CSC 241- Lec 5 (July 15th 1996)


Encapsulation+Inheritance

The major advantages to Object Orientation are encapsulation and inheritance.

We have already discussed encapsulation when we discussed the class Tank. An object or an instance of class tank not only has the content of a tank held inside it, it has the methods that operate on or query the content inside it also. This way, no matter where in the application that uses tanks trys to add or remove liquid, we can be sure content is never left in an abnormal state. Remember that we throw exceptions if the application attempts to over fill the tank or take too much out of the tank. This simple solution saves a great deal of our debugging time, as we would not have designed tank this way if we were programming imperatively with side-effects. We would have probably had a content variable for each tank maintained and would have increased or decreased its content with assignment statements right in the application. So, say the content has become negative due to a bad operation, most often, we wouldn't figure that out until much later. These bugs used to be difficult to find and very time consuming. The way that tank is designed here enables us to detect the error where it happens. I want to make it clear that this doesn't mean we were stupid before and with object orientation we have become smart. It simply means that we avoid the whole issue of forgetting to check for content having become instable, the object itself won't let you forget by throwing an exception.

So, it is very useful for tank objects to protect their content. We have achieved that by encapsulating data and methods, including exception handling mechanism that we created for tank in each instance/variable/object (any of those names are ok here) of class Tank.

We first talked about Inheritance when we discussed the exception handling features of Java. Recall that in Java we create exception classes such as TankOverFlowException that are subclasses of a super class called Exception. It is not hard to think of why super classes such as Exception are useful. For one thing, it provides uniformity to the way we handle abnormal cases. Any of our methods that throw an exception force the application that invokes them to catch that exception. This is something that if the application programmer forgets to deal with, he or she will be reminded with a compilation error. Since all exceptions are designed in similar fashion, we have a sense of familiarity when we look at new classes that throw them. That feeling of familiarity is part of the draw for object orientation, it prevails all over the place.

A class hierarchy for Queues

We first looked at queues in stringQueue.

      public abstract class stringQueue {
         public abstract boolean empty();
         public abstract boolean full();
         public abstract void enqueue(String x);
         public abstract String dequeue();
      }

      

A couple of things to note here, the word abstract implies that this class won't have any variables (i.e. won't hold any data that changes) and its methods won't have any code. Note that I have left off a front method that would return the string in front of the queue without removing it which is sometimes useful. Also that no exception is thrown for dequeuing when empty. These can certainly be added in a lab at a later time.

What stringQueue does for our applications is that we can declare an object of it with a statement such as, stringQueue q;. Understand the difference between declaring and constructing. When we declare q, we simply saying that it will be a stringQueue. You can see that there is no constructor in stringQueue, so, you can't say something like stringQueue q = new stringQueue(...);. If you think about it, it all makes sense--if you don't have any variables in this class then what could you construct?

Obviously, we need code in order for our q variable to actually behave like a queue. We also need it to have some way of storing the strings that we queue into it. This happens in concrete classes that inherit stringQueue. Concrete refers to classes that have code along with structures and variables that hold data.
fixedStringQueue class

fixedStringQueue class is provides us with one implementation for stringQueue. This class will hold the strings in an array which is of a fixed length determined when a queue is constructed. There are plenty of applications where with reasonable certainty we can choose a fixed size for our queue. For example, in an operating system where we allow for time sharing, but have a limit on how many processes can actually run at the same time. It makes sense to have a fixed length queue for processes that are waiting to run. As processes come in, they go into the queue, when it is their turn, they get taken off the queue and get processed. If a process needs more time than the system is willing to give, they get processed for the duration that they can get, and go back to the queue and wait to be processed again. A queue implemented with an array works great in this case since we have a max limit on how many processes are allowed in the system at the same time, so you never have to worry about going beyond the fixed size.

There are several variations to array implementation of a queue. The easiest is to maintain a last index that holds the position of the last element. Each time enqueue is invoked, we increment last and place the new element there. When dequeue is invoked, we remove what is in the zero cell and move all elements back by one cell. However, we won't employ this method. We are going use a circular queue. In the previous method, note that the front of the queue is stationary, always in position zero. In a circular queue, we use a last and a first index. Each time enqueue is invoked, we change last to point to the next available space and place the new element there. Note that I said point to the next available space. When dequeue is invoked, we remove the element in front and move first to next occupied cell. The obvious difference between the simple version of queue and the circular one that in the simple version, every time you dequeue an element, the rest have to move back. No such movement is needs with circular queues and thus significantly more efficient.

Moving to the next cell, in either enqueue or dequeue may require to circle around to the beginning. Let me demonstrate with an example. Assume This array of elements represents our queue with existing content and values for the first and last indexes.


  first=0                    last=3
 --------------------------------------------
| jones  | morris | vax    | fays   |        |
 --------------------------------------------
    0        1        2        3        4

Lets enqueue "new" into this queue.


  first=0                             last=4
 --------------------------------------------
| jones  | morris | vax    | fays   | new    |
 --------------------------------------------
    0        1        2        3        4

Lets dequeue one element from this queue. Note that you don't have to physically remove "jones" from the array. You can assume that "jones" is the string that we return from dequeue.


           first=1                    last=4
 --------------------------------------------
| jones  | morris | vax    | fays   | new    |
 --------------------------------------------
    0        1        2        3        4

Lets enqueue "circle" into this queue. Notice that we rap around since we have no more cells on the right, but there is available space at the beginning. It is strange to see last be behind first, but that is how circular queues work.


  last=0   first=1                    
 --------------------------------------------
| circle | morris | vax    | fays   | new    |
 --------------------------------------------
    0        1        2        3        4

It is simple to implement, The private method next is written to accomplish this for either enqueue or dequeue.


  private int next (int pos) {
    if (pos == size-1)
       return 0;      // allow rap-around
    else
       return pos+1;
  }

How is full determined? next(last) == first simply says that if last is moved to the next cell, it will run into the first.

There are a couple of other things for you to note. first and last are set to -1 when queue is empty. The constructor for the fixedStringQueue is parameterized to let the application set the size that it wants. The statement q=new String[size]; creates the array q at run time using the size provided by the application. Of course, this size is fixed for the life of the object created.