Migrate to argonaut-codecs from foreign-generic #212
Conversation
|
|
||
| -- | The range of text associated with an error | ||
| newtype ErrorPosition = ErrorPosition | ||
| type ErrorPosition = |
There was a problem hiding this comment.
All of these types can be generically encoded and decoded via Argonaut without requiring newtypes and generic instances, so I've returned them to be raw records. However, if we really want those newtypes (for example, for more readable type errors) then I can reinstate them.
|
I think we can also remove the |
|
Good point on the |
| requestBody = AXRB.String code | ||
| requestBody = Just $ AXRB.Json $ encodeJson code |
There was a problem hiding this comment.
Wondering if the CORS issue reported in #215 is (unexpectedly) related to using AXRB.Json instead of AXRB.String.
There was a problem hiding this comment.
Wow, great find -- it totally is. Changing this to AXRB.String $ unsafeCoerce (encodeJson code) works as expected. I don't know why -- going to look a little deeper.
There was a problem hiding this comment.
Well, on reflection the code isn't actually JSON. It's just a string. Encoding it runs through a stringify call, which isn't what we want; I still don't know why this would cause a CORS issue (perhaps it's a red herring?) but regardless this wasn't a correct change to make in the first place.
This PR continues #211 by migrating from
foreign-generictoargonaut-codecsfor JSON encoding and decoding. It also updates the tests and ensures they're exercised in CI.