Class Instance Examples

More Getters and Setters

As an exercise you could write a simple class Example, with

  1. int instance variable n and double instance variable d

  2. a simple constructor with parameters to set instance variables n and d

  3. getters and setters for both instance variables (4 methods in all)

  4. a ToString that prints a line with both instance variables labeled.

Compare yours to example_class/example_class.cs.

We include a testing class at the end of this file.

class ExampleTest
   public static void Main()
      Example e = new Example(1, 2.5); // use constructor
      // this creates the first Example object with reference e
      Console.WriteLine("e.n = {0}", e.GetN()); // prints 1
      Console.WriteLine("e.d = {0}", e.GetD()); // prints 2.5
      Console.WriteLine(e); // prints Example: n = 1, d = 2.5

      Console.WriteLine("e.n = {0}", e.GetN()); // prints 25
      Console.WriteLine("e.d = {0}", e.GetD()); // prints 3.14159
      Console.WriteLine(e); // prints Example: n = 25, d = 3.14159

      Example e2 = new Example(3, 10.5);
      // this creates the second Example object with reference e2
      Console.WriteLine(e2); // prints Example: n = 3, d = 10.5

      e = e2; // now both e and e2 reference the second object
      // the first Example object is now no longer referenced
      // and its memory can be reclaimed at runtime if necessary
      Console.WriteLine(e); // prints Example: n = 3, d = 10.5

      e2.SetN(77);  // symbolism uses e2, not e
      Console.WriteLine(e2); // prints Example: n = 77, d = 10.5
      Console.WriteLine(e); // prints Example: n = 77, d = 10.5
      // but e is the same object - so its fields are changed
} // end of class ExampleTest

Besides the obvious tests, we also use the fact that an Example object is mutable to play with References and Aliases: In the last few lines of Main, after e becomes an alias for e2, we change an object under one name, and it affect the alias the same way. Check this by running the program!

Make sure you can follow the code and the output from running.

Beyond Getters and Setters

Thus far we have had very simple classes with instance variables and getter and setter methods, and maybe a ToString method. These classes were designed to introduce the basic syntax for classes with instances. The classes have just been containers for data that we can read back, and change if there are setter methods - pretty boring and limited. Many objects have more extensive behaviors, so we will take a small step and imagine a somewhat more complicated Averager class:

  • A new Averager starts with no data acknowledged.

  • Be able to enter data values one at a time (method AddDatum). We will use double values.

  • At any point be able to return the average of the numbers entered so far (method GetAverage). The average is the sum of all the values divided by number of values. With double values we assume a double average. This does not make sense if there are no values so far, but with double type we can use the value NaN (Not a Number) in this case.

  • Be able to return the number of data elements entered so far (method GetDataCount)

  • Be able to clear the Averager, going back to no data elements considered yet, like in a new Averager (method Clear)

  • It is good to have a ToString method. We choose to have it label the number of data items and the average of the items.

The object starts from a fixed state (no data) so we do not need any constructor parameters.

We can imagine a demonstration class AveragerDemo with a Main method containing

         Averager a = new Averager();
         Console.WriteLine("New Averager: " + a);
         foreach (double x in new[] {5.1, -7.3, 11.0, 3.7}) {
            Console.WriteLine ("Next datum " + x);
            a.AddDatum (x);
            Console.WriteLine("average {0} with {1} data values",
                              a.GetAverage(), a.GetDataCount());
         Console.WriteLine("After clearing:");
         Console.WriteLine("average {0} with {1} data values",
                           a.GetAverage(), a.GetDataCount());

It should print

New Averager: items: 0; average: NaN
Next datum 5.1
average 5.1 with 1 data values
Next datum -7.3
average -1.1 with 2 data values
Next datum 11
average 2.93333333333333 with 3 data values
Next datum 3.7
average 3.125 with 4 data values
After clearing:
average NaN with 0 data values

Now that we have a clear idea of the proposed outward behavior, we can consider how to implement the Averager class.

We could store a list of all the data values entered in any instance, requiring a large amount of memory for a long list. However, this functionality was built into early calculators, with very limited memory. How can we do it without remembering the whole list? Consider a concrete example:

If I have entered numbers 2.1, 4.5 and 5.4, and want the average, it is


If I want to include a further number 5.0, the average becomes


Note the relationship to the previous average expression:


In the numerator we have added the latest value to the previous sum, and in the denominator the count of data items is increased by one. All we need to remember to go on to include the next item is the sum of values so far and the number of values so far. We can just have instance variables sum and dataCount.

You might think how to create this class….

The full Averager code follows:

using System;
namespace IntroCS
/// a class that is more than a container
   class Averager
      private int dataCount;
      private double sum;

      /// new Averager with no data
      public Averager()

      public void AddDatum(double value)
         sum += value;

      public int GetDataCount()
         return dataCount;

      /// Gets the average of the data
      ///   or NaN if no data.
      public double GetAverage()
         return sum/dataCount;  // is NaN if dataCount is 0

      public void Clear()
         sum = 0.0;
         dataCount = 0;

      public override string ToString ()
         return string.Format("items: {0}; average: {1}",
                              GetDataCount(), GetAverage());

Several things to note:

  • We show that a constructor, like an instance method, can include a call to a further instance method. Though we illustrate this idea, the constructor code is actually unneeded. See the Unneeded Constructor Code Exercise below.

  • We have methods that are not ToString or mere getters or setters of instance variables. The logic of the class requires more.

  • The GetAverage method does return data obtained by reading instance variable, but it does a calculation using the instance variables first. See Alternate Internal State Exercise.

The code for both classes is in project averager.

Statistics Exercise

Modify the project averager so the Averager class is converted to Statistics. Besides having methods to calculate the count of data items and average, also calculate the standard deviation with a method StandardDeviation. It turns out that the only other instance variable needed is the sum of the squares of the data items, call it sumOfSqr. Before calculating the standard deviation, suppose we assign the current average to a local variable avg. Then the handiest form of expression for the standard deviation is

Math.Sqrt((sumOfSqr - avg*avg)/dataCount)

Modify the demonstration program to show the standard deviation, too.

Unneeded Constructor Code Exercise

Recall that objects are always initialized. Each instance variable has a default value assigned before a constructor is even run. The default value for numeric instance variables is 0, so the call to Clear in the constructor could be left out, leaving the body of the constructor empty! Try commenting that line out and test that the behavior of demo program is the same.

Emphasizing the fact that you are thinking about the initial values of instance variables is not a bad idea. Hence a common practice is to explicitly assign even the default values in the constructor, as we did initially with the call to Clear inside the constructor.

If no constructor definition is explicitly provided at all, the compiler implicitly creates one with no parameters and an empty body. This means that the entire constructor in Averager could be omitted. Comment the whole constructor out and check.

There is a defensive programming reason to provide even the do-nothing constructor explicitly: If you use the implicit constructor and then decide to add a constructor with parameters, the implicit constructor is no longer defined by the compiler, so any remaining call to it in your code becomes an error.

Alternate Internal State Exercise

The way we represent the internal state for an Averager is the best probably, but if it turns out that the GetAverage method is called a lot more often than a method that changes the state, we could avoid repeated division by saving the average as an instance variable. We could keep that instead of sum (and still keep dataCount). We can manage to keep the same public interface to the methods. With these alternate instance variables how would you change the implementation code and not change the method headings or meanings? If we keep the assumption that the average of 0 items is double.Nan, you will need to treat adding the first datum as a special case. The code is simpler if we change the outward assumptions enough to consider the average of 0 items to be 0. Try it either way.

In the version with NaN you can avoid testing for NaN, but if you choose to test for NaN, you need the boolean Double.IsNaN function, because the expression double.NaN == double.NaN is false!

Converting A Static Game To A Game Instance

For a comparison of procedural and object-oriented coding, consider converting Number Guessing Game Lab so that a GuessGame is an object, an instance of the GuessGame class.

While our earlier example, Contact, is a simple but practical use of object-oriented programming, GuessGame is somewhat more artificial. We create it hoping that highlighting the differences between procedural and object-oriented presentation is informative. Also, we will see with Interfaces that there are C# language features that require an object rather than just a function and data. In IGame Interface Exercise you can use a Game object.

Here is a procedural game version, example file static_version/static_version.cs

   public class Game
      private static Random r = new Random();

      public static void Main()
         int big = UI.PromptInt("Enter a secret number bound: ");

      public static void Play(int big)
         int secret = r.Next(big);
         int guesses = 1;
         Console.WriteLine("Guess a number less than {0}.", big);
         int guess = UI.PromptInt("Next guess: ");
         while (secret != guess) {
            if (guess < secret) {
               Console.WriteLine("Too small!");
            } else {
               Console.WriteLine("Too big!");

            guess = UI.PromptInt("Next guess: ");
         Console.WriteLine("You won on guess {0}!", guesses);

The project also refers to the library class UI, with the functions we use for safe keyboard input. It is all static methods.

Is there any reason to make this UI class have its own own instances?

No. There is no state to remember between UI method calls. What comes in through the keyboard goes out through a return value, and then you are completely done with it. A simple static function works fine each time. Do not get fancy for nothing.

What state would a game hold? We might set it up so the user chooses the size of the range of choices just once, and remember it for possibly multiple plays of the GuessGame. The variable was big before, we can keep the name. If we are going to remember it inside our GuessGame instance, then big needs to become an instance variable, and it will be something we can set in a constructor.

What actions/methods will this object have? Only one - playing a GuessGame. The GuessGame could be played multiple times, and that action, play, makes sense as a method, Play, which will look a lot like the current static function.

In the procedural version there are several other important variables:

  • Random rand: That was static before, for good reason: We only need one Random number generator for the whole time the program is running, so one static variable makes sense.

  • The central number in the procedural Game and our future Play method is secret. Should that be an instance variable? It would work, but it would be unhelpful and misleading: Secret is reset every time the game is played, and it has no meaning after a Play function would be finished. There is nothing to remember between time you Play. This is the perfect place for a local variable as we have now.

A common newbie error is to make things into instance variables, just because you can, when an old-fashioned local variable is all that you need. It is good to have variables leave the programmer’s consciousness when they are no longer needed, as a local variable does. An instance variable lingers on, leaving extra places to make errors.

This introductory discussion could get you going, making a transformation. Go ahead and make the changes as far as you can: create project GuessGame inside the current solution. Have a class GuessGame for the GuessGame instance, with instance variable big and method Play.

You still need a static Main method to first create the GuessGame object. You could prompt the user for the value for big to send to the constructor. Once you have an object, you can call instance method Play. What about parameters? What needs to change from the procedural version?

There is also a video for this section that follows all the way through the steps. A possible final result is in instance_version/guess_game.cs.

Animal Class Lab

Objectives: Complete a simple (silly) class, with constructor and methods, including a ToString method, and a separate testing class.

Make an animal_lab project in your solution, and copy in the files from the example project animal_lab_stub. Then modify the two files as discussed below.

  1. Complete the simple class Animal in your copy of the file animal.cs. The bullets below name and describe the instance variables, constructor, and methods you need to write:

    • An Animal has a name and a gut. In our version the gut is a List of strings describing the contents, in the order eaten. A newly created Animal gets a name from a parameter passed to the constructor, while the gut always starts off empty.

    • An Animal has a Greet method, so an animal named “Froggy” would say (that is, print)

      Hello, my name is Froggy.

    • An Animal can Eat a string naming the food, adding the food to the gut. If Froggy eats “worm” and then “fly”, its gut list contains “worm” and “fly”.

    • An Animal can Excrete (removing and printing what was first in the gut List). Recall the method RemoveAt in List Syntax. Print the empty string, “”, if the gut was already empty. Following the Froggy example above, Froggy could Excrete, and “worm” would be printed. Then its gut would contain only “fly”.

    • A ToString method: Remember the override keyword. Make it return a string in the format shown below for Froggy, including the Animal’s name:

      “Animal: Froggy”

      Try this first, and note the extra credit version below.

    • All the methods that print should be void. Which need a parameter, of what type?

  2. Complete the file test_animal.cs with its class TestAnimal containing the Main method, testing the class Animal: Create a couple of Animals and visibly test all the methods, with enough explanation that someone running the test program, but not looking at the code of either file, can see that everything works.

  3. 20% EXTRA CREDIT: Elaborate ToString so if Froggy had “worm”, “fly” and “bug” in the gut, the string would be:

    “Animal: Froggy ate worm, fly and bug”

    with a comma separated list of the gut contents, except use proper English, so the last separator is ” and “, not “, “. If the gut has nothing in it, list the contents as “nothing”:

    “Animal: Froggy ate nothing”

Clock Example

Consider the logic for a digital 24 hour clock object, type Clock, that shows hours and minutes, so 03:45 would be three forty-five. Note that there is no AM or PM: The hours go from 00, starting at midnight, through hour 23, the 11PM hour, so 23:59 would be a minute before midnight, and 13:00 would be 1PM.

Assume there is some attached circuit to signal when a new minute starts.

This class could have just a few methods: Tick, called when a new minute is signaled, and GetTimeString to return the time in the format illustrated above, and SetTime specifying new values for the hours and minutes. We can start from a constructor that just sets the clock’s time to midnight.

We can imagine a demonstration class ClockDemo with a Main method containing

         Clock c = new Clock();
         Console.WriteLine("Midnight " + c.GetTimeString());
         c.SetTime(23, 58);
         Console.WriteLine("Before midnight " + c.GetTimeString());
         for (int i = 0; i < 4; i++) {
            c.Tick ();
            Console.WriteLine ("Tick " + c.GetTimeString());

It should print

Midnight 00:00
Before midnight 23:58
Tick 23:59
Tick 00:00
Tick 00:01
Tick 00:02

A Clock object will need instance variables. One obvious approach would be to have int instance variables for the hours and minutes. Both can be set and can advance and will need ot be read.

These actions are common to both the hours and minutes, so we might think how we can avoid writing some things twice. There is one main difference: The minutes roll over at 60 while the hours roll over at 24. Though the limits are different, they are both numbers, so we can store the limit for each, 60 or 24. Then the same code could be used to advance each one, just using a different value for the rollover limit.

How would we neatly code this in a way that reuses code? The most significant thing to notice is that dealing with minutes involves data (the current count and the limit 60) and associated actions: being set, advanced and read. The same is true for the hours. The combination of data and tightly associated actions, particularly used in more than one situation, suggests a new class of objects, say RolloverCounter.

Notice the shift in this approach: The instance variables for hours and minutes would become instances of the RolloverCounter class. A RolloverCounter should know how to advance itself. Hence the logic for advancing a counter, sometimes rolling it over, would not be directly in the Clock class, but in the RolloverCounter class.

So let’s think more about what we would want in the RolloverCounter class. What instance variables? Of course we have the current count, and since we want the same class to work for both minutes and hours, we also need to have the rollover limit. They are both integers.

The limit should just be set once for a particular counter, presumably when the object is created. For simplicity we can just assume the count is 0 when a RolloverCounter is first created. Of course we must have a method to let the count advance, rolling over back to 0 when the limit it reached.

Throw in a getter and a setter for the count and we can have the following class:

using System;
namespace IntroCS
/// class used twice in Clock
   class RolloverCounter
      private int limit, count;

      public RolloverCounter(int limit)
         this.limit = limit;
         count = 0;  //for clarity - this is the default value

      public int GetCount()
         return count;

      public void SetCount(int count)
         this.count = count;

      /// advance by one time tick
      /// eventually roll over at limit
      public void Advance()
         count = (count + 1) % limit;

Note how concise the Advance method is! With the remainder operation, we do not need an if statement. Check examples by hand if this seems strange.

Finally we introduce the Clock class itself. We display the entire code first, and follow it with comments about a number of new features.

using System;
namespace IntroCS
/// class with instance variables of another user class
   class Clock
      private RolloverCounter hours, minutes;

      /// new Clock set to midnight
      public Clock()
         hours = new RolloverCounter(24);
         minutes = new RolloverCounter(60);

      ///  new Clock set to specified time
      public Clock(int nHours, int nMinutes)
         hours = new RolloverCounter(24);
         minutes = new RolloverCounter(60);
         SetTime (nHours, nMinutes);

      public void SetTime(int nHours, int nMinutes)

      /// advance by one time tick
      public void Tick()
         if (minutes.GetCount() == 0) {

      /// Always 2 digits for both hours and minute with colon in middle
      public string GetTimeString()
         return string.Format("{0:D2}:{1:D2}",
            hours.GetCount(), minutes.GetCount());
  1. First the principal reason for this example: We illustrate writing a class where the instance variables are objects of a different user-defined type. Because the instance variables hours and minutes are objects, we must initialize them using the new syntax.

  2. Skip over the second constructor for now, and see the SetTime method: We call the appropriate method to update the individual RolloverCounter instances.

  3. Now go back to the second constructor. This is not really necessary: With the first constructor the calling code could just have one more SetTime line any time you want to to create a clock with a time other than midnight. We can make a case for this being so common, that we want to do it in just one line, with a constructor that sets a specified time. However, the main excuse was really to illustrate that constructors can be overloaded, like methods: You can have separate constructors with distinct signatures. In this case versions with no parameters vs. two int parameters.

  4. The Tick method has a bit of logic to it: while the minutes always advance, the hours only advance when the minutes roll over to 0.

  5. Finally the GetTimeString method illustrates a new integer string formatting mode: The D2 format specifier applies to an integer, and displays it as a minimum of 2 digits, padding on the left with 0’s as necessary. This is just what we want here. In general the 2 could be replaced by another literal integer, so D6 would force at least 6 digits: With the D6 format specifier 12 would be formatted as 000012, and the longer 1234567 would add no extra 0’s: still 1234567.

The code for all the classes is in project clock.

Admittedly, with this exact functionality and such a concise line to advance a count, it would actually have shorter to have done everything inside the Clock class, with no RolloverCounter, but we were looking for a simple illustration of combining user-defined types this way, and a RolloverCounter is a clear unified concept that can be used in other situations. See an upcoming exercise.

Alternate Clock Constructor Exercise

Make a small change to clock/clock_demo.cs, so the second constructor is tested.

Clock With Seconds Exercise

Modify the project clock, assuming the Tick is for each second, and the time also show the seconds, like 55 seconds before midnight would be 23:59:05.

Twelve Hour Time Exercise

Modify the project clock so a GetTimeString12 method returns the 12 hour time with AM or PM, like 11:05PM or 3:45AM. (The hours do not have a leading 0 in this format.) This could be done modifying a lot of things: keeping the actual hours and minutes that you will display and remembering AM or PM (with the hours being more complicated, not starting at 0). We suggest something else instead:

This is a good place to note a very useful pattern for programming, called model-view-controller. The model is the way chosen to store the state internally. The controller has the logic to modify the model as it needs to evolve. A view of a part of the model is something shown to the user that does not need to be in the exact same form as the model itself: A view just needs to be something that can be easily calculated from the model, and presumably is desired by the user.

In this case a simple (and already coded!) way to store and control the time model data is the minutes and up to 23 hours that do happen to directly correspond to the 24 hour clock view.

The main control is to advance the time, and with just two 0-based counts we have the very simple remainder formulas.

So the suggestion is to keep the internal data the same way as before. Just in the method to create the desire 12-hour view have the logic to do the conversion of the internal 24-hour model data.

You could leave in the method to provide the time in the 24 hour format, giving the Clock class user the option to use either view of the shared model data. To be symmetrical in the naming, you might change the original name GetTimeString to GetTimeString24.