Avoid depending on free and distributive

[vmchale]

Unfortunately, free and distributive (and all that depend on them) break cross-compilation via cabal-install due to ekmett/distributive#53 (comment)

Would it be possible to avoid depending on them? They introduce a large dependency footprint.

This would entail (directly) not depending on free, either, and recursion-schemes.

[sjakobi]

@vmchale Note that dhall plans on using cabal-doctest too!

Regarding the dependencies: In principle we could put them behind feature-flags.

[Gabriella439]

@vmchale: Do you have any issue with a dependency on the data-fix package?

[vmchale]

@sjakobi Oh hmm. I'll comment over there. This is ostensibly a cabal-install problem but I'm guessing it will take a lot of work to fix.

@Gabriel439 nope. Nothing wrong with recursion schemes :)

[quasicomputational]

cabal-install bugs aside, this issue would also be fixed by Cabal learning how to drive doctest properly. It's a very worthy long-term goal, but it's non-trivial. For the time being I am quite definitely -1 on excising cabal-doctest because it really is carrying weight - plumbing the correct dependency information around without it is somewhere between very hard and actively impossible to maintain.

Also, on removing distributive, that entails no profunctor, which would mean no prisms or isos (or, at least, none that will interoperate with the rest of the Kmettverse). Not that there are currently any of those at present, but foreclosing the possibility is another thing weighing against this in my mind.

[Gabriella439]

I think the thing I can definitely do is remove the dependency on free and recursion-schemes. Those dependencies are very easy to excise and I'll have a pull request ready soon. I've actually been meaning to do that anyway and this gives me an excuse to do so.

Removing other dependencies will be trickier and I'd like to first explore other alternatives

[Gabriella439]

The fix to remove the free/recursion-schemes dependencies is up here: #1298

[Gabriella439]

This was fixed by #1298