main reason for this was the desire to pass Lazy value into the foreign function, which is not supported (for good reason, I think), so it was made a primitive.
But I think that this was, basically, the XY-problem -- we actually wanted to pass not a Lazy value of type a, but a function that returns a when demanded -- in general case, semantics of Lazy values may differ, but semantics of this call does not. Also, it is pretty disgusting that we have a compiler primitive that is excersised only in contrib. Also, this siituation leads to problems like #3292.
So, I propose the change which makes both basic functions in this module be declared as %foreign, thus removing the MakeFuture primitive from the compiler.
Should this change go in the CHANGELOG?
[x] If this is a fix, user-facing change, a compiler change, or a new paper
implementation, I have updated CHANGELOG_NEXT.md (and potentially also
CONTRIBUTORS.md).
Description
System.Futureincontribis strange, especially because it relies on two backend-specific functions,prim__makeFutureandprim__awaitFuturewhich are very different: the first one is a special primitive function that compiler is aware of, when the second one is a normal%foreignfunction. The main reason for this was the desire to passLazyvalue into the foreign function, which is not supported (for good reason, I think), so it was made a primitive.But I think that this was, basically, the XY-problem -- we actually wanted to pass not a
Lazyvalue of typea, but a function that returnsawhen demanded -- in general case, semantics ofLazyvalues may differ, but semantics of this call does not. Also, it is pretty disgusting that we have a compiler primitive that is excersised only incontrib. Also, this siituation leads to problems like #3292.So, I propose the change which makes both basic functions in this module be declared as
%foreign, thus removing theMakeFutureprimitive from the compiler.Should this change go in the CHANGELOG?
CHANGELOG_NEXT.md(and potentially alsoCONTRIBUTORS.md).