Java - binary compatibility of abstract class & subclasses - java

In Java, I define an abstract class with both concrete and abstract methods in it, and it has to be subclassed independently by third-party developers. Just to be sure: are there any changes I could make to the abstract class that are source compatible with their classes but not binary compatible? In other words: after they have compiled their subclasses, could I change the abstract class - apart from e.g. adding an abstract method to it or removing a protected method from it that is called by subclasses, which are of course source incompatible - in a way that could force them to recompile their subclasses?

If it isn't too late to change your system, I would suggest that you do that. Overriding is usually not a good way to customize functionality, as it is incredibly fragile. For example, if you later use a method name that your clients have used (which they are now unintentionally automatically overriding), then it is possible that the override will completely break the invariants of your class. A usually better way of providing customization is to give your clients an interface which is limited to just the customized behavior, and then you have a fully concrete class that depends on an instance of this interface, and delegates appropriately to the interface when it needs to use the customized behaviors. This way, your code and your client's code are completely separated, and they won't interfere with each other.

I am assuming that you are using "binary incompatibility" in the technical sense; e.g. where the classloader detects the incompatibility and refuses to load the classes.
Binary incompatibility could also be introduced if you added a visible method and declared it final, and that method collided with the signature of some existing method in a third-party subclass. However, if the method is non-final, the existing method will turn into an override of your (new) method which might cause problems ... but not binary incompatibility.
Likewise, adding new visible fields will result in hiding, may result in confusing behavior and will break object serialization. But this will not result in binary incompatibility.
In general this points to the fact that you need to consider application semantic issues as well as simple binary compatibility. And the Java type system won't help you there.
For completeness, there are a other things that you could do in your code that would break binary compatibility for the 3rd party classes:
reduce the visibility of your abstract class and/or its methods,
change the signatures of other classes used as parameter result and exception types,
change the chain of superclasses that your abstract class extends, or make an incompatible change in those classes, or
change the tree of interfaces that your abstract class implements, or make an incompatible change in those interfaces.

Sure.
You can accidently use a method name that they've used, which is now suddenly overridden, with perhaps dramatically different results.
You can add fields to the class which mess up serialization etc.

Related

Java inheritance, two-step inheritance

I have the main abstract class that is a base for bunch of classes. Some of them does not need all the fields and methods from the main abstract class, so I have created second abstract class and splitted main abstract class into two parts. The main abstract class contains, for example, a, x fields and their getters/setters, the second abstract class inherits from the main and contains additional b, c fields and their getter/setters. There are simple classes that are inheriting from the main class,and more complicated are inheriting from the second class. I want to create objects of each class as instances of the main class. Is it right way to do that? I have to type check and cast when I want to use methods from the second abstract class. It makes my code complicated. How can I solve this problem?
MainAbstractClass ---> SecondAbstractClass ---> MyComplicatedClasses
|
|
V
MySimpleClasses
One of the OO principles is Favor composition over inheritance.
This means that common behavior is not provided through base classes but via Component classes which are passed in via dependency injection (preferably as constructor parameters.
The answer depends on your actual needs.
You can instead choose to store the extended abstract class specific fields in a class that does not implement your base class and make it a member of more complicated classes.
You can choose to keep everything in a single base class and nothing forces you to use all the fields of an interface in every class that implemented your interface.
You can also keep using your approach but since you store the classes as an instance of the base class, it will be hard to read.
I believe that if you think code does not look very good, it is probably not good. However, there is usually no single answer to this kind of design questions and the best solution is relative to your preferences.
I think this need of type cast is a smell of fragile design. Here when we assume MyComplicatedClass ISA KIND OF MainAbstractClass as shown by TJ Crowder then object must behave as MainAbstractClass (meaning it can honor only API of MainAbstractClass). If it expects special treatment as MyComplicatedClass its false commitment and will need Casting. Such casting (by identifying type) goes against OO principles and kills polymorphism. Later this will end up in Ladder of InstanceOf and type casts as in the scenarios rightly pointed out by T.J. Crowder.
I would suggest readdress the design. e.g. though our all user defined type instances ARE KIND OF Object, but we use Object API only for methods defined in Object class. We do not use Object o = new MyClass(). There are occasions in frameworks or like Object.equals() method where type cast is needed as API is defined before even concrete extension is written. But it is not a good idea for such simple complete (without open hooks for extensions) Hierarchies.

Is it mandatory utility class should be final and private constructor?

By making private constructor, we can avoid instantiating class from anywhere outside. and by making class final, no other class can extend it. Why is it necessary for Util class to have private constructor and final class ?
This is not a mandate from a functional point of view or java complication or runtime. However, it's a coding standard accepted by the wider community. Even most static code review tools, like checkstyle, check that such classes have this convention followed.
Why this convention is followed is already explained in other answers and even OP covered that, but I'd like to explain it a little further.
Mostly utility classes are a collection of methods/functions which are independent of an object instance. Those are kind of like aggregate functions as they depend only on parameters for return values and are not associated with class variables of the utility class. So, these functions/methods are mostly kept static. As a result, utility classes are, ideally, classes with only static methods. Therefore, any programmer calling these methods doesn't need to instantiate the class. However, some robo-coders (maybe with less experience or interest) will tend to create the object as they believe they need to before calling its method. To avoid that, we have 3 options:
Keep educating people to not instantiate it. (No sane person can keep doing it.)
Mark the utility class as abstract: Now robo-coders will not create the object. However, reviewers and the wider java community will argue that marking the class as abstract means you want someone to extend it. So, this is also not a good option.
Private constructor: Not protected because it'll allow a child class to instantiate the object.
Now, if someone wants to add a new method for some functionality to the utility class, they don't need to extend it: they can add a new method as each method is independent and has no chance of breaking other functionalities. So, no need to override it. Also, you are not going to instantiate it, so no need to subclass it. Better to mark it final.
In summary, instantiating a utility class (new MyUtilityClass()) does not make sense. Hence the constructors should be private. And you never want to override or extend it, so mark it final.
It's not necessary, but it is convenient. A utility class is just a namespace holder of related functions and is not meant to be instantiated or subclassed. So preventing instantiation and extension sends a correct message to the user of the class.
There is an important distinction between the Java Language, and the Java Runtime.
When the java class is compiled to bytecode, there is no concept of access restriction, public, package, protected, private are equivalent. It is always possible via reflection or bytecode manipulation to invoke the private constructor, so the jvm cannot rely on that ability.
final on the other hand, is something that persists through to the bytecode, and the guarantees it provides can be used by javac to generate more efficient bytecode, and by the jvm to generate more efficient machine instructions.
Most of the optimisations this enabled are no longer relevant, as the jvm now applies the same optimisations to all classes that are monomorphic at runtime—and these were always the most important.
By default this kind of class normally is used to aggregate functions who do different this, in that case we didn't need to create a new object

Groovy's extension-module compared to java's inheritance

Is groovy's extension module feature a hybrid form of java's inheritance feature? Why are the extension-module needs to be declared as static?
Short answer is I think yes. It is a bit difficult to answer clearly, since the inheritance for the extension methods is done completely by the runtime (and the static compiler). As such it has nothing to do with how Java does inheritance.
To answer the second question... They are static, because for situations in which you need state you usually use the meta class. Extension methods are initially thought of as convenience methods or to make the API more Groovy. As such, they are a special form of methods added to the meta class. You can see them as simplified version. But that also means they don't have all the abilities. Implementing extension methods, that can keep local state on a per "self"-object basis (basically what fields/properties would do) is actually difficult to do efficient... but you could always use per instance meta classes for this.
For all extensive purposes they are syntactic sugar so that a static method appears to be more OOP like. There is no inheritance because static methods in Java and Groovy do not participate in inheritance (that is classes do not inherit static methods).
The methods need to be static because the compiler does not know how to instantiate the surrounding class of the extension method.
However I believe there are languages that do allow for methods to be defined outside of the enclosing class and do some sort inheritance but not many do not (I believe CLOS and Dylan do). Also they are many languages that appear to allow methods to be added but the type of "object" is actually changed/hidden to some other type. This is called adhoc polymorphism (e.g. Clojure, Haskell, sort of Golang and sort of Scala) but again has nothing to do with inclusional polymorphism (Java inheritance).
Unfortunately the reference documentation and other docs don't define the semantics of extension methods:
Q. Can they override instance methods?
I tested extension methods via use Category methods and metaClass expando methods. Neither approach overrides instance methods. I didn't test extension modules installed via module descriptor.
Q. Can they be overridden by extension methods on subclasses?
I tested that, too. use methods and metaClass extension methods don't get overridden by extension methods on subclasses.
Q. Can they call inherited super methods?
No, since they're implemented via static methods.
Q. Can they call private methods?
Experiments showed that they can, surprisingly.
Q. Can they access private instance variables?
No, since they're implemented via static methods.
Q. Are they callable from Java methods?
Maybe, if the extension module is on the classpath when compiling the calling code. I didn't test it.
Conclusion: Extension methods are not a form of inheritance. They seem to be a dynamic form of Universal Function Call Syntax (UFCS), that is, when the language can't find a method variable.foo(arguments) it looks for a static extension function foo(variable, arguments) to call. [Please correct my hypothesis if wrong!]
You asked why they're defined as static. That seems to match the semantics: A static function that's not involved in inheritance, although its calling syntax makes it look like a convenient method call.
You can write an extension method like an instance method using the #groovy.lang.Category annotation. That does AST transformations at compile time to turn it into a suitable static method.
Also see Groovy traits. That is a form of (mixin) inheritance.

Check if object is instanceof a protected class

Say I am using a Java library that has the following method
public static SomeInterface foo();
The interface SomeInterface has multiple implementations, some of which are protected within the library's package. One of these implementation is TheProtectedClass
What would be the best way to check if the object returned by foo() is an instance of TheProtectedClass?
My current plan is to create an Utils class that lives within my project but in the same package as the protected class. This Utils can refer to TheProtectedClass since it is in the same package and thus it can check if an object is instanceof TheProtectedClass.
Any other ideas?
EDIT: Some people are asking "why" so here is more context.
I am using jOOQ and in some part of my code, I want to know if the Field instance that I have is an instance of Lower.
Currently, I use field.getName().equals("lower") but this isn't as robust as I'd like it to be.
I realize that since Lower is a protected class, it isn't part of the API and that it can change but I am ok with that.
Class.forName("TheProtectedClass").isAssignableFrom(foo())
although it is a bad idea for many reasons. You're breaking the encapsulation and the abstraction here. If it's package-private, you shouldn't have to concern with it outside. If it's protected, you should explicitly inherit from it and use the API provided by class for this case.
The less obvious but more correct solution is to get an instance of TheProtectedClass, and compare it by
guaranteedTPCInstance.getClass().isAssignableFrom(foo())
, while still being kind of hacky, at least is more portable and OOPy IMO.
As to your idea of creating a class in the same package as TheProtectedClass to avoid being package-private - it's a viable solution, but a) it breaks the basic principle of encapsulation and the programming contract of the TPC class; packaging is done by library/class authors for a reason - to prevent irresponsible data access and using private API or undocumented proprietary methods, b) it's not always possible (and shouldn't be possible in case of properly designed library classes), since those classes can be not only package-private, but final or effectively final (anonymous inner classes etc) - for the reasons described by Bloch in EJ 2nd, "favor composition over inheritance" item, see also Good reasons to prohibit inheritance in Java? Use of final class in Java etc c) you can't do it with some Java library classes, as you can't define your class to be and use e.g. java.lang package. As such, the only "portable" solution is through reflection and through what I described.
tl;dr The fact you can piggyback another package by mimicking its package definition is an obvious C-style deficiency of Java's syntax (allowing programmer to do what he shouldn't be able to normally do; same goes with some specific reflection methods); hacks made this way are neither maintainable nor safe.
NOTE: If you you expect to do something in a internal implementation-dependent and, at the same time, portable and maintainable (e.g. impervious to implementation changes/class name changes etc) way, you're obviously expecting the impossible.
It appears that the best solution is to create a package in your project that has the same package as the package-private class and either expose TheProtectedClass.class as a Class<?> or simply add a simple method that checks if your Object is instanceof TheProtectedClass.
This does not require reflection, it is fast and relatively safe (compilation will break if the package-private class changes name).

Why avoid the final keyword?

In java, is there ever a case for allowing a non-abstract class to be extended?
It always seems to indicate bad code when there are class hierarchies. Do you agree, and why/ why not?
There are certainly times when it makes sense to have non-final concrete classes. However, I agree with Kent - I believe that classes should be final (sealed in C#) by default, and that Java methods should be final by default (as they are in C#).
As Kent says, inheritance requires careful design and documentation - it's very easy to think you can just override a single method, but not know the situations in which that method may be called from the base class as part of the rest of the implementation.
See "How do you design a class for inheritance" for more discussion on this.
I agree with Jon and Kent but, like Scott Myers (in Effective C++), I go much further. I believe that every class should be either abstract, or final. That is, only leaf classes in any hierarchy are really apt for direct instantiation. All other classes (i.e. inner nodes in the inheritance) are “unfinished” and should consequently be abstract.
It simply makes no sense for usual classes to be further extended. If an aspect of the class is worth extending and/or modifying, the cleaner way would be to take that one class and separate it into one abstract base class and one concrete interchangeable implementation.
there a good reasons to keep your code non-final. many frameworks such as hibernate, spring, guice depend sometimes on non-final classes that they extends dynamically at runtime.
for example, hibernate uses proxies for lazy association fetching.
especially when it comes to AOP, you will want your classes non-final, so that the interceptors can attach to it.
see also the question at SO
This question is equally applicable to other platforms such as C# .NET. There are those (myself included) that believe types should be final/sealed by default and need to be explicitly unsealed to allow inheritance.
Extension via inheritance is something that needs careful design and is not as simple as just leaving a type unsealed. Therefore, I think it should be an explicit decision to allow inheritance.
Your best reference here is Item 15 of Joshua Bloch's excellent book "Effective Java", called "Design and document for inheritance or else prohibit it". However the key to whether extension of a class should be allowed is not "is it abstract" but "was it designed with inheritance in mind". There is sometimes a correlation between the two, but it's the second that is important. To take a simple example most of the AWT classes are designed to be extended, even those that are not abstract.
The summary of Bloch's chapter is that interaction of inherited classes with their parents can be surprising and unpredicatable if the ancestor wasn't designed to be inherited from. Classes should therefore come in two kinds a) classes designed to be extended, and with enough documentation to describe how it should be done b) classes marked final. Classes in (a) will often be abstract, but not always. For
I disagree. If hierarchies were bad, there'd be no reason for object oriented languages to exist. If you look at UI widget libraries from Microsoft and Sun, you're certain to find inheritance. Is that all "bad code" by definition? No, of course not.
Inheritance can be abused, but so can any language feature. The trick is to learn how to do things appropriately.
In some cases you want to make sure there's no subclassing, in other cases you want to ensure subclassing (abstract). But there's always a large subset of classes where you as the original author don't care and shouldn't care. It's part of being open/closed. Deciding that something should be closed is also to be done for a reason.
I couldn't disagree more. Class hierarchies make sense for concrete classes when the concrete classes know the possible return types of methods that they have not marked final. For instance, a concrete class may have a subclass hook:
protected SomeType doSomething() {
return null;
}
This doSomething is guarenteed to be either null or a SomeType instance. Say that you have the ability to process the SomeType instance but don't have a use case for using the SomeType instance in the current class, but know that this functionality would be really good to have in subclasses and most everything is concrete. It makes no sense to make the current class an abstract class if it can be used directly with the default of doing nothing with its null value. If you made it an abstract class, then you would have its children in this type of hierarchy:
Abstract base class
Default class (the class that could have been non-abstract, only implements the protected method and nothing else)
Other subclasses.
You thus have an abstract base class that can't be used directly, when the default class may be the most common case. In the other hierarchy, there is one less class, so that the functionality can be used without making an essentially useless default class because abstraction just had to be forced onto the class.
Default class
Other subclasses.
Now, sure, hierarchies can be used and abused, and if things are not documented clearly or classes not well designed, subclasses can run into problems. But these same problems exist with abstract classes as well, you don't get rid of the problem just because you add "abstract" to your class. For instance, if the contract of the "doSomething()" method above required SomeType to have populated x, y and z fields when they were accessed via getters and setters, your subclass would blow up regardless if you used the concrete class that returned null as your base class or an abstract class.
The general rule of thumb for designing a class hierarchy is pretty much a simple questionaire:
Do I need the behavior of my proposed superclass in my subclass? (Y/N)
This is the first question you need to ask yourself. If you don't need the behavior, there's no argument for subclassing.
Do I need the state of my proposed superclass in my subclass? (Y/N)
This is the second question. If the state fits the model of what you need, this may be a canidate for subclassing.
If the subclass was created from the proposed superclass, would it truly be an IS-A relation, or is it just a shortcut to inherit behavior and state?
This is the final question. If it is just a shortcut and you cannot qualify your proposed subclass "as-a" superclass, then inheritance should be avoided. The state and logic can be copied and pasted into the new class with a different root, or delegation can be used.
Only if a class needs the behavior, state and can be considered that the subclass IS-A(n) instance of the superclass should it be considered to inherit from a superclass. Otherwise, other options exist that would be better suited to the purpose, although it may require a little more work up front, it is cleaner in the long run.
There are a few cases where we dont want to allow to change the behavior. For instance, String class, Math.
I don't like inheritance because there's always a better way to do the same thing but when you're making maintenance changes in a huge system sometimes the best way to fix the code with minimum changes is to extend a class a little. Yes, it's usually leads to a bad code but to a working one and without months of rewriting first. So giving a maintenance man as much flexibility as he can handle is a good way to go.

Categories

Resources