In Java, we have seen and developed methods that took in no arguments or several arguments.  But we have not seen any examples of methods that take in other methods as arguments.  This is because Java does not allow this.  Methods can only take objects and primitives as arguments.  Scheme, on the other hand, does not have this restriction.  As it would be very useful to somehow pass methods to methods, we need to come up with a way to get around this restriction.  The solution is to design classes whose sole purpose is to hold the methods that we want to pass to other methods.  In the object-oriented design literature, this technique is called the command pattern.  Here we are using this pattern in a purely functional fashion.  (Source: www.cs.rice.edu/~cork/teachjava/2002/notes/current/node52.html)     

PART 1:

You will see examples of two methods, mapIntToInt and orMapInt that require some method be passed to them.  When you have finished PART 1, you should understand how interfaces and concrete subclass are created to represent the method that needs to be passed as an argument to other methods.

The method mapIntToInt requires as an argument a method of type (int àint) while orMapInt requires a method of type      (int à boolean).  Since these two arguments are different types, this means that we will need two separate interfaces:     

1) Open the Lab 7 – Part 1 project.

2) The method mapIntToInt belongs to the AListInt class.  Its purpose is to use this (which is a list of int) and the passed class (which represents a method) to produce a new list of int.  The new list of int that is produced depends entirely on the class (which represents a method) passed.  For example, suppose we wanted to take a list of int and pass a method (via a class) that multiplies each int in the list by 3.  The list (2, -5, 0, 8) would give us the new list (6, -15, 0, 24).  Still another example would be to pass a method (via a class) that squares each int in the list.  Our list example above would give us the new list (4, 25, 0, 64).  There are many other examples like this that you can think of.  For this lab, we will work with the two classes (which ultimately will represent methods) called AddOne and AbsVal.  AddOne will represent the method that adds 1 to each member of the int list.  AbsVal will represent the method that produces a similar list to the original, except each int is replaced by its absolute value.

3) Open only the interface called IInt2Int, the family of AListInt classes and the class called AddOne and AbsVal.   Since AddOne and AbsVal will represent methods of type (int à int), we created an interface called IInt2Int [which you can think of as standing for “Interface for methods of type (int à int)”].  This interface will hold only one method called apply.  Notice that apply is also of type (int à int).  It is this method that AddOne and AbsVal will implement differently to achieve the effect of “adding one” and “absolute value”.  Think of apply as a place holder for any method of type (int à int). 

In AddOne, we implement IInt2Int interface and write up the method apply as required by implementing IInt2Int.  In this version of apply, we see the code that actually adds 1 to an integer.  In AbsVal, everything is exactly the same as AddOne except the code for apply.  This version of apply uses the Java built in method called Math.abs to return the absolute value of an int.

4) In the class ConsInt, the method mapIntToInt takes in an object of type IInt2Int (which will be either AddOne or AbsVal) and then uses the apply method to process recursively this (which is a AListInt) to get a new list which depends entirely on the argument of type IInt2Int.  Let’s take a closer look at the body of the mapIntToInt:

                                     return new ConsInt(i2i.apply(this.first), this.rest.mapIntToInt(i2i));

i2i is the name of the object of type IInt2Int that was passed to the method mapIntToInt.  It will be either AddOne or AbsVal.  apply is a method in i2i, so i2i.apply is the apply method of class AddOne or AbsVal and (this.first) is the argument to that method.  this.rest.mapIntToInt(i2i) is the recursive call to mapIntToInt(i2i) using this.rest and we put this altogether with new ConsInt because we want to return a list of integers.

5) In the class EmptyInt, the method mapIntToInt has only an empty list to work with, so the only sensible thing to return is this (the EmptyInt class).

6) Open the class called AListIntTest.  In section A2 you will find an instance of the class AddOne called Add1 and in section A3, you will find an instance of the class AbsVal called AbsValue.  These are used in the test cases of section A5.  Go ahead and run these two test cases.

7) You will now design a class called SubOne which will resemble AddOne except that its version of the method apply will return a number one less than its input.  The outline of the class SubOne already exists, so you only need to replace the bogus code that is sitting in the apply method.  Is the SubOne class already implementing IIint2Int?  If not, then make it so. 

8) Create an example of SubOne as indicated in section A4.  Comment out the existing test cases of section A5 of the class AListIntTest and create 2 test cases for Sub1 as indicated in A5.  Go ahead and run these test cases to see if you get the correct results. 

9) Leave open the family of AListInt classes and AListIntTest class. Close all others.  Open the interface called IInt2Bool and the class called IsEven.  

10) The method orMapInt also belongs to the AListInt class.  Its purpose is to use this (which is a list of int) and the passed method (via the class) to produce a boolean.  The boolean that is returned depends entirely on the class (which represents the method) that is passed.  For example, suppose we have a method that inputs a int and returns true if the int is greater than 6 and false otherwise.   Then our method orMapInt would use that method (via the class) and this to return true if ANY of the int are greater than 6.  Since the method (via the class) that needs to be passed to orMapInt is of type (int à boolean), we develop an interface called IInt2Bool [which you can think of as standing for “Interface for methods of type (int à boolean)”].  As we have seen before, this interface will hold only one method called apply.  Notice that apply is also of type (int à boolean).  It is this method that IsEven will implement to achieve the effect of “checking for evenness”.  Again, think of apply as a place holder for any method of type (int à boolean). 

In IsEven, we implement IInt2Bool interface and write up the method apply as required by implementing IInt2Bool.  In this version of apply, we see the code that actually checks if an int is even.  This code just checks whether the reminder after division by two is equal to zero.

11) In the class ConsInt, the method orMapInt takes in an object of type IInt2Bool (which is IsEven) and then uses the apply method to process recursively this (which is a AListInt) to get a boolean which depends entirely on the argument of type IInt2Bool.  Let’s take a close look at the body of the orMapInt:

                                     return i2b.apply(this.first) || this.rest.orMapInt(i2b));

i2b is the name of the object of type IInt2Bool that was passed to the method orMapInt.  It will be IsEven.  apply is a method in i2b, so i2b.apply is the apply method of class IsEven and (this.first) is the argument to that method.  this.rest.orMapInt(i2b) is the recursive call to orMapInt(i2b) using this.rest and we put this altogether with || because we want to return a boolean if any i2b.apply(this.first) is ever true.

12) In the class EmptyInt, the method orMapInt needs to return either true or false.  Since orMapInt will return true if it finds at least one true, then the answer for EmptyInt should be one that does not affect any of the booleans returned recursively in the ConsInt list.  So the value should be false.  If this is not clear, then think about what the answer should be for the EmptyInt if the ConsInt consisted of all negative int and i2b.apply was the method that checked for non-negative int.  The answer to orMapInt should be false because no int on this list is 0 or positive.  But orMapInt will return true if the answer for EmptyInt is true because it takes only one true to make an or statement true.

13) Open the class called AListIntTest.  In section B2 you will find an instance of the class IsEven called AllEven.  This is used in the test cases of section B4.  Go ahead and run these three test cases.

14) You will now design a class called IsPositive which will resemble IsEven except that its version of the method apply will return a boolean if any the int are greater than 0.  The outline of the class IsPositive already exists, so you only need to replace the bogus code that is sitting in the apply method.  Is the IsPositive class already implementing IIint2Bool?  If not, then make it so. 

15) Create an example of IsPositive as indicated in section B3.  Comment out the existing test cases of section B4 of the class AListIntTest and create 3 test cases for AnyPositive as indicated in B4.  Go ahead and run these test cases to see if you get the correct results.

 

PART 2:

We saw in PART 1 that before we could test orMapInt and mapIntToInt we needed to create instances of the IsEven, AddOne and AbsVal classes.  But this in turn required the creation of the concrete classes IsEven, AddOne and AbsVal classes.  Is there anyway to simplify this?  When you have finished PART 2, you will see that we can indeed simplify this requirement of creating concrete classes by the use of anonymous inner classes.

16) Open the Lab 7 – Part 2 project.

17)  You will notice that all of the previous classes are still the same EXCEPT that IsEven, AddOne and AbsVal classes are missing!  If these classes are missing, how can we write something like:                                            

AbsVal   AbsValue  = new AbsVal()

AddOne   Add1 = new AddOne()

IsEven   AnyEven = new IsEven()

in the AListIntTest class?

Before we can explain how to actually do this, we need to explain how we create instances of anonymous classes.  Suppose we have an interface called Command with one method called execute of type (String à boolean) that does something with a string, but we don’t have any concrete classes that actually implement Command.  When we want to create an instance of an object of type Command where execute returns true if the given String has length less than 6, we do it as follows:

Command  IsLessThan6  =  new Command( ) {

                                                 boolean execute (String givenString) {

                                                     return (givenString.length( ) < 6);

                                                    }

                                                 }

Now you can you use IsLessThan6 as if it was an object in a concrete class with some name that you had created, which extends the interface Command.

So how do we create an instance of IInt2Int that is just like AbsVal?

IInt2Int  AbsValue =  new IInt2Int( ) {

                                                 int apply (int number) {

                                                     return Math.abs(number);

                                                    }

                                                 }

Where did we get the code for the body of apply?  From the method apply in the concrete class called AbsVal from PART 1!  Go ahead write this up in A3 and run the test cases in A5 to see if you get the correct results.

18)  In A2 and A4, write up instances of IInt2Int that are just like A2 and A4 of PART 1.  Go ahead and run the test cases in A5 to see if you get the correct results.

19) In B2 and B3, write up instances of IInt2Bool that are just like B2 and B3 of PART 1.  Go ahead and run the test cases in B4 to see if you get the correct results.

20) Would you believe that we can do even better than this?  If we are not interested in reusing the anonymous class we create, then we can forgo giving it a name and use it directly in a test case:

actual (List4.mapIntToInt (new IInt2Int( ) {

               int apply (int number) {

                   return Math.abs(number);

               }

          }) );

21) Redo all your test cases from section A and section B so that all of your test cases look like number 20) above.  This means commenting all anonymous inner classes that have names assigned to them [as in number 17) above].

22) Can you make up an instance of IInt2Int and IInt2Bool in which the body of apply is different that anything you have seen today?  Here are some possible ideas for IInt2Int: doubling each number, reversing the sign of each number, cubing each number and then subtracting one.  Some ideas for IInt2Bool: is each number in the list odd, is each number in the list a perfect square, is each number in the list 1 more than a multiple of 4?

23) Can you make up an interface called IString2Bool of type (String à boolean) and a method called mapStringToBool in a family of classes called AListString that is similar to what we have seen with IInt2Bool, mapIntToBool and the AListInt family of classes?