Why is it allowed to declare properties in categories when neither they nor their accessor methods are synthesized? Is there any performance overhead involved?
Is categorisation purely a compiler technique?
I'm trying to understand how categories work. This just explains what to do and what not to do. Are there any sources which go into more detail?
EDIT : I know that I can use associated references. Thats not what I'm asking for. I want to know why are the properties not synthesised? Is there a performance issue or a security issue if the compiler synthesises them? If there is I want to know what and how?
Why is it allowed to declare properties in categories [...] ?
Properties have many aspects (during compile- and runtime).
They always declare one or two accessor methods on the class.
They can change the selector when the compiler transforms dot notation to messages.
In combination with the #synthesize directive (or by default) they can make the compiler synthesize accessor methods and optionally ivars.
They add introspection information to the class which is available during runtime.
Most of these aspects are still useful when declaring properties in categories (or protocols) and synthesizing is not available.
Is categorisation purely a compiler technique?
No. Categories, as properties, have both compile time as well as runtime aspects.
Categories, for example, can be loaded from dynamic libraries at a later time. So there might already be instances of a class that suddenly gets new methods added. That's one of the reasons categories cannot add ivars, because old objects would be missing these ivars and how should the runtime tell if an object has been created before or after the category has been added.
Before you go into categories, please reconsider the concept of properties in Obj-C: A property is something you can write and read to in an abstract sense, using accessors. Usually, there is an instance variable assigned to it, but there is no need to do so.
A property may also be useful e.g., to set a number of different instance variables in a consistent way, or to read from severals variables or do some calulation.
The crucial fact here: there is no need to have an instance variable assigned to a property.
A category serves as an extensiton of an object's behavior, i.e., to extend its set of methods, without changing the data. If you see a property in it abstract sense, then it add accessors, thus it matches the idea of a category.
But if you synthesize it, an instance variable would be generated what contradicts the idea of a category.
Thus, a property in a category makes only sense if you use it in the uncommon, abstract way, and #synthesize is to ease the common way.
You may want to read NSHipster about how to implement properties storage in categories.
Quoting from the article: "Why is this useful? It allows developers to add custom properties to existing classes in categories, which is an otherwise notable shortcoming for Objective-C."
#synthesize informs the compiler to go ahead and provide a default implementation for the setter and the getter.
Said default setters/getters rely on the existence of some kind of storage inside the object.
Categories do not offer any extra storage, so default setters/getters would have no place to store into, or read from.
An alternative is to use:
#dynamic
and then provide your own implementation and own storage for the said properties.
One way is to use associated objects.
Another would be to store into/read from some completely unrelated place, such as some accessible dictionary of NSUserDefaults or ...
In some cases, for read only properties, you can also reconstruct/compute their values at runtime without any need to store them.
Related
We have an app that makes fairly extensive use of TIniFile. In the past we created our own descendant class, let's call it TMyIniFile, that overrides WriteString. We create one instance of this that the entire app uses. That instance is passed all around through properties and parameters, but the type of all of these is still TIniFile, since that is what it was originally. This seems to work, calling our overridden method through polymorphism, even though all the variable types are still TIniFile. This seems to be proper since we descend from TIniFile.
Now we are making some changes where we want to switch TMyIniFile to descend from TMemIniFile instead of TIniFile. Those are both descendants of TCustomIniFile. We'll also probably be overriding some more methods. I'm inclined to leave all the declarations as TIniFile even though technically our class is no longer a descendant of it, just to avoid having to change a lot of source files if I don't need to.
In every tutorial example of polymorphism, the variable is declared as the base class, and an instance is created of the descendant class and assigned to the variable of the base class. So I assume this is the "right" way to do it. What I'm looking at doing now will end up having the variables declared as, what I guess you'd call a "sibling" class, so this "seems wrong". Is this a bad thing to do? Am I asking for trouble, or does polymorphism actually allow for this sort of thing?
TIniFile and TMemIniFile are distinct classes that do not derive from each other, so you simply cannot create a TMemIniFile object and assign it to a TIniFile variable, and vice versa. The compiler won't let you do that. And using a type-cast to force it will be dangerous.
You will just have to update the rest of your code to change all of the TIniFile declarations to TCustomIniFile instead, which is the common ancestor for both classes. That is the "correct" thing to do.
The compiler is your friend - why would you lie to it by using the wrong type ... and if you do lie to it why would you expect it to know what you want it to do?
You should use a base class that you derive from, like TCustomIniFile. I would expect compile issues if you are trying to make assignments which are known at compile time to be wrong.
The different classes have different signatures so the compiler needs to know which class it is using to call the correct method or access the correct property. With virtual methods the different classes setup their own implementation of those methods so that the correct one is called - so using a pointer to a base type when you call the virtual method it calls that method in the derived type because it is in the class vtable.
So if the code does compile, it's very likely that the compiler will not be doing the right thing ...
I just want to ask, we are having one class, in which we have two private data members say:
class Employee{
private int empid;
private String empname;
}
I am declaring them private that means I can use them in Employee class only. So what is the need to create getter setter for both the data members and making them public.
Hope you got my problem.
This is an excellent question. Often you see code examples that make members private but exposing them via a getter/setter pair, without the getter/setter doing anything else than setting the corresponding member.
In my book this is not encapsulation at all. You are no better of than just making the members public. Although a lot of people are uneasy to do that, they would happily provide accessors automatically for all their members.
One reason to do provide accessors is the ability to do input validation. E.g. if you empIds have a checksum, you could enforce it in the setter. Something that is not possible with direct access to the member.
In my opinion it would be better to think about the role this object will play and see how it can achieve that role with a minimum of accessors. Otherwise your code will probably violate the Law of Demeter.
You are absolutely right, that creating a setter/getter or making the fields public both violate encapsulation in the same way, therefore if you do want to encapsulate your private fields, presumably because you are working in an Object-Oriented environment, you do not want to use either of those things.
To your question why there is a need to have create setter/getters: Most projects (consciously or unconsciously) do not base their designs on Object-Orientation. There are other paradigms, where data and function are separated, thus encapsulation plays a minor role if any.
In the Java world it is common to have pure (or very close to pure) data structures (Beans), and Services/Components/EJBs/etc that work on these Beans (have access to all fields basically). Often these architectures split the function part further into topics like Presentation, Business, Persistence (3-tier architectures), or create explicit control procedures that has access to all the relevant fields (like how MVC is often done).
Whether one approach is better than the other would be a subjective discussion probably, but the short answer is: It's a different paradigm usually (not OO), that is why setters/getters get created.
In Apple's Programming with Objective-C the section on Encapsulating Data states that:
You Can Define Instance Variables without Properties
It’s best practice to use a property on an object any time you need to keep track of a value or another object.
In other words they are strongly recommending that you use private properties rather than instance variables for any private object state.
I am wondering why this might be the case? I understand that properties have features such as KVO, and attributes (strong, weak ...), but in many cases I do not need these features and an instance variable will work just fine.
Are there any good reasons why instance variables might not be considered best practice?
Even though right now your private variable might work as a plain variable, you may decide later that some of properties 'goodies' are useful:
observing
atomic accessors
custom accessors
logging on access
access from subclasses
If you only access your variables as properties, you don't add much overhead (except in tight cycles) and leave room for reaping any of those benefits described above.
Basically, properties are useful even if you don't ever plan on making them public.
Of course, there are places where using an instance variable is still 'more natural', for example if you reimplement Foundation classes in a class cluster (such as NSArray), it's expected that your implementation is private and performant, so I don't use properties then.
Also, I don't think you should read too much into the advice. To me it reads more like "If you've only learned 5 minutes ago about properties and instance variables, let's start with using properties first".
People who are new to the language can go quite far without knowing what the instance variables are.
In other words they are strongly recommending that you use private properties rather than instance variables for any private object state.
Where did you read that they are recommending private properties? I think they mean public variables/properties.
And of course using properties instead of public instance variables has a lots of advantages:
encapsulation and custom getters/setters
memory management
KVO
binary compatibility
and so on
But in my opinion using private properties in general has no advantages and it's much easier to use private instance variables. The only reason I can imagine is to make custom getters/setters for such variables in future, but I don't think that it's a "best practice".
The point is underlaying storage abstraction. So simple yet very powerful.
This question is specifically focused around static libraries / frameworks; in other words, code that other people will eventually touch.
I'm fairly well versed in properties, since I started iOS development when iOS 6 was released. I have used hidden properties declared in interface extensions to do all of my "private" property work, including using readonly on public facing properties I don't want others to modify and readwrite within interface extensions.
The important thing is that I do not want other people who are using these static libraries / frameworks to be accessing these properties if I don't allow it, nor writing these properties if I let them read it.
I've known for a while that they could theoretically create their own interface extension and make my readonly properties readwrite themselves, or guess the names of hidden properties.
If I want to prevent this, should I be using ivars with the #private tag with directly declared ivars? Are there potential downfalls to doing it this way? Does it actually get me an additional measure of security, or is it a red herring?
Under ARC the only mode supported by properties and not instance variables is copy - so if you need copy use a property.
If you declare your private instance variables in the #implementation section:
#implementation MyClass
{
// private instance vars
}
then it takes serious effort to access them from outside the class. As you say accessing a "private" property just takes guessing its name - or using the library calls which tell you.
Is it worth it for security? YMMV. But its a good coding practice regardless.
Addendum
As the comment trail shows there has been much discussion over my use of serious effort.
First let's be clear: Objective-C is in the C family of languages, they all allow the programmer to just about anything they choose while staying within the language[*] - these are not the languages of choice if you want strong typing, access restrictions, etc., etc. within your code.
Second, "effort" is not an absolute measure! So maybe I should have chosen the word "obvious" to qualify it rather than "serious". To access a private property just requires the use of a standard method call where the object has type id - there is little clue in the code that the method being called is hidden. To access a private variable requires either an API call (a runtime function or KVC call) or some pointer manipulation - the resultant code looks nothing like a standard variable assignment. So its more obvious.
That said, apart from uses requiring copy, under ARC there is no good reason to use a private property when a private instance variable will do. For a private variable fred compare:
self.fred = 42; // property access, may involve a call (if not optimised out)
_fred = 42; // common way to bypass the accessors and get at the underlying var
fred = 42; // direct access
Take your pick, there is no right answer, but there isn't a wrong one either - this is the realm of opinion (and that is of course an opinion ;-)). I would often pick the last one, private variable - clean & simple. However #RobNapier in his answer prefers the use of properties.
[*] Note: once you consider linking to external code, say written in assembler, all bets are of in any language. At that point you have to look at the "hardware" (real or virtual) and/or "OS" to provide protection.
You should use private ("hidden") properties here. There is no "security" risk. The "attacker" in this scenario is the caller. The caller has complete access to all memory in the process. She can access anything in your framework she wants and there is absolutely nothing you can do to stop that (nor should you). This is true in any language. You can bypass "private:" designations in C++ as well if you know what you're doing. It's all just memory at the end of the day.
It is not your job to protect yourself or your framework from the caller. You both have the same goal: correct program behavior. Your goal is to protect callers from themselves. Make it difficult for them to use your framework incorrectly and easy to use it correctly.
So, you should use the tool that leads to the most correct code. And that tool is properties, and avoiding directly ivar access except in init and dealloc.
Is it ok to create dependencies between categories in Objective C? Also between the categories and their base class?
I know that there should be no difference at runtime, they probably are just merged together at compile-time. For instance, let us say i break down my class B into:
B(base class)
B+categ1
B+categ2
B+categ3
My question is, is it wrong to either:
a) import B+categ2 and B+categ3 in B.m
b) import B+categ1 in B+categ3?
I'm asking both performance-wise and conceptually.
EDIT:
What would you suggest for a single screen app? Categories or Extending that class?
There's nothing deeply wrong with it, but it may suggest an overuse of categories. While they are a powerful tool for certain problems, and definitely can create some conveniences, I generally wouldn't build a complex system on them.
I usually find that overuse of categories is based on too much IS-A thinking rather than HAS-A thinking. In other words, if you're putting a lot of categories on an class to act as a fancy kind of subclassing, you may be better off using composition instead. Rather than adding lots of category methods to NSArray (as an example), you would want an data object that has an NSArray property and provides the interface you want.
But this is just advice if you're overusing categories. There's no fundamental problem with having categories import other categories. However, this claim is incorrect:
I know that there should be no difference at runtime, they probably are just merged together at compile-time.
Categories are resolved at runtime, not compile-time. The one major danger of that is that if two categories implement the same method, then the behavior is undefined. This is why you should never use categories to override methods, only to add them.
Avoid having a base class depend on a category of itself. Exceptions to this rule are made for private categories (use extensions instead) and categories intended to be used by subclasses of the base class. These are specialized exceptions and should not be thought of as a proper general purpose solution.
For categories depending on categories, if you make sure the dependency graph never has a cycle, then you should be fine.
As a final bit of advice, be explicit about dependencies.
// A+c2
#import "A+c1.h" // A(c2) relies on A(c1) declared methods/properties.
…
// A+c3
#import "A+c1.h" // A(c3) relies on A(c1) declared methods/properties.
#import "A+c2.h" // A(c3) relies on A(c2) declared methods/properties.
Even though A(c3) is implicitly including A(c1) by importing "A+c2.h", I still explicitly import it. This will save frustration as code changes in the future.