(Data models renamed to preserve anonymity. :D )
So, I have a situation where on a Thing, whose configuration is defined by its own special object, though that object can be one of 2 class types.
I would like to be able to do Fetch Requests on this model by asking if the Thing's Configuration is of a certain known subclass, and if so, does it have a few specific flags set in its bitmask. If this is true, I define this as being "not special". Therefore, if it also has the other type of subclass as it's Configuration, it is also "not special".
My questions are these:
Have I got my syntax right? (In my app I get faults)
Why does it want to return an NSArray? (I thought its traversing a to-one relationship on a single object. How would I therefore interpret this NSArray?)
I should add that I only know Core Data via MagicalRecord more or less. And I use mogenerator too.
Related
Is there a convention for naming Core Data entity? The argument I heard for not prefixing Core Data entity is because there is no chance they will be collision since they only need to be unique within a model, which is not true because the NSManagedObject subclass generated may still collide with existing Objective-C classes.
So it seems logical for me to do two things to Core Data Entity: Prefix it with my project class prefix, and suffix it with Entity. This way, I know it's a Core Data entity, and its name will never collide with any other classes.
I have actually seen both prefixing with the project class prefix and without. I have never seen a suffix being added. I prefer without the project prefix, since if you have a remote database that you are syncing up with, I would use the same entity names. And then if you ever release a public API, do you really want your project prefix all over the place? For example, Stripe's entities are Customer, Card, etc. They use the prefix in the unique identifiers, which I like. Also, if you are using the project prefix for your other classes, you do not run the risk of overwriting, like you mention in your OP.
From here https://developer.apple.com/library/archive/releasenotes/ObjectiveC/RN-TransitioningToARC/Introduction/Introduction.html
You cannot give an accessor a name that begins with new. This in turn means that you can’t, for example, declare a property whose name begins with new unless you specify a different getter
And as far as I know also copy keyword cant be used as suffix.
As a complete novice programmer I am trying to populate my neo4j DB with data from heterogeneous sources. For this I am trying to use the Neo4jClient C# API. The heterogeneity of my data comes from a custom, continuously evolving DSL/DSML/metamodel that defines the possible types of elements, i.e. models, thus creating classes for each type would not be ideal.
As I understand, my options are the following:
Have a predefined class for each type of element: This way I can easily serialize my objects that is if all properties are primitive types or arrays/lists.
Have a base class (with a Dictionary to hold properties) that I use as an interface between the models that I'm trying to serialize and neo4j. I've seen an example for this at Can Neo4j store a dictionary in a node?, but I don't understand how to use the converter (defined in the answer) to add a node. Also, I don't see how an int-based dictionary would allow me to store Key-Value pairs where the keys (that are strings) would translate to Property names in neo4j.
Generate a custom query dynamically, as seen at https://github.com/Readify/Neo4jClient/wiki/cypher#manual-queries-highly-discouraged. This is not recommended and possibly is not performant.
Ultimately, what I would like to achieve is to avoid the need to define a separate class for every type of element that I have, but still be able to add properties that are defined by types in my metamodel.
I would also be interested to somehow influencing the serializer to ignore non-compatible properties (similarly to XmlIgnore), so that I would not need to create a separate class for each class that has more than just primitive types.
Thanks,
J
There are 2 problems you're trying to solve - the first is how to program the C# part of this, the second is how to store the solution to the first problem.
At some point you'll need to access this data in your C# code - unless you're going fully dynamic you'll need to have some sort of class structure.
Taking your 3 options:
Please have a look at this question: neo4jclient heterogenous data return which I think covers this scenario.
In that answer, the converter does the work for you, you would create, delete etc as before, the converter just handles the IDictionary instance in that case. The IDictionary<int, string> in the answer is an example, you can use whatever you want, you could use IDictionary<string, string> if you wanted, in fact - in that example, all you'd need to do would be changing the IntString property to be an IDictionary<string,string> and it should just work.
Even if you went down the route of using custom queries (which you really shouldn't need to) you will still need to bring back objects as classes. Nothing changes, it just makes your life a lot harder.
In terms of XmlIgnore - have you tried JsonIgnore?
Alternatively - look at the custom converter and get the non-compatible properties into your DB.
I'm in the process of removing RestKit from our iOS app. I'm able to get things that I want into Core Data, but they're not really connected.
For example, we have one network call that returns a list of "Category"s (which have a "categoryID" and a "categoryName"; "Category"s also map to-many "StoreLocation"s). We then have another network call that returns a list of "StoreLocation"s (which, among other things, have a "storeName", "storeID", "storeCategoryIDs"; "StoreLocation"s also map to-many "Category"s).
With RestKit, I could use a RKConnectionDescription to describe that "storeCategoryIDs" drove the relationship to-many "Category"s. With that, if I had a given Category object, I could easily determine which StoreLocations belonged to that category.
I'm struggling to see how to accomplishing this without any RestKit dependencies. I suppose I could, whenever I'm about to insert a new Category or new StoreLocation, fetch all of the opposite managed objects and do this manually, but I seem to be missing some component of Core Data that can do.
The main part you're missing is the predicate applied to the fetch and which uses the identification attributes to find the appropriate existing objects. You do need to run your own fetch as core data will not magically update one object if you create a different new object and insert it.
I use Core Data within my iOS 7 app to handle the editing and creation of entities. The entities have relationships between them, which all have inverses (as Apple advises).
For the sake of this question, let's pick any one of these interrelated entities and call it the Root entity: the thing that I want to encode with; the thing that logically lives on the 'top' of the hierarchy. I will call this the 'object graph'.
The question is:
What's the easiest way of encoding and decoding such an object graph to and from NSData?
The reason I want to do this is that I'd like my Core Data object graph to be persisted onto a cloud service, without the need of writing my own NSIncrementalStore subclass (it's a bit involved...!).
AutoCoding together with HRCoder almost looks like it could do the job, but I've experimented with this combination and it doesn't quite work with NSManagedObjects at the time of writing.
Still, I'm seeking alternatives. There can't only be one way to do this, surely.
It doesn't have to be JSON, but it'd be nice. Binary would be fine.
It seems to me you do not need to subclass NSIncrementalStore. You can create records and save them to your store with a plain vanilla store created via addPersistentStoreWithType:... with a NSPersistentStoreCoordinator.
The straight-forward way is to handle the incoming JSON by simply taking the data and copying it to the properties of your NSManagedObject subclasses, like this:
object.title = jsonDictionary[#"title"];
object.numericAttribute = [jsonDictionary[#"numericAttribute] integerValue];
If you take care about naming the attribute and entity names exactly the same you can maybe use some shortcuts using KVC, like
[object setValue:jsonDictionary[key] forKey:key];
I once did the above for a large legacy project where it was not feasible to repeat the old attribute names, so I used a custom property list (plist) to match around 800 attribute names.
When getting an object from the DB, should the object's properties also be loaded? There seems to be these approaches.
Create well-formed, fully-loaded objects.
Pro: No need to check if a property has been loaded; it has. Pass it around and don’t worry about parts of the object not being there.
Con: Where do you stop? If an object has an object, and that object has an object, and that object has 40 objects as properties, etc… Do you load the whole DB? Or do you make a decision in the BLL as to what constitutes a well-formed object, and load those properties?
Don’t load any properties that are other objects.
Pro: Quick, no loading unnecessary properties.
Con: Code has to be constantly written to check if properties are populated.
Lazy-loading: only load properties when they are first used.
Pro/Con: Not sure what to say about this approach. It seems intuitively wrong.
Is there another approach? What approach is the best?
And finally, what about properties that can be null? For example, a car may not have a PreviousOwner object. Do you set it to null? An empty PreviousOwner object? Does that property belong in another class then?
There's no easy answer to your question because it depends on what you're trying to achieve.
It looks like you expect a more or less complete object graph to be loaded from the database (i.e. with relationships between multiple object types and the objects themselves stored in the database).
If this is the case, I would look into using the Object Relationship Mapper that's convenient in my language of choice.
As to how much of the object graph is being loaded, the model employed by Apple CoreData's system is objects not yet retrieved are marked as faulty (they call the concept "faulting" - it's described in Limiting the Size of the Object Graph: Faulting. This is a play on the lazy loading concept you described yourself.