A Swift library for working with JSON:API APIs. It supports mapping to custom model classes, fetching, advanced querying, linking and persisting.
Swift Other
Latest commit ee1ad2b Jan 18, 2017 @wvteijlingen committed on GitHub Merge pull request #150 from jensgrud/swiftyjson-3.1.4
SwiftyJSON compatibility update


Build Status Join the chat at https://gitter.im/wvteijlingen/Spine


Spine is a Swift library for working with APIs that adhere to the jsonapi.org standard. It supports mapping to custom model classes, fetching, advanced querying, linking and persisting.


This library was born out of a hobby project. Some things are still lacking, one of which is test coverage. Beware of this when using Spine in a production app!

Table of Contents

Supported features

Feature Supported Note
Fetching resources Yes
Creating resources Yes
Updating resources Yes
Deleting resources Yes
Top level metadata Yes
Top level errors Yes
Top level links Yes
Top level JSON API Object Yes
Client generated ID's Yes
Resource metadata Yes
Custom resource links No
Relationships Yes
Inclusion of related resources Yes
Sparse fieldsets Partially Fetching only, all fields will be saved
Sorting Yes
Filtering Yes Supports custom filter strategies
Pagination Yes Offset, cursor and custom pagination strategies
Bulk extension No
JSON Patch extension No



Add github "wvteijlingen/Spine" "master" to your Cartfile. See the Carthage documentation for instructions on how to integrate with your project using Xcode.


Add pod 'Spine', :git => 'https://github.com/wvteijlingen/Spine.git' to your Podfile. The spec is not yet registered with the Cocoapods repository, because the library is still in flux.


Defining resource types

Every resource is mapped to a class that inherits from Resource. A subclass should override the variables resourceType and fields. The resourceType should contain the type of resource in plural form. The fields array should contain an array of fields that must be persisted. Fields that are not in this array are ignored.

Each class must be registered using the Spine.registerResource method.

Defining resource fields

You need to specify the fields that must be persisted using an array of Fields. These fields are used when turning JSON into resources instances and vice versa. The name of each field corresponds to a variable on your resource class. This variable must be specified as optional.

Field name formatters

By default, the key in the JSON will be the same as your field name or serialized field name. You can specify a different name by using serializeAs(name: String). The name or custom serialized name will be mapped to a JSON key using a KeyFormatter. You can configure the key formatter using the keyFormatter variable on a Spine instance.

Spine comes with three key formatters: AsIsKeyFormatter, DasherizedKeyFormatter, UnderscoredKeyFormatter.

// Formats a field name 'myField' to key 'MYFIELD'.
public struct AllCapsKeyFormatter: KeyFormatter {
    public func format(field: Field) -> String {
        return field.serializedName.uppercaseString

spine.keyFormatter = AllCapsKeyFormatter()

Built in attribute types


An attribute is a regular attribute that can be serialized by NSJSONSerialization. E.g. a String or NSNumber.


An url attribute corresponds to an NSURL variable. These are represented by strings in the JSON document. You can instantiate it with a baseURL, in which case Spine will expand relative URLs from the JSON relative to the given baseURL. Absolute URLs will be left as is.


A date attribute corresponds to an NSDate variable. By default, these are represented by ISO 8601 strings in the JSON document. You can instantiate it with a custom format, in which case that format will be used when serializing and deserializing that particular attribute.


A to-one relationship corresponds to another resource. You must instantiate it with the type of the linked resource.


A to-many relationship corresponds to a collection of other resources. You must instantiate it with the type of the linked resources. If the linked types are not homogenous, they must share a common ancestor as the linked type. To many relationships are mapped to LinkedResourceCollection objects.

Custom attribute types

Custom attribute types can be created by subclassing Attribute. A custom attribute type must have a registered transformer that handles serialization and deserialization.

Transformers are registered using the registerTransformer method. A transformer is a class or struct that implements the Transformer protocol.

public class RomanNumeralAttribute: Attribute { }

struct RomanNumeralValueFormatter: ValueFormatter {
    func unformat(value: String, attribute: RomanNumeralAttribute) -> AnyObject {
        let integerRepresentation: NSNumber = // Magic...
        return integerRepresentation

    func format(value: NSNumber, attribute: RomanNumeralAttribute) -> AnyObject {
        let romanRepresentation: String = // Magic...
        return romanRepresentation

Example resource class

// Resource class
class Post: Resource {
    var title: String?
    var body: String?
    var creationDate: NSDate?
    var author: User?
    var comments: LinkedResourceCollection?

    override class var resourceType: ResourceType {
        return "posts"

    override class var fields: [Field] {
        return fieldsFromDictionary([
            "title": Attribute(),
            "body": Attribute().serializeAs("content"),
            "creationDate": DateAttribute(),
            "author": ToOneRelationship(User),
            "comments": ToManyRelationship(Comment)



Fetching resources

Resources can be fetched using find methods:

// Fetch posts with ID 1 and 2
spine.find(["1", "2"], ofType: Post).onSuccess { resources, meta, jsonapi in
  println("Fetched resource collection: \(resources)")
.onFailure { error in
  println("Fetching failed: \(error)")

spine.findAll(Post) // Fetch all posts
spine.findOne("1", ofType: Post)  // Fetch a single posts with ID 1

Alternatively, you can use a Query to perform a more advanced find:

var query = Query(resourceType: Post)
query.include("author", "comments", "comments.author") // Sideload relationships
query.whereProperty("upvotes", equalTo: 8) // Only with 8 upvotes
query.addAscendingOrder("creationDate") // Sort on creation date

spine.find(query).onSuccess { resources, meta, jsonapi in
  println("Fetched resource collection: \(resources)")
.onFailure { error in
  println("Fetching failed: \(error)")

All fetch methods return a Future with onSuccess and onFailure callbacks.

Saving resources

spine.save(post).onSuccess { _ in
    println("Saving success")
.onFailure { error in
    println("Saving failed: \(error)")

Extra care MUST be taken regarding related resources. Saving does not automatically save any related resources. You must explicitly save these yourself beforehand. If you added a new create resource to a parent resource, you must first save the child resource (to obtain an ID), before saving the parent resource.

Deleting resources

spine.delete(post).onSuccess {
    println("Deleting success")
.onFailure { error in
    println("Deleting failed: \(error)")

Deleting does not cascade on the client.

Loading and reloading resources

You can use the Spine.load methods to make sure resources are loaded. If it is already loaded, it returns the resource as is. Otherwise it loads the resource using the passed query.

The Spine.reload method works similarly, except that it always reloads a resource. This can be used to make sure a resource contains the latest data from the server.


You can fetch next and previous pages of collections by using: Spine.loadNextPageOfCollection and Spine.loadPreviousPageOfCollection.

JSON:API is agnostic about pagination strategies. Because of this, Spine by default only supports two pagination strategies:

  • Page based pagination using the page[number] and page[size] parameters
  • Offset based pagination using the page[offset] and page[limit] parameters

You can add a custom filter strategy by creating a new type that conforms to the Pagination protocol, and then subclassing the built in Router class and overriding the queryItemsForPagination(pagination: Pagination) method.

Example: implementing 'cursor' based pagination

In this example, cursor based pagination is added a using the page[limit], and either a page[before] or page[after] parameter.

public struct CursorBasedPagination: Pagination {
    var beforeCursor: String?
    var afterCursor: String?
    var limit: Int
class CustomRouter: Router {
    override func queryItemsForPagination(pagination: Pagination) -> [NSURLQueryItem] {
        if let cursorPagination = pagination as? CursorBasedPagination {
            var queryItems = [NSURLQueryItem(name: "page[limit]", value: String(cursorPagination.limit))]

            if let before = cursorPagination.beforeCursor {
                queryItems.append(NSURLQueryItem(name: "page[before]", value: before))
            } else if let after = cursorPagination.afterCursor {
                queryItems.append(NSURLQueryItem(name: "page[after]", value: after))

            return queryItems
        } else {
            return super.queryItemsForPagination(pagination)


JSON:API is agnostic about filter strategies. Because of this, Spine by default only supports 'is equal to' filtering in the form of ?filter[key]=value.

You can add a custom filter strategy by subclassing the built in Router class and overriding the queryItemForFilter(filter: NSComparisonPredicate) method. This method takes a comparison predicate and returns a matching NSURLQueryItem.

Example: implementing a 'not equal to' filter

In this example, a switch statement is used to add a 'not equal filer in the form of ?filter[key]=!value.

class CustomRouter: Router {
    override func queryItemForFilter(field: Field, value: AnyObject, operatorType: NSPredicateOperatorType) -> NSURLQueryItem {
        switch operatorType {
        case .NotEqualToPredicateOperatorType:
            let key = keyFormatter.format(field)
            return NSURLQueryItem(name: "filter[\(key)]", value: "!\(value)")
            return super.queryItemForFilter(filter)

let baseURL = NSURL(string: "http://api.example.com/v1")
let spine = Spine(baseURL: baseURL, router: CustomRouter())


Spine uses a NetworkClient to communicate with the remote API. By default it uses the HTTPClient class which performs request over the HTTP protocol.

Customising HTTP headers of HTTPClient

The HTTPClient class supports setting HTTP headers as follows:

(spine.networkClient as! HTTPClient).setHeader("User-Agent", to: "My App")
(spine.networkClient as! HTTPClient).removeHeader("User-Agent")

Using a custom network client

You can use a custom client by subclassing HTTPClient or by creating a class that implements the NetworkClient protocol. Pass an instance of this class when instantiating a Spine:

var customClient = CustomNetworkClient()
var spine = Spine(baseURL: NSURL(string:"http://example.com")!, networkClient: customClient)


Spine comes with a rudimentary logging system. Each logging domain can be configured with a certain log level:

Spine.setLogLevel(.Debug, forDomain: .Spine)
Spine.setLogLevel(.Info, forDomain: .Networking)
Spine.setLogLevel(.Warning, forDomain: .Serializing)

These levels are global, meaning they apply to all Spine instances.

Log domains

  • Spine: The main Spine component.
  • Networking: The networking component, requests, responses etc.
  • Serializing: The (de)serializing component.

Log levels

  • Debug
  • Info
  • Warning
  • Error
  • None

Custom loggers

The default ConsoleLogger logs to the console using the Swift built in print command. You can assign a custom logger that implements the Logger protocol to the static Spine.logger variable.

Memory management

Spine suffers from the same memory management issues as Core Data, namely retain cycles for recursive relationships. These cycles can be broken in two ways:

  1. Declare one end of the relationship as weak or unowned.
  2. Use a Resource's unload method to break cycles when you are done with the resource.

Using the serializer separately

You can also just use the Serializer to (de)serialize to and from JSON:

let serializer = Serializer()

// Register resources

// Optional configuration
serializer.keyFormatter = DasherizedKeyFormatter()

// Convert NSData to a JSONAPIDocument struct
let data = fetchData()
let document = try! serializer.deserializeData(data)

// Convert resources to NSData
let data = try! serializer.serializeResources([post])

// Convert resources to link data
let data = try! serializer.serializeLinkData(post)
let data = try! serializer.serializeLinkData([firstPost, secondPost])