In this assignment you will build a custom data structure named QuadTree. The data structure you will build for this homework is similar to the classic quad tree, octree, k-d tree, and binary space partition data structures from computational geometry. These structures are used to improve the performance of applications that use large spatial data sets including: ray tracing in computer graphics, collision detection for simulation and gaming, motion planning for robotics, nearest neighbor calculation, image processing, and many, many others. Our QuadTree implementation will share some of the framework of the ds_set implementation we have seen in lecture and lab. You are encouraged to carefully study that implementation as you work on this homework.

The diagrams below illustrate the incremental construction of the QuadTree data structure. In this example, we add the 21 two-dimensional points shown in the first image to the tree structure. We will add them in the alphabetical order of their labels. Each time a point is added we locate the rectangular region containing that point and subdivide that region into 4 smaller rectangles using the x,y coordinates of that point as the vertical and horizontal dividing lines.

Each 2D coordinate (x,y) is stored in the Point class. In these plots (0,0) is in the upper left corner. The x axis runs horizontally, with increasing values to the right. The y axis runs vertically with increasing values in the downward direction.

Like an STL map and STL set, inserting a new Point into the QuadTree or querying (using find) whether a Point is already in the structure is fast, and can be completed in O(log n), where n is the number of Points in the tree. However, unlike the set and map structures which are based on a binary tree, having two subtrees per node, the QuadTree has 4 children at each node!

The data in our QuadTree can be visualized using two different output routines, plot and print_sideways. You will also implement two different methods for iterating over the tree: depth-first (specifically pre-order) or breadth-first. To do this, you will write two helper classes DepthIterator and BreadthIterator which will be typedef-ed within the QuadTree class.

Here is a diagram showing the relationships between the different classes you will implement for this assignment. Be sure to draw plenty of your own diagrams as you work, and be prepared to show these diagrams when you come to office hours or ALAC tutoring to ask for help on the assignment.

Note that the QuadTree is templated over two types, the number type and the label type. In this example, the number type is int and the label type is char. This is the detailed internal memory representation after adding the first 3 points in the example on the previous page. We also show a sample DepthIterator object initially attached to the root node of this tree. You should follow this diagram exactly, using these specific class and member variable types and names and placeholder template type names. We do not present the details of the BreadthIterator class in this diagram. You may design the internal representation of that class to complete the specified functionality.

Implementation

Your task for this homework is to implement the structure diagrammed above. We recommend that you begin your implementation by following the structure of the ds_set class we studied in lecture and lab. You will need to make a number of significant changes to the code, but the overall design: a manager class (QuadTree), the Node class, and tree_iterator classes is similar.

We provide the Point class and the implementation of the two QuadTree member functions for printing: plot and print_sideways. You will build the rest of your implementation around this starter code. Remember that because this is a templated class, you will not separate the implementation into .h and .cpp files. Keep all of your implementation in a single file named quad_tree.h, but make sure it is well-organized and appropriately commented.

The provided code in main.cpp illustrates the basic functionality of the QuadTree class including the QuadTree functions: size, insert, find, height, begin, end, bf_begin, and bf_end, and iterator functions: operator++ (both pre- and post- increment), operator* (dereference), getLabel, and getDepth. Study these examples carefully to deduce the expected argument and return types of the functions. As this is a class with dynamically-allocated memory, you will also need to implement, test, and debug the copy constructor, assignment operator, and destructor. The homework server will compile and run your quad_tree.h

2

file with the instructors solution to test your implementation of these functions. It will test your code with Dr. Memory and your program must be memory error and memory leak free for full credit.

We encourage you to work through the test cases in the provided main.cpp step-by-step, uncommenting and debugging one test at a time. The provided test cases do not adequately test all corner cases, so as you work, add your own test cases to the student_tests function. Be sure to test your templated implementation with other number types (e.g., float, double) and other label/data types (e.g., STL string, int, etc.). Note that the ASCII art plot function is only designed to work with small non-negative integers and char labels. Furthermore, the plot function assumes that no two points have the same x coordinate or the same y coordinate.

Extra Credit: Tree Balancing

How does the point insertion order affect the shape of the resulting QuadTree object? What are specific examples of the worst case and best case? Discuss in your README.txt file. For extra credit, you can implement a function named BalanceTree to re-order a point collection before inserting the data into the tree to improve the quality of the resulting tree. QuadTree quality can be defined with various metrics including: minimal tree height, more equal partitioning of data into the 4 subtrees (approximately the same number of elements in each child tree), and rectangular subregions that have approximately equal area or that have more equal ratio of height to width. Certainly for some input collections you cannot simultaneously satisfy all of these properties! You may define how to prioritize these metrics. Be sure to write challenging test cases that show off your implementation.

Performance

Assuming our tree holds n points that are well-distributed and have been inserted into the structure in a random order to produce a generally well balanced tree, what is the order notation for running time and memory usage of the different operations of the QuadTree? Put your answers and a short justification for each answer in your README.txt file.