Interpret instance for unnamed data fields

[bosu]

Maps unnamed data fields to the positional record keys. For example:

data Foo = Foo Bool Integer Natural deriving (Generic, Interpret)

is mapped to:

Record { _1 : Bool, _2 : Integer, _3 : Natural }

Dhall data type.

[Gabriel439]

I think this is a great idea! If we do this then we can eliminate the need for R2. Just a few comments

[Gabriel439]

I think this is a great idea! If we do this then we can eliminate the need for R2. Just a few comments

[Gabriel439]

Shouldn't this line be:

let Type extractR expectedR = evalState (genericAutoWith options) 1

[Gabriel439]

Same comment here. Shouldn't this be:

let Type extractL expectedL = evalState (genericAutoWith options) 1

[Gabriel439]

Since we're turning genericAutoWith into a monadic function, we can also automatically thread the InterpretOptions through the code by changing the type to:

genericAutoWith :: ReaderT InterpretOptions (State Int (Type (f a)))

... although it might be slightly easier to work with this type:

genericAutoWith :: StateT Int (Reader InterpretOptions (Type (fa)))

[bosu]

The amended commit much simplified the diff. IMHO Reader may be unnecessary anymore...

[Gabriel439]

Note that TypeApplications only works for GHC 8.0 and later, but the Travis test suite still supports GHC 7.8 (in spirit; the test is currently failing due to cabal solver issues)

However, I'd be willing to drop GHC 7.8 support because even Debian stable now supports GHC 8.0

[bosu]

+1 for dropping GHC 7.8

[Gabriel439]

Minor suggestion: you can simplify this a little bit by relocating expected and extract inside the do block, like this:

name <- case ...
let extract (RecordLit m) = ...
let expected = ...
pure (Type {..})

[Gabriel439]

I believe Test.Tasty.Hunit also provides assertEqual, so you don't need to pull in the additional HUnit dependency

(To be honest, I'm not even sure what's the difference between the two libraries. I didn't originally author the test suite)

[Gabriel439]

You can lower the lower bound on containers to >= 0.5.0.0 (consistent with the containers dependency for the main library) since that version still has Data.Map.fromList

[bosu]

Sorry for losing the remarks by using amended push request. All of them were addressed. Now I see the advantages of additional commits :)

Regarding R2 removal: it is still needed by Inject instance... Probably the same kind of refactoring can happen there.

[Gabriel439]

Thanks! :)