Decoding

For typed values, implement FromJson.

import Std.Fail (Fail)
import SagaJson as J
import SagaJson.Decode as D
import SagaJson.Decode (FromJson)

record Point {
  x: Int,
  y: Int,
} deriving (Debug)

impl FromJson for Point needs {Fail J.Error} {
  from_json j = Point {
    x: D.at "x" D.int j,
    y: D.at "y" D.int j,
  }
}

D.deserialize """{"x":1,"y":2}""" : Result Point J.Error

Primitive decoders:

D.string
D.int
D.float
D.bool

Navigation and composition:

D.field "name" j
D.at "name" D.string j
D.list_of from_json j
D.nullable D.string j

The FromJson trait already has impls for primitives, List a, and Maybe a, so nested fields can often use from_json directly.

record Drawing {
  name: String,
  points: List Point,
  note: Maybe String,
} deriving (Debug)

impl FromJson for Drawing needs {Fail J.Error} {
  from_json j = Drawing {
    name: D.at "name" D.string j,
    points: D.at "points" from_json j,
    note: D.at "note" from_json j,
  }
}

D.at prefixes errors with the field path. For example, decoding {"address":{"zip":123}} with D.at "address" (D.at "zip" D.string) returns a path like ["address", "zip"].

Use D.refine for validation that JSON shape alone cannot express.

fun nonnegative : Point -> Point needs {Fail J.Error}
nonnegative p =
  if p.x >= 0 && p.y >= 0 then p
  else fail! (J.InvalidShape "nonnegative point" "negative coordinate" [])

fun decode_nonnegative_point : J.Json -> Point needs {Fail J.Error}
decode_nonnegative_point = D.refine nonnegative from_json