Updates for persistent-2.5 and base-4.9

[paul-rouse]

I've built this and run the tests (sqlite only) with lts-6.2, lts-6.2 plus forced persistent*-2.5.X, and nightly-2016-06-08 to bring GHC 8 into the mix.

I included a major version bump because of the type constraint on deleteKey changing.

Something has changed in the way persistent exports count, so I had to hide it explicitly, but I didn't get to the bottom of why.

Hope it helps!

[paul-rouse]

Just to be clear - this differs from #134 and #139 in that it is still compatible with previous persistent releases.

[tebello-thejane]

So... will this be getting released soon?

[bitemyapp]

Keep in mind the people that prompted the changes in Persistent 2.5 already have a variant of Esqueleto updated to work with them: https://github.com/pseudonom/esqueleto/tree/split-db

If you don't mind I'll put a PR of @pseudonom's changes in and ping the owner.

[tebello-thejane]

As @paul-rouse said above

Just to be clear - this differs from #134 and #139 in that it is still compatible with previous persistent releases.

@pseudonom's changes do not allow esqueleto to work with versions of persistent < 2.5, thereby drastically increasing the possibility that cabal won't be able to compile complicated projects which use esqueleto.

This current PR seems to be the most appropriate one but it isn't getting merged?

[bitemyapp]

do not allow Esqueleto to work with versions of persistent < 2.5

Ah yeah, that's an ache and a good point.

[pseudonom]

The thing I like about my PR is that it annotates select as a read-only operation and allows you to use it against a read DB. Should we use CPP to come up with a PR that satisfies both criteria?

[bitemyapp]

@pseudonom that was originally why I wanted your version. If you could reconcile the benefits of both, even with CPP, that would be great.

[paul-rouse]

Of course this PR uses CPP too, in just one file. It would be great if we can get the best of both worlds - I didn't attempt to make use of the new read-only functionality. Give me a shout if I can help, for example with testing.

[paul-rouse]

I suggest we close this in favour of #144