Using codable with value that is sometimes an Int and other times a String

Question

I have an API that will sometimes return a specific key value (in this case id) in the JSON as an Int and other times it will return that same key value as a String. How do I use codable to parse that JSON?

struct GeneralProduct: Codable {
    var price: Double!
    var id: String?
    var name: String!

    private enum CodingKeys: String, CodingKey {
        case price = "p"
        case id = "i"
        case name = "n"
    }

    init(price: Double? = nil, id: String? = nil, name: String? = nil) {
        self.price = price
        self.id = id
        self.name = name
    }
}

I keep getting this error message: Expected to decode String but found a number instead. The reason that it returns a number is because the id field is empty and when the id field is empty it defaults to returning 0 as an ID which codable identifies as a number. I can basically ignore the ID key but codable does not give me the option to ignore it to my knowledge. What would be the best way to handle this?

Here is the JSON. It is super simple

Working

{
  "p":2.12,
  "i":"3k3mkfnk3",
  "n":"Blue Shirt"
}

Error - because there is no id in the system, it returns 0 as a default which codable obviously sees as a number opposed to string.

{
  "p":2.19,
  "i":0,
  "n":"Black Shirt"
}

Answer 1

struct GeneralProduct: Codable {
    var price: Double?
    var id: String?
    var name: String?
    private enum CodingKeys: String, CodingKey {
        case price = "p", id = "i", name = "n"
    }
    init(price: Double? = nil, id: String? = nil, name: String? = nil) {
        self.price = price
        self.id = id
        self.name = name
    }
    init(from decoder: Decoder) throws {
        let container = try decoder.container(keyedBy: CodingKeys.self)
        price = try container.decode(Double.self, forKey: .price)
        name = try container.decode(String.self, forKey: .name)
        do {
            id = try String(container.decode(Int.self, forKey: .id))
        } catch DecodingError.typeMismatch {
            id = try container.decode(String.self, forKey: .id)
        }
    }
}

---

let json1 = """
{
"p":2.12,
"i":"3k3mkfnk3",
"n":"Blue Shirt"
}
"""

let json2 = """
{
"p":2.12,
"i":0,
"n":"Blue Shirt"
}
"""

---

do {
    let product = try JSONDecoder().decode(GeneralProduct.self, from: Data(json2.utf8))
    print(product.price ?? "nil")
    print(product.id ?? "nil")
    print(product.name ?? "nil")
} catch {
    print(error)
}

---

edit/update:

You can also simply assign nil to your id when your api returns 0:

do {
    let value = try container.decode(Int.self, forKey: .id)
    id = value == 0 ? nil : String(value)
} catch DecodingError.typeMismatch {
    id = try container.decode(String.self, forKey: .id)
}

Answer 2

This is a possible solution with MetadataType, the nice thing is that can be a general solution not for GeneralProduct only, but for all the struct having the same ambiguity:

struct GeneralProduct: Codable {
  var price:Double?
  var id:MetadataType?
  var name:String?

  private enum CodingKeys: String, CodingKey {
    case price = "p"
    case id = "i"
    case name = "n"
  }

  init(price:Double? = nil, id: MetadataType? = nil, name: String? = nil) {
    self.price = price
    self.id = id
    self.name = name
  }
}

enum MetadataType: Codable {
  case int(Int)
  case string(String)

  init(from decoder: Decoder) throws {
    let container = try decoder.singleValueContainer()
    do {
      self = try .int(container.decode(Int.self))
    } catch DecodingError.typeMismatch {
      do {
        self = try .string(container.decode(String.self))
      } catch DecodingError.typeMismatch {
        throw DecodingError.typeMismatch(MetadataType.self, DecodingError.Context(codingPath: decoder.codingPath, debugDescription: "Encoded payload not of an expected type"))
      }
    }
  }

  func encode(to encoder: Encoder) throws {
    var container = encoder.singleValueContainer()
    switch self {
    case .int(let int):
      try container.encode(int)
    case .string(let string):
      try container.encode(string)
    }
  }
}

this is the test:

let decoder = JSONDecoder()
var json =  "{\"p\":2.19,\"i\":0,\"n\":\"Black Shirt\"}"
var product = try! decoder.decode(GeneralProduct.self, from: json.data(using: .utf8)!)
if let id = product.id {
  print(id) // 0
}

json =  "{\"p\":2.19,\"i\":\"hello world\",\"n\":\"Black Shirt\"}"
product = try! decoder.decode(GeneralProduct.self, from: json.data(using: .utf8)!)
if let id = product.id {
  print(id) // hello world
}

Answer 3

Seamlessly decoding from either Int or String into the same property requires writing some code.

However, thanks to a (somewhat) new addition to the language,(property wrappers), you can make it quite easy to reuse this logic wherever you need it:

// note this is only `Decodable`
struct GeneralProduct: Decodable {
    var price: Double
    @Flexible var id: Int // note this is an Int
    var name: String
}

The property wrapper and its supporting code can be implemented like this:

@propertyWrapper struct Flexible<T: FlexibleDecodable>: Decodable {
    var wrappedValue: T
    
    init(from decoder: Decoder) throws {
        wrappedValue = try T(container: decoder.singleValueContainer())
    }
}

protocol FlexibleDecodable {
    init(container: SingleValueDecodingContainer) throws
}

extension Int: FlexibleDecodable {
    init(container: SingleValueDecodingContainer) throws {
        if let int = try? container.decode(Int.self) {
            self = int
        } else if let string = try? container.decode(String.self), let int = Int(string) {
            self = int
        } else {
            throw DecodingError.dataCorrupted(.init(codingPath: container.codingPath, debugDescription: "Invalid int value"))
        }
    }
}

---

Original answer

You can use a wrapper over a string that knows how to decode from any of the basic JSON data types: string, number, boolean:

struct RelaxedString: Codable {
    let value: String
    
    init(_ value: String) {
        self.value = value
    }
    
    init(from decoder: Decoder) throws {
        let container = try decoder.singleValueContainer()
        // attempt to decode from all JSON primitives
        if let str = try? container.decode(String.self) {
            value = str
        } else if let int = try? container.decode(Int.self) {
            value = int.description
        } else if let double = try? container.decode(Double.self) {
            value = double.description
        } else if let bool = try? container.decode(Bool.self) {
            value = bool.description
        } else {
            throw DecodingError.typeMismatch(String.self, .init(codingPath: decoder.codingPath, debugDescription: ""))
        }
    }
    
    func encode(to encoder: Encoder) throws {
        var container = encoder.singleValueContainer()
        try container.encode(value)
    }
}

You can then use this new type in your struct. One minor disadvantage would be that consumer of the struct will need to make another indirection to access the wrapped string. However that can be avoided by declaring the decoded RelaxedString property as private, and use a computed one for the public interface:

struct GeneralProduct: Codable {
    var price: Double!
    var _id: RelaxedString?
    var name: String!
    
    var id: String? {
        get { _id?.value }
        set { _id = newValue.map(RelaxedString.init) }
    }

    private enum CodingKeys: String, CodingKey {
        case price = "p"
        case _id = "i"
        case name = "n"
    }

    init(price: Double? = nil, id: String? = nil, name: String? = nil) {
        self.price = price
        self._id = id.map(RelaxedString.init)
        self.name = name
    }
}

Advantages of the above approach:

1. no need to write custom init(from decoder: Decoder) code, which can become tedious if the number of properties to be decoded increase
2. reusability - RelaxedString can be seamlessly used in other structs
3. the fact that the id can be decoded from a string or an int remains an implementation detail, consumers of GeneralProduct don't know/care that the id can come from a string or an int
4. the public interface exposes string values, which keeps the consumer code simple as it will not have to deal with multiple types of data

Answer 4

I created this Gist which has a ValueWrapper struct that can handle the following types

case stringValue(String)
case intValue(Int)
case doubleValue(Double)
case boolValue(Bool)

https://gist.github.com/amrangry/89097b86514b3477cae79dd28bba3f23

Answer 5

Based on @Cristik 's answer, I come with another solution using @propertyWrapper.

@propertyWrapper
struct StringForcible: Codable {
    
    var wrappedValue: String?
    
    enum CodingKeys: CodingKey {}
    
    init(from decoder: Decoder) throws {
        let container = try decoder.singleValueContainer()
        if let string = try? container.decode(String.self) {
            wrappedValue = string
        } else if let integer = try? container.decode(Int.self) {
            wrappedValue = "\(integer)"
        } else if let double = try? container.decode(Double.self) {
            wrappedValue = "\(double)"
        } else if container.decodeNil() {
            wrappedValue = nil
        }
        else {
            throw DecodingError.typeMismatch(String.self, .init(codingPath: container.codingPath, debugDescription: "Could not decode incoming value to String. It is not a type of String, Int or Double."))
        }
    }
    
    func encode(to encoder: Encoder) throws {
        var container = encoder.singleValueContainer()
        try container.encode(wrappedValue)
    }
    
    init() {
        self.wrappedValue = nil
    }
    
}

And usage is

struct SomeDTO: Codable {
   @StringForcible var id: String? 
}

Also works like -I think-

struct AnotherDTO: Codable {
    var some: SomeDTO?
}