DataKernel is a minimalistic wrapper around CoreData stack to ease persistence operations. It is heavily inspired by SugarRecord but have no external dependencies (except Cocoa of course) and with some refinements. It is covered with unit tests.
If you have any propositions please file issue. If you need usage examples - see unit test, it is very straightforward
- Swift (tested in XCode 9.1)
- Protocols based design
- Fully tested
- Actively supported
- Rich set of operations (but I think something may be missed)
- No iCloud yet (but planned)
- context.wipe uses NSBatchDeleteRequest on iOS 9 and OSX 10.11, it gives big performance improvements, but can cause problems if you use patterns like wipe all and create new, because it acts directly on persistent store (see Apple WWDC 2015 video or slides)
- Install CocoaPods. You can do it with
gem install cocoapods
- Edit your
Podfile
file and add the following linepod 'DataKernel'
(you have to use use_frameworks!, because this is a Swift pod) - Update your pods with the command
pod install
- Open the project from the generated workspace (
.xcworkspace
file).
Note: You can also test the last commits by specifying it directly in the Podfile line
- Install Carthage on your computer using
brew install carthage
- Edit your
Cartfile
file adding the following linegithub "mrdekk/DataKernel"
- Update and build frameworks with
carthage update
- Add generated frameworks to your app main target following the steps here
- Link your target with CoreData library (from Build Phases)
A storage is a general wrapper around CoreData PersistentStoreCoordination and PersistentStores. The first step you need to get started is to create CoreDataLocalStorage
.
let store = StoreRef.Named("test1")
let bundle = NSBundle(forClass: self.classForCoder)
let model = ModelRef.Merged([bundle])
let storage = try! CoreDataLocalStorage(store: store!, model: model!, migration: true)
Storage offer access to uiContext (NSManagedObjectContext on main thread to GET data from storage). All modification operations on CoreData entities should be performed with perform
operation on storage.
let cars: [Cars] = try! storage.uiContext.fetch(Request<Car>().filtered("mark", equalTo: "Honda"))
let cars: [Cars] = try! storage.uiContext.fetch(Request<Car>())
let cars: [Cars] = try! storage.uiContext.fetch(Request<Car>().sorted("model", ascending: true))
let predicate = NSPredicate(format: "model == %@", "CRZ")
let crz: Car? = try! storage.uiContext.fetch(Request<Car>(predicate: predicate)).first
All modification operations should be performed under perform
operation due to it handles all core data context/threading things internally and you can safely use modification operations under them.
save
operation perform recursive save with nesting context on core data stack. So, if you change entities that already loaded in uiContext, they will be updated.
Note the first parameter ephemeral
on perform
function. If it is true, new context for this operation will be created, and then after save succeeded it will be removed. Creation of contexts in CoreData is rather cheap operations, so don't worry. If you don't need to save data in ephemeral context, just don't call save
. If you set ephemeral = false
then a pre-created special save context will be used for all ephemeral = false
operations.
do {
storage.perform(true) { (context, save) throws -> Void in
// do you unit of work here
save()
}
}
catch {
// There was an error in the operation
}
You can use the create()
for initializing and inserting in the context in the same operation:
do {
storage.perform(true) { (context, save) throws -> Void in
let newCar: Car = try! context.create()
newCar.model = "Honda"
newCar.mark = "CRZ"
save()
}
}
catch {
// There was an error in the operation
}
Upsert - update or insert first search the entity in storage, and then if it is not found creates it. Just call the acquire
function.
Important: to get this feature work, you should add model property pk
and set it to name of the field, that is primary key of this entity. Moreover, this property should be indexed. It is needed to achieve needed level of performance.
do {
storage.perform(true) { (context, save) throws -> Void in
let car: Car = try! context.acquire("CRZ")
}
}
catch {
// There was an error in the operation
}
In a similar way you can use the remove()
method from the context passing the objects you want to remove from the database:
do {
storage.perform(true) { (context, save) throws -> Void in
let car: Car? = try! context.fetch(Request<Car>.filtered("model", equalTo: "CRZ")).first
if let car = car {
try! context.remove([car])
save()
}
}
}
catch {
// There was an error in the operation
}
If you want to communicate any issue, suggestion or even make a contribution, you have to keep in mind the flow bellow:
- If you need help, ask your doubt in Stack Overflow using the tag 'datakernel'
- If you want to ask something in general, use Stack Overflow too.
- Open an issue either when you have an error to report or a feature request.
- If you want to contribute, submit a pull request, and remember the rules to follow related with the code style, testing, ...
This project adheres to the [Open Code of Conduct][code-of-conduct]. By participating, you are expected to honor this code. [code-of-conduct]: http://todogroup.org/opencodeofconduct/#DataKernel/[email protected]
The MIT License (MIT)
Copyright (c) <2017>
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.