Mapper is a tiny yet very powerful library which allows you to create custom strongly-typed instances from any kind of structured data (JSON and other data interchange formats, for example) with only a single initializer. And vice versa - with only a single method.
Mapper extensively uses power of Swift generics and protocols, dramatically reducing the boilerplate you have to write. With Mapper, mapping is a breeze.
The maing advantage of Mapper is that you don't need to write multiple initializers to support mapping from different formats (if you've done it before - you know what I mean), thus eliminating the boilerplate and leaving only the core logic you need. With Mapper your code is safe and expressive.
And while reducing boilerplate, Mapper is also amazingly fast. It doesn't use reflection, and generics allows the compiler to optimize code in the most effective way.
Mapper itself is just a core mapping logic without any implementations. To actually use Mapper, you also have to import one of Mapper-conforming libraries. You can find a current list of them here. If you want to support Mapper for your data types, checkout Adopting Mapper short guide.
Mapper is deeply inspired by Lyft's Mapper. You can learn more about the concept behind their idea in this talk.
struct City : InMappable, OutMappable {
let name: String
let population: Int
enum MappingKeys : String, IndexPathElement {
case name, population
}
init<Source : InMap>(mapper: InMapper<Source, MappingKeys>) throws {
self.name = try mapper.map(from: .name)
self.population = try mapper.map(from: .population)
}
func outMap<Destination : OutMap>(mapper: inout OutMapper<Destination, MappingKeys>) throws {
try mapper.map(self.name, to: .name)
try mapper.map(self.population, to: .population)
}
}
enum Gender : String {
case male
case female
}
// Mappable = InMappable & OutMappable
struct Person : Mappable {
let name: String
let gender: Gender
let city: City
let identifier: Int
let isRegistered: Bool
let biographyPoints: [String]
enum MappingKeys : String, IndexPathElement {
case name, gender, city, identifier, registered, biographyPoints
}
init<Source : InMap>(mapper: InMapper<Source, MappingKeys>) throws {
self.name = try mapper.map(from: .name)
self.gender = try mapper.map(from: .gender)
self.city = try mapper.map(from: .city)
self.identifier = try mapper.map(from: .identifier)
self.isRegistered = try mapper.map(from: .registered)
self.biographyPoints = try mapper.map(from: .biographyPoints)
}
func outMap<Destination : OutMap>(mapper: inout OutMapper<Destination, Person.MappingKeys>) throws {
try mapper.map(self.name, to: .name)
try mapper.map(self.gender, to: .gender)
try mapper.map(self.city, to: .city)
try mapper.map(self.identifier, to: .identifier)
try mapper.map(self.isRegistered, to: .registered)
try mapper.map(self.biographyPoints, to: .biographyPoints)
}
}
// in-mapping
let jessy = Person(from: json)
let messi = Person(from: messagePack)
let michael = Person(from: mongoBSON)
// out-mapping
let json: JSON = try jessy.map()
let messi: MessagePack = try messi.map()
// and so on...- Add
Mapperto yourPackage.swift
import PackageDescription
let package = Package(
dependencies: [
.Package(url: "https://github.com/Zewo/Mapper.git", majorVersion: 0, minor: 14),
]
)Mapper allows you to map data in both ways, and so it has two major parts: in mapping (for example, JSON -> your model) and out mapping (your model -> JSON). So the two main protocols of Mapper is InMappable and OutMappable.
To use Mapper in it's full glory, first you need to define nested MappingKeys enum. MappingKeys are needed to represent keys from/to which your properties will be mapped. Using nested MappingKeys is a win for type-safety and can save you from some painful typos:
struct City {
let name: String
let population: Int
enum MappingKeys : String, IndexPathElement {
case name, population
}
}Make sure to declare MappingKeys as IndexPathElement!
Now we're going to write mapping code. Let's start with in mapping:
extension City : InMappable {
init<Source : InMap>(mapper: InMapper<Source, MappingKeys>) throws {
self.name = try mapper.map(from: .name)
self.population = try mapper.map(from: .population)
}
}
let city = try City(from: json)Actually, that's it! Now your City can be created from JSON, BSON, MessagePack and a whole range of other data formats. And that's all thanks to the amazing power of generics. As you see, that's why your initializer is generic. And from: .name is actually where your MappingKeys are used.
Each call to mapper is marked with try because, obviously, it can fail. In this case initializer will throw with InMapperError. If one of your properties is optional, you can just write try?.
Let's continue with out mapping:
extension City : OutMappable {
func outMap<Destination : OutMap>(mapper: inout OutMapper<Destination, MappingKeys>) throws {
try mapper.map(self.name, to: .name)
try mapper.map(self.population, to: .population)
}
}
let json: JSON = city.map()As you see, the code is pretty similar, easy to reason about, and very expressive.
As you see, both mappers have two generic arguments: Source/Destination, which is the structured data format, and MappingKeys, which is specific MappingKeys defined for your model.
Actually, if you don't want to write that MappingKeys, we made BasicInMappable/BasicOutMappable just for you.
struct Planet : BasicInMappable, BasicOutMappable {
let radius: Int
init<Source : InMap>(mapper: BasicInMapper<Source>) throws {
self.radius = try mapper.map(from: "radius")
}
func outMap<Destination : OutMap>(mapper: inout BasicOutMapper<Destination>) throws {
try mapper.map(radius, to: "radius")
}
}You can map array just like anything else -- simply by using .map.
struct Album : Mappable {
let songs: [String]
enum MappingKeys : String, IndexPathElement {
case songs
}
init<Source : InMap>(mapper: InMapper<Source, MappingKeys>) throws {
self.songs = try mapper.map(from: .songs)
}
func outMap<Destination : OutMap>(mapper: inout OutMapper<Destination, Album.MappingKeys>) throws {
try mapper.map(self.songs, to: .songs)
}
}Mapper can also automatically map enums with raw values, which is neat.
enum Wood : String {
case mahogany
case koa
case cedar
case spruce
}
enum Strings : Int {
case four = 4
case six = 6
case seven = 7
}
struct Guitar : Mappable {
let wood: Wood
let strings: Strings
enum MappingKeys : String, IndexPathElement {
case wood, strings
}
init<Source : InMap>(mapper: InMapper<Source, MappingKeys>) throws {
self.wood = try mapper.map(from: .wood)
self.strings = try mapper.map(from: .strings)
}
func outMap<Destination : OutMap>(mapper: inout OutMapper<Destination, Guitar.MappingKeys>) throws {
try mapper.map(self.wood, to: .wood)
try mapper.map(self.strings, to: .strings)
}
}Cool thing about Mapper is that you can easily map instances which are itself Mappable:
struct Sport : Mappable {
let name: String
enum MappingKeys : String, IndexPathElement {
case name
}
init<Source : InMap>(mapper: InMapper<Source, MappingKeys>) throws {
self.name = try mapper.map(from: .name)
}
func outMap<Destination : OutMap>(mapper: inout OutMapper<Destination, Sport.MappingKeys>) throws {
try mapper.map(self.name, to: .name)
}
}
struct Team : Mappable {
let sport: Sport
let name: String
let foundationYear: Int
enum MappingKeys : String, IndexPathElement {
case sport
case name
case foundationYear = "foundation-year"
}
init<Source : InMap>(mapper: InMapper<Source, MappingKeys>) throws {
self.sport = try mapper.map(from: .sport)
self.name = try mapper.map(from: .name)
self.foundationYear = try mapper.map(from: .foundationYear)
}
func outMap<Destination : OutMap>(mapper: inout OutMapper<Destination, Team.MappingKeys>) throws {
try mapper.map(self.sport, to: .sport)
try mapper.map(self.name, to: .name)
try mapper.map(self.foundationYear, to: .foundationYear)
}
}(Advanced topic)
There are some situations when you need to describe more than one way of mappings. It can be for several reasons - different sources/destinations of data, architectural restrictions and so on. For this situation we have "contextual mapping".
Let's start with something called InMappableWithContext:
enum SuperContext {
case json
case mongo
case gordon
}
struct SuperheroHelper {
let name: String
let id: Int
enum MappingKeys : String, IndexPathElement {
case name
case id, identifier, g_id
}
typealias MappingContext = SuperContext
}
extension SuperheroHelper : InMappableWithContext {
init<Source : InMap>(mapper: ContextualInMapper<Source, MappingKeys, MappingContext>) throws {
self.name = try mapper.map(from: .name)
switch mapper.context {
case .json:
self.id = try mapper.map(from: .id)
case .mongo:
self.id = try mapper.map(from: .identifier)
case .gordon:
self.id = try mapper.map(from: .g_id)
}
}
}
extension SuperheroHelper : OutMappableWithContext {
func outMap<Destination : OutMap>(mapper: inout ContextualOutMapper<Destination, SuperheroHelper.MappingKeys, MappingContext>) throws {
try mapper.map(self.name, to: .name)
switch mapper.context {
case .json:
try mapper.map(self.id, to: .id)
case .mongo:
try mapper.map(self.id, to: .identifier)
case .gordon:
try mapper.map(self.id, to: .g_id)
}
}
}
// now we can do
let robin = try BatmanHelper(from: robinJSON, withContext: .json)
// or
let robin = try BatmanHelper(from: robinMongo, withContext: .mongo)
// or whatever that is
let robin = try BatmanHelper(from: robinGordon, withContext: .gordon)
// And also
let robinJSON: JSON = try robin.map(withContext: .json)Now let's get to something even more cool:
struct Superhero {
let name: String
let helper: SuperheroHelper
enum MappingKeys : String, IndexPathElement {
case name, helper
}
typealias MappingContext = SuperContext
}
extension Superhero : InMappableWithContext {
init<Source : InMap>(mapper: ContextualInMapper<Source, MappingKeys, MappingContext>) throws {
self.name = try mapper.map(from: .name)
self.helper = try mapper.map(from: .helper)
}
}
extension Superhero : OutMappableWithContext {
func outMap<Destination : OutMap>(mapper: inout ContextualOutMapper<Destination, MappingKeys, MappingContext>) throws {
try mapper.map(self.name, to: .name)
try mapper.map(self.helper, to: .helper)
}
}
let batman = try Superhero(from: batJSON, withContext: .json)Noticed something strange? Yes, self.helper is mapped even though we didn't specify the context! That's because Superhero has, actually, the same context as SuperheroHelper, and is also InMappableWithContext/OutMappableWithContext, and so the right context is passed to SuperheroHelper automatically. This is context infering.
If you don't want that, you can specify the context manually:
// in
self.helper = try mapper.map(from: .helper, withContext: .json)
// out
try mapper.map(self.helper, to: .helper, withContext: .json)Tutorial by example: making Foundation's Date conform to Mappable.
extension Date : Mappable {
public init<Source : InMap>(mapper: PlainInMapper<Source>) throws {
let interval: TimeInterval = try mapper.map()
self.init(timeIntervalSince1970: interval)
}
public func outMap<Destination : OutMap>(mapper: inout PlainOutMapper<Destination>) throws {
try mapper.map(self.timeIntervalSince1970)
}
}Mappers take variadic parameter as index path, so it's possible to pass no index path at all. We call it "plain mapping".
And using MappableWithContext, we can even do something like that:
public enum DateMappingContext {
case timeIntervalSince1970
case timeIntervalSinceReferenceDate
}
extension Date : InMappableWithContext {
public init<Source : InMap>(mapper: PlainContextualInMapper<Source, DateMappingContext>) throws {
let interval: TimeInterval = try mapper.map()
switch mapper.context {
case .timeIntervalSince1970:
self.init(timeIntervalSince1970: interval)
case .timeIntervalSinceReferenceDate:
self.init(timeIntervalSinceReferenceDate: interval)
}
}
}
extension Date : OutMappableWithContext {
public func outMap<Destination : OutMap>(mapper: inout PlainContextualOutMapper<Destination, DateMappingContext>) throws {
switch mapper.context {
case .timeIntervalSince1970:
try mapper.map(self.timeIntervalSince1970)
case .timeIntervalSinceReferenceDate:
try mapper.map(self.timeIntervalSinceReferenceDate)
}
}
}Mapper can work only with four basic "primitive" types: Int, Double, Bool, String (these four are expected to work with any Mapper-conforming type). But, of course, you can map other, more specific primitive types that your format supports. In order to do that, you should use .unsafe_map and .unsafe_mapArray methods:
struct TeamStat : Mappable {
let rate: Int32
let goals: [Int32]
enum MappingKeys : String, IndexPathElement {
case rate, goals
}
init<Source : InMap>(mapper: InMapper<Source, MappingKeys>) throws {
self.rate = try mapper.unsafe_map(from: .rate)
self.goals = try mapper.unsafe_mapArray(from: .goals)
}
func outMap<Destination : OutMap>(mapper: inout OutMapper<Destination, TeamStat.MappingKeys>) throws {
try mapper.unsafe_map(self.rate, to: .rate)
try mapper.unsafe_mapArray(self.goals, to: .goals)
}
}
// `BSON` supports `Int32` directly, so
let stat = try TeamStat(from: mongoDocument)If you have some classes that you don't have direct access to (for example, Cocoa classes), and you want to make them Mappable for some reason, you should use BasicInMappable /BasicOutMappable with this approach:
extension BasicInMappable where Self : NSDate {
public init<Source : InMap>(mapper: BasicInMapper<Source>) throws {
let interval: TimeInterval = try mapper.map()
self.init(timeIntervalSince1970: interval)
}
}
extension NSDate : BasicInMappable { }
extension BasicOutMappable where Self : NSDate {
public func outMap<Destination : OutMap>(mapper: inout BasicOutMapper<Destination>) throws {
try mapper.map(self.timeIntervalSince1970)
}
}
extension NSDate : BasicOutMappable { }Now NSDate can be mapped as usual.
Mapper boosts your productivity dramatically, but we didn't stop there. We have prepared for you Xcode code snippets located in CodeSnippets/ directory that will speed you up even more. Just put them in ~/Library/Developer/Xcode/UserData/CodeSnippets/ and restart Xcode. Then you can just type mappable and Xcode will give you this:
Or type context mappable and you see this:
- Zewo/Map
- Zewo/JSON
Swift is in process of discovering its own native patterns. One thing for sure -- enum-based structured data types are among them. The problem is that they are completely disconnected from each other, although they are similar in many ways. Mapper's mission is to build bridges between these types.
There are two main protocols with which you should work: InMap (for in mapping) and OutMap (for out mapping).
Let's work from example. Imagine that we have some pretty typical enum-based type:
public enum MapperMap {
case int(Int)
case double(Double)
case string(String)
case bool(Bool)
case array([MapperMap])
case dictionary([String: MapperMap])
}Let's look at InMap protocol:
public protocol InMap {
func get(at indexPath: IndexPathValue) -> Self?
func get(at indexPath: [IndexPathValue]) -> Self?
func asArray() -> [Self]?
func get<T>() -> T?
var int: Int? { get }
var double: Double? { get }
var bool: Bool? { get }
var string: String? { get }
}Where IndexPathValue is:
public enum IndexPathValue {
case index(Int)
case key(String)
}Also it's worth mentioning, that get(at indexPath: [IndexPathElement]) method has default implementation, so you don't have to write it yourself.
So, here is our MapperMap : InMap implementation:
extension MapperMap : InMap {
public func get(at indexPath: IndexPathValue) -> MapperMap? {
switch (indexPath, self) {
case (.key(let key), .dictionary(let dict)):
return dict[key]
case (.index(let index), .array(let array)):
if array.indices.contains(index) {
return array[index]
}
return nil
default:
return nil
}
}
public func get<T>() -> T? {
switch self {
case .int(let int as T):
return int
case .double(let double as T):
return double
case .string(let string as T):
return string
case .bool(let bool as T):
return bool
case .array(let array as T):
return array
case .dictionary(let dict as T):
return dict
default:
return nil
}
}
public var int: Int? {
if case .int(let value) = self {
return value
}
return nil
}
public var double: Double? {
if case .double(let value) = self {
return value
}
return nil
}
public var bool: Bool? {
if case .bool(let value) = self {
return value
}
return nil
}
public var string: String? {
if case .string(let value) = self {
return value
}
return nil
}
public func asArray() -> [MapperMap]? {
if case .array(let array) = self {
return array
}
return nil
}
}So here what's going on: in get<T>() you just try to return in any of the type you directly support, that's it.
That seems nice. But not always we have situation that ordinary. Let's pretend that now we have new format:
public enum MapperNeomap {
case bool(Bool)
case int32(Int32)
case uint(UInt)
case uint8(UInt8)
case string(String)
case float(Float)
case array([MapperNeomap])
case dictionary([String: MapperNeomap])
}Here, it's important to mention one thing.
All Mapper-compatible libraries are expected to work with next "primitive" types:
IntDoubleStringBool
As you see, our MapperNeomap doesn't support Int and Double. So we should do this:
extension MapperNeomap : InMap {
public func get<T>() -> T? {
switch self {
case .bool(let value as T): return value
case .int32(let value as T): return value
case .uint(let value as T): return value
case .uint8(let value as T): return value
case .string(let value as T): return value
case .float(let value as T): return value
case .array(let value as T): return value
case .dictionary(let value as T): return value
default:
return nil
}
}
public var int: Int? {
switch self {
case .int32(let value): return Int(value)
case .uint(let value): return Int(value)
case .uint8(let value): return Int(value)
default:
return nil
}
}
public var double: Double? {
if case .float(let value) = self {
return Double(value)
}
return nil
}
public var bool: Bool? {
if case .bool(let value) = self {
return value
}
return nil
}
public var string: String? {
if case .string(let value) = self {
return value
}
return nil
}
}Pretty verbose, yes, but necessary. Again, Int, Double, String and Bool are expected to work, even if they are not directly available in your data format.
Now let's look at OutMap:
public protocol OutMap {
static var blank: Self { get }
mutating func set(_ map: Self, at indexPath: IndexPathValue) throws
mutating func set(_ map: Self, at indexPath: [IndexPathValue]) throws
static func fromArray(_ array: [Self]) -> Self?
static func from<T>(_ value: T) -> Self?
static func from(_ int: Int) -> Self?
static func from(_ double: Double) -> Self?
static func from(_ bool: Bool) -> Self?
static func from(_ string: String) -> Self?
}Pretty much the same, but reversed. Again, set(_ map: Self, at indexPath: [IndexPathElement]) has default implementation.
So, let's look at MapperMap : OutMap:
extension MapperMap : OutMap {
public static var blank: MapperMap {
return .dictionary([:])
}
public mutating func set(_ map: MapperMap, at indexPath: IndexPathValue) throws {
switch (indexPath, self) {
case (.key(let key), .dictionary(var dict)):
dict[key] = map
self = .dictionary(dict)
case (.index(let index), .array(var array)):
array[index] = map
self = .array(array)
default:
throw MapperMapOutMapError.incompatibleType
}
}
public static func fromArray(_ array: [MapperMap]) -> MapperMap? {
return .array(array)
}
public static func from<T>(_ value: T) -> MapperMap? {
if let int = value as? Int {
return .int(int)
}
if let double = value as? Double {
return .double(double)
}
if let string = value as? String {
return .string(string)
}
if let bool = value as? Bool {
return .bool(bool)
}
if let array = value as? [MapperMap] {
return .array(array)
}
if let dict = value as? [String: MapperMap] {
return .dictionary(dict)
}
return nil
}
public static func from(_ int: Int) -> MapperMap? {
return MapperMap.int(int)
}
public static func from(_ double: Double) -> MapperMap? {
return MapperMap.double(double)
}
public static func from(_ bool: Bool) -> MapperMap? {
return MapperMap.bool(bool)
}
public static func from(_ string: String) -> MapperMap? {
return MapperMap.string(string)
}
}And our neomap counterpart:
extension MapperNeomap : OutMap {
public static func from<T>(_ value: T) -> MapperNeomap? {
if let string = value as? String {
return .string(string)
}
if let bool = value as? Bool {
return .bool(bool)
}
if let i32 = value as? Int32 {
return .int32(i32)
}
if let uint = value as? UInt {
return .uint(uint)
}
if let uint8 = value as? UInt8 {
return .uint8(uint8)
}
if let float = value as? Float {
return .float(float)
}
if let array = value as? [MapperNeomap] {
return .array(array)
}
if let dict = value as? [String: MapperNeomap] {
return .dictionary(dict)
}
return nil
}
public static func from(_ int: Int) -> MapperNeomap? {
return .int32(Int32(int))
}
public static func from(_ double: Double) -> MapperNeomap? {
return .float(Float(double))
}
public static func from(_ bool: Bool) -> MapperNeomap? {
return .bool(bool)
}
public static func from(_ string: String) -> MapperNeomap? {
return .string(string)
}
}Well, that's it! Now one can easily map strongly-typed instances from and to your data type!
let user = try User(from: mapperMap)
let userMap: MapperMap = try user.map()If you need any help you can join our Slack and go to the #help channel. Or you can create a Github issue in our main repository. When stating your issue be sure to add enough details, specify what module is causing the problem and reproduction steps.
The entire Zewo code base is licensed under MIT. By contributing to Zewo you are contributing to an open and engaged community of brilliant Swift programmers. Join us on Slack to get to know us!
This project is released under the MIT license. See LICENSE for details.