Add some hashed operations from classes that cannot be implemented

[phadej]

Especially traverseHashed which is a Lens (Hashed a) (Hashed b) a b can prove to be useful (there could be Iso, but we'd need a profunctors for that, which we cannot have here).

[tibbe]

This one feels a bit odd. Do you have an example of how you would use it?

[phadej]

Well, I don't know, added it for "completeness". Not insisting on it to be included.

[phadej]

Removed joinHashed

[phadej]

rebased

[RyanGlScott]

@phadej: Sadly, fixing the tests won't be easy because of the reasons I mention in #124 (comment).

[phadej]

@RyanGlScott well, the tests are run with newer GHCs at least, which is better than nothing.

As you can see, I found an issues with in the library itself, when build on older GHC (e.g. DeriveDataTypeable and Proxy), so having at least somehow working CI should be top priority. (EDIT and keeping it green!)

Now the travis is finally green, with tests run with GHC7.10 and GHC8.0.

[RyanGlScott]

@phadej, I totally agree that having green Travis builds is a must. Although at the moment, we're cheating a bit by selectively disabling tests, aren't we? It would be great if you could run the hashable tests on old GHCs yourself to verify that they do indeed work (after manually working around the dependency cycle by preinstalling all dependencies, then configuring).

Our task would probably be a lot simpler if QuickCheck had a -flight flag as proposed in nick8325/quickcheck#118 (comment).

[phadej]

@RyanGlScott I'd rather fix cabal. Using cabal-head is an option, would rather fix the issue (cabal is dumb) than try to work around it. This issue hits us in many packages, especially their benchmarks.

[RyanGlScott]

If you can fix the issue by depending on a newer version of cabal-install, I'm all for it :)

[phadej]

@RyanGlScott, will do, but separately.