More BackendCompatible generalizations

[parsonsmatt]

This PR includes some additional generalizations. Unfortunately, my work project does not use the SqlReadT or SqlWriteT features, and it was only when the code was tested on those did the issues with Esqueleto come up.

This PR generalizes SqlReadT so that Esqueleto can be updated in a way that doesn't cause any breaking changes.

If any folks are using SqlReadT, I'd appreciate testing this PR/commit out with this Esqueleto PR to verify that it doesn't break any corner cases I've missed 😄

[parsonsmatt]

These functions needed to be generalized to satisfy Esqueleto.

[parsonsmatt]

This import is necessary for the BackendCompatible SqlBackend SqlBackend instance.

[dariooddenino]

Compilation of persistent-mysql-2.6.2 fails with this branch. persistent-mysql-2.6.1 works fine.

Progress: 2/3
--  While building package persistent-mysql-2.6.2 using:
      /home/dario/.stack/setup-exe-cache/x86_64-linux-nopie/Cabal-simple_mPHDZzAJ_1.24.2.0_ghc-8.0.2 --builddir=.stack-work/dist/x86_64-linux-nopie/Cabal-1.24.2.0 build --ghc-options " -ddump-hi -ddump-to-file"
    Process exited with code: ExitFailure 1
    Logs have been written to: /home/dario/dev/blue-moon/.stack-work/logs/persistent-mysql-2.6.2.log

    Configuring persistent-mysql-2.6.2...
    Building persistent-mysql-2.6.2...
    Preprocessing library persistent-mysql-2.6.2...
    [1 of 1] Compiling Database.Persist.MySQL ( Database/Persist/MySQL.hs, .stack-work/dist/x86_64-linux-nopie/Cabal-1.24.2.0/build/Database/Persist/MySQL.o )
    
    /tmp/stack4584/persistent-mysql-2.6.2/Database/Persist/MySQL.hs:1188:17: error:
        • Could not deduce (BackendCompatible r0 backend)
            arising from a use of ‘projectBackend’
          from the context: (backend ~ PersistEntityBackend record,
                             BackendCompatible SqlBackend backend,
                             PersistEntity record,
                             MonadIO m)
            bound by the type signature for:
                       insertManyOnDuplicateKeyUpdate :: (backend
                                                          ~
                                                          PersistEntityBackend record,
                                                          BackendCompatible SqlBackend backend,
                                                          PersistEntity record, MonadIO m) =>
                                                         [record]
                                                         -> [SomeField record]
                                                         -> [Update record]
                                                         -> ReaderT backend m ()
            at Database/Persist/MySQL.hs:(1175,1)-(1185,27)
          The type variable ‘r0’ is ambiguous
          Relevant bindings include
            insertManyOnDuplicateKeyUpdate :: [record]
                                              -> [SomeField record]
                                              -> [Update record]
                                              -> ReaderT backend m ()
              (bound at Database/Persist/MySQL.hs:1186:1)
          These potential instances exist:
            instance BackendCompatible SqlBackend SqlBackend
              -- Defined in ‘persistent-2.7.1:Database.Persist.Sql.Orphan.PersistStore’
            instance BackendCompatible SqlBackend SqlReadBackend
              -- Defined in ‘persistent-2.7.1:Database.Persist.Sql.Orphan.PersistStore’
            instance BackendCompatible SqlBackend SqlWriteBackend
              -- Defined in ‘persistent-2.7.1:Database.Persist.Sql.Orphan.PersistStore’
        • In the first argument of ‘withReaderT’, namely ‘projectBackend’
          In the first argument of ‘(.)’, namely ‘withReaderT projectBackend’
          In the expression: withReaderT projectBackend . uncurry rawExecute
    
    /tmp/stack4584/persistent-mysql-2.6.2/Database/Persist/MySQL.hs:1189:15: error:
        • Could not deduce (BackendCompatible SqlBackend r0)
            arising from a use of ‘rawExecute’
          from the context: (backend ~ PersistEntityBackend record,
                             BackendCompatible SqlBackend backend,
                             PersistEntity record,
                             MonadIO m)
            bound by the type signature for:
                       insertManyOnDuplicateKeyUpdate :: (backend
                                                          ~
                                                          PersistEntityBackend record,
                                                          BackendCompatible SqlBackend backend,
                                                          PersistEntity record, MonadIO m) =>
                                                         [record]
                                                         -> [SomeField record]
                                                         -> [Update record]
                                                         -> ReaderT backend m ()
            at Database/Persist/MySQL.hs:(1175,1)-(1185,27)
          The type variable ‘r0’ is ambiguous
          These potential instances exist:
            instance BackendCompatible SqlBackend SqlBackend
              -- Defined in ‘persistent-2.7.1:Database.Persist.Sql.Orphan.PersistStore’
            instance BackendCompatible SqlBackend SqlReadBackend
              -- Defined in ‘persistent-2.7.1:Database.Persist.Sql.Orphan.PersistStore’
            instance BackendCompatible SqlBackend SqlWriteBackend
              -- Defined in ‘persistent-2.7.1:Database.Persist.Sql.Orphan.PersistStore’
        • In the first argument of ‘uncurry’, namely ‘rawExecute’
          In the second argument of ‘(.)’, namely ‘uncurry rawExecute’
          In the expression: withReaderT projectBackend . uncurry rawExecute

[parsonsmatt]

Good catch @dariooddenino The generalizations I'm making here make the preemptive persistent-mysql generalizations unnecessary :)

[paul-rouse]

@parsonsmatt This is looking good! Could you please rebase to the latest master, and make the changelog message describe the whole PR (include a link to the PR too) instead of saying it fixes a compilation error (which was just an intermediate step!)? Then, assuming travis is still OK, I can merge it.

[parsonsmatt]

Weird spurious build failure: https://travis-ci.org/yesodweb/persistent/jobs/296983378 Tests passed and all :\

[paul-rouse]

Looked like a last-minute glitch. I restarted that test and it is fine now.

Thanks for all of this - I'll go ahead and merge it.