Simpler Core Data with RubyMotion Stefán Hafliðason http://stefan.haflidason.com @styrmis 

Why RubyMotion? • Promises increased developer productivity • Brings the flexibility of Ruby to iOS and OSX development • Bridges directly to Obj-C libraries: no intermediate glue code • A REPL for working with your app live! • Make tweaks quickly • Build whole views programmatically on the fly 

Why Core Data? • Optimised for low-memory/embedded (iOS) devices • Mature data access/persistence framework • Also available on OSX • Works with iCloud—free cloud syncing for your app 

Core Data is Difficult • Provided boilerplate code unnecessarily complex • An object graph that’s persisted to an SQLite database • Suggests relational access, which is not quite the case • Typical patterns for working with relational data are not optimal here 

RubyMotion is “Easy” • Friendliness of Ruby • An ARC equivalent is included • Lots of work done to abstract complexity away • More concepts similar to other OO languages 

Core Data and RubyMotion • No equivalent of Xcode’s visual data modeller • How do I define my data model?! • What about versioning?! • How will I handle migrations? 

What we need • Our data model (NSEntityDescriptions + NSRelationshipDescriptions forming our NSManagedObject) • A Core Data Stack (NSManagedObjectModel + NSPersistentStoreCoordinator + NSManagedObjectContext) • A workflow for versioning and migrating between versions 

Defining Our Data Model • We would normally do this in Xcode • Visual Editor for .xcdatamodel bundles • Integrated handling of versioning and custom migration code • Automatic lightweight (schema) migrations • How do we achieve this with RubyMotion? 

Options for RubyMotion • Handle everything programmatically (low level) • Use Xcode to work with .xcdatamodel files, copy in each time • Use a Ruby library for creating .xcdatamodel files 

Handling Everything Programmatically entity = NSEntityDescription.alloc.init entity.name = 'Task' entity.managedObjectClassName = 'Task' entity.properties = [ 'task_description', NSStringAttributeType, 'completed', NSBooleanAttributeType ].each_slice(2).map do |name, type| property = NSAttributeDescription.alloc.init property.name = name property.attributeType = type property.optional = false property end 

Handling Everything Programmatically entity = NSEntityDescription.alloc.init entity.name = 'Task' entity.managedObjectClassName = 'Task' entity.properties = [ 'task_description', NSStringAttributeType, 'completed', NSBooleanAttributeType ].each_slice(2).map do |name, type| property = NSAttributeDescription.alloc.init property.name = name property.attributeType = type property.optional = false property end Not all that bad, but we want to use .xcdatamodel files 

.xcdatamodel files are just XML ! ! ! ! ! ! ! ! ! ! 

Using a library to generate .xcdatamodel files (ruby-xcdm) 1 schema "001" do! 2 entity "Article" do! 3 string :body, optional: false! 4 integer32 :length! 5 boolean :published, default: false! 6 datetime :publishedAt, default: false! 7 string :title, optional: false! 8 ! 9 belongs_to :author! 10 end! 11 ! 12 entity "Author" do! 13 float :fee! 14 string :name, optional: false! 15 has_many :articles! 16 end! 17 end 

Workflow • Create schema file in schemas directory, e.g. schemas/001_initial.rb • Build the schema • Add a new schema version, e.g. 002_add_new_fields.rb • Rebuild the schema • That’s it! 

Workflow $ echo "gem 'ruby-xcdm', '0.0.5'" >> Gemfile $ bundle install $ rake schema:build Generating Data Model learn-xcdm Loading schemas/001_initial.rb Writing resources/learn-xcdm.xcdatamodeld/1.xcdatamodel/ contents $ rake # The default rake task is to run the app in the simulator (main)> mom = NSManagedObjectModel.mergedModelFromBundles(nil) => # (main)> mom.entities.count => 2 (main)> mom.entities.first.name => "Article" (main)> mom.entities.first.propertiesByName => {"body"=>#, "title"=>#} 

Advantages of using ruby- xcdm • No magic: generates XML from a schema • Schema versions are fully text-based and readable, making them well-suited to version control • Can compile our versions into .xcdatamodeld bundles, completely removing dependence on Xcode 

Basic Core Data Stack 1 model = NSManagedObjectModel.mergedModelFromBundles(nil) 2 3 store = NSPersistentStoreCoordinator.alloc.initWithManagedObjectModel(model) 4 store_path = File.join(NSHomeDirectory(), 'Documents', 'LearnXcdm.sqlite') 5 store_url = NSURL.fileURLWithPath(store_path) 6 7 options = { NSMigratePersistentStoresAutomaticallyOption => true, 8 NSInferMappingModelAutomaticallyOption => true } 9 10 error_ptr = Pointer.new(:object) 11 12 unless store.addPersistentStoreWithType(NSSQLiteStoreType, 13 configuration: nil, 14 URL: store_url, 15 options: options, 16 error: error_ptr) 17 raise "[ERROR] Failed to create persistent store: #{error_ptr[0].description}" 18 end 19 20 @context = NSManagedObjectContext.alloc.init 21 @context.persistentStoreCoordinator = store 

Core Data Query • From the developers of ruby-xcdm • Abstracts away much of the complexity of Core Data • All you need is your .xcdatamodeld bundle 

Core Data Query in Action # app/models/task.rb class Task < CDQManagedObject end ! # app/app_delegate.rb class AppDelegate include CDQ ! def application(application, didFinishLaunchingWithOptions:launchOptions) cdq.setup true end end 

Core Data Query in Action (main)> Task.count => 0 (main)> t1 = Task.create(task_description: "Complete presentation") (main)> t2 = Task.create(task_description: "File tax return") (main)> cdq.save => true (main)> exit $ rake ... (main)> Task.count => 2 (main)> t1, t2 = Task.all.array (main)> t1.task_description => "Complete chapter" (main)> t2.task_description => "File tax return" (main)> t2.destroy => # (main)> cdq.save => true (main)> Task.count => 1 

Core Data in Action Author.where(:name).eq("Emily") Author.where(:name).not_equal("Emily") Author.limit(1) Author.offset(10) Author.where(:name).contains("A").offset(10).first ! # Conjuctions Author.where(:name).contains("Emily").and.contains("Dickinson") Author.where(:name).starts_with("E").or(:pub_count).eq(1) ! # Nested Conjuctions Author.where(:name).contains("Emily").and(cdq(:pub_count).gt(100).or.lt(10) ) ! # Relationships Author.first.publications.offset(2).limit(1) cdq(emily_dickinson).publications.where(:type).eq('poetry') ! class Author < CDQManagedObject scope :prolific, where(:pub_count).gt(50) end 

Takeaways • Don’t be put off by the Xcode boilerplate: Core Data doesn’t have to be that hard • With CDQ, Core Data is arguably easier to use with RubyMotion rather than harder • XCDM, CDQ and RubyMotion Query (all by Infinitered) are all worth taking a look at 

Next Steps • In the coming weeks I’ll be researching and writing about: • How to best handle heavyweight/data migrations in RubyMotion • Deconstructing the ‘magic’ in Core Data Query • RubyMotion development best practices Stefán Hafliðason http://stefan.haflidason.com @styrmis