Is it possible to make the -init method private in Objective-C? - objective-c

I need to hide (make private) the -init method of my class in Objective-C.
How can I do that?

NS_UNAVAILABLE
- (instancetype)init NS_UNAVAILABLE;
This is a the short version of the unavailable attribute. It first appeared in macOS 10.7 and iOS 5. It is defined in NSObjCRuntime.h as #define NS_UNAVAILABLE UNAVAILABLE_ATTRIBUTE.
There is a version that disables the method only for Swift clients, not for ObjC code:
- (instancetype)init NS_SWIFT_UNAVAILABLE;
unavailable
Add the unavailable attribute to the header to generate a compiler error on any call to init.
-(instancetype) init __attribute__((unavailable("init not available")));
If you don't have a reason, just type __attribute__((unavailable)), or even __unavailable:
-(instancetype) __unavailable init;
doesNotRecognizeSelector:
Use doesNotRecognizeSelector: to raise a NSInvalidArgumentException. “The runtime system invokes this method whenever an object receives an aSelector message it can’t respond to or forward.”
- (instancetype) init {
[self release];
[super doesNotRecognizeSelector:_cmd];
return nil;
}
NSAssert
Use NSAssert to throw NSInternalInconsistencyException and show a message:
- (instancetype) init {
[self release];
NSAssert(false,#"unavailable, use initWithBlah: instead");
return nil;
}
raise:format:
Use raise:format: to throw your own exception:
- (instancetype) init {
[self release];
[NSException raise:NSGenericException
format:#"Disabled. Use +[[%# alloc] %#] instead",
NSStringFromClass([self class]),
NSStringFromSelector(#selector(initWithStateDictionary:))];
return nil;
}
[self release] is needed because the object was already allocated. When using ARC the compiler will call it for you. In any case, not something to worry when you are about to intentionally stop execution.
objc_designated_initializer
In case you intend to disable init to force the use of a designated initializer, there is an attribute for that:
-(instancetype)myOwnInit NS_DESIGNATED_INITIALIZER;
This generates a warning unless any other initializer method calls myOwnInit internally. Details will be published in Adopting Modern Objective-C after next Xcode release (I guess).

Apple has started using the following in their header files to disable the init constructor:
- (instancetype)init NS_UNAVAILABLE;
This correctly displays as a compiler error in Xcode. Specifically, this is set in several of their HealthKit header files (HKUnit is one of them).

Objective-C, like Smalltalk, has no concept of "private" versus "public" methods. Any message can be sent to any object at any time.
What you can do is throw an NSInternalInconsistencyException if your -init method is invoked:
- (id)init {
[self release];
#throw [NSException exceptionWithName:NSInternalInconsistencyException
reason:#"-init is not a valid initializer for the class Foo"
userInfo:nil];
return nil;
}
The other alternative — which is probably far better in practice — is to make -init do something sensible for your class if at all possible.
If you're trying to do this because you're trying to "ensure" a singleton object is used, don't bother. Specifically, don't bother with the "override +allocWithZone:, -init, -retain, -release" method of creating singletons. It's virtually always unnecessary and is just adding complication for no real significant advantage.
Instead, just write your code such that your +sharedWhatever method is how you access a singleton, and document that as the way to get the singleton instance in your header. That should be all you need in the vast majority of cases.

You can declare any method to be not available using NS_UNAVAILABLE.
So you can put these lines below your #interface
- (instancetype)init NS_UNAVAILABLE;
+ (instancetype)new NS_UNAVAILABLE;
Even better define a macro in your prefix header
#define NO_INIT \
- (instancetype)init NS_UNAVAILABLE; \
+ (instancetype)new NS_UNAVAILABLE;
and
#interface YourClass : NSObject
NO_INIT
// Your properties and messages
#end

That depends on what you mean by "make private". In Objective-C, calling a method on an object might better be described as sending a message to that object. There's nothing in the language that prohibits a client from calling any given method on an object; the best you can do is not declare the method in the header file. If a client nevertheless calls the "private" method with the right signature, it will still execute at runtime.
That said, the most common way to create a private method in Objective-C is to create a Category in the implementation file, and declare all of the "hidden" methods in there. Remember that this won't truly prevent calls to init from running, but the compiler will spit out warnings if anyone tries to do this.
MyClass.m
#interface MyClass (PrivateMethods)
- (NSString*) init;
#end
#implementation MyClass
- (NSString*) init
{
// code...
}
#end
There's a decent thread on MacRumors.com about this topic.

If you are talking about the default -init method then you can't. It's inherited from NSObject and every class will respond to it with no warnings.
You could create a new method, say -initMyClass, and put it in a private category like Matt suggests. Then define the default -init method to either raise an exception if it's called or (better) call your private -initMyClass with some default values.
One of the main reasons people seem to want to hide init is for singleton objects. If that's the case then you don't need to hide -init, just return the singleton object instead (or create it if it doesn't exist yet).

Put this in header file
- (id)init UNAVAILABLE_ATTRIBUTE;

well the problem why you can't make it "private/invisible" is cause the init method gets send to id (as alloc returns an id) not to YourClass
Note that from the point of the compiler (checker) an id could potencialy respond to anything ever typed (it can't check what really goes into the id at runtime), so you could hide init only when nothing nowhere would (publicly = in header) use a method init, than the compile would know, that there is no way for id to respond to init, since there is no init anywhere (in your source, all libs etc...)
so you cannot forbid the user to pass init and get smashed by the compiler... but what you can do, is to prevent the user from getting a real instance by calling a init
simply by implementing init, which returns nil and have an (private / invisible) initializer which name somebody else won't get (like initOnce, initWithSpecial ...)
static SomeClass * SInstance = nil;
- (id)init
{
// possibly throw smth. here
return nil;
}
- (id)initOnce
{
self = [super init];
if (self) {
return self;
}
return nil;
}
+ (SomeClass *) shared
{
if (nil == SInstance) {
SInstance = [[SomeClass alloc] initOnce];
}
return SInstance;
}
Note : that somebody could do this
SomeClass * c = [[SomeClass alloc] initOnce];
and it would in fact return a new instance, but if the initOnce would nowhere in our project be publicly (in header) declared, it would generate a warning (id might not respond ...) and anyway the person using this, would need to know exactly that the real initializer is the initOnce
we could prevent this even further, but there is no need

I have to mention that placing assertions and raising exceptions to hide methods in the subclass has a nasty trap for the well-intended.
I would recommend using __unavailable as Jano explained for his first example.
Methods can be overridden in subclasses. This means that if a method in the superclass uses a method that just raises an exception in the subclass, it probably won't work as intended. In other words, you've just broken what used to work. This is true with initialization methods as well. Here is an example of such rather common implementation:
- (SuperClass *)initWithParameters:(Type1 *)arg1 optional:(Type2 *)arg2
{
...bla bla...
return self;
}
- (SuperClass *)initWithLessParameters:(Type1 *)arg1
{
self = [self initWithParameters:arg1 optional:DEFAULT_ARG2];
return self;
}
Imagine what happens to -initWithLessParameters, if I do this in the subclass:
- (SubClass *)initWithParameters:(Type1 *)arg1 optional:(Type2 *)arg2
{
[self release];
[super doesNotRecognizeSelector:_cmd];
return nil;
}
This implies that you should tend to use private (hidden) methods, especially in initialization methods, unless you plan to have the methods overridden. But, this is another topic, since you don't always have full control in the implementation of the superclass. (This makes me question the use of __attribute((objc_designated_initializer)) as bad practice, although I haven't used it in depth.)
It also implies that you can use assertions and exceptions in methods that must be overridden in subclasses. (The "abstract" methods as in Creating an abstract class in Objective-C )
And, don't forget about the +new class method.

Related

Can NSUUID be extended by inheritance? How?

Recently (reviewing some code) I stumbled upon an oddity that results in a bug in our program.
An API we are using has the following implementation (that I am going to write in Swift, even though the original code is in Objective-C)
internal class MyUUID: NSUUID { }
Which is completely useless as it always returns an empty instance.
I am going to paste the code from my playground here for explanation purposes.
For example: creating a simple NSUUID would be something like this:
let a = NSUUID()
a.description //this creates a valid uuid
While creating a MyUUID should be similar
let b = MyUUID()
b.description //it returns an instance, but is completely empty.
But it doesn't work.
Inspecting a little bit more, reveals the NSUUID initialiser creates a __NSConcreteUUID instance, while MyUUID doesn't and it doesn't matter what I try to do, it won't create an appropriate UUID.
So, my question: Is it possible to be able to create a child implementation of NSUUID?
Your evidence would appear empirically to answer your own question: it's not possible. NSUUID would appear to be a class cluster rather than a single class, which effectively prevents subclassing.
An alternative idea to Aaron's:
Implement an object that has an NSUUID rather than that is one. Implement -forwardingTargetForSelector: and return your instance of NSUUID. Consider overriding -isKindOfClass:, but ideally don't unless you have to. Then you should be able to pass your class as though it were an NSUUID to anyone that expects one without their knowing the difference.
Given that the solution depends upon the fallback mechanism built into dynamic messaging, I suspect there's no Swift equivalent; however if you define your class as Objective-C then it should be equally usable from Swift.
You could use class_setSuperclass to change the superclass of MyUUID at runtime. This approach would be illegal in Swift, due to type safety, but you could still do it in Objective-C.
Depending on your actual goals you may be able to use CFUUIDRef instead.
As requested, here's an example of the class_setSuperclass approach. Just drop this in to a new single view project.
#import <objc/runtime.h>
#interface MyUUID : NSUUID
- (void) UUIDWithHello;
#end
#implementation MyUUID
- (void) UUIDWithHello {
NSLog(#"Hello! %#", self.UUIDString);
}
#end
#interface ViewController ()
#end
#implementation ViewController
- (void)viewDidLoad {
[super viewDidLoad];
// Do any additional setup after loading the view, typically from a nib.
// Make a UUID that you want to subclass
NSUUID *uuid = [[NSUUID alloc] init];
NSLog(#"Initial UUID: %#", uuid.UUIDString);
// Ignore deprecation warnings, since class_setSuperclass is deprecated
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wdeprecated-declarations"
// Change MyUUID to inherit from the NSUUID's hidden subclass instead of NSUUID
class_setSuperclass([MyUUID class], [uuid class]); // [uuid class] is __NSConcreteUUID
// Turn deprecation warnings back on
#pragma GCC diagnostic pop
// Make a new myUUID and print it
MyUUID *myUuid = [[MyUUID alloc] init];
[myUuid UUIDWithHello];
}
#end
Note that this is a bit dangerous. If whatever secret subclass NSUUID has additional instance variables, it will require more memory, which [MyUUID alloc] won't request. This could cause a crash later when something requests these instance variables.
To get around this, you could instead instantiate your MyUUID instance like this:
NSLog(#"Initial UUID's class: %#", NSStringFromClass(uuid.class));
Class topSecretUUIDSubclass = uuid.class; // __NSConcreteUUID
MyUUID *myUuid2 = [[topSecretUUIDSubclass alloc] init];
[myUuid2 UUIDWithHello];
object_setClass(myUuid2, [MyUUID class]);
Basically this will make myUuid2 a __NSConcreteUUID and then change it to a MyUUID. However, this will only work if MyUUID doesn't add any instance variables.
If MyUUID does need to add its own instance variables, it will need to override +alloc to provide additional memory for these instance variables, using class_createInstance().

Accessing NSObject's (superclass) version of allocWithZone while bypassing overriding (subclass) version of allocWithZone

In iOS Programming Book from Big Nerd Ranch (3rd ed) they say on pg.194
..a knowledgeable programmer could still create an instance of BNRItemStore via allocWithZone:, which would bypass our sneaky alloc trap.To prevent this possibility, override allocWithZone: in BNRItemStore.m to return the single BNRItemStore instance.
+(id) allocWithZone:(NSZone *)zone
{
return [self sharedStore];
}
This statement seems confusing to me. Doesn't this following code not prove this wrong in a way-
#import <Foundation/Foundation.h>
#interface BNRItemStore : NSObject
+(BNRItemStore *)sharedStore;
+(id)retrieveObject;
#end
#implementation BNRItemStore
+(BNRItemStore *)sharedStore{
static BNRItemStore *sharedStore=nil;
if (!sharedStore){
NSLog(#"Test2");
sharedStore= [[super allocWithZone:nil] init];
}
NSLog(#"sharedStore-> %#",sharedStore);
return sharedStore;
}
+(id)allocWithZone:(NSZone *)zone{
NSLog(#"Test1");
return [self sharedStore];
}
+(id)alloc{
NSLog(#"Retrieving super object");
NSLog(#"%#", [super allocWithZone:nil]);//Bypassing the subclass version of allocWithZone.
return [super allocWithZone:nil];
}
#end
int main(){
[[BNRItemStore alloc] init]; //the alloc message triggers a call to the subclass (overriding) version of +(id)alloc method
}
The output is:
2013-10-18 18:24:40.132 BNRItemStore[381:707] Retrieving super object
2013-10-18 18:24:40.134 BNRItemStore[381:707] BNRItemStore:0x7f8c72c091e0
If the call [super allocWithZone:nil] inside of subclass 'alloc' method would have triggered a call to subclass allocWithZone,the console would be logging "Test1" and "Test2" and finally would lead to static pointer getting allocated. But this did not happen.
This means that if we directly call [NSObject allocWithZone:nil] or [super allocWithZone:nil], the message would not redirect to the overriding version (subclass version) of allocWithZone but will give direct access to NSAllocateObject() function which does the actual allocation.
The code of +(id)allocWithZone in NSObject must look somewhat like this-
+(id)allocWithZone:(NSZone *)zone{
return NSAllocateObject();
}
Had this implementation(NSObject's allocWithZone:) included something like [self allocWithZone], the message dispatch mechanism would have included the subclass version of allocWithZone which would then make us go through the "sneaky" trap involving a call to sharedStore method.Following is the case that I'm talking about. Now if this were the case the code would definitely have infinite-looped.Clearly this isn't the case.
+(id)allocWithZone:(NSZone *)zone{
if([self allocWithZone:zone]) //this would trigger a call to subclass ver. which would call sharedStore method which would then have [super allocWithZone:nil].Infinite Loop
return NSAllocateObject();
}
So can someone clear up this query about this so called "sneaky" trap. Was the trap meant for blocking anyone from instantiating separately .i.e not being able to use NSObject's allocWithZone except when inside of sharedStore method ? Pls clarify..
The first, most important lesson here is that you should not override +allocWithZone:. I know the BNR book describes it (and the BNR book is generally very good). You shouldn't do it. I know that Apple includes some example code that does it. You shouldn't do it. (And Apple notes in the explanation that it is rare to need this.) Singletons should be created with the dispatch_once pattern.
You don't give the initial code, but I suspect that their example code overrides alloc, but not allocWithZone:. They're simply saying that if the caller uses allocWithZone:, it won't go through alloc, so they've also overridden alloc to catch that. (Of course the right answer would be just to override allocWithZone: and not alloc. But you shouldn't be overriding these methods in any case.)
EDIT:
I believe you are misunderstanding what "our sneaky alloc trap" means here. The author is assuming the following code at this point in the text:
#interface BNRItemStore : NSObject
+(BNRItemStore *)sharedStore;
#end
#implementation BNRItemStore
+(BNRItemStore *)sharedStore{
static BNRItemStore *sharedStore=nil;
if (!sharedStore){
sharedStore = [[super allocWithZone:nil] init];
}
return sharedStore;
}
#end
That's it; no +alloc overrides at all. It then points out "to enforce the singleton status…you must ensure that another instance of BNRItemStore cannot be allocated." (*)
The author goes on to suggest that we might enforce the singleton status by overriding +alloc, but immediately notes that this is insufficient, since the caller can use +allocWithZone: instead. Since it is documented that [NSObject alloc] calls [self allocWithZone:], it is necessary and sufficient to override +allocWithZone: and unnecessary and insufficient to override +alloc.
What you've done in your code is demonstrate that you can modify BNRItemStore to call [super allocWithZone:] in +alloc. That is not the point. If you can modify BNRItemStore, you could also make it a non-singleton. The point is whether an outside caller (main() in your case) can bypass the singleton instantiation, which she cannot. (**)
(*) The point it doesn't make at this point, and probably should, is that it is generally a bad idea to "enforce the singleton status" by quietly returning a singleton when the callers asked you to allocate a new object. If you need to enforce the singleton status, it is better IMO to do so with an assertion in init, since the request for a second allocation represents a programming error. That said, there are times when "transparent" singletons of immutable objects can be useful for performance reasons, such as the special singletons NSNumber provides for certain common integers, and this technique is appropriate in those cases. (By "transparent," I mean that the singleton-ness is an implementation detail that the caller should never worry about. This presumes at a minimum that the object is immutable.)
(**) Actually she can if she is determined to do so. She could always call NSAllocateObject() herself, bypassing +alloc entirely, and then call -init. This would of course be insane, and there is no reason to "protect" her from herself in doing this. It is not the job of an SDK to protect itself from the caller. It is only the job of an SDK to protect a caller from likely mistakes. The caller is never the enemy.
i'm not sure if this quite answers your question or not, but "allocWithZone:" was used back in the day to be able to partition the memory allocated. apple has since moved away from this concept and expects everything to be allocated in the same heap space. "allocWithZone:" does not even function the way it used to, and apple specifically says not to use it.

Release method not recognized at runtime (non Cocoa)

I'm a newbie to Objective-C but have extensive experience in C and C++. The first thing I have noticed is that there is a real void in basic tutorials out there as all assume you are developing for the iPhone or Mac and using Cocoa. I'm not using Cocoa or Gnustep. To the point:
As a simple example to get started I'm trying to wrap the C's File I/O functionality. My code starts as
File.h
#include <objc/Object.h>
#include <stdio.h>
#interface File:Object
{
FILE *pFile;
char *path;
}
#property FILE *pFile;
#property char *path;
- (void)new;
- (void)OpenReadText:(const char*)var1;
- (void)release;
#end
And File.m
#include "File.h"
#implementation File
#synthesize pFile, path;
- (void)new
{
self = [super init];
}
- (void)release
{
fclose(pFile);
[super release];
}
- (void)OpenReadText:(char*)var1
{
path = var1;
pFile = fopen(path,"r");
}
#end
Then main.m
#include <stdio.h>
#import <objc/Object.h>
#include "File.h"
int main(void) {
File *Fileobj = [File new];
[Fileobj OpenReadText:"File.h"];
[Fileobj release];
}
The compiler gives me a warning that my object "may not respond to '-release'". Then when running the program is results in a runtime error: "does not recognize release. This application has requested the Runtime to terminate" .. and so on.
I'm guessing I'm making a simple newbie error, but where? Or perhaps there is something missing? I'm hoping someone can point me in the right direction here. Thanks.
If this qst has been asked already then a reference would do too. I did try to find a reference but no luck.
FOLLOW UP:
changed release method to
- (void)release
{
fclose(pFile);
[super free];
}
and it appeared to work. Apparently free is recognized in object.h.
As others have said it is unusual to use Objective-C without the Foundation frameworks. However, the Object class should implement release, retain etc. The Object class included (but not used) in Apple's Objective-C Runtime certainly contains these basic methods.
Assuming your Object class does contain these basic methods there are a couple of problems with your class as implemented.
First, you have created a new instance method which simply calls [super init]. The new method by convention is a class method which is shorthand for calling alloc and init to create and initialise an object. new is defined in Apple's Object class. It is implemented as:
+ (id)new
{
id newObject = (*_alloc)((Class)self, 0);
Class metaClass = self->isa;
if (class_getVersion(metaClass) > 1)
return [newObject init];
else
return newObject;
}
Note that this method is a class method, signified by the + instead of the -. GNUStep implements new as follows:
+ new
{
return [[self alloc] init];
}
The idiomatic way to use new would be:
File *obj = [File new];
This is in fact what you have done, however, this is calling the class method new not your instance method new.
If you wanted to call your new method you'd have to call:
File *obj = [[File alloc] new];
but as others have stated you'd need to return your object. Removing your new method would have no effect on your implementation as it isn't currently being called.
Secondly, you have placed your call to fclose in your overriden release method. This is wrong, certainly in Apple's implementation of Object anyway, GNUstep appears to be different. release could get called multiple times on a single instance of an object. retain and release are used to increment/decrement the objects retain count. Only when the retain count reaches zero should the file handle be closed. Normally, within Foundation you'd place the call to fclose in a dealloc method. dealloc is Objective-C's destructor method. The dealloc should look something like:
- (void)dealloc
{
fclose(pFile);
[super dealloc];
}
However, dealloc doesn't appear to be implemented in either Apple's or GNUstep's Object class. There is, as you point out in your question a free method which seems to be a destructor.
It would appear that replacing the above dealloc method with an equivalent free method would work as a destructor, e.g.:
- (void)free
{
fclose(pFile);
[super free];
}
Apple's implementation of Object contains retain and release methods but the GNUstep implementation does not. Neither implementation contains a dealloc method.
The implementations of Object.m and NSObject.m for Apple and GNUstep can be found at the following locations:
Apple Object.m: http://opensource.apple.com/source/objc4/objc4-532.2/runtime/Object.m
GNUstep Object.m: https://github.com/gnustep/gnustep-libobjc/blob/master/Object.m
Apple NSObject.mm: http://opensource.apple.com/source/objc4/objc4-532.2/runtime/NSObject.mm
GNUstep NSObject.m: https://github.com/gnustep/gnustep-base/blob/master/Source/NSObject.m
Is release defined on class Object? If it is not, then your call to
[super release];
will not work. (In cocoa, release is a member of NSObject; your Object class may or may not have it, and in fact the retain/release reference counting might not be there at all.)
You should confirm that your base class includes all methods called via super.
As #xlc0212 pointed out, the reference counting style of memory management is included in NSObject.
NSObject is a part of CoreFoundation library for Cocoa, CocoaTouch and GnuStep. I would say you need to link to CoreFoundation.
One book that I've read and focuses on pure Objective-C (not necessarily Cocoa) is "Programming in Objective-C 2.0" by Steven G Kochan.

Objective C - How do I inherit from another class?

- (id) init
{
[super init];
//initialitation of the class
return self;
}
I know that when I am inheriting from another class I am suppose to call super.init
Does this apply to "inheriting from NSObject" ?
Yes, usually you have something like:
- (id) init
{
if (self = [super init]) {
// instantiation code
}
return self;
}
Technically, yes, because Apple's documentation says init... methods should always include a call to super. However at present, the NSObject implementation of -init does nothing, so omitting the call wouldn't prevent your code from working.
The downside of omitting the call to super is that your code wouldn't be as robust to future changes; for example if you later changed the inheritance, or if (God forbid) Apple changed NSObject's -init method so that it actually did something essential.

Creating an abstract class in Objective-C

I'm originally a Java programmer who now works with Objective-C. I'd like to create an abstract class, but that doesn't appear to be possible in Objective-C. Is this possible?
If not, how close to an abstract class can I get in Objective-C?
Typically, Objective-C class are abstract by convention only—if the author documents a class as abstract, just don't use it without subclassing it. There is no compile-time enforcement that prevents instantiation of an abstract class, however. In fact, there is nothing to stop a user from providing implementations of abstract methods via a category (i.e. at runtime). You can force a user to at least override certain methods by raising an exception in those methods implementation in your abstract class:
[NSException raise:NSInternalInconsistencyException
format:#"You must override %# in a subclass", NSStringFromSelector(_cmd)];
If your method returns a value, it's a bit easier to use
#throw [NSException exceptionWithName:NSInternalInconsistencyException
reason:[NSString stringWithFormat:#"You must override %# in a subclass", NSStringFromSelector(_cmd)]
userInfo:nil];
as then you don't need to add a return statement from the method.
If the abstract class is really an interface (i.e. has no concrete method implementations), using an Objective-C protocol is the more appropriate option.
No, there is no way to create an abstract class in Objective-C.
You can mock an abstract class - by making the methods/ selectors call doesNotRecognizeSelector: and therefore raise an exception making the class unusable.
For example:
- (id)someMethod:(SomeObject*)blah
{
[self doesNotRecognizeSelector:_cmd];
return nil;
}
You can also do this for init.
Just riffing on #Barry Wark's answer above (and updating for iOS 4.3) and leaving this for my own reference:
#define mustOverride() #throw [NSException exceptionWithName:NSInvalidArgumentException reason:[NSString stringWithFormat:#"%s must be overridden in a subclass/category", __PRETTY_FUNCTION__] userInfo:nil]
#define methodNotImplemented() mustOverride()
then in your methods you can use this
- (void) someMethod {
mustOverride(); // or methodNotImplemented(), same thing
}
Notes: Not sure if making a macro look like a C function is a good idea or not, but I'll keep it until schooled to the contrary. I think it's more correct to use NSInvalidArgumentException (rather than NSInternalInconsistencyException) since that's what the runtime system throws in response to doesNotRecognizeSelector being called (see NSObject docs).
The solution I came up with is:
Create a protocol for everything you want in your "abstract" class
Create a base class (or maybe call it abstract) that implements the protocol. For all the methods you want "abstract" implement them in the .m file, but not the .h file.
Have your child class inherit from the base class AND implement the protocol.
This way the compiler will give you a warning for any method in the protocol that isn't implemented by your child class.
It's not as succinct as in Java, but you do get the desired compiler warning.
From the Omni Group mailing list:
Objective-C doesn't have the abstract compiler construct like Java at
this time.
So all you do is define the abstract class as any other normal class
and implement methods stubs for the abstract methods that either are
empty or report non-support for selector. For example...
- (id)someMethod:(SomeObject*)blah
{
[self doesNotRecognizeSelector:_cmd];
return nil;
}
I also do the following to prevent the initialization of the abstract
class via the default initializer.
- (id)init
{
[self doesNotRecognizeSelector:_cmd];
[self release];
return nil;
}
Instead of trying to create an abstract base class, consider using a protocol (similar to a Java interface). This allows you to define a set of methods, and then accept all objects that conform to the protocol and implement the methods. For example, I can define an Operation protocol, and then have a function like this:
- (void)performOperation:(id<Operation>)op
{
// do something with operation
}
Where op can be any object implementing the Operation protocol.
If you need your abstract base class to do more than simply define methods, you can create a regular Objective-C class and prevent it from being instantiated. Just override the - (id)init function and make it return nil or assert(false). It's not a very clean solution, but since Objective-C is fully dynamic, there's really no direct equivalent to an abstract base class.
This thread is kind of old, and most of what I want to share is already here.
However, my favorite method is not mentioned, and AFAIK there’s no native support in the current Clang, so here I go…
First, and foremost (as others have pointed out already) abstract classes are something very uncommon in Objective-C — we usually use composition (sometimes through delegation) instead. This is probably the reason why such a feature doesn’t already exist in the language/compiler — apart from #dynamic properties, which IIRC have been added in ObjC 2.0 accompanying the introduction of CoreData.
But given that (after careful assessment of your situation!) you have come to the conclusion that delegation (or composition in general) isn’t well suited to solving your problem, here’s how I do it:
Implement every abstract method in the base class.
Make that implementation [self doesNotRecognizeSelector:_cmd];…
…followed by __builtin_unreachable(); to silence the warning you’ll get for non-void methods, telling you “control reached end of non-void function without a return”.
Either combine steps 2. and 3. in a macro, or annotate -[NSObject doesNotRecognizeSelector:] using __attribute__((__noreturn__)) in a category without implementation so as not to replace the original implementation of that method, and include the header for that category in your project’s PCH.
I personally prefer the macro version as that allows me to reduce the boilerplate as much as possible.
Here it is:
// Definition:
#define D12_ABSTRACT_METHOD {\
[self doesNotRecognizeSelector:_cmd]; \
__builtin_unreachable(); \
}
// Usage (assuming we were Apple, implementing the abstract base class NSString):
#implementation NSString
#pragma mark - Abstract Primitives
- (unichar)characterAtIndex:(NSUInteger)index D12_ABSTRACT_METHOD
- (NSUInteger)length D12_ABSTRACT_METHOD
- (void)getCharacters:(unichar *)buffer range:(NSRange)aRange D12_ABSTRACT_METHOD
#pragma mark - Concrete Methods
- (NSString *)substringWithRange:(NSRange)aRange
{
if (aRange.location + aRange.length >= [self length])
[NSException raise:NSInvalidArgumentException format:#"Range %# exceeds the length of %# (%lu)", NSStringFromRange(aRange), [super description], (unsigned long)[self length]];
unichar *buffer = (unichar *)malloc(aRange.length * sizeof(unichar));
[self getCharacters:buffer range:aRange];
return [[[NSString alloc] initWithCharactersNoCopy:buffer length:aRange.length freeWhenDone:YES] autorelease];
}
// and so forth…
#end
As you can see, the macro provides the full implementation of the abstract methods, reducing the necessary amount of boilerplate to an absolute minimum.
An even better option would be to lobby the Clang team to providing a compiler attribute for this case, via feature requests. (Better, because this would also enable compile-time diagnostics for those scenarios where you subclass e.g. NSIncrementalStore.)
Why I Choose This Method
It get’s the job done efficiently, and somewhat conveniently.
It’s fairly easy to understand. (Okay, that __builtin_unreachable() may surprise people, but it’s easy enough to understand, too.)
It cannot be stripped in release builds without generating other compiler warnings, or errors — unlike an approach that’s based on one of the assertion macros.
That last point needs some explanation, I guess:
Some (most?) people strip assertions in release builds. (I disagree with that habit, but that’s another story…) Failing to implement a required method — however — is bad, terrible, wrong, and basically the end of the universe for your program. Your program cannot work correctly in this regard because it is undefined, and undefined behavior is the worst thing ever. Hence, being able to strip those diagnostics without generating new diagnostics would be completely unacceptable.
It’s bad enough that you cannot obtain proper compile-time diagnostics for such programmer errors, and have to resort to at-run-time discovery for these, but if you can plaster over it in release builds, why try having an abstract class in the first place?
Using #property and #dynamic could also work. If you declare a dynamic property and don't give a matching method implementation, everything will still compile without warnings, and you'll get an unrecognized selector error at runtime if you try to access it. This essentially the same thing as calling [self doesNotRecognizeSelector:_cmd], but with far less typing.
In Xcode (using clang etc) I like to use __attribute__((unavailable(...))) to tag the abstract classes so you get an error/warning if you try and use it.
It provides some protection against accidentally using the method.
Example
In the base class #interface tag the "abstract" methods:
- (void)myAbstractMethod:(id)param1 __attribute__((unavailable("You should always override this")));
Taking this one-step further, I create a macro:
#define UnavailableMacro(msg) __attribute__((unavailable(msg)))
This lets you do this:
- (void)myAbstractMethod:(id)param1 UnavailableMacro(#"You should always override this");
Like I said, this is not real compiler protection but it's about as good as your going to get in a language that doesn't support abstract methods.
The answer to the question is scattered around in the comments under the already given answers. So, I am just summarising and simplifying here.
Option1: Protocols
If you want to create an abstract class with no implementation use 'Protocols'. The classes inheriting a protocol are obliged to implement the methods in the protocol.
#protocol ProtocolName
// list of methods and properties
#end
Option2: Template Method Pattern
If you want to create an abstract class with partial implementation like "Template Method Pattern" then this is the solution.
Objective-C - Template methods pattern?
Another alternative
Just check the class in the Abstract class and Assert or Exception, whatever you fancy.
#implementation Orange
- (instancetype)init
{
self = [super init];
NSAssert([self class] != [Orange class], #"This is an abstract class");
if (self) {
}
return self;
}
#end
This removes the necessity to override init
(more of a related suggestion)
I wanted to have a way of letting the programmer know "do not call from child" and to override completely (in my case still offer some default functionality on behalf of the parent when not extended):
typedef void override_void;
typedef id override_id;
#implementation myBaseClass
// some limited default behavior (undesired by subclasses)
- (override_void) doSomething;
- (override_id) makeSomeObject;
// some internally required default behavior
- (void) doesSomethingImportant;
#end
The advantage is that the programmer will SEE the "override" in the declaration and will know they should not be calling [super ..].
Granted, it is ugly having to define individual return types for this, but it serves as a good enough visual hint and you can easily not use the "override_" part in a subclass definition.
Of course a class can still have a default implementation when an extension is optional. But like the other answers say, implement a run-time exception when appropriate, like for abstract (virtual) classes.
It would be nice to have built in compiler hints like this one, even hints for when it is best to pre/post call the super's implement, instead of having to dig through comments/documentation or... assume.
If you are used to the compiler catching abstract instantiation violations in other languages, then the Objective-C behavior is disappointing.
As a late binding language it is clear that Objective-C cannot make static decisions on whether a class truly is abstract or not (you might be adding functions at runtime...), but for typical use cases this seems like a shortcoming. I would prefer the compiler flat-out prevented instantiations of abstract classes instead of throwing an error at runtime.
Here is a pattern we are using to get this type of static checking using a couple of techniques to hide initializers:
//
// Base.h
#define UNAVAILABLE __attribute__((unavailable("Default initializer not available.")));
#protocol MyProtocol <NSObject>
-(void) dependentFunction;
#end
#interface Base : NSObject {
#protected
__weak id<MyProtocol> _protocolHelper; // Weak to prevent retain cycles!
}
- (instancetype) init UNAVAILABLE; // Prevent the user from calling this
- (void) doStuffUsingDependentFunction;
#end
//
// Base.m
#import "Base.h"
// We know that Base has a hidden initializer method.
// Declare it here for readability.
#interface Base (Private)
- (instancetype)initFromDerived;
#end
#implementation Base
- (instancetype)initFromDerived {
// It is unlikely that this becomes incorrect, but assert
// just in case.
NSAssert(![self isMemberOfClass:[Base class]],
#"To be called only from derived classes!");
self = [super init];
return self;
}
- (void) doStuffUsingDependentFunction {
[_protocolHelper dependentFunction]; // Use it
}
#end
//
// Derived.h
#import "Base.h"
#interface Derived : Base
-(instancetype) initDerived; // We cannot use init here :(
#end
//
// Derived.m
#import "Derived.h"
// We know that Base has a hidden initializer method.
// Declare it here.
#interface Base (Private)
- (instancetype) initFromDerived;
#end
// Privately inherit protocol
#interface Derived () <MyProtocol>
#end
#implementation Derived
-(instancetype) initDerived {
self= [super initFromDerived];
if (self) {
self->_protocolHelper= self;
}
return self;
}
// Implement the missing function
-(void)dependentFunction {
}
#end
Probably this kind of situations should only happen at development time, so this might work:
- (id)myMethodWithVar:(id)var {
NSAssert(NO, #"You most override myMethodWithVar:");
return nil;
}
You can use a method proposed by #Yar (with some modification):
#define mustOverride() #throw [NSException exceptionWithName:NSInvalidArgumentException reason:[NSString stringWithFormat:#"%s must be overridden in a subclass/category", __PRETTY_FUNCTION__] userInfo:nil]
#define setMustOverride() NSLog(#"%# - method not implemented", NSStringFromClass([self class])); mustOverride()
Here you will get a message like:
<Date> ProjectName[7921:1967092] <Class where method not implemented> - method not implemented
<Date> ProjectName[7921:1967092] *** Terminating app due to uncaught exception 'NSInvalidArgumentException', reason: '-[<Base class (if inherited or same if not> <Method name>] must be overridden in a subclass/category'
Or assertion:
NSAssert(![self respondsToSelector:#selector(<MethodName>)], #"Not implemented");
In this case you will get:
<Date> ProjectName[7926:1967491] *** Assertion failure in -[<Class Name> <Method name>], /Users/kirill/Documents/Projects/root/<ProjectName> Services/Classes/ViewControllers/YourClass:53
Also you can use protocols and other solutions - but this is one of the simplest ones.
Cocoa doesn’t provide anything called abstract. We can create a class abstract which gets checked only at runtime, and at compile time this is not checked.
I usually just disable the init method in a class that I want to abstract:
- (instancetype)__unavailable init; // This is an abstract class.
This will generate an error at compile time whenever you call init on that class. I then use class methods for everything else.
Objective-C has no built-in way for declaring abstract classes.
Changing a little what #redfood suggested by applying #dotToString's comment, you actually have the solution adopted by Instagram's IGListKit.
Create a protocol for all the methods that make no sense to be defined in the base (abstract) class i.e. they need specific implementations in the children.
Create a base (abstract) class that does not implement this protocol. You can add to this class any other methods that make sense to have a common implementation.
Everywhere in your project, if a child from AbstractClass must be input to or output by some method, type it as AbstractClass<Protocol> instead.
Because AbstractClass does not implement Protocol, the only way to have an AbstractClass<Protocol> instance is by subclassing. As AbstractClass alone can't be used anywhere in the project, it becomes abstract.
Of course, this doesn't prevent unadvised developers from adding new methods referring simply to AbstractClass, which would end up allowing an instance of the (not anymore) abstract class.
Real world example: IGListKit has a base class IGListSectionController which doesn't implement the protocol IGListSectionType, however every method that requires an instance of that class, actually asks for the type IGListSectionController<IGListSectionType>. Therefore there's no way to use an object of type IGListSectionController for anything useful in their framework.
In fact, Objective-C doesn't have abstract classes, but you can use Protocols to achieve the same effect. Here is the sample:
CustomProtocol.h
#import <Foundation/Foundation.h>
#protocol CustomProtocol <NSObject>
#required
- (void)methodA;
#optional
- (void)methodB;
#end
TestProtocol.h
#import <Foundation/Foundation.h>
#import "CustomProtocol.h"
#interface TestProtocol : NSObject <CustomProtocol>
#end
TestProtocol.m
#import "TestProtocol.h"
#implementation TestProtocol
- (void)methodA
{
NSLog(#"methodA...");
}
- (void)methodB
{
NSLog(#"methodB...");
}
#end
A simple example of creating an abstract class
// Declare a protocol
#protocol AbcProtocol <NSObject>
-(void)fnOne;
-(void)fnTwo;
#optional
-(void)fnThree;
#end
// Abstract class
#interface AbstractAbc : NSObject<AbcProtocol>
#end
#implementation AbstractAbc
-(id)init{
self = [super init];
if (self) {
}
return self;
}
-(void)fnOne{
// Code
}
-(void)fnTwo{
// Code
}
#end
// Implementation class
#interface ImpAbc : AbstractAbc
#end
#implementation ImpAbc
-(id)init{
self = [super init];
if (self) {
}
return self;
}
// You may override it
-(void)fnOne{
// Code
}
// You may override it
-(void)fnTwo{
// Code
}
-(void)fnThree{
// Code
}
#end
Can't you just create a delegate?
A delegate is like an abstract base class in the sense that you say what functions need to be defined, but you don't actually define them.
Then whenever you implement your delegate (i.e abstract class) you are warned by the compiler of what optional and mandatory functions you need to define behavior for.
This sounds like an abstract base class to me.