Is declare new property in class extension a bad practice in Objective-C?

Question

One strong advantage of class extension is that with class extension you can declare a readonly property in the header file and override this property in class extension as readwrite property. Like below :

//SomeClass.h

@interface SomeClass : NSObject
{
    NSInteger someInt;   //with modern runtime you can omit this line
}
@property (readonly) NSInteger someInt;
@end

//SomeClass.m
@interface SomeClass ()
@property (readwrite) NSInteger someInt;
@end

@implementation SomeClass
@synthesize someInt;
@end

But if you use a modern runtime ,you can also declare a totally new property in the class extension (which also generate an iVar for that property if there isn't).

//SomeClass.h

@interface SomeClass : NSObject
{
}
@end

//SomeClass.m
@interface SomeClass ()
@property (readwrite) NSInteger someInt;
@end

@implementation SomeClass
@synthesize someInt;
@end

Here's my question : I think declare a totally new property in class extention is somehow has some side effects. Because class extension my not be in the header file and someone else who subclass the class may not know about that "secret property". And if he declare a property with the same name of that "secret property". And this new property's getter and setter method will override the super class's. Isn't this a problem?And why would modern runtime allow such thing happen?

EDIT I posted another question about this topic , please check it out: The risk of declare new propery in class extension (Ojbective-C) , how to solve it?

Answer 1

I don't think it's bad practice to declare a new property in a class extension. I do this with some frequency. The only reason to include the readonly property in the header in the first place is to allow other classes to get the value, while only you are allowed to modify it. Quite often, that ivar should be of no concern to other classes, and is an implementation detail only. As such, it has no place in the header file.

Implementing this ivar as a private property (a new property only declared in the Class Extension) is still useful, because of the convenient memory management boilerplate code it can abstract for you. Unfortunately, name collisions are just a fact of life in Objective C. Apple lays out some pretty clear naming conventions for you to follow (or not follow) to prevent collisions with their method names. If you're worried about collisions with the getters and setters you've invisibly created with that private property, just adopt and obsessively follow some naming convention for those private property names that you only ever use when implementing a private property. That's the best you're going to do with Objective C, but I personally think the benefits outweigh the risks.