Replace Word32 with a newtype

[michaelpj]

This time with the correct (strange) bounds.

[michaelpj]

Builds are generally passing in haskell/haskell-language-server#2494, so I think this is safe to merge, and then we should be okay to do a release and then fix up HLS.

[Bodigrim]

Introducing OddWord dependency is likely to cause a trouble: it was not updated in years and OddWord-1.0.2.0 already does not build with GHC 9.2.

[michaelpj]

Introducing OddWord dependency is likely to cause a trouble: it was not updated in years and OddWord-1.0.2.0 already does not build with GHC 9.2.

Ugh, how annoying. I was hoping that it was a simple enough package that it would just "keep working".

I was really hoping to avoid writing the plumbing for a 31-bit word myself...

[Bodigrim]

Unexpectedly, older releases of OddWord do work (I guess they contain less type level magic), and since I've revised OddWord-1.0.2.0 with an appropriate base bound, cabal should now be able to derive a working build plan for GHC 9.2. Dunno, still does not sound sustainable to me, but I've been just passing by.

[michaelpj]

Maybe I'll ask if I can take over OddWord. It's a small package.

[jneira]

looks good to me, do you think we can continue using OddWord?
Well if it becomes a blocker we can always do the boilerplate

[michaelpj]

Yes, ripping it out later would be a localized change. Maybe the LSP people will hear my plea and change the upper bound (yeah right).

[Bodigrim]

Any particular reason to prefer OddWord to any well-maintained package for modular arithmetic? Something like https://hackage.haskell.org/package/mod-0.1.2.2/docs/Data-Mod-Word.html would do, for instance. OddWord is not even on Stackage.

[michaelpj]

Thanks, it didn't occur to me to look for a modular arithmetic package, that makes a lot of sense in fact!

Mod is missing a bunch of the numeric instances, but that might be okay. Let me see.