`primitive` library

[konsumlamm]

I plan to make a MicroHs version of the primitive package. It mostly consists of wrappers around primitive types and functions (arrays, mutable references, pointers). Since the implementation is very compiler dependent, I don't think it makes much sense to modify the original version. However, MicroHs should already provide all the necessary tools for a reimplementation. Now my question is: Should this be included with MicroHs itself (like bytestring etc.) or in a separate primitive-mhs package?

[augustss]

I think it should be in primitive-mhs. I will probably move bytestring&co out of base to bytestring-mhs etc.

[konsumlamm]

Ok. Should we make a GitHub organization for that?

[augustss]

I don't know how GitHub organisations work. But it sounds good. 🙂

[augustss]

How are you planning to deal with the fact that the package is full of MagicHash?

[konsumlamm]

I don't know how GitHub organisations work. But it sounds good. 🙂

An organization can have multiple members and repositories under a shared namespace. You can configure the permissions of each member to have access to all organization repositories or only specific ones. See https://docs.github.com/en/organizations/collaborating-with-groups-in-organizations/about-organizations.

How are you planning to deal with the fact that the package is full of MagicHash?

I plan to not include those parts. MagicHash is only used for unlifted/unboxed types and functions on unlifted/unboxed types, most of which are GHC primitives, which MicroHs doesn't support anyway.

There is a problem however. I noticed that the PrimMonad class, which is used throughout the primitive package, has a type family, which MicroHs doesn't support yet, so I guess this project will have to wait.

[augustss]

I think we should do primitive-mhs. MicroHs now has the MagicHash extension and pretends to have unboxed tuples and sums. It can also fake having data Int = I# Int#, which the new version of the ghc-compat library will do.

There's still no TypeFamilies, but I think I'll do that next.