Minborg

Minborg
Minborg

Tuesday, October 21, 2014

Protect Your Immutable Object Invariants in More Complex Java Objects

Background

Duke and Spire Locking in Objects
By using Immutable Objects, which are objects that can not be observed to change once they are created, you gain a number of advantages including inherent thread safety, improved run time efficiency and code robustness. But how do you protect your object's invariants? This post shows some of the fundamental schemes that can be used to ensure that your objects remain immutable. The examples in this post requires Java 8 but it is relatively easy to rewrite them in Java 7 or lower.

Immutable objects are used extensively in the open-source project Speedment that I am contributing to. With Speedment we can view database tables as standard Java 8 streams. Check out Speedment on GitHub.

Simple Immutable Objects

In my Previous post, I talked a lot about how one can create immutable objects using the Builder Pattern. In this post I will use a more simple approach to create the objects, because I want to focus on another aspect of the immutable objects.

Consider the following Object:

public class Author {

    private final String name;
    private final int bornYear;

    public Author(final String name, final int bornYear) {
        this.name = name;
        this.bornYear = bornYear;
    }

    public String getName() {
        return name;
    }

    public int getBornYear() {
        return bornYear;
    }

}

The object's invariants are protected by the final declarations (that make it impossible to write code in the class that directly changes the properties of the Object), by the private declarations (that make sure that no other class can gain access to the fields) and by the fact that there are no setters for the object's properties (it would in fact, not be possible to write setters because the fields are declared final). Now is a good time to mention that your objects can, in theory, be modified anyhow, for example using Java Reflection. However, this is considered "cheating" ...


Now if we run the following test program we get exactly what one would expect.

public class Main {

    public static void main(String[] args) {
        final Author author = new Author("William Shakespeare", 1564);

        System.out.println(author.getName() + " was born in " + author.getBornYear());

    }

}

William Shakespeare was born in 1564

N.B. Even though author is declared final, this does not affect how methods can be called on the object itself. It only says that the object reference variable author can be assigned only once.

Complex Immutable Objects

Some objects contain more complex properties such a Maps, Sets, Lists, Collections and the likes. Consider the following Author object with an added property consisting of a List of the author's works.

import java.util.List;

public class Author {

    private final String name;
    private final int bornYear;
    private final List<String> works;

    public Author(final String name, final int bornYear, final List<String> works) {
        this.name = name;
        this.bornYear = bornYear;
        this.works = works;
    }

    public String getName() {
        return name;
    }

    public int getBornYear() {
        return bornYear;
    }

    public List<String> getWorks() {
        return works;
    }

}

If we run the following test program, we expose a problem with the "immutable" object that really makes it mutable.

import java.util.List;
import java.util.stream.Collectors;
import java.util.stream.Stream;

public class Main {

    public static void main(String[] args) {
        final List<String> works = Stream.of("Hamlet", "Othello", "Macbeth")
              .collect(Collectors.toList());
        final Author author = new Author("William Shakespeare", 1564, works);
        println(author);

        // NOT GOOD! We can add things to the list after the object is created!
        author.getWorks().add("Harry Potter");
        println(author);
    }

    private static void println(final Author author) {
        System.out.println(author.getName() + " was born in " + author.getBornYear()
                + " and wrote " + author.getWorks().stream().collect(Collectors.joining(", ")));
    }

}

William Shakespeare was born in 1564 and wrote Hamlet, Othello, Macbeth
William Shakespeare was born in 1564 and wrote Hamlet, Othello, Macbeth, Harry Potter

The getWorks() method returns a reference to the same List that we used to construct the Author. Because the original list used during construction was writable, we can now change this List, for example we can add "Harry Potter" to William Shakespeare's list of works! Not good! Back to the drawing board!

UnmodifiableList

By using a static method from the Collections class, we can create a view of an existing List that prevents the underlying List from being modified. This is nice and thus we make a new attempt to fix the problem:

import java.util.Collections;
import java.util.List;

public class Author {

    private final String name;
    private final int bornYear;
    private final List<String> works;

    public Author(final String name, final int bornYear, final List<String> works) {
        this.name = name;
        this.bornYear = bornYear;
        this.works = Collections.unmodifiableList(works);
    }

    public String getName() {
        return name;
    }

    public int getBornYear() {
        return bornYear;
    }

    public List<String> getWorks() {
        return works;
    }

}

And here is the corresponding test program:
import java.util.List;
import java.util.stream.Collectors;
import java.util.stream.Stream;

public class Main {

    public static void main(String[] args) {
        final List<String> works = Stream.of("Hamlet", "Othello", "Macbeth").collect(Collectors.toList());
        final Author author = new Author("William Shakespeare", 1564, works);
        println(author);

        // We failed again because we can modify the works List 
        // and it reflects in the Author after creation
        works.add("Harry Potter");
        println(author);
        
        // This works though!
        author.getWorks().add("Harry Potter 2");
        
    }

    private static void println(final Author author) {
        System.out.println(author.getName() + " was born in " + author.getBornYear()
                + " and wrote " + author.getWorks().stream().collect(Collectors.joining(", ")));
    }

}

William Shakespeare was born in 1564 and wrote Hamlet, Othello, Macbeth
William Shakespeare was born in 1564 and wrote Hamlet, Othello, Macbeth, Harry Potter
Exception in thread "main" java.lang.UnsupportedOperationException
 at java.util.Collections$UnmodifiableCollection.add(Collections.java:1115)
 at com.blogspot.minborgsjavapot.immutables._4unmod2.Main.main(Main.java:20)

Again we fail, because even though we now cannot change the List by the reference returned by the getWorks() method, we can still use the original List, used during construction of the Author object, to change the works list after the immutable is created. This is a clear violation against the definition of an immutable object (remeber, no observable change shall be detected after an immutable object is created).

Defensive Copying

By employing Defensive Copying we can protect the immutable object's more complex invariants as shown in the example below:

import java.util.ArrayList;
import java.util.Collections;
import java.util.List;

public class Author {

    private final String name;
    private final int bornYear;
    private final List<String> works;

    public Author(final String name, final int bornYear, final List<String> works) {
        this.name = name;
        this.bornYear = bornYear;
        // Now we make a new List that is a copy of the provided works list
        this.works = Collections.unmodifiableList(new ArrayList<>(works));
    }

    public String getName() {
        return name;
    }

    public int getBornYear() {
        return bornYear;
    }

    public List<String> getWorks() {
        return works;
    }

}

Finally, we can not change the List of works after object creation. The price is that we need to make a new internal copy of the list that is provided during object creation. This can sometimes be a good thing, since we can select the implementation of the internal List in a way that it can be more efficient than the original List. For example, if the List only consists of one element, one can create a defensive copy using the Collections.singletonList() that creates a specialized implementation of a List with exactly one element, potentially much more efficient than a general List. If the List is empty, one can use the Collections.emptyList() that is even more efficient.

The Collections class

There are several useful methods in the Collections class with respect to protecting immutable objects including unmodifiableCollection(), unmodifiableList(), unmodifiableMap() and unmodifiableSet() and more. Use them in your immutable classes!

A Final Warning

In the examples above, we had a list of Strings and we draw to our mind that the class String, for good reasons, is immutable. But suppose that we had a List of some mutable objects such as StringBuilders or other Lists. Then we would have to make defensive copies of them too, recursively until we finally arrive at an immutable object...

Tip: If you only work with immutable objects within your immutable objects then you are better off...

Try it!

Take the code for a spin using "git clone https://github.com/minborg/javapot.git"

Remember "To Protect and to Serve"






Wednesday, October 15, 2014

New Java 8 "Object Support Mixin Pattern" for overriding Object.equals(), hashCode(), toString() and compareTo()

Preface

You have, with a probability infinitely close to 1, made one or several errors when overriding basic Object methods like equals() and hashCode()! When I discovered a new class off error in one of my client's classes, I started to look in the entire module and discovered that not a single class was correct. Then I checked the entire project and discovered that every class had the same type of error. Panic! Things went from bad to worse when I checked my own Java code written over the last decade. Just one single class was correct in a strict sense. The errors did not surface, but still, they were there, lurking to appear in the future. So far I have seen 49513 faulty classes and one correct making 99.998% of the classes faulty... Read this post to see how you can avoid these errors!

Overview

In this post, I will disclose a new pattern that I have called the "Object Support Mixin Pattern" and that can be used for automatically overriding the Object methods equals(), hashCode() and toString() in a correct way. I will also show how the related Comparable.compareTo() can be implemented in a similar way.

Using this pattern, that relies on Java 8's new default interface methods, it is possible to automatically mix in support methods without scarifying the ability to inherit from another super class. I will show several variants of the pattern that, depending on the circumstances, can be used directly in your own code. This is a long and sometimes complicated post but please read it thoroughly and you will end up being a better programmer!

When you have read this post to the very end, you will understand why the following class will automatically have correct equals(), hashCode() and toString() methods that will consider the bean properties name, email and born:

public class Person extends ReflectionObjectSupport<Person> {

    private final String name;
    private final String email;
    private final int born;

    public Person(String name, String email, int born) {
        this.name = name;
        this.email = email;
        this.born = born;
    }

    public String getName() {
        return name;
    }

    public String getEmail() {
        return email;
    }

    public int getBorn() {
        return born;
    }

 }
The superclass ReflectionObjectSupport will override the equals(), hashCode(), toString() and compareTo() methods and, based on reflection, provide fully automatic methods.

If you have classes that already have inherited from another super class (as you probably know, Java can, for good reasons, only inherit from one super class) you can use the Object Support Mixin Pattern. Below, the same Person class is depicted using the Object Support Mixin Pattern. This class also implements the Comparable.compareTo() method.
public class Person implements Comparable<Person>, ReflectionObjectMixin<Person> {

    private final String name;
    private final String email;
    private final int born;

    public Person(String name, String email, int born) {
        this.name = name;
        this.email = email;
        this.born = born;
    }

    public String getName() {
        return name;
    }

    public String getEmail() {
        return email;
    }

    public int getBorn() {
        return born;
    }

    @Override
    public Comparable[] compareToMembers() {
        return mkComparableArray(getName());
    }

    @Override
    public int hashCode() {
        return _hashCode();
    }

    @Override
    public boolean equals(Object obj) {
        return _equals(obj);
    }

    @Override
    public String toString() {
        return _toString();
    }

    @Override
    public int compareTo(Person o) {
        return _compareTo(o);
    }

}
It is difficult to imagine shorter override methods for the Object methods and no support methods are needed in the class except for compareTo() where we need to make a human decision on how we shall order the object. Note that the class is free to inherit from another class, since it is only implementing interfaces and does not extend any class right now.

The problem

While working for some clients, I have often noticed how here are errors in the fundamental Object methods equals() and hashCode(). More suprisingly on my own part, I have also discovered major errors related to the same methods in code that I have written myself. This seams to be a ubiquitous problem and I have not seen a really good pattern to eliminate this plague. Most solution relies on hand coded methods.

There are several articles on the net describing how to write the equals() and hashCode() methods and both methods are related to each other. In short, it is imperative that these methods fulfill their contract or else the class will fail! Failures can occur, for example, when a faulty object is put into a Map or in other types of Collections that are so commonly used in Java code.

The equals() contract

Let us start with the equals() method. The contract requires the following properties for any non-null object "a":

A) It is reflexive, meaning that a.equals(a)
B) It is symmetric, meaning that if a.equals(b) then b.equals(a)
C) It is transient, meaning that if a.equals(b) and b.equals(c) then a.equals(c)
D) It is consistent, meaning that if a.equals(b) this must always be true unless a and/or b change.
E) a.equals(null) shall always be false

What does this mean in pure English? Let me use an analogy with a "dime" that is a ten-cent coin, one tenth of a United States dollar.

A) unsurprisingly says "a dime is equal to a dime". It is an axiomatic definition that seams reasonable. Otherwise equals() would really not make much sense, would it?

B) says "if a dime is 10 cent, then 10 cent is a dime" which seams reasonable. If a thing is equal to something else, the latter thing must also be equal to the first thing.

C) says "if a dime is 10 cent and 10 cent is 0.1 buck, then a dime is 0.1 buck". This rule allows us to infer equality between several objects. It really says that if there are bunch of equal objects and another object is equal to one of these objects, then that object is also equal to all objects in the bunch.

D) says "if a dime is 10 cent now, it must always remain so (unless we change a dime to say 8 cent or so)". Perhaps politicians can navigate around this rule, but we programmers must not!

E) says that "a dime is never equal to nothing". Let us hope that even politicians will adhere to this rule!

The hashCode() contract

If we look at the contract of the hashCode() we conclude that it must:

  A) be consistent, meaning that a.hashCode() should always return the same integer unless a is changed. This is similar to item D for equals().
  B) if a.equals(b) then a.hashCode() and b.hashCode() must return the same integer value.

A is easy to understand but B is a somewhat confusing statement: if a.equals(b) is false, then a.hashCode() and b.hashCode() may or may not return the same integer. If this was not the case, then the number of integers would quickly run out. So, new Point(100,234).hashCode() might produce the same result as new Color(240, 240, 240).hashCode() for example, even though they certainly are not equal.

The default equals() method inherited from the Object class relies on comparing the references to the objects. If they refer to the same object, then the objects are considered to be equal, otherwise they are not equal. This is how it looks:

public boolean equals(Object obj) {
    return (this == obj);
}

The simplest imaginable hashCode() method would be:
 // Legal but do not use this method!!
public int hashCode() {
    return 0;
}
Albeit legal, this method would lead to very poor hash performance. For a Map, all keys would hash to the same bucket, effectively turning the Map into a List. 0 fulfills A because it is consistently 0 and it also fulfills B because all objects would have the hashCode of 0, regardless if they are equal or not. So those objects that are in fact equal, would undoubtedly have the same hashCode.

The hashCode() inherited from the Object class is better than this and will compute an integer based on the numerical value of the object's reference. This is actually a good starting point. It fulfills the contract of the methods (check this for youself!) and for classes that are generally not comparable, such as Thread and Random, they work like a charm. For a typical beans however, where we want to compare the bean properties to asses equality, the situation is not so good.

Common error types

I have notices four distinct errors commonly made in this area:

i) One common mistake is that a programmer overrides one of the methods and not the other. This will, in almost any situation, break the contract of rule B for the hashCode() method! If the programmer overrides equals() and not hashCode() then a.equals(b) will behave differently than before but hashCode() will behave the same. If the programmer on the other hand overrides hashCode() and not equals() then a.equals(b) will behave differently but the hashCode() will remain the same. Always override hashCode() and equals() at the same time!

ii) Another common error is that you overide both methods but regard different bean properties in the methods. Suppose that, if you have a class similar to the Person class above, you use all the properties: getName(), getEmail() and getAge() but in the hashCode() you only use getName() and getEmail() because you added age later and forgot to add it in the hashCode().

iii) A third more subtile error is that you create a (perhaps anonymous) class that overrides one of the getters. In your equals() and hashCode() you use the member variables directly to compare classes and not the corresponding getters. Now your bean will expose the overridden property for the getter but equals() and hashCode() will use the member variable that is not exposed any more.

iv) Class types are compared using instanceof instead of comparing class objects directly, resulting in an asymmetry for overridden classes visa vi their parent class. This will break B and/or C. For example, consider two classes Person and FemalePerson where the latter extends the former. If Person is using (female instanceof Person) and FemalePerson is using (person instanceof FemalePerson) in their equals() methods, (female instanceof Person) is true while (person instanceof FemalePerson) certainly is false.

although strictly not an error but more of an inconvenience, I would like to add another problem commonly appearing in classes:

v) In an attempt to avoid iv) errors, class types are compared using the getClass() method effectively prohibiting derived classes (such as anonymous or inherited classes) to be equal to their base classes.

Asymmetry

I will elaborate more on type iv) errors because they are a bit more difficult to explain. Suppose that we have a bean class named "A" with a bean property named "value" like this:

public class A {

    private final int val;

    public A(int val) {
        this.val = val;
    }

    @Override
    public int hashCode() {
        int hash = 7;
        hash = 13 * hash + this.getVal();
        return hash;
    }

    @Override
    public boolean equals(Object obj) {
        if (obj == null) {
            return false;
        }
        if (!(obj instanceof A)) {
            return false;
        }
        final A other = (A) obj;
        if (this.getVal() != other.getVal()) {
            return false;
        }
        return true;
    }

    public int getVal() {
        return val;
    }

}

Then we extend class A with another class with the somewhat expected name "B", where B introduces an additional bean property named "anotherValue" like this:
public class B extends A {

    private final int anotherValue;

    public B(int val, int anotherValue) {
        super(val);
        this.anotherValue = anotherValue;
    }

    @Override
    public int hashCode() {
        int hash = 7;
        hash = 67 * hash + this.getValue();
        hash = 67 * hash + this.getAnotherValue();
        return hash;
    }

    @Override
    public boolean equals(Object obj) {
        if (obj == null) {
            return false;
        }
        if (!(obj instanceof B)) {
            return false;
        }
        final B other = (B) obj;
        if (this.getAnotherValue() != other.getAnotherValue()) {
            return false;
        }
        if (this.getValue() != other.getValue()) {
            return false;
        }
        return true;
    }

    public int getAnotherValue() {
        return anotherValue;
    }

}

We also create a test class to test the A and B classes:

public class Main {

    public static void main(String[] args) {
        A a = new A(1);
        B b = new B(1, 2);

        System.out.println("a.equals(b) is " + a.equals(b));  // true
        System.out.println("b.equals(a) is " + b.equals(a));  // false
    }

}

As can be seen, we get the disappointing result that "a.equals(b) is true" whereas "b.equals(a) is false". A clear violation of the symmetry rule B!


Preface and Overall Solution Strategy

Before we start with the solution, I want to draw to the attention that equals() and hashCode() basically has the same skeleton. They should both iterate over the (same) bean properties and produce a result. equals() shall return a boolean value if all the bean properties are equal whereas hashCode() shall return an integer that depends on all the bean properties. It would appear rational if both methods got their input properties from the same source using their getters. This way, we avoid mistake i), ii) and iii).

So, imagine that we have an interface with the method Object[] members() that the class has to implement and that it is used both in the equals() and hashCode() method.

Before we lay out a solution, we start with looking at what most IDE:s will generate if you request that they shall generate equals() and hashCode() methods. This allows us to understand the problem a bit more.


The IDE Bean Pattern

Most IDEs have built-in functions for generating the equals() and hashCode(). My IDE makes mistake iii) but otherwise it looks reasonable. After fixing this and some other minor flaws it looks something like this:

public class Person implements Comparable&ltPerson&gt {

    private final String name;
    private final String email;
    private final int born;

    public Person(String name, String email, int born) {
        this.name = Objects.requireNonNull(name);
        this.email = email;
        this.born = born;
    }

    public String getName() {
        return name;
    }

    public String getEmail() {
        return email;
    }

    public int getBorn() {
        return born;
    }

    @Override
    public int hashCode() {
        int hash = 7;
        hash = 61 * hash + Objects.hashCode(getName());
        hash = 61 * hash + Objects.hashCode(getEmail());
        hash = 61 * hash + getBorn();
        return hash;
    }

    @Override
    public boolean equals(Object obj) {
        if (obj == null) {
            return false;
        }
        if (getClass() != obj.getClass()) {
            return false;
        }
        final Person other = (Person) obj;
        if (!Objects.equals(this.getName(), other.getName())) {
            return false;
        }
        if (!Objects.equals(this.getEmail(), other.getEmail())) {
            return false;
        }
        if (this.getBorn() != other.getBorn()) {
            return false;
        }
        return true;
    }

    @Override
    public String toString() {
        return "Person{" + "Name=" + getName() + ", Email=" + getEmail() + ", Born=" + getBorn() + '}';
    }

    @Override
    public int compareTo(Person that) {
        int nameCompareTo = this.getName().compareTo(that.getName());
        if (nameCompareTo != 0) {
            return nameCompareTo;
        }
        return Integer.valueOf(this.getBorn()).compareTo(that.getBorn());
    }

}

The science field for computing hash codes is quit broad and falls outside the scope of this post. I will perhaps come back on this subject in a later post. Here, one apparently starts with a prime number (7) and then iteratively multiplies the interim result with another prime (61) and adds the hash of the bean property. This progresses until all bean properties has been used.

The equals() is similar in the way that it iterates over the properties, but it progressively checks if they are equal. As soon as one property is determined to be un-equal, the method aborts and returns false. If all bean properties are equal, then the beans are also equal which seams reasonable. Again, note the mistake of using if (!(obj instanceof Person)) instead of if (getClass() != obj.getClass()). Can you see why this will lead to problems when Person is overriden? If we override the Person class with FemalePerson and we let FemalePerson do if (!(obj instanceof FemalePerson)) while the Person still will retain its if (!(obj instanceof Person)) then we will violate contract items B and C! Why is that? A FemalePerson is an instance of both Person and FemalePerson while a Person is an instance of only Person and not FemalePerson. So female.equals(person) might be true while person.equals(female) is false at the same time! The symmetry becomes broken.
The solution depicted above further has the disadvantage that the getClass():es must return exactly the same class which is not good when we, for instance, create anonymous classes. We will see how to fix this problem later on in this post.

I have also shown the toString() method here. It also iterates over the bean values and produces a string which contains all the bean properties.

At the end, there is also a compareTo() method that is similar to the equals() method. However, It iterates over a subset of the bean properties and compares them. If two properties are not equal, the compareTo value of them are returned, otherwise the method progresses over the properties until the last properties and the compareTo of them are returned.

The observant reader is now seeing a pattern here. We only need two "picker" methods that will select the bean properties for equals(), hashCode(), toString() on one hand and for compareTo() on the other hand. If we want to implement the toString() method, we also need a supplementary third picker method that returns the name of each bean property.

Before we continue, I would like to show one way of extending the Person class:

public class FemalePerson extends Person {

    private final String handbagBrand;

    public FemalePerson(String handbagBrand, String name, String email, int born) {
        super(name, email, born);
        this.handbagBrand = handbagBrand;
    }

    public String getHandbagBrand() {
        return handbagBrand;
    }

    @Override
    public int hashCode() {
        return 61 * super.hashCode() + Objects.hashCode(getHandbagBrand());
    }

    @Override
    public boolean equals(Object obj) {
        if (!super.equals(obj)) {
            return false;
        }
        final FemalePerson other = (FemalePerson) obj;
        if (!Objects.equals(this.getHandbagBrand(), other.getHandbagBrand())) {
            return false;
        }
        return true;
    }

    @Override
    public String toString() {
        return "FemalePerson{" + "Name=" + getName() + ", Email=" + getEmail() + ", Born=" + getBorn() + ", HandbagBrand=" + getHandbagBrand() + '}';
    }

}
We have added an important property of females, namely the "handbagBrand". Both the hashCode() and the equals() method calls their respective super methods after which the new bean property is added. The toString() method is rewritten from scratch.

Testing the Beans

Now we can test our new bean. We will use the same test for all the implementation variants of Person and FemalePerson throughout this post. Here it is:

public class Test {

    public static void main(String[] args) {
        final Person adam = new Person("Adam", "adam@mail.com", 1966);
        final Person adam2 = new Person("Adam", "adam@mail.com", 1966);
        final Person adamYoung = new Person("Adam", "adam_88@mail.com", 1988);
        final Person bert = new Person("Bert", "bert@mail.com", 1979);
        final Person bert2 = new Person("Bert", "bert@mail.com", 1979) {

            @Override
            public String toString() {
                return "Strange:" + super.toString();
            }

        };
        final Person cecelia = new FemalePerson("Guchi", "Cecelia", "cecelia@mail.com", 1981);
        final Person ceceliaPro = new FemalePerson("Guchi Pro", "Cecelia", "cecelia@mail.com", 1981);
        final Person cecelia2 = new Person("Cecelia", "cecelia@mail.com", 1981);

        printEquals(adam, adam2);
        printEquals(adam2, adam);
        printEquals(bert, bert2);
        printEquals(bert2, bert);
        printEquals(cecelia, cecelia2);
        printEquals(cecelia2, cecelia);

        final List l = Arrays.asList(cecelia, adamYoung, ceceliaPro, cecelia2, adam, bert, adam2, bert2);

        System.out.println("*** Initial order");
        l.forEach(System.out::println);
        System.out.println("*** Sorted order");
        Collections.sort(l);
        l.forEach(System.out::println);

    }

    private static void printEquals(Person p1, Person p2) {
        System.out.println("It is " + p1.equals(p2) + " that " + p1 + " equals " + p2
                + ". hashCode()s are " + ((p1.hashCode() == p2.hashCode()) ? "equals" : "are different"));
    }

}

We first create two adams with identical bean properties (adam and adam2) followed by a younger version of adam called youngAdam. Then we create two berts with the same bean properties but with different toString() methods, just to illustrated what happens with anonymous class overrides. Then we have three incarnations of cecilias: two FemalePersons with different HandbagBrands, then a cecilia that is just a Person. The objects are rigged to demonstrate different behavior of the implementations depicted in this post.

The test program then prints out how some of the objects that are related in terms of equality and then a list with the person in a mixed order is created. This list is printed, then the list is sorted (to test the compareTo() method) and is then printed again.

When run, the following result will show:

It is true that Person{Name=Adam, Email=adam@mail.com, Born=1966} equals Person{Name=Adam, Email=adam@mail.com, Born=1966}. hashCode()s are equals
It is true that Person{Name=Adam, Email=adam@mail.com, Born=1966} equals Person{Name=Adam, Email=adam@mail.com, Born=1966}. hashCode()s are equals
It is false that Person{Name=Bert, Email=bert@mail.com, Born=1979} equals Strange:Person{Name=Bert, Email=bert@mail.com, Born=1979}. hashCode()s are equals
It is false that Strange:Person{Name=Bert, Email=bert@mail.com, Born=1979} equals Person{Name=Bert, Email=bert@mail.com, Born=1979}. hashCode()s are equals
It is false that FemalePerson{Name=Cecelia, Email=cecelia@mail.com, Born=1981, HandbagBrand=Guchi} equals Person{Name=Cecelia, Email=cecelia@mail.com, Born=1981}. hashCode()s are are different
It is false that Person{Name=Cecelia, Email=cecelia@mail.com, Born=1981} equals FemalePerson{Name=Cecelia, Email=cecelia@mail.com, Born=1981, HandbagBrand=Guchi}. hashCode()s are are different
*** Initial order
FemalePerson{Name=Cecelia, Email=cecelia@mail.com, Born=1981, HandbagBrand=Guchi}
Person{Name=Adam, Email=adam_88@mail.com, Born=1988}
FemalePerson{Name=Cecelia, Email=cecelia@mail.com, Born=1981, HandbagBrand=Guchi Pro}
Person{Name=Cecelia, Email=cecelia@mail.com, Born=1981}
Person{Name=Adam, Email=adam@mail.com, Born=1966}
Person{Name=Bert, Email=bert@mail.com, Born=1979}
Person{Name=Adam, Email=adam@mail.com, Born=1966}
Strange:Person{Name=Bert, Email=bert@mail.com, Born=1979}
*** Sorted order
Person{Name=Adam, Email=adam@mail.com, Born=1966}
Person{Name=Adam, Email=adam@mail.com, Born=1966}
Person{Name=Adam, Email=adam_88@mail.com, Born=1988}
Person{Name=Bert, Email=bert@mail.com, Born=1979}
Strange:Person{Name=Bert, Email=bert@mail.com, Born=1979}
FemalePerson{Name=Cecelia, Email=cecelia@mail.com, Born=1981, HandbagBrand=Guchi}
FemalePerson{Name=Cecelia, Email=cecelia@mail.com, Born=1981, HandbagBrand=Guchi Pro}
Person{Name=Cecelia, Email=cecelia@mail.com, Born=1981}  


As can be seen, it works as expected. As we allredy are aware of, bert2 will not equal bert even though their bean properties are the same because they are not of the same class. There are no occurrences of two objects being equal but at the same time having different hashCodes. The mixed list is sorted in correct order.

Abstract Object Support Class

Consider the following abstract object support class:
public abstract class AbstractObjectSupport<T extends AbstractObjectSupport<T>> implements Comparable<T> {
    protected abstract Object[] members();
    protected abstract Object[] names();
    protected abstract Comparable<?>[] compareToMembers();
    protected Object[] mkArray(final Object... members) {
        return members;
    }

    protected Comparable<?>[] mkComparableArray(final Comparable<?>... members) {
        return members;
    }

    protected Object[] exArray(final Object[] originalMembers, final Object... newMembers) {
        final Object[] result = Arrays.copyOf(originalMembers, originalMembers.length + newMembers.length);
        for (int i = originalMembers.length, n = 0; i < result.length; i++, n++) {
            result[i] = newMembers[n];
        }
        return result;
    }

    protected Comparable<?>[] exComparableArray(final Comparable<?>[] originalMembers, final Comparable<?>... newMembers) {
        final Comparable<?>[] result = (Comparable<?>[]) exArray(originalMembers, (Object[]) newMembers);
        return result;
    }

    @Override
    public int hashCode() {
        return Objects.hash(members());

    }
    @Override
    public boolean equals(final Object obj) {
        if (obj == null) {
            return false;
        }
        if (getClass() != obj.getClass()) {
            return false;
        }
        @SuppressWarnings("rawtypes")
        // Must be an AbstractObjectSupport since the class is the same as this class
        final AbstractObjectSupport thatAbstractObjectSupport = (AbstractObjectSupport) obj; 
        return Arrays.equals(members(), thatAbstractObjectSupport.members());
    }

    @Override
    public String toString() {
        final String className = getClass().getSimpleName().isEmpty() ? getClass().getName() : getClass().getSimpleName();
        final StringJoiner sj = new StringJoiner(", ", className + "{", "}");
        final Object[] members = members();
        final Object[] names = names();
        final int n = Math.min(members.length, names.length);
        for (int i = 0; i < n; i++) {
            final StringJoiner msj = new StringJoiner("=");
            msj.add(Objects.toString(names[i]));
            msj.add(Objects.toString(members[i]));
            sj.merge(msj);
        }
        return sj.toString();
    }

    @Override
    public int compareTo(T that) {
        @SuppressWarnings("rawtypes")
        final Comparable[] thisComparables = this.compareToMembers();
        @SuppressWarnings("rawtypes")
        final Comparable[] thatComparables = that.compareToMembers();

        final int n = Math.min(thisComparables.length, thatComparables.length);
        for (int i = 0; i < n; i++) {
            @SuppressWarnings("unchecked")
            final int result = thisComparables[i].compareTo(thatComparables[i]);
            if (result != 0) {
                return result;
            }
        }
        return 0; // They are equal
    }

}

When subclassing from this class, a new concrete class must implement the three support methods:

- members() that will return an ordered array with all the bean properties.
- names() that will return an ordered array of the corresponding bean propoerty names. These names are used in the toString() function only.
- compareToMembers() that will return an ordered array with all the (Comparable) bean properties that shall be used in the compareTo() method.

Then the class implements the equals(), hashCode(), toString() and compareTo() methods by first using the corresponding support methods and then performing some logic on the results. Note how simple equals() and hashCode() are implemented using the Objects and Arrays classes. Note also the use of StringJoiner in the toString() method.

Now we can create our Person and FemalePerson classes very easily like this:
public class Person extends AbstractObjectSupport<Person> {

    private final String name;
    private final String email;
    private final int born;

    public Person(String name, String email, int born) {
        this.name = name;
        this.email = email;
        this.born = born;
    }

    public String getName() {
        return name;
    }

    public String getEmail() {
        return email;
    }

    public int getBorn() {
        return born;
    }

    @Override
    public Object[] members() {
        return mkArray(getName(), getEmail(), getBorn());
    }

    @Override
    public Object[] names() {
        return mkArray("Name", "Email", "Born");
    }

    @Override
    public Comparable<?>[] compareToMembers() {
        return mkComparableArray(getName());
    }

}


and

public class FemalePerson extends Person {

    private final String handbagBrand;

    public FemalePerson(String handbagBrand, String name, String email, int born) {
        super(name, email, born);
        this.handbagBrand = handbagBrand;
    }

    @Override
    public Object[] members() {
        return exArray(super.members(), getHandbagBrand());
    }

    @Override
    public Object[] names() {
        return exArray(super.names(), "handbagBrand");
    }

    public String getHandbagBrand() {
        return handbagBrand;
    }

}

We are now certain that the equals() and hashCode() methods are using the same bean properties and thus we know that they will fulfill their contracts. If we run out test program, we will get the same result as before which is encouraging.

This pattern can be used if you have not inherited from a super class before, but since Java objects can only have one super class, you can not use this pattern when you want to inherit from another class. In the next chapters, we will learn how we can mix in these methods while still being able to inherit from another class.

The Object Support Mixin Pattern

Java 8 provides default methods in interfaces. This functionality was needed to extend existing classes (such as the Collection classes) while retaining compatibility with old code. New methods can be added to interface without the need to implement these methods in the implementing classes. This feature can also be used for other purposes. Now is a good time to mention that some people are against the use of interfaces as carrying any form of logic. According to them, interfaces shall only describe what can be done, not how! I will not engage in this philosophic discussion now. As a marker that this is not just any interface, I have chosen to name the interface to ObjectMixin where the suffix "Mixin" is intended to indicate that it is more than just an interface: methods will be mixed in (not inherited) from the interface. The ObjectMixin looks very similar to the AbstractObjectSupport class:

public interface ObjectMixin<T extends ObjectMixin<T>> {

    Object[] members();

    Object[] names();

    Comparable<?>[] compareToMembers();

    default Object[] mkArray(final Object... members) {
        return members;
    }

    default Comparable<?>[] mkComparableArray(final Comparable<?>... members) {
        return members;
    }

    default Object[] exArray(final Object[] originalMembers, final Object... newMembers) {
        final Object[] result = Arrays.copyOf(originalMembers, originalMembers.length + newMembers.length);
        for (int i = originalMembers.length, n = 0; i < result.length; i++, n++) {
            result[i] = newMembers[n];
        }
        return result;
    }

    default Comparable<?>[] exComparableArray(final Comparable<?>[] originalMembers, final Comparable<?>... newMembers) {
     
        final Comparable<?>[] result = (Comparable<?>[])exArray(originalMembers, (Object[])newMembers);
        return result;
    }

    default int _hashCode() {
        return Objects.hash(members());
    }

    default boolean _equals(final Object obj) {
        if (obj == null) {
            return false;
        }
        if (getClass() != obj.getClass()) {
            return false;
        }
        @SuppressWarnings("rawtypes")
        // Must be an AbstractObjectSupport since the class is the same as this class
        final AbstractObjectSupport thatAbstractObjectSupport = (AbstractObjectSupport) obj; 
        return Arrays.equals(members(), thatAbstractObjectSupport.members());
    }

    default String _toString() {
        final String className = getClass().getSimpleName().isEmpty() ? getClass().getName() : getClass().getSimpleName();
        final StringJoiner sj = new StringJoiner(", ", className + "{", "}");
        final Object[] members = members();
        final Object[] names = names();
        final int n = Math.min(members.length, names.length);
        for (int i = 0; i < n; i++) {
            final StringJoiner msj = new StringJoiner("=");
            msj.add(Objects.toString(names[i]));
            msj.add(Objects.toString(members[i]));
            sj.merge(msj);
        }
        return sj.toString();
    }

    default int _compareTo(T obj) {
        @SuppressWarnings("rawtypes")
        final Comparable[] thisComparables = compareToMembers();
        @SuppressWarnings("rawtypes")
        final Comparable[] thatComparables = obj.compareToMembers();

        final int n = Math.min(thisComparables.length, thatComparables.length);
        for (int i = 0; i < n; i++) {
            @SuppressWarnings("unchecked")
            final int result = thisComparables[i].compareTo(thatComparables[i]);
            if (result != 0) {
                return result;
            }
        }
        return 0; // They are equal
    }

}


If we let our Person class implement this interface, it can look like this:

public class Person implements Comparable<Person>, ObjectMixin<Person> {

    private final String name;
    private final String email;
    private final int born;

    public Person(String name, String email, int born) {
        this.name = name;
        this.email = email;
        this.born = born;
    }

    public String getName() {
        return name;
    }

    public String getEmail() {
        return email;
    }

    public int getBorn() {
        return born;
    }

    @Override
    public Object[] members() {
        return mkArray(getName(), getEmail(), getBorn());
    }

    @Override
    public Object[] names() {
        return mkArray("Name", "Email", "Born");
    }

    @Override
    public Comparable<?>[] compareToMembers() {
        return mkComparableArray(getName());
    }

    @Override
    public int hashCode() {
        return _hashCode();
    }

    @Override
    public boolean equals(Object obj) {
        return _equals(obj);
    }

    @Override
    public String toString() {
        return _toString();
    }

    @Override
    public int compareTo(Person o) {
        return _compareTo(o);
    }

}

This way, we do not need to scarify the inheritance and but can still gain all the benefits that the AbstractObjectSupport gave us. The only disadvantage is that we need to explicitly override the equals(), hashCode(), toString() and compareTo() methods and delegate to the ObjectMixin methods. As you might be aware of, an interface can neither introduce new bean properties nor can it override existing methods. When we run the test program, we still get the same output as before.

Classes can easily be extended just as in the previous chapter where we saw FemalePerson being declared.

The Standard Object Support Mixin

We still have the nuisance that overridden classes like anonymous classes are not equal to seemingly equal classes. For example, bert and bert2 are not equal even though they have the same bean properties. Remember that only the toString() differs and this should not make them different. By introducing a new method called compareClass() we can use this class instead of the getClass() and compare them. Now we are in charge what class we elect to return and can set a new "watermark" whenever we think that an inherited class shall never be equal to its super class. The neat thing with the solution below is that we will also have a default compareClass() that automatically will determine the highest class that also is an ObjectMixin. So, you get the initial base class compareClass() for free. Note how the defaultBaseCompareObjectMixinClass() is hid in the inner class MethodUtil so it will not be exposed to the implementing class. The  defaultBaseCompareObjectMixinClass() recursively inspects super classes and when a super class does not implement ObjectMixin, it returns.

public interface ObjectMixin<T extends ObjectMixin<T>> {

    Object[] members();

    Object[] names();

    Comparable<?>[] compareToMembers();

    default Class<? extends ObjectMixin<T>> compareClass() {
        return MethodUtil.defaultBaseCompareObjectMixinClass((Class<T>) getClass());
    }

    default Object[] mkArray(final Object... members) {
        return members;
    }

    default Comparable<?>[] mkComparableArray(final Comparable<?>... members) {
        return members;
    }

    default Object[] exArray(final Object[] originalMembers, final Object... newMembers) {
        final Object[] result = Arrays.copyOf(originalMembers, originalMembers.length + newMembers.length);
        for (int i = originalMembers.length, n = 0; i < result.length; i++, n++) {
            result[i] = newMembers[n];
        }
        return result;
    }

    default Comparable<?>[] exComparableArray(final Comparable<?>[] originalMembers, final Comparable<?>... newMembers) {

        final Comparable<?>[] result = (Comparable<?>[]) exArray(originalMembers, (Object[]) newMembers);
        return result;
    }

    default int _hashCode() {
        return Objects.hash(members());
    }

    default boolean _equals(final Object obj) {
        if (!(obj instanceof ObjectMixin)) {
            return false;
        }
        @SuppressWarnings("rawtypes")
        final ObjectMixin thatObjectMixin = (ObjectMixin) obj;
        if (this.compareClass() != thatObjectMixin.compareClass()) {
            return false;
        }
        return Arrays.equals(members(), thatObjectMixin.members());
    }

    default String _toString() {
        final String className = getClass().getSimpleName().isEmpty() ? getClass().getName() : getClass().getSimpleName();
        final StringJoiner sj = new StringJoiner(", ", className + "{", "}");
        final Object[] members = members();
        final Object[] names = names();
        final int n = Math.min(members.length, names.length);
        for (int i = 0; i < n; i++) {
            final StringJoiner msj = new StringJoiner("=");
            msj.add(Objects.toString(names[i]));
            msj.add(Objects.toString(members[i]));
            sj.merge(msj);
        }
        return sj.toString();
    }

    default int _compareTo(T obj) {
        @SuppressWarnings("rawtypes")
        final Comparable[] thisComparables = compareToMembers();
        @SuppressWarnings("rawtypes")
        final Comparable[] thatComparables = obj.compareToMembers();

        final int n = Math.min(thisComparables.length, thatComparables.length);
        for (int i = 0; i < n; i++) {
            @SuppressWarnings("unchecked")
            final int result = thisComparables[i].compareTo(thatComparables[i]);
            if (result != 0) {
                return result;
            }
        }
        return 0; // They are equal
    }

    static abstract class MethodUtil {

        public static <T extends ObjectMixin> Class<T> defaultBaseCompareObjectMixinClass(Class<T> clazz) {
            final Class<? super T> superClazz = clazz.getSuperclass();
            if (!ObjectMixin.class.isAssignableFrom(superClazz)) {
                return clazz;
            }
            @SuppressWarnings("unchecked")
            final Class<T> objectMixinSuperClazz = (Class<T>) superClazz;
            return defaultBaseCompareObjectMixinClass(objectMixinSuperClazz);
        }

    }

}

Please note that the equals() method now considers the compareClass() instead of just the getClass() and that we have full control of the compareClass() method as opposed to the getClass() method.

When we now run the test program we get the following result (shortened listing):

It is true that Person{Name=Adam, Email=adam@mail.com, Born=1966} equals Person{Name=Adam, Email=adam@mail.com, Born=1966}. hashCode()s are equals

It is true that Person{Name=Adam, Email=adam@mail.com, Born=1966} equals Person{Name=Adam, Email=adam@mail.com, Born=1966}. hashCode()s are equals

It is true that Person{Name=Bert, Email=bert@mail.com, Born=1979} equals Strange:com.blogspot.minborgsjavapot.objectmixin._4interface_class.Test$1{Name=Bert, Email=bert@mail.com, Born=1979}. hashCode()s are equals

It is true that Strange:com.blogspot.minborgsjavapot.objectmixin._4interface_class.Test$1{Name=Bert, Email=bert@mail.com, Born=1979} equals Person{Name=Bert, Email=bert@mail.com, Born=1979}. hashCode()s are equals

It is false that FemalePerson{Name=Cecelia, Email=cecelia@mail.com, Born=1981, handbagBrand=Guchi} equals Person{Name=Cecelia, Email=cecelia@mail.com, Born=1981}. hashCode()s are are different

It is false that Person{Name=Cecelia, Email=cecelia@mail.com, Born=1981} equals FemalePerson{Name=Cecelia, Email=cecelia@mail.com, Born=1981, handbagBrand=Guchi}. hashCode()s are are different


Now, bert and bert2 are equal just as we would expect! Great progress!

When FemalePerson inherit from Person, we also set a new watermark to ensure that FemalePerson are never equal to Person() as shown in this class:

public class FemalePerson extends Person {

    private final String handbagBrand;

    public FemalePerson(String handbagBrand, String name, String email, int born) {
        super(name, email, born);
        this.handbagBrand = handbagBrand;
    }

    @Override
    public Object[] members() {
        return exArray(super.members(), getHandbagBrand());
    }

    @Override
    public Object[] names() {
        return exArray(super.names(), "handbagBrand");
    }

    public String getHandbagBrand() {
        return handbagBrand;
    }

    @Override
    public Class<FemalePerson> compareClass() {
        return FemalePerson.class;
    }

}

The Reflection Object Support Mixin

The pattern can be simplified even more for the implementing classes. It is possible for the interface to provide default support methods for the menbers() and names() methods, eliminating the need for implementing these method by hand. This can be done by extending the previous ObjectMixin inteface as shown hereunder:

public interface ReflectionObjectMixin<T extends ReflectionObjectMixin<T>> extends ObjectMixin<T> {

    @Override
    default Object[] members() {
        return new MethodUtil(getClass()) {

            @Override
            protected Object onMethod(final Method method) {
                try {
                    return method.invoke(ReflectionObjectMixin.this, (Object[]) null);
                } catch (IllegalAccessException | IllegalArgumentException | InvocationTargetException e) {
                    throw new IllegalStateException("Unexpected invocation error", e);
                }
            }
        }.toObjects();

    }

    @Override
    default Object[] names() {

        return new MethodUtil(getClass()) {

            @Override
            protected Object onMethod(final Method method) {
                return method.getName().substring(MethodUtil.INGRESS.length());
            }
        }.toObjects();

    }

    static abstract class MethodUtil {

        public static final String INGRESS = "get";
        public static final Set<String> EXCLUDED_METHODS = new HashSet<>(Arrays.asList("getClass"));
        private final Class<?> clazz;

        private MethodUtil(Class<?> clazz) {
            this.clazz = clazz;
        }

        private static List<Method> obtainGetMethods(Class<?> clazz) {
            final List<Method> result = new ArrayList<>();
            final Method[] methods = clazz.getMethods();
            for (final Method method : methods) {
                final String methodName = method.getName();
                if (methodName.startsWith(INGRESS) && method.getParameterCount() == 0 && !EXCLUDED_METHODS.contains(methodName)) {
                    result.add(method);
                }
            }
            Collections.sort(result, METHOD_COMPARATOR);
            return result;
        }

        protected abstract Object onMethod(Method method);

        public Object[] toObjects() {
            final List<Object> result = new ArrayList<>();
            for (final Method method : MethodUtil.obtainGetMethods(clazz)) {
                result.add(onMethod(method));
            }
            return result.toArray();
        }

        private final static MethodComparator METHOD_COMPARATOR = new MethodComparator();

        private static class MethodComparator implements Comparator<Method> {

            @Override
            public int compare(Method o1, Method o2) {
                int classCompare = o1.getDeclaringClass().getName().compareTo(o2.getDeclaringClass().getName());
                if (classCompare != 0) {
                    return classCompare;
                }
                return o1.getName().compareTo(o2.getName());
            }
        }
    }
}

Both the new default methods members() and names() will iterate over all methods that starts with "get" (except the getClass()) and that does not take any parameters. These are all assumed to be bean properties as dictated by the Bean Pattern. For the names() method, we will just cut out the name of the bean property as the name of the getter excluding the "get" prefix (e.g. "getName" becomes "Name"). For the members() method, we will iterate over the same methods but instead we will invoke the method for the bean and save the resulting result in the result array. The clazz.getMethods() will return the classes methods in any order, so we will sort the methods in class declaration order (name of the class it is declared in) and then in alphabetic order of the method name.

The implementing class is now shorter since we got rid of the members() and names() method declaration:

public class Person implements Comparable<Person>, ReflectionObjectMixin<Person> {

    private final String name;
    private final String email;
    private final int born;

    public Person(String name, String email, int born) {
        this.name = name;
        this.email = email;
        this.born = born;
    }

    public String getName() {
        return name;
    }

    public String getEmail() {
        return email;
    }

    public int getBorn() {
        return born;
    }

    @Override
    public Comparable<?>[] compareToMembers() {
        return mkComparableArray(getName());
    }

    @Override
    public int hashCode() {
        return _hashCode();
    }

    @Override
    public boolean equals(Object obj) {
        return _equals(obj);
    }

    @Override
    public String toString() {
        return _toString();
    }

    @Override
    public int compareTo(Person o) {
        return _compareTo(o);
    }

}

When we run the test program we get the following output:

It is true that Person{Born=1966, Email=adam@mail.com, Name=Adam} equals Person{Born=1966, Email=adam@mail.com, Name=Adam}. hashCode()s are equals
It is true that Person{Born=1966, Email=adam@mail.com, Name=Adam} equals Person{Born=1966, Email=adam@mail.com, Name=Adam}. hashCode()s are equals
It is true that Person{Born=1979, Email=bert@mail.com, Name=Bert} equals Strange:javapot.objectmixin._5interface_reflection.Test$1{Born=1979, Email=bert@mail.com, Name=Bert}. hashCode()s are equals
It is true that Strange:javapot.objectmixin._5interface_reflection.Test$1{Born=1979, Email=bert@mail.com, Name=Bert} equals Person{Born=1979, Email=bert@mail.com, Name=Bert}. hashCode()s are equals
It is false that FemalePerson{HandbagBrand=Guchi, Born=1981, Email=cecelia@mail.com, Name=Cecelia} equals Person{Born=1981, Email=cecelia@mail.com, Name=Cecelia}. hashCode()s are are different
It is false that Person{Born=1981, Email=cecelia@mail.com, Name=Cecelia} equals FemalePerson{HandbagBrand=Guchi, Born=1981, Email=cecelia@mail.com, Name=Cecelia}. hashCode()s are are different
*** Initial order
FemalePerson{HandbagBrand=Guchi, Born=1981, Email=cecelia@mail.com, Name=Cecelia}
Person{Born=1988, Email=adam_88@mail.com, Name=Adam}
FemalePerson{HandbagBrand=Guchi Pro, Born=1981, Email=cecelia@mail.com, Name=Cecelia}
Person{Born=1981, Email=cecelia@mail.com, Name=Cecelia}
Person{Born=1966, Email=adam@mail.com, Name=Adam}
Person{Born=1979, Email=bert@mail.com, Name=Bert}
Person{Born=1966, Email=adam@mail.com, Name=Adam}
Strange:javapot.objectmixin._5interface_reflection.Test$1{Born=1979, Email=bert@mail.com, Name=Bert}
*** Sorted order
Person{Born=1988, Email=adam_88@mail.com, Name=Adam}
Person{Born=1966, Email=adam@mail.com, Name=Adam}
Person{Born=1966, Email=adam@mail.com, Name=Adam}
Person{Born=1979, Email=bert@mail.com, Name=Bert}
Strange:javapot.objectmixin._5interface_reflection.Test$1{Born=1979, Email=bert@mail.com, Name=Bert}
FemalePerson{HandbagBrand=Guchi, Born=1981, Email=cecelia@mail.com, Name=Cecelia}
FemalePerson{HandbagBrand=Guchi Pro, Born=1981, Email=cecelia@mail.com, Name=Cecelia}
Person{Born=1981, Email=cecelia@mail.com, Name=Cecelia}

Nice! The only thing that differs is the order of the bean properties in the toString() method.

It becomes even better when we are considering the class FemalePerson, which now looks like this:

public class FemalePerson extends Person {

    private final String handbagBrand;

    public FemalePerson(String handbagBrand, String name, String email, int born) {
        super(name, email, born);
        this.handbagBrand = handbagBrand;
    }

    public String getHandbagBrand() {
        return handbagBrand;
    }

}

It is almost magical, you now get everything for free! The equals(), compareTo() and toString() automatically adjusts to the newly introduced bean property.


The Annotated Object Support Mixin

We can also decide what methods shall be used in the members() and names() method by using annotations. We start by creating our own annotation class named EqualsAndHashCode:

@Retention(value = RetentionPolicy.RUNTIME)
@Target(ElementType.METHOD)
public @interface EqualsAndHashCode {

}

The intention now is that we should simply be able to annotate our methods that we want to "mark" as being in the members() and names() function using this annotation. To allow this we create yet another variant of the ObjectMixin as follows:

public interface AnnotationObjectMixin<T extends AnnotationObjectMixin<T>> extends ObjectMixin<T> {

    @Override
    default Object[] members() {
        return new MethodUtil(getClass(), EqualsAndHashCode.class) {

            @Override
            protected Object onMethod(final Method method) {
                try {
                    return method.invoke(AnnotationObjectMixin.this, (Object[]) null);
                } catch (IllegalAccessException | IllegalArgumentException | InvocationTargetException e) {
                    throw new IllegalStateException("Unexpected invocation error", e);
                }
            }
        }.toObjects();

    }

    @Override
    default Object[] names() {

        return new MethodUtil(getClass(), EqualsAndHashCode.class) {

            @Override
            protected Object onMethod(final Method method) {
                final String methodName = method.getName();
                if (methodName.startsWith(INGRESS)) {
                    return methodName.substring(ReflectionObjectMixin.MethodUtil.INGRESS.length());
                } else {
                    return methodName;
                }
            }
        }.toObjects();

    }

    static abstract class MethodUtil {

        public static final String INGRESS = "get";
        private final Class<?> clazz;
        private final Class annotationClass;

        private MethodUtil(Class<?> clazz, Class annotationClass) {
            this.clazz = clazz;
            this.annotationClass = annotationClass;
        }

        private static List<Method> obtainGetMethods(Class<?> clazz, Class annotationClass) {
            final List<Method> result = new ArrayList<>();
            final Method[] methods = clazz.getMethods();
            for (final Method method : methods) {
                if (method.getParameterCount() == 0 && (method.getAnnotation(annotationClass) != null)) {
                    result.add(method);
                }
            }
            Collections.sort(result, METHOD_COMPARATOR);
            return result;
        }

        protected abstract Object onMethod(Method method);

        public Object[] toObjects() {
            final List<Object> result = new ArrayList<>();
            for (final Method method : MethodUtil.obtainGetMethods(clazz, annotationClass)) {
                result.add(onMethod(method));
            }
            return result.toArray();
        }

        private final static MethodComparator METHOD_COMPARATOR = new MethodComparator();

        private static class MethodComparator implements Comparator<Method> {

            @Override
            public int compare(Method o1, Method o2) {
                int classCompare = o1.getDeclaringClass().getName().compareTo(o2.getDeclaringClass().getName());
                if (classCompare != 0) {
                    return classCompare;
                }
                return o1.getName().compareTo(o2.getName());
            }
        }
    }
}

Now we are able just to "mark" our implementing class methods with @EqualsAndHashCode as shown here:

public class Person implements Comparable<Person>, AnnotationObjectMixin<Person> {

    private final String name;
    private final String email;
    private final int born;

    public Person(String name, String email, int born) {
        this.name = name;
        this.email = email;
        this.born = born;
    }

    @EqualsAndHashCode
    public String getName() {
        return name;
    }

    @EqualsAndHashCode
    public String getEmail() {
        return email;
    }

    @EqualsAndHashCode
    public int getBorn() {
        return born;
    }

    @Override
    public Comparable<?>[] compareToMembers() {
        return mkComparableArray(getName());
    }

    @Override
    public int hashCode() {
        return _hashCode();
    }

    @Override
    public boolean equals(Object obj) {
        return _equals(obj);
    }

    @Override
    public String toString() {
        return _toString();
    }

    @Override
    public int compareTo(Person o) {
        return _compareTo(o);
    }

}

The extending FemalePerson class can now look like this:

public class FemalePerson extends Person {

    private final String handbagBrand;

    public FemalePerson(String handbagBrand, String name, String email, int born) {
        super(name, email, born);
        this.handbagBrand = handbagBrand;
    }

    @EqualsAndHashCode
    public String getHandbagBrand() {
        return handbagBrand;
    }

}

Performance

The equals() and hashCode() methods call the members() method which converts any primitive bean properties (such as int) to their corresponding wrapper classes (e.g. Integer) by means of auto-boxing. This leads to unnecisary creation of short lived objecs compare to hand coded equals() and hashCode() methods where the primitives can be used directly for comparison.

The performance of reflection is relatively poor, so if you use the ReflectionObjectMixin or the AnnotationObjectMixin you will notice reduced performance. A large part of this performance drop can be regained by caching the reflection calls using a ConcurrentHashMap as shown in the following snippet, form a performance optimized ReflectionObjectMixin class:

        private static final Map<Class<?>, List<Method>> methodCache = new ConcurrentHashMap<>();

        private static List<Method> obtainGetMethods(Class<?> clazz) {

            List<Method> cacheResult = methodCache.get(clazz);
            if (cacheResult != null) {
                return cacheResult;
            } else {
                final List<Method> result = new ArrayList<>();
                final Method[] methods = clazz.getMethods();
                for (final Method method : methods) {
                    final String methodName = method.getName();
                    if (methodName.startsWith(INGRESS) && method.getParameterCount() == 0 && !EXCLUDED_METHODS.contains(methodName)) {
                        result.add(method);
                    }
                }
                Collections.sort(result, METHOD_COMPARATOR);
                methodCache.put(clazz, result);
                return result;
            }
        }
The performance of the compareClass() can also be improved in the same way using a static lookup Map.


Interface Wrapper Class

If you can use inheritance, you can create a small interface wrapper class that you can inherit from to save the work of overriding the Object methods like this:

public class ReflectionObjectSupport<T extends ReflectionObjectMixin<T>> implements ReflectionObjectMixin<T>, Comparable<T> {

    @Override
    public Comparable<?>[] compareToMembers() {
        throw new UnsupportedOperationException("Override this method in your class to implement comapreTo() support.");
    }

    @Override
    public boolean equals(Object obj) {
        return _equals(obj);
    }

    @Override
    public int hashCode() {
        return _hashCode();
    }

    @Override
    public String toString() {
        return _toString();
    }

    @Override
    public int compareTo(T o) {
        return _compareTo(o);
    }

}


Now your Person class can look just as promised at the top of this post!

Conclusions

You should develop a strategy on how to override the equals() and hashCode() methods that ensures that you will use the same bean properties for them both. You should also make sure that, when you override classes, their equals() and hashCode() should work as expected.

The Object Support Mixin Pattern ensures that the contract of the equals() and hashCode() are fulfilled. Furthermore, it makes coding of these method much easier and less error prone. The Object Support Mixin Pattern allows you to extend a different super class and just mix in the functionality you need without scarifying the single class inheritance. The Object Support Mixin Pattern also allows easy subclassing, both with normal classes and anonymous classes. One drawback with the pattern is that its performance is less than their hand coded counter parts.

Future Improvements

All the mixin methods and support methods are exposed as public methods. Perhaps it is possible to move the methods to an inner class so that they are not seen directly in the implementing class.

Bean properties stored using primitive classes (such as ints and longs) can perhaps be handled by separate member() methods to eliminate auto-boxing overhead. Perhaps, these primitive bean properties shall be compared before the wrapper class bean properties since, presumably, they are faster to compare.

Good luck with improving your basic Object methods!