Weirdness in Equated Java Arrays: References vs. Pointers

Having a problem understanding what's going on n the code below. The behavior of arrays c and d is what I would expect. But what's going on with a and b? (I also tried this with normal, scalar variables, and nothing surprising happens in either case.)

The output is copied to the RH comments.

import java.util.Arrays;

public class ArraysParadox {

    public static void main(String[] args) {

        int[] c = {1, 2, 3};
        int[] d = {6, 5, 4, 3};


        System.out.print("c:       ");
        System.out.println(Arrays.toString(c)); // c:       [1, 2, 3]

        System.out.print("d:       ");
        System.out.println(Arrays.toString(d)); // d:       [6, 5, 4, 3]

        System.out.println("--- swap ---");
        int[] tmp = c;
        c = d;
        d = tmp;    // <----- Magic?


        System.out.print("c' (=d): ");
        System.out.println(Arrays.toString(c)); // c' (=d): [6, 5, 4, 3]

        System.out.print("d' (=c): ");
        System.out.println(Arrays.toString(d)); // d' (=c): [1, 2, 3]

        System.out.println("--- c = 0 ---");
        Arrays.fill(c, 0);
        System.out.print("c (=0):  ");

        System.out.println(Arrays.toString(c)); // c (=0):  [0, 0, 0, 0]

        System.out.print("d (=c):  ");
        System.out.println(Arrays.toString(d)); // d (=c):  [1, 2, 3]

        System.out.println("--- d = 1 ---");
        Arrays.fill(d, 1);
        System.out.print("c (=d):  ");
        System.out.println(Arrays.toString(c)); // c (=d):  [0, 0, 0, 0]


        System.out.print("d (=1):  ");
        System.out.println(Arrays.toString(d)); // d (=1):  [1, 1, 1]

        System.out.println("================");

        int[] a = {1, 2, 3};
        int[] b = {6, 5, 4, 3};

        System.out.print("a:       ");
        System.out.println(Arrays.toString(a)); // a:       [1, 2, 3]


        System.out.print("b:       ");
        System.out.println(Arrays.toString(b)); // b:       [6, 5, 4, 3]

        a = b;
        System.out.print("a (=b):  ");
        System.out.println(Arrays.toString(a)); // a (=b):  [6, 5, 4, 3]

        System.out.println("--- α = 0 ---");
        Arrays.fill(a, 0);

        System.out.print("a (=0):  ");
        System.out.println(Arrays.toString(a)); // a (=0):  [0, 0, 0, 0]
        System.out.print("b (=a?): ");
        System.out.println(Arrays.toString(b)); // b (=a?): [0, 0, 0, 0]    ???

        System.out.println("--- b = 1 ---");
        Arrays.fill(b, 1);
        System.out.print("b (=1):  ");
        System.out.println(Arrays.toString(b)); // b (=1):  [1, 1, 1, 1]
        System.out.print("a (=b?): ");

        System.out.println(Arrays.toString(a)); // a (=b?): [1, 1, 1, 1]
    }
}

The swapability of c and d indicates pass-by-value according to this post: Java is Pass-by-Value, Dammit!. (I also looked at java array pass by reference does not work?, but I can't understand the asker's English, and the method call obscures the example.)

Notice that with the line d = tmp; commented out, c and d exhibit the same odd behavior as a and b. Still I don't know what to make of it.

Can anyone explain how a and b's behavior can be explained with pass-by-value?

It turns out the main issue in my post is not pass-by-value, but aliasing. To be clear about the distinction between pass-by-value and pointers, I added the following method to my code and used it to (try to) swap c and d (suggested by an article linked by JavaDude's article linked above).

static  void swap (T c, T d) {
    T tmp = c;
    c = d;
    d = tmp;
}

The result is that c and d come back unchanged. This would have worked if Java (like C) passed along pointers to c and d to the method, but instead it simply passes their values, leaving the original variables unchanged.

Changing a = b to a = b.clone(); or to a = Arrays.copyOf(b, b.length); gives the behavior I was expecting. This code also works:

    int[] tmp = new int[b.length];
    System.arraycopy( b, 0, tmp, 0, b.length );
    a = tmp;

Relative timing descried here.

Answer

There is nothing "weird" going on here: array variables are references to the actual arrays (also known as pointers in other languages). When you manipulate array variables, all you do is manipulating pointers.

When you assign an array variable to another one, you create an alias to the array pointed to by the variable you assign, and make the array previously pointed to by the variable being assigned eligible for garbage collection. Because the assignment a = b makes a an alias of b, filling b with data acts exactly the same as filling a with data: once the assignment is complete, a and b are merely two different names for the same thing.

As far as pass by value is concerned, none of it is going on in your example: the concept of passing by value applies only when you pass objects as parameters to the methods that you call. In your example, variables a, b, c, and d are not method parameters, they are local variables. You do pass them by reference to methods toString and fill (or more precisely, you pass by value the references to your objects to toString and fill, because in Java everything is passed by value), that is why modifications to your arrays done by fill are visible upon the return from the method.