# Sudoku-4: Logical Structure CSCI 4526 / 6626

 To use an STL vector.
 To use an enumerated type.
 To learn about tightly coupled classes.
 To learn about circular dependencies.
 To use the this pointer.
 To model the logical structure of a Sudoku puzzle
2 Modeling the Sudoku Board
Together, the classes Board and Square model the physical structure of a Sudoku puzzle. We need one more class to model the logical structure of the puzzles: a class that represents the \row" or \column" or \box" relationship among a set of squares. We will call this class Cluster. The diagram below shows how these three classes are related.
Board
Cluster
Square
main
creates 81
9
27
3
State
A UML class diagram shows the relationships among all the classes in an application. In the diagram above, you see that main() creates the Board class. The black diamond between main() and Board indicates that main aggregates Board, and that they are allocated together and deallocated together. Similarly, Board aggregates an array of 81 Squares, so when you deallocate Board, the Square destructor is automatically called 81 times. Thus, Board is connected in the diagram to Square with a black diamond.
A Board also contains 27 Clusters. The white diamond indicates that the Clusters are created and attached to the Board after the Board is allocated. The Clusters have a shorter lifetime than the Board. They are born after the Board and die before the Board dies.
A cluster is an array of 9 pointers to Squares { all of the squares in a row or a column or a box.
It stores pointers to squares, not the squares themselves, because each Square is part of multiple clusters. Whenever the Solver gives a value to a square, the clusters are used to efficiently a nd all related squares so that the value can be removed from their possibility lists.
Each Square is part of 3 clusters (4 or 5 in some Sudoku variations) and each Cluster points to 9 squares. When two classes are mutually dependent in this way, we say they are tightly-coupled.
This is shown in the diagram above by a yellow area surrounding Square and Cluster. To compile such a pair, each class must #include the header le for the other class. However, if we put the
two #includes in the normal place at the top of the two .hpp les, this creates a circular #include pattern and the program will not compile.
We satisfy this paradox by using a partial class declaration called a forward declaration. Putthis statement at the top of Square.hpp:
class Cluster; This noti es the compiler that Cluster is a class name and allows it to compile references to Cluster pointers. The line:
#include Cluster.hpp
must then be placed at the top of Square.cpp so that the Square functions can call Cluster functions.
This is the only time you will put two include statements in the .cpp le for a class.
3 Using vectors
In a traditional Sudoku puzzle, three clusters are related to each Square. In some Sudoku variations, a square can be part of more clusters. Therefore, we need to model the relationship between a Square and a variable number of Clusters. Happily, this is easy to do using the STL vector class.
A vector is an array of objects or pointers that will grow as long as needed to contain the data you put into it. Each vector knows how many items are stored in it.
To access the data in a vector, you can use subscript, as if it were an ordinary array. We can also use vectors through variables called iterators that are very much like pointers. Here is the syntax:
 To use the vector class: #include <vector
 To declare a vector variable named \clues": vector<Cluster* clues;
 To put a new Cluster* into the vector<Cluster* named clues:
clues.push_back( aClusterPointer );
 To access an element of the array inside the vector: clues[k]. Note that you do not need iterators in this program if you use subscripts.
 To process all the data stored in a vector: for (Cluster* cl : clues) cout << cl;
4 Using enumerations
There are three types of clusters in a traditional Sudoku puzzle (more in some of the variations).
In order to make debugging easier, each cluster will store its own type (row, column, or box) as an enumeration constant. Enumeration constants are a great way to make code easier to read and less error prone. They have one great fault: you cannot input or output them directly.
The easiest way to input an enumeration constant is to input a char and use the char in a switch that stores the right enum constant in your enum variable.
If you output an enumeration constant directly using <<, you see an integer (the internal code for the constant). These integers are not very useful to someone trying to read or debug the output.
The civilized way to do the output is to de ne an array of strings and use the enum constant to subscript the array. For example, the following two declarations work properly together:
enum Colorf RED, BLUE, YELOg;
static const char* colorStrings[3];
The enum declaration belongs near the top of the .hpp le of the class that will use Colors. (Suppose that class is named Palette.) The data declaration goes inside the Palette class, and the static
Sudoku-4: Logical Structure CSCI 4526 / 6626 Fall 2016 3
initializer for it must go into a .cpp le. This can be placed at the top of Palette.cpp:
const char* Palette::colorStrings[] = f"red", "blue", "yellow"g;
5 Modify the Square Class.
 Add a vector of Cluster*.
 To Square.hpp, add a forward declaration for class Cluster.
 Add a data member: a short int for the possibility count.
 In Square.cpp, #include Cluster.hpp.
 In Square, make a public inline function, addCluster( Cluster* ) that pushes the parameter into the Square's vector.
 Add a move() function by adding a loop that will cycle through all the Clusters in the vector.
For each cluster in the vector, call the Cluster::shoop() function to remove the new digit from the possibility lists of all the neighboring squares. Delegate the actual move action to the
State class.
 Add a function: void turnO ( int n); See discussion below. Turn o position n in the square's possibility list and decrement the possibility count if position n changed. (Use case 3.) This will be called from Cluster::shoop(). The player may also turn on possibility
when he deduces that it cannot occur.
 Add code to State::print() to print the possibility count. My output looks like this:
Square [1, 1] Value: 4 Possibilities = 5: 12-45--8-
5.1 The State of a Square
You have built and tested the skeleton two classes: Square and Board This week we will add data
and function members to Square to model the relationships among squares.
Possibilities. Each square either contains a digit from 1 to 9, represented as a character, or is
empty (represented on the output as a dash). At any moment, an empty square can legally hold
only a subset of the nine digits. (It cannot hold the same digit as any other square in the same
row, column, or box.) We will model this situation by adding a data member to the State class:
the number of digits that are still possible in that square.
1 0 1 0 0 0 1 1 1
9 8 7 6 5 4 3 2 1
possible
Changing the Possibility List. Each square on the board is part of three clusters of squares,
which we will call its neighbors. These clusters represent the row, column, and box of the square.
When we mark() a square with a value (use-case 2 in the overview), the possibility lists of all its
neighbors must be adjusted (use-case 3) by turning o the bit that corresponds to the value we just
marked. I call this shooping the neighbors. (Shoop is a made-up word, a combination of swoop
and shoot.)
Sudoku-4: Logical Structure CSCI 4526 / 6626 Fall 2016 4
Did it change? When we shoop a digit from a cluster of squares, some of the squares in the
cluster will remain unchanged because that digit had been previously shooped. Other squares
will be changed, and it is essential to know which case happens. If a possibility list changes, the
associated count must be decremented. If the bit was already turned o , a decrement should not
happen. To know whether a change occurred, save the possibility list before the turn-o operation
and compare it to the result after the operation.
6 Modify the Board Class.
 At the top of Board.hpp, before the beginning of the class de nition, declare an enumeration
called ClusterT for the three types of clusters. (More types will be added later.)
 On the next line, declare a static array of three const char*, to be used in the output when you
print the cluster. Initialize this static array at the top of Board.cpp, following the example
code given above.
 Add a data member to the Board class: a vector of Cluster*.
 Modify the Board constructor to build 27 clusters. As each one is constructed, push it into
the vector. See discussion below for how to construct a cluster.
 Modify Board::print() to print the 27 clusters using Cluster::print();
6.1 Creating the Clusters
This is a long, long function. De ne it as a private helper function to be called from the Board
constructor.
To create one cluster, Board needs to create an array of nine Square*'s for the nine squares in
a row, column, or box. Then call the Cluster constructor with this array as a parameter.
Selecting the Square*s is easy for rows: write a loop to ll the parameter array with pointers
to 9 consecutive squares in the Board array, then call the Cluster constructor with this array. The
Board will use an outer loop to do this 9 times, once for each of the 9 rows.
The loop to create the parameter array for a column is only a little harder, since the subscript
of the next Square will increase by 9 each time around the inner loop. Again, you need an outer
loop to create clusters for all 9 columns.
However, for the boxes, the code will be a pain. Basically, you will need a nest of for loops.
The two outer loops will jump by 3 each time, and locate the upper-left corner of each box. The
inner loop will collect pointers to the squares in each of the 3 rows of the boxes. For those three
squares, either write a 4th level loop or write three assignment statements.
7 The Cluster Class
Implement a Cluster class with these parts:
 #include Square.hpp at the top of Cluster.hpp because a Cluster will contain an array of
Square pointers.
 The Cluster class should have these data members:
{ A const char* to store the print name of the cluster type
{ An array of 9 Square*
Sudoku-4: Logical Structure CSCI 4526 / 6626 Fall 2016 5
 The Cluster constructor will create a tightly cross-linked and ecient data structure for mak-
ing the necessary calculations during the game. It must have two parameters, a ClusterType
enum code and an array of 9 Square pointers. Use the parameters to initialize the data mem-
bers of the cluster. Then, for each Square* you add to the Cluster, use the Square* to call
Square::addCluster() to add the current cluster to the square's vector. You will need to
use the keyword this.
 De ne a print function that prints the type of the cluster followed by the 9 squares in that
cluster, one per line, with a blank line after the 9th square. Delegate the task of printing a
square to the print function in the Square class.
 Declare an inline method for the output operator for Clusters.
 De ne a function void Cluster::shoop( char val ); that will be called to eliminate
possibilities each time the human Solver moves a value in a Square. This is the heart of the
application. Do the following
{ Use character subtraction to convert the char parameter to an int value.
{ For each of the cluster's nine Square pointers, use the Square* to turn o the bit cor-
responding to the parameter value. Do this for all Squares in the Cluster, even if they
are xed or equal to the Square that initiated the move. This will help to make the
possibility display useful.
8 SOMETHING is Due October 10, the rest by October 17.
If you cannot nish this on time, submit what you have on the 10th and submit the rest on the
17th. There will be late penalties if I do not get some substantial e ort by the 10th.
Update your test plan and unit tests for Square and Board and construct a test plan and unit
test for Cluster. To test the new parts of Board and Square, it is enough to print the 27 clusters
on the Board.
To test Cluster, it is adequate to show that shoop works for rows, columns, and boxes.
Call all the test functions from main(). Turn in a zipped folder containing copies of your test