API to construct InputTypes for Record types.

[eamsden]

Adds the RecordInputType type and associated functions inputFieldWith, inputField, and inputRecord.

These inputField{,With} functions and the Contravariant and Divisible operations create RecordInputType values which describe how a Haskell record type may be encoded into a Dhall expression. The inputRecord function converts a RecordInputType to an InputType.

An example (using the same type as the example for RecordType) is given in the documentation comment for `RecordInputType.

See #525

[Gabriella439]

Looks great! Just two small comments

[Gabriella439]

I believe the type applications are not necessary since they should be inferred by the context

[eamsden]

I think you are right. And actually, given the way the API is written, they should either be inputFieldWith or the second parameter should be dropped entirely. The example as written woudn't typecheck!

[Gabriella439]

Note that we have doctest enabled on this module, so if you transform the example into a doctest then it will be checked 🙂

[eamsden]

I might do that in a separate PR, as it would involve modifying the $setup block to bring the Project type into scope.

[Gabriella439]

Yeah, it's not urgent to do in this PR

[Gabriella439]

I just realized that the >$< operator already exists in the contravariant package:

http://hackage.haskell.org/package/contravariant-1.5/docs/Data-Functor-Contravariant.html

... so we can reuse that

[eamsden]

I can drop that definitiion and just rexport the contravariant definition. Give me a few minutes.

[quasicomputational]

Can I bikeshed the naming here? We already have an input* family of functions that reads expressions from strings in various ways; at a glance, because of the naming, these look like members of that family, but they're not, really. So maybe recordFieldWith, recordField and recordInputType would be better?

[eamsden]

@quasicomputational maybe we should use the inject prefix? I was naming after InputType, which they are intended to aid construction of.

calling them recordField{,With} doesn't really give an idea of their distinction from field

[eamsden]

I think this should be infixl 4 >*< to work properly with the >$< exported from contravariant, right?

[Gabriella439]

It should still be infixr if you want the tuple nesting to be (a, (b, c)), but the precedence needs to be elevated to 5

[eamsden]

@Gabriel439 does this need anything else before it can be merged?

[Gabriella439]

@eamsden: No, nothing needs to be done. I just thought you had additional changes planned, but if you think this is ready then I'll go ahead and merge