Why is there no Integer 8 NSAttributeDescription attributeType in Core Data?

Question

Core Data's NSAttributeDescription has integer types for 16-bit, 32-bit, and 64-bit numbers, but not for 8-bit numbers. Why is that? Is the recommended way to store 8-bit numbers in an Integer 16 type?

It seems wasteful from a storage perspective to double the data size (by using 16 bits to store the 8-bit number). Also, what happens if, due to programmer error, a number out of the range of an 8-bit number is stored in that Integer 16? Then any function/method that takes int8_t could be passed the wrong number. For example:

NSManagedObject *object = // fetch from store
int16_t value = object.value.intValue;

[otherObject methodThatTakesInt8:(int8_t)value]; // bad things happen if value isn't within an 8-bit range

Answer 1

I don't think the answer as to why NSAttributeDescription doesn't offer an 8-bit number is any more complicated than that Core Data doesn't have an 8-bit storage type. Which is probably a circular argument. Probably Apple just didn't see the worth.

As to your other concerns: what if the programmer wanted to store a 12-bit number? What if they wanted to store a 24-bit number? It seems odd to pull out 8 bits as a special case in the modern world. But the problem is easily solved: you can implement -willSave on any NSManagedObject subclass to validate data before it is committed to the store. Or you could implement your own custom setter (ultimately calling -setPrimitiveValue:forKey:) similarly to validate immediately upon set. In either case you can implement whatever strategy you want for an out-of-bounds number: raise an exception, saturate, whatever.

Answer 2

In addition to @Tommy's answer, if you're using a SQLite persistent store (which nearly everyone does when using Core Data), it's not actually wasteful from a storage perspective. SQLite uses dynamic typing, meaning that any column can contain a value of any type. Size requirements are determined based on the value being saved. If you tell Core Data that you want a 64-bit integer attribute but all values of that attribute would fit in 8 bits, you haven't actually wasted 7/8 of the space used for that attribute.