When the above code is run, arrayMaster changes as well as arrayChanges, contrary to my intentions.
The line
static private int[][] arrayChanges = arrayMaster;
is the culprit. This line makes arrayChanges and arrayMaster point to the same object, so a change to either one is visible when you access the object from either.
EDIT: What happens whenever you clone one dimension of a multidimensional array
As Eric Lippert explains, an array is conceptually a list of variables. If you just assign another variable to point to the same array a la static private int[][] arrayChanges = arrayMaster;, you haven’t changed the set of variables at all. You haven’t created any new variables except for arrayChanges, so you haven’t gotten more memory from the operating system/JVM, so any change you make to arrayMaster is applied to arrayChanges and vice versa.
Now let’s look at a two-dimensional array. In Java, a two-dimensional array is a list of variables that happens to have the property that each one of these variables refers to a one-dimensional array. So, whenever you clone a two-dimensional array, you create a new list of variables, each pointing in the same place that the old variables pointed in. So, you have gained a little in that you can safely write arrayChanges[0] = new int[10] without affecting arrayMaster, but as soon as you start referencing arrayChanges[i][j] you are still referencing the same second-level arrays that arrayMaster references. What you really want in order to deep-copy a two-dimensional array of ints is
public static int[][] deepCopyIntMatrix(int[][] input) {
if (input == null)
return null;
int[][] result = new int[input.length][];
for (int r = 0; r < input.length; r++) {
result[r] = input[r].clone();
}
return result;
}
To those who may look at this answer in the future: yes, it is better replace int with T here and make the method generic, but for this purpose a more concrete deep copy method is simpler to explain well.