When using Swift Codable you can specify a keyDecodingStrategy that will convert snake_case to camelCase. This works great for keys in a dictionary, but is there any solution for decoding values in a dictionary in a similarly clean way?
I have an enumeration that is used as a key in one place and as a value in another:
enum Foo: String, Codable, CodingKey, CaseIterable {
case bar
case bazQux // baz_qux in JSON, say
}
Then this is used as a key like so:
struct MyStruct: Codable {
enum Keys: String, CodingKey {
case myKey
}
let myProperty: [Bool]
public init(from decoder: Decoder) throws {
let container = try decoder.container(keyedBy: Foo.self)
myProperty = Foo.allCases.map {
try container.decode(Bool.self, forKey: $0)
}
}
}
But it is also used as a value like so:
struct Buz: Decodable {
enum CodingKeys: String, CodingKey {
case foo
}
// Note this is called using a decoder with keyDecodingStrategy = .convertFromSnakeCase
init(from decoder: Decoder) throws {
let container = try decoder.container(keyedBy: CodingKeys.self)
let foo = try container.decode(Foo.self, forKey: .foo)
}
}
The JSON for example might include a line like this:
"foo": "baz_qux"
The value decoding only works if I set the raw value of the bazQux case to be baz_qux. But then that breaks the key decoding. It would also be annoying to have to make two separate enums for the same thing just to avoid this problem.
So I can then initialize the right model corresponding to the value.
I'd also like to avoid employing a "hacky" solution… Is there something reasonably elegant that works well with Codable?
Wound up adding a custom decoding initializer on Foo:
public init(from decoder: Decoder) throws {
let rawValue = try decoder.singleValueContainer().decode(String.self)
if let foo = Self.init(rawValue: rawValue.snakeToCamelCase) {
self = foo
} else {
throw CodingError.unknownValue
}
}
snakeToCamelCase is a simple extension method on String I added locally. Plus need to add the enum CodingError: Error to Foo if you want some error handling.
This isn't ideal but at least it's not too complicated. I would rather rely on built-in case conversion methods but I don't see a way to do that here.
Related
I'm having an inconsistent API that might return either a String or an Number as a part of the JSON response.
The dates also could be represented the same way as either a String or a Number, but are always an UNIX timestamp (i.e. timeIntervalSince1970).
To fix the issue with the dates, I simply used a custom JSONDecoder.DateDecodingStrategy:
decoder.dateDecodingStrategy = JSONDecoder.DateDecodingStrategy.custom({ decoder in
let container = try decoder.singleValueContainer()
if let doubleValue = try? container.decode(Double.self) {
return Date(timeIntervalSince1970: doubleValue)
} else if let stringValue = try? container.decode(String.self),
let doubleValue = Double(stringValue) {
return Date(timeIntervalSince1970: doubleValue)
}
throw DecodingError.dataCorruptedError(in: container,
debugDescription: "Unable to decode value of type `Date`")
})
However, no such customization is available for the Int or Double types which I'd like to apply it for.
So, I have to resort to writing Codable initializers for each of the model types that I'm using.
The alternative approach I'm looking for is to subclass the JSONDecoder and override the decode<T>(_ type: T.Type, from data: Data) throws -> T where T : Decodable method.
In that method I'd like to "inspect" the type T that I'm trying to decode to and then, if the base implementation (super) fails, try to decode the value first to String and then to the T (the target type).
So far, my initial prototype looks like this:
final class CustomDecoder: JSONDecoder {
override func decode<T>(_ type: T.Type, from data: Data) throws -> T where T : Decodable {
do {
return try super.decode(type, from: data)
} catch {
if type is Int.Type {
print("Trying to decode as a String")
if let decoded = try? super.decode(String.self, from: data),
let converted = Int(decoded) {
return converted as! T
}
}
throw error
}
}
}
However, I found out that the "Trying to decode as a String" message is never printed for some reason, even though the control reaches the catch stage.
I'm happy to have that custom path only for Int and Double types, since the T is Codable and that doesn't guarantee ability to initialize a value with the String, however, I of course welcome a more generalized approach.
Here's the sample Playground code that I came up with to test my prototype. It can be copy-pasted directly into the Playground and works just fine.
My goal is to have both jsonsample1 and jsonsample2 to produce the same result.
import UIKit
final class CustomDecoder: JSONDecoder {
override func decode<T>(_ type: T.Type, from data: Data) throws -> T where T : Decodable {
do {
return try super.decode(type, from: data)
} catch {
if type is Int.Type {
print("Trying to decode as a String")
if let decoded = try? super.decode(String.self, from: data),
let converted = Int(decoded) {
return converted as! T
}
}
throw error
}
}
}
let jsonSample1 =
"""
{
"name": "Paul",
"age": "38"
}
"""
let jsonSample2 =
"""
{
"name": "Paul",
"age": 38
}
"""
let data1 = jsonSample1.data(using: .utf8)!
let data2 = jsonSample2.data(using: .utf8)!
struct Person: Codable {
let name: String?
let age: Int?
}
let decoder = CustomDecoder()
let person1 = try? decoder.decode(Person.self, from: data1)
let person2 = try? decoder.decode(Person.self, from: data2)
print(person1 as Any)
print(person2 as Any)
What could be the reason for my CustomDecoder not working?
The primary reason that your decoder doesn't do what you expect is that you're not overriding the method that you want to be: JSONDecoder.decode<T>(_:from:) is the top-level method that is called when you call
try JSONDecoder().decode(Person.self, from: data)
but this is not the method that is called internally during decoding. Given the JSON you show as an example, if we write a Person struct as
struct Person: Decodable {
let name: String
let age: Int
}
then the compiler will write an init(from:) method which looks like this:
init(from decoder: Decoder) throws {
let container = try decoder.container(keyedBy: CodingKeys.self)
name = try container.decode(String.self, forKey: .name)
age = try container.decode(Int.self, forKey: .age)
}
Note that when we decode age, we are not calling a method on the decoder directly, but on a KeyedCodingContainer that we get from the decoder — specifically, the Int.Type overload of KeyedDecodingContainer.decode(_:forKey:).
In order to hook into the methods that are called during decode at the middle levels of a Decoder, you'd need to hook into its actual container methods, which is very difficult — all of JSONDecoder's containers and internals are private. In order to do this by subclassing JSONDecoder, you'd end up needing to pretty much reimplement the whole thing from scratch, which is significantly more complicated than what you're trying to do.
As suggested in a comment, you're likely better off either:
Writing Person.init(from:) manually by trying to decode both Int.self and String.self for the .age property and keeping whichever one succeeds, or
If you need to reuse this solution across many types, you can write a wrapper type to use as a property:
struct StringOrNumber: Decodable {
let number: Double
init(from decoder: Decoder) throws {
let container = try decoder.singleValueContainer()
do {
number = try container.decode(Double.self)
} catch (DecodingError.typeMismatch) {
let string = try container.decode(String.self)
if let n = Double(string) {
number = n
} else {
throw DecodingError.dataCorruptedError(in: container, debugDescription: "Value wasn't a number or a string...")
}
}
}
}
struct Person: Decodable {
let name: String
let age: StringOrNumber
}
You can also write StringOrNumber as an enum which can hold either case string(String) or case number(Double) if knowing which type of value was in the payload was important:
enum StringOrNumber: Decodable {
case number(Double)
case string(String)
init(from decoder: Decoder) throws {
let container = try decoder.singleValueContainer()
do {
self = try .number(container.decode(Double.self))
} catch (DecodingError.typeMismatch) {
let string = try container.decode(String.self)
if let n = Double(string) {
self = .string(string)
} else {
throw DecodingError.dataCorruptedError(in: container, debugDescription: "Value wasn't a number or a string...")
}
}
}
}
Though this isn't as relevant if you always need Double/Int access to the data, since you'd need to re-convert at the use site every time (and you call this out in a comment)
I have an api response in the following shape -
{
"textEntries":{
"summary":{
"id":"101e9136-efd9-469e-9848-132023d51fb1",
"text":"some text",
"locale":"en_GB"
},
"body":{
"id":"3692b0ec-5b92-4ab1-bc25-7711499901c5",
"text":"some other text",
"locale":"en_GB"
},
"title":{
"id":"45595d27-7e06-491e-890b-f50a5af1cdfe",
"text":"some more text again",
"locale":"en_GB"
}
}
}
I'd like to decode this via JSONDecoder so I can use the properties. The challenge I have is the keys, in this case summary,body and title are generated elsewhere and not always these values, they are always unique, but are based on logic that takes place elsewhere in the product, so another call for a different content article could return leftBody or subTitle etc.
The model for the body of these props is always the same however, I can expect the same fields to exist on any combination of responses.
I will need to be able to access the body of each key in code elsewhere. Another API response will tell me the key I need though.
I am not sure how I can handle this with Decodable as I cannot type the values ahead of time.
I had considered something like modelling the body -
struct ContentArticleTextEntries: Decodable {
var id: String
var text: String
var locale: Locale
}
and storing the values in a struct like -
struct ContentArticle: Decodable {
var textEntries: [String: ContentArticleTextEntries]
private enum CodingKeys: String, CodingKey {
case textEntries
}
init(from decoder: Decoder) throws {
let values = try decoder.container(keyedBy: CodingKeys.self)
self.textEntries = try values.decode(ContentArticleTextEntries.self, forKey: .textEntries)
}
}
I could them maybe use a subscript elsewhere to access property however I do not know how to decode into this shape as the above would not work.
So I would later access like textEntries["body"] for example.
I also do no know if there is a better way to handle this.
I had considered converting the keys to a 'type' using an enum, but again not knowing the enum cases ahead of time makes this impossible.
I know textEntries this does not change and I know id, text and locale this does not change. It is the keys in between this layer I do not know. I have tried the helpful solution posted by #vadian but cannot seem to make this work in the context of only needing 1 set of keys decoded.
For the proposed solution in this answer the structs are
struct ContentArticleTextEntries: Decodable {
let title : String
let id: String
let text: String
let locale: Locale
enum CodingKeys: String, CodingKey {
case id, text, locale
}
init(from decoder: Decoder) throws {
self.title = try decoder.currentTitle()
let container = try decoder.container(keyedBy: CodingKeys.self)
self.id = try container.decode(String.self, forKey: .id)
self.text = try container.decode(String.self, forKey: .text)
let localeIdentifier = try container.decode(String.self, forKey: .locale)
self.locale = Locale(identifier: localeIdentifier)
}
}
struct ContentArticle: TitleDecodable {
let title : String
var elements: [ContentArticleTextEntries]
}
struct Container: Decodable {
let containers: [ContentArticle]
init(from decoder: Decoder) throws {
self.containers = try decoder.decodeTitledElements(ContentArticle.self)
}
}
Then decode Container.self
If your models are like,
struct ContentArticle: Decodable {
let textEntries: [String: ContentArticleTextEntries]
}
struct ContentArticleTextEntries: Decodable {
var id: String
var text: String
var locale: String
}
Then, you can simply access the data based on key like,
let response = try JSONDecoder().decode(ContentArticle.self, from: data)
let key = "summary"
print(response.textEntries[key])
Note: No need to write enum CodingKeys and init(from:) if there is no special handling while parsing the JSON.
Use "decodeIfPresent" variant method instead of decode method also you need to breakdown the ContentArticleTextEntries dictionary into individual keys:
struct ContentArticle: Decodable {
var id: String
var text: String?
var locale: String?
private enum CodingKeys: String, CodingKey {
case id, text, locale
}
init(from decoder: Decoder) throws {
let container = try decoder.container(keyedBy: CodingKeys.self)
self.id = try container.decodeIfPresent(String.self, forKey: .id) ?? ""
self.text = try container.decodeIfPresent(String.self, forKey: .text)
self.locale = try container.decodeIfPresent(String.self, forKey: .locale)
}
}
The short question is how can I make a type conditionally conform to a protocol with two conditions when either one of them is met?
I have a generic type NetworkResponse<Data>. It represents server response. Here is how it is defined:
enum NetworkResponse<Data> {
case success(Data)
case error(ServerError)
}
I want to make NetworkResponse to conform Decodable. Here is my server response format:
{
"data": {
"someKey": "someValue",
"anotherKey": 15
},
"meta": {
"returnCode": 0,
"returnMessage": "operation is successful"
}
}
The data part depends on what request is made. The meta part represents some meta data about response. Like whether it is success or if not what is the error.
So here is how I implemented Decodable:
extension NetworkResponse: Decodable where Data: Decodable {
enum CodingKeys: CodingKey {
case meta
case data
}
init(from decoder: Decoder) throws {
let container = try decoder.container(keyedBy: CodingKeys.self)
let metaValue = try container.decode(ServerError.self, forKey: .meta)
if metaValue.code != 0 {
self = .error(metaValue)
} else {
self = .success(try container.decode(Data.self, forKey: .data))
}
}
}
So far so good. But here is my problem. For some apis which don't need to return any data the data part in response is omitted. In this case my response would look like this:
{
"meta": {
"returnCode": 0,
"returnMessage": "operation is successful"
}
}
In this case I want to decode the response json as NetworkResponse<Void>. But since Void can not conform to Decodable (since it is non nominal type) compiler gives error.
To overcome this I tried to create more specialized extension of Decodable where Data is Void like this:
extension NetworkResponse: Decodable where Data == Void {
init(from decoder: Decoder) throws {
let container = try decoder.container(keyedBy: CodingKeys.self)
let metaValue = try container.decode(AppErrors.Server.self, forKey: CodingKeys.meta)
if metaValue.code != 0 {
self = .error(metaValue)
} else {
self = .success(())
}
}
}
But still compiler complines like: Conflicting conformance of 'NetworkResponse<Data>' to protocol 'Decodable'; there cannot be more than one conformance, even with different conditional bounds.
So how can I create seperate init(from:) function that is used for when Data is Void?
I strongly urge you to change the generic parameter's type name away from Data since it's too easily confused with Foundation.Data which is widely used throughout Swift.
On to the problem itself, you can make an empty struct to represent "void" and add a new case to your NetworkResponse:
struct EmptyData: Decodable {}
enum NetworkResponse<T> {
case success(T)
case successWithEmptyData
case error(ServerError)
}
extension NetworkResponse: Decodable where T: Decodable {
private enum CodingKeys: CodingKey {
case meta
case data
}
init(from decoder: Decoder) throws {
let container = try decoder.container(keyedBy: CodingKeys.self)
let metaValue = try container.decode(ServerError.self, forKey: .meta)
if metaValue.code != 0 {
self = .error(metaValue)
} else if T.self == EmptyData.self {
self = .successWithEmptyData
} else {
self = .success(try container.decode(T.self, forKey: .data))
}
}
}
let response = try! JSONDecoder().decode(NetworkResponse<EmptyData>.self, from: jsonData)
Alternatively, you can make your .success case contains a T? when you expect no data in return and decode accordingly.
I'm trying to understand how could I parse this multiple container JSON to an object. I've tried this approach (Mark answer), but he explain how to solve it using one-level container. For some reason I can't mimic the behaviour for multiple containers.
{
"graphql": {
"shortcode_media": {
"id": "1657677004214306744",
"shortcode": "BcBQHPchwe4"
}
}
}
class Post: Decodable {
enum CodingKeys: String, CodingKey {
case graphql // The top level "user" key
case shortcode_media
}
enum PostKeys: String, CodingKey {
case id
}
required init(from decoder: Decoder) throws {
let values = try decoder.container(keyedBy: CodingKeys.self)
let post = try values.nestedContainer(keyedBy: PostKeys.self, forKey: .shortcode_media)
self.id = try post.decode(String.self, forKey: .id)
}
var id: String
}
I'm getting:
Swift.DecodingError.Context(codingPath: [], debugDescription: "Cannot get KeyedDecodingContainer<PostKeys> -- no value found for key \"shortcode_media\"", underlyingError: nil))
Any help will be much appreciated, thank you!
As vadian notes, you haven't matched the JSON structure. There is no shortcode_media key at the top level like you've encoded in CodingKeys.
In order to decode this with a custom decoder, you will need to walk through each level and deal with it.
class Post: Decodable {
enum CodingKeys: String, CodingKey {
case graphql
}
enum GraphQLKeys: String, CodingKey {
case shortcode_media
}
enum PostKeys: String, CodingKey {
case id
}
required init(from decoder: Decoder) throws {
// unload the top level
let container = try decoder.container(keyedBy: CodingKeys.self)
// Unload the graphql key
let graphql = try container.nestedContainer(keyedBy: GraphQLKeys.self, forKey: .graphql)
// unload the shortcode_media key
let post = try graphql.nestedContainer(keyedBy: PostKeys.self, forKey: .shortcode_media)
// Finally, unload the actual object
self.id = try post.decode(String.self, forKey: .id)
}
var id: String
}
Please read the JSON.
Any opening { is quasi a separator. The indentation of the JSON indicates also the hierarchy.
For clarity I removed all coding keys and left the variable names – which should be camelCased – unchanged.
struct Root : Decodable {
let graphql : Graph
// to access the `Media` object declare a lazy instantiated property
lazy var media : Media = {
return graphql.shortcode_media
}()
}
struct Graph : Decodable {
let shortcode_media : Media
}
struct Media : Decodable {
let id: String
let shortcode : String
}
let jsonString = """
{
"graphql": {
"shortcode_media": {
"id": "1657677004214306744",
"shortcode": "BcBQHPchwe4"
}
}
}
"""
do {
let data = Data(jsonString.utf8)
var result = try decoder.decode(Root.self, from: data)
print(result.media)
} catch {
print("error: ", error)
}
Writing a custom initializer with nestedContainer is more effort than creating the actual hierarchy.
Please paste the entire code in a Playground and check it out.
I have a json in such format:
{
"route":{
"1":"Atrakcyjno\u015b\u0107 przyrodnicza",
"2":"Atrakcyjno\u015b\u0107 kulturowa",
"3":"Dla rodzin z dzie\u0107mi",
"5":"Dla senior\u00f3w",
"6":"Dla or\u0142\u00f3w",
"8":"Niepe\u0142nosprawni"
},
"apartments":{
"1":"WifI",
"4":"Gastronomia",
"5":"Parking",
"6":"Dla niepe\u0142nosprawnych",
"7":"Dla rodzin z dzie\u0107mi",
"8":"Dla senior\u00f3w"
},
"levels":{
"1":"\u0141atwy",
"2":"\u015aredni",
"3":"Trudny",
"4":"Bardzo trudny"
}
}
I would like to decode it as simple as possible, but I don't know how to decode these sub dictionaries. These are dicts, but it should be array instead. Can I somehow write something, that will make it decode in special way, so that I'll get arrays? So far I have something like this:
struct PreferencesList: Decodable {
private enum CodingKeys: String, CodingKey {
case routes = "route"
case apartments
case levels
}
let routes: [Preference]
let apartments: [Preference]
let levels: [Preference]
}
struct Preference: Decodable {
let id: Int
let name: String
}
I guess you need to do this step by step.
Keep your struct like it is. Preference doesn't need to be Decodable. Then, override init(from decoder: Decoder) throws function like this.
init(from decoder: Decoder) throws {
let container = try decoder.container(keyedBy: CodingKeys.self)
let routes = try container.decode([String: String].self, forKey: .routes)
self.routes = []
for (key, value) in routes {
self.routes.append(Preference(id: key, name: value))
}
// Do the same for other var...
}
I hope it helps.