Realm in multithreaded environments - swift

Question

I have a tracking app working in the background that uses Realm for persistency. I noticed the problem that sometimes the received locations are not saved in Realm an I think this could happen because of multithreading.
Here is my architecture:
The LocationLogger has an instance of CLLocationManager and an instance of my class for persistency: LocationModel. LocationManager is of type BaseModel and this one has an instance of Realm. In BaseModel realm loads an instance of Realm from my class RealmProvider:
lazy var realm: Realm = {
return RealmProvider().loadRealm()
}()!
And this code of RealmProvider is this:
class RealmProvider {
private var realm: Realm?
private let currentSchemaVersion: UInt64 = 8
func loadRealm() -> Realm? {
if let realm = self.realm {
return realm
}
do {
if let _ = NSClassFromString("XCTest"){
realm = try Realm(configuration: Realm.Configuration(fileURL: nil, inMemoryIdentifier: "test", syncConfiguration: nil, encryptionKey: nil, readOnly: false, schemaVersion: currentSchemaVersion, migrationBlock: nil, deleteRealmIfMigrationNeeded: true, objectTypes: nil))
} else {
realm = try Realm(configuration: Realm.Configuration(encryptionKey: nil, readOnly: false, schemaVersion: currentSchemaVersion, migrationBlock: nil, deleteRealmIfMigrationNeeded: true, objectTypes: nil))
}
}
catch {
logger.error("eror loading Realm!")
}
return realm
}
}
The reason why I have a RealmProvider is to have the configuration like versioning of the schema in one place.
Can you imagine why this isn't working always? Perhaps when LocationLogger is created in Thread A und the location callback comes in thread B? Other ideas?
Is it better to create an instance of Realm everytime a new location is reported? How can I do the schema configuration then?
Suggestion for the final solution:
class RealmProvider {
static private let currentSchemaVersion: UInt64 = 8
private lazy var configuration: Realm.Configuration = {
if let _ = NSClassFromString("XCTest") {
return Realm.Configuration(inMemoryIdentifier: "test", schemaVersion: currentSchemaVersion, deleteRealmIfMigrationNeeded: true)
}
return Realm.Configuration(schemaVersion: currentSchemaVersion, deleteRealmIfMigrationNeeded: true)
}()
public var realm: Realm {
var tempRealm: Realm?
do {
tempRealm = try Realm(configuration: configuration)
}
catch {
logger.error("eror loading Realm!")
}
if let tempRealm = tempRealm{
return tempRealm
}
return self.realm
}
}

Answer 1

Yes. It is better to create instances of Realm on demand rather than pre-emptively creating an instance ahead of time and retaining it. Since Realm objects are thread-confined, creating a Realm on Thread A, and then trying to interact with it on Thread B will cause an exception.
When calling Realm(configuration:), Realm itself will internally cache that object and will return the same object on each subsequent call. The object remains in the cache until the autoreleasepool it was created in is drained.
Like I said above, Realm objects are thread-confined. It's necessary to create a new instance of Realm when running on a new thread. Trying to pass an instance of a Realm from another thread will trigger an exception. This is the main reason why it's recommended not to retain an instance of Realm, especially if background operations are present.
Realm Configuration objects, as long as they aren't modified after their creation are thread-safe. So it's completely reasonable to create and retain a Configuration instance, and then re-use this instance to create Realm instances on any subsequent thread.
So to modify your above example:
class RealmProvider {
private let currentSchemaVersion: UInt64 = 8
private lazy var configuration: Configuration = {
if let _ = NSClassFromString("XCTest") {
return Realm.Configuration(inMemoryIdentifier: "test", schemaVersion: currentSchemaVersion,deleteRealmIfMigrationNeeded: true)
}
return Realm.Configuration(schemaVersion: currentSchemaVersion, deleteRealmIfMigrationNeeded: true)
}()
public var realm: Realm? {
var realm: Realm?
do {
realm = try Realm(configuration: configuration)
}
catch {
logger.error("eror loading Realm!")
}
return realm
}
}