Close
Register
Close Window

CS5020 Software Design & Data Structures

Chapter 5 Introduction to Pointers in Java

Show Source |    | About   «  5.2. Basic References Part 1   ::   Contents   ::   5.4. Pointers Syntax  »

5.3. Basic References Part 2

5.3.1. Sharing

Two references which both refer to a single pointee are said to be “sharing”. Sometimes we say that each is an alias for the other, because we can refer to the referenced object through either name. That two or more references can cooperatively share a single memory structure is a key advantage of references. References second and empRef in the above example both share the same object, so either can modify the object’s value. Reference manipulation is just technique—sharing is often the real goal. Later we will see how sharing can be used to provide efficient communication between parts of a program.

5.3.2. Shallow and Deep Copying

5.3.2.1. What does it mean to copy?

An important use of sharing is to enable communication between two methods. One method passes a reference to the object of interest to another method. Both methods can access the object of interest, but the object of interest itself is not copied. This communication is called shallow copy since, instead of making and sending a (large) copy of the object of interest, a (small) reference is sent and the object of interest is shared. The recipient needs to understand that they have a shallow copy, so that they know not to change or delete it accidentally, since it is shared. The alternative—where a complete copy is made and sent—is known as a deep copy. Deep copies are simpler in a way, since each method can change their copy without interfering with the other copy. But deep copies run slower because of all the copying. And if the second method was meant to modify the copy for every user of the object, then deep copy won’t let this happen. The drawing below shows shallow and deep copying between two methods, A() and B(). In the shallow case, the smiley face is shared by passing a reference between the two. In the deep case, the smiley face is copied, and each method gets their own.


Settings

Proficient Saving... Error Saving
Server Error
Resubmit

5.3.3. Shallow and Deep Comparing

Related to copying correctly (shallow means to copy the reference itself, deep means to copy the object being referenced), there are different levels that we might use to compare things when objects are involved. One possibility is that we just want to know if two reference variables are referencing the same object. We could do this with code like the following to see if ref1 and ref2 are referring to the same object:

if (ref1 == ref2)

So here, we are comparing the value of the arrows, that both are pointing to the same box. While sometimes this is worth doing, it actually is not all that common. Any use of == with reference variables should be treated with extra suspicion.

Quite often, we want to know if the values of the two objects are the same. This is especially common when the objects are strings. We might want to know if two strings are the same, meaning they have the same characters in the same order, not whether two string reference variables are referencing the same string object. If we want to know whether the contents of two strings are the same, then in most languages we do not simply compare the two reference variables for equality. Instead, we need to do a “deep comparison” where we are looking at the characters in the strings themselves. Most languages have methods to do this for you. In Java, we can see if two strings are the same with the .equals() method of the String class, like this:

if (ref1.equals(ref2))

In Java, another option for comparing the values of two string objects is the .compareTo() method. This will return -1 if the first is less than the second in alphabetical order, 0 if they are the same, and 1 if the first is greater than the second. The following is equivalent to the use of .equals() shown above:

if (ref1.compareTo(ref2) == 0)

5.3.4. Bad References

When a reference is first allocated, it does not have a pointee. The reference is uninitialized or simply “bad”. In Java, references are actually initialized to the value null, while in some other languages they are literally of unknown value. Either way, dereferencing a bad or null reference value is a serious runtime error. The dereference operation will crash or halt immediately. Each reference must be assigned a pointee before it can support dereference operations. Before that, the reference is bad and must not be used. In our memory drawings, the bad reference value is drawn as a diagonal line between the corners of the reference variable’s box.

Bad references are common. In fact, every reference starts out with a bad value. Correct code overwrites the bad value with a correct reference to a pointee, and thereafter the reference works fine. There is nothing automatic that gives a reference a valid pointee. You just have to program carefully. Fortunately, the Java compiler will usually spot when a reference variable has not been initialized, and generate an error (meaning that you cannot even complete compiling and run the program).

On the other hand, that does not stop you from assigning the variable to something that will lead to a serious problem later. If your code is crashing, a bad reference should be your first suspicion. In Java, the run-time system checks each time that a reference variable is dereferenced to see if it is null. So code can still exhibit reference bugs, but the kind that dereferences a null value will at least halt politely on the offending line. A runtime error called NullPointerException will occur and the program will stop. As a result, it is much easier to locate and fix reference bugs in Java. Such run-time checks are also a reason why Java tends to run at least a little slower than a language like C or C++.

One way to think about reference code is that it operates at two levels—reference level and pointee level. The trick is that both levels need to be initialized and connected for things to work. (1) The reference must be allocated, (2) the pointee must be allocated, and (3) the reference must be assigned to point to the pointee. It’s rare to forget step (1). But forget (2) or (3), and the whole thing will blow up at the first dereference. For example, a popular mistake is to declare a string variable, but then never assign it an actual string before tyring to print or otherwise use it. Remember to account for both levels. Making a memory drawing during your design can help to make sure that it’s right.

Code with the most common sort of reference bug will look correct, but without the middle step where the references are assigned pointees. The bad code will compile fine, but at run-time, each dereference with a bad reference will raise NullPointerException and the program will crash. It is up to you to ensure that each reference is assigned a pointee before it is used. Here is a simple example of bad code, and a visualization of how memory would react if this code were executed.

        Employee badPointer; // Allocate the reference ...
        // badPointer = new Employee("Sam", 1000); ... but forget the pointee
        System.out.println(badPointer.getName());  // This line causes a NullPointerException
        Employee badPointer; // Allocate the reference ...
        // badPointer = new Employee("Sam", 1000); ... but forget the pointee
        System.out.println(badPointer.getName());  // This line causes a NullPointerException

Settings

Proficient Saving... Error Saving
Server Error
Resubmit

5.3.4.1. Why Are Bad Reference Bugs So Common?

In the badPointer example above, the compiler would actually catch the mistake above before it is allowed to even run, because the unitialized reference is being dereferenced. But the exact same result would happen if your program had for some reason set the value of badPointer to null. The compiler cannot catch that for you.

There must be a reason why Java cares so much about dereferencing null pointers, that its always watching out for it. Why? Because it happens in a lot of programs.

Why is it so often the case that programmers will allocate a reference, but forget to set it to refer to a pointee? Or, why will a programmer set the value of a reference to be null, and then dereference it? The rules for references do not seem that complex, yet every programmer makes this error repeatedly. Why? One explanation is that we are trained by the tools that we use. Simple variables don’t require any extra setup. You can allocate a simple variable, such as int, and use it immediately. You can change it to whatever you want, and the value won’t typically make the program crash. Try to remember not to dereference a null pointer value. But don’t be surprised when it happens, and your program breaks.

   «  5.2. Basic References Part 1   ::   Contents   ::   5.4. Pointers Syntax  »

nsf
Close Window