In Core Data, why does a non-optional String attribute generate a String? property, while an optional Integer 16 attribute generates a plain Int16?
answer
- two meanings of optional
- checked when the context saves
- objects can be nil before save
- a scalar has no nil
- error 1570, the required value
basics
~20 sCore Data's Optional box is a save-time validation rule, not the Swift type. Object attributes generate Swift optionals because they can be nil before save; scalars generate plain numbers, so an unset optional Int16 reads 0.
solid answer
~40 sCore Data's **Optional** flag means "may be nil when the context saves": a non-optional attribute left nil makes `save()` throw `NSValidationMissingMandatoryPropertyError` (1570), not the assignment. Because a new object starts nil until you set it, Xcode generates object-typed attributes (`String`, `Date`, `Data`, `UUID`, `URL`, `NSDecimalNumber`) as Swift optionals either way. Numbers and Booleans default to **Use Scalar Type**, generating `Int16`, `Double` or `Bool`, which cannot represent nil, so an unset optional `Int16` reads as `0` while `value(forKey:)` returns nil. To tell "never fertilised" from zero, turn the scalar off and get `NSNumber?`. A model **default value**, such as a watering interval of 7, is applied when the object is inserted.
code
swift · 14 linesimport CoreData
@objc(Plant)
public class Plant: NSManagedObject {}
// Generated by Xcode for the Plant entity (abridged).
extension Plant {
@NSManaged public var nickname: String? // String, Optional on
@NSManaged public var speciesName: String? // String, Optional off
@NSManaged public var wateringInterval: Int16 // Integer 16, scalar, default 7
@NSManaged public var fertiliserDose: Double // Double, scalar, Optional on
@NSManaged public var lastWatered: Date? // Date, scalar off
@NSManaged public var photo: Data? // Binary Data, external storage
}go deeper
Recall that the model's Optional box is checked when the context saves, and that a missing required value fails the save with a validation error.
Explain how the attribute type and Use Scalar Type choose the generated Swift type, and why an unset optional scalar reads as zero.
Spot data bugs where zero and missing were conflated, and choose defaults or NSNumber? before release so the fix needs no migration.
Set a team rule for attribute review: types, optionality and defaults are schema decisions that cost a migration to change later.
## Two different meanings of optional In a Core Data model every attribute has an **Optional** box (`isOptional` on `NSPropertyDescription`, `true` by default in code). The header describes it as whether the value "can be nil or not (before an object can be persisted)". It is a **validation rule checked at save**, not a type. A non-optional `speciesName` can be nil all the way from `Plant(context:)` to the save; only then does `save()` throw an `NSError` in `NSCocoaErrorDomain` with code `NSValidationMissingMandatoryPropertyError` (1570), whose message calls the property "a required value". Swift's optional, by contrast, is a compile-time promise about every read. Because a fresh managed object really does hold nil in its non-optional attributes until you set them, Xcode cannot honestly generate `String` for them, and emits `String?`. ## How the generated Swift type is chosen The generated property type depends on the **attribute type** and its **Use Scalar Type** box, not on Optional: | Attribute type | Scalar type | Scalar by default | Without scalar | |---|---|---|---| | Integer 16 / 32 / 64 | `Int16` / `Int32` / `Int64` | yes | `NSNumber?` | | Double, Float | `Double`, `Float` | yes | `NSNumber?` | | Boolean | `Bool` | yes | `NSNumber?` | | Date | `TimeInterval` | no | `Date?` | | String, Binary Data, URI | none | n/a | `String?`, `Data?`, `URL?` | | Decimal, UUID | same class, no separate scalar | no | `NSDecimalNumber?`, `UUID?` | | Transformable | none | n/a | the custom class, or `NSObject?` | A scalar cannot be nil, so: - an **optional scalar** that was never set reads as `0` (or `false`) through the generated property, while `value(forKey:)` still returns nil; - "never fertilised" and "fertilised with 0 ml" become indistinguishable through `fertiliserDose: Double`; - the fix is to clear Use Scalar Type for that attribute and work with `NSNumber?`, or to make the attribute non-optional with a meaningful default. ## Default values An attribute's **default value** (`defaultValue` in code) is applied when the object is inserted, so a new `Plant` reads `wateringInterval == 7` before any code touches it. Two details matter: 1. The model **retains** a default value object rather than copying it, so never mutate an object after setting it as a default. 2. A default makes a non-optional scalar safe to leave untouched, which is usually better than an optional scalar that silently reads 0. 3. A default is part of the model, not of the stored rows: setting a new default value for an attribute leaves existing stores openable, while adding the attribute itself does not, as the model documentation notes. ## The plant-care Plant entity - `nickname`: String, Optional on, so `String?` and nil is a legal saved state. - `speciesName`: String, Optional off, still `String?` in Swift; a nil value fails the save with 1570. - `wateringInterval`: Integer 16, scalar, non-optional, default 7, so `Int16`. - `fertiliserDose`: Double, Optional on, scalar on, so `Double` that reads 0 when unset; switch the scalar off if nil must stay visible. - `lastWatered`: Date with the scalar off, so `Date?`; the scalar form would be a `TimeInterval`. - `photo`: Binary Data with **Allows External Storage** (`allowsExternalBinaryDataStorage`), which lets Core Data keep a large value in a file outside the store while the property is still `Data?`. ## Why the choice comes back later The type and optionality of an attribute describe the store's schema, and the Optional flag is one of the values that make up a property's version hash. Changing an attribute's type, or its optionality, after release changes the schema the store was created with, so the store needs a migration, a subject covered with model versioning and migration. Choosing the type carefully the first time (Integer 16 is enough for a day count, Integer 64 is not needed) saves that work. If the generated `String?` is unwelcome, Manual/None codegen lets you declare `@NSManaged var speciesName: String` yourself; that only changes the Swift surface, and the value can still be nil until it is set.
- In Core Data, when exactly does a missing non-optional attribute produce an error?When the context validates the object during `save()`. Assigning nil, inserting an object without setting the attribute, and fetching are all allowed; the save throws `NSValidationMissingMandatoryPropertyError` (1570), and `validateForInsert()` lets you run the same check earlier.
- When should a Core Data Binary Data attribute use Allows External Storage?For large values such as a plant photo. With `allowsExternalBinaryDataStorage` set, Core Data may keep the bytes in a file outside the persistent store itself, so the store stays small; the attribute's Swift type is still `Data?`.
- How do you keep nil visible for an optional Core Data number such as a fertiliser dose?Clear Use Scalar Type for that attribute so Xcode generates `NSNumber?`, or read it with `value(forKey:)`. With the scalar on, the generated `Double` reports 0 for a value that was never set.
- Treating the model's Optional box as the same thing as a Swift optional
- Expecting assigning nil to a non-optional attribute to throw immediately
- Using an optional scalar and reading 0 as a real recorded value
- Assuming a non-optional String attribute generates a non-optional String property
- Changing an attribute's type after release without planning a migration