Changes required for Opaleye 0.9.1.0

[tomjaguarpaw]

Opaleye 0.9.1.0 has fairly aggressive simplification of QueryArr internals. This PR is required for rel8 to build against it (0.9.1.0 is not internals-compatible with 0.9.0.0).

I took the opportunity to expose the rebind and distinct things you need in Opaleye's public API, so they won't break in future internal-only changes.

(I'm not completely certain that c47317e is correct. It doesn't seem to be tested. You can drop it if you find it's not correct. All the other commits are essential though.)

[ocharles]

Thanks @tomjaguarpaw, I really appreciate your help when there are breaking changes (for us) in Opaleye! I've just bumped Haskell.nix so CI can run, and I've pinged @shane-circuithub to give me a hand reviewing.

[tomjaguarpaw]

You're welcome. I tend to think that it's easier for me to make the fixes, since I know exactly the things that have changed in Opaleye.

[tomjaguarpaw]

One thing to bear in mind when reviewing is that the new PrimQueryArr monoid composes in the opposite direction that the PrimQuery function builder used to compose, that is, the instance is

instance Semigroup PrimQueryArr where
  PrimQueryArr f1 <> PrimQueryArr f2 = PrimQueryArr (\lat -> f2 lat . f1 lat)

and what use to be written, for example, as

\lat -> Rebind True p lat . restrict e lat`

is now written in the "opposite" order

aRestrict e <> aRebind p

[ocharles]

Good to know!

We've had a chat about this and the changes look good. We're going to deploy them at CircuitHub for the week just to double check, and if everything looks good I'll cut a new release on Friday. It's unfortunate I can't take enough confidence just from test results, but if anything comes up I can hopefully add tests in.

[shane-circuithub]

This looks good to me, and I've tested the alignBy change and it seems to still work.