This is a bit hard to explain, but my idea is to spark a discussion about how much graphql-api could benefit from using a proper Record library in some way (it's not clear how to me yet).
So I recently realized that, an object definition in a graphql-api schema is never inhabited with a value per-say. Like, we cannot create a value out of nowhere for it:
type Plant = Object "Plant" '[]
'[ Field "name" Text ]
-- this does not work =>
let test = Plant "Olala"
-- this does work, but we do not know it's a subtype of Plant though:
let (Just foo) = objectFromList [("name", toValue @Text "Olala")]
Of course we can create a resolver (that is a Handler in graphql-api) which can "sort of" create a graphql-api's Value's AST which will correspond to a "subtype" of Plant here.
Now this behavior is actually the very essence of graphql, so no problem. My point is, there should be a way to get back the fact that it's a Plant type, and a way to create an ad-hoc plant-type value. Probably the best way is to finalize the implementation of #192 . But I feel like this should be kept at the type level on haskell's side, and perhaps records or some kind of subtyping trick could do...
This is a bit hard to explain, but my idea is to spark a discussion about how much graphql-api could benefit from using a proper Record library in some way (it's not clear how to me yet).
So I recently realized that, an object definition in a graphql-api schema is never inhabited with a value per-say. Like, we cannot create a value out of nowhere for it:
Of course we can create a resolver (that is a Handler in graphql-api) which can "sort of" create a graphql-api's Value's AST which will correspond to a "subtype" of Plant here. Now this behavior is actually the very essence of graphql, so no problem. My point is, there should be a way to get back the fact that it's a Plant type, and a way to create an ad-hoc plant-type value. Probably the best way is to finalize the implementation of #192 . But I feel like this should be kept at the type level on haskell's side, and perhaps records or some kind of subtyping trick could do...