$id: antiquoter should expect an `Id`, not a `String`

[mainland]

$id: antiquoter should expect an `Id`, not a `String`

[mainland]

You're right, it is inconsistent. I have in the past considered making the change you are suggesting, but decided not to because it would require changing a lot of client code.

As for $idname:x, is x the Name of a Haskell variable that contains a value of type Id? Are you generating Haskell code that generates C code? Or is the Name supposed to be converted to a value of type Id?

Let me propose a different approach that I think would address all of these issues. There is currently a ToExp class in Language.C.Quote, and any value with a type that has a visible instance of ToExp can appear as the argument to an $exp antiquotation, i.e., $exp:e is valid for any e that has a type that is an instance of ToExp. When the antiquotation is expanded, toExp e is spliced in. We could add a ToIdent class and define instances for both String and Id. That would retain backwards compatibility, support antiquoting values of type Id, avoid adding yet another type of antiquotation, and allow us (or library clients) to add new instances in the future---like yours for Name.

Sound good?

[mainland]

$id:x now allows x to have type Id.

[mchakravarty]

Great — thanks!