## Table of Contents

- Introduction
- Background
- Code Description
- Class Usage
- Tree Load Balancing
- Sample Demo
- Source Code Files
- References
- Thanks to...

## Introduction

In this article, I will show you how to use the Binary Search Tree to store data. And I am using templates for keys and data values to simplify usage of any data types. If you have good experience about the theory behind the binary search trees, you can skip the next section (**Background**).

## Background

[1] Binary search trees are data structures which support many dynamic-set operations including **search, insert, delete, minimum, maximum, predecessor, **and** successor**. Basic operations on binary search tree take time proportional to the height of the tree. For a complete binary tree with n nodes, such operations take O(log n) time in the worst-case, as the height of randomly built binary search tree is proportional to log n. In some cases, operations take O(n) time in case of sorted input, so the tree will be like a sorted list, so I used the `Red-Black`

tree balancing technique, as described in the Tree Load Balancing section

A binary search tree is organized in a binary tree as shown in figure 1. Such a tree can be represented by a linked data structure in which each node is an object. In addition to the `key`

field, each node contains fields `left`

, `right`

, and `parent`

that point to the nodes corresponding to its left child, its right child, and its parent, respectively. If a child or the parent is missing, the appropriate field contains a value `NIL`

.

I will not dig in all **Binary Tree** functions here, I will just refer to the `search`

function, and you can check reference [1] for more details, and I will talk about some of them in the code description, as many of them are so sophisticated like the `Delete`

function.

### Searching

The most common operation performed in binary search tree is searching for a key stored in the tree, which can be done in two ways:

- Recursion:
Node Search(Node, key)
{
if Node = NIL or key = Node.key
return Node
if key < Node.key
then return Search(Node.Left, key)
else return Search(Node.Right, key)
}

`While`

loop:
Node Search(key)
{
Node = root
while Node != NIL and key != Node.key
do if key < Node.key
then Node = Node.Left
else Node = Node.Right
return Node
}

The two ways begin search at the root and trace a path downward in the tree till they find the `key`

or return `NIL`

(not found case). First way code is more simple, but second way optimizes stack usage, which I am preferring in all my cases to deal with huge data without affecting the stack and the performance.

## Code Description

All `Tree`

functions encapsulated in a template class `CBinaryTree`

and `CBinaryTreeNode`

.

`GetCount` | returns tree nodes count with repetition (repeated key will assigned to one node) |

`RemoveAll` | removes all tree nodes |

`Insert` | inserts new `key` in the tree |

`Search` | searches for `key` in the tree |

`Min` | gets minimum node key under the input node |

`Max` | gets maximum node key under the input node |

`Successor` | gets node successor (node which comes next in the overall tree sorting) |

`Predecessor` | gets node predecessor (node which comes previous in the overall tree sorting) |

`Delete` | deletes node from tree and adjusts its childs' nodes |

`Save` | saves all tree nodes' order in a vector of integers |

In all my code here, I am avoiding recursion functions to avoid stack overflow, as I am dealing with huge data, so you will find all my code using `while`

loops, sometimes code becomes complicated like the `RemoveAll`

function:

### RemoveAll Function

This function removes all tree nodes, it does that by order to each node to delete its `left`

child then its `right`

child, then delete itself. This can be done in two ways:

`While`

loop:
void RemoveAll()
{
TREENODE *node = Root, *pTemp;
while(node != Nil)
{
if(node->Left != Nil)
node = node->Left;
else if(node->Right != Nil)
node = node->Right;
else
{
pTemp = node;
if(node->Parent != Nil)
node->Parent->Childs[node != node->Parent->Left] = Nil;
node = node->Parent;
delete pTemp;
}
}
Count = Serial = 0;
Root = Nil;
Modified = false;
}

As you can see, there is some complication at this section (node has no children).

- Recursion way:
It simply puts `delete left`

and `right`

nodes at each node destructor and call `delete`

for tree root.

~CBinaryTreeNode()
{
if(Childs[0])
delete Childs[0];
if(Childs[1])
delete Childs[1];
}
delete Tree.Root;

All the nodes will be deleted automatically through stack usage, but if you read the comments in the first way code, you will find it easy to understand it, and avoid stack usage.

### Min Function

The Minimum of a node `x`

can be found by following `left`

child from the node x until a `Nil`

is encountered, as in the following code:

TREENODE* Min(TREENODE* node) const
{
while(node != Nil && node->Left != Nil)
node = node->Left;
return node;
}

### Successor Function

The successor of a node `x`

is the node with the smallest key greater than `key[x]`

. For example, if you add keys (`C I D H B F E G A J K`

) in sequence and will be as shown in the following figure:

Each arrow represents the direction from each node to its successor. The code in the `Successor `

function takes two paths:

- If the node has right child, then return the left most node in the right sub tree.
- Else, it goes up from node until we find a node that is to the left of its parent.

TREENODE* Successor(TREENODE* node) const
{
if(node->Right != Nil)
return Min(node->Right);
TREENODE* Parent = node->Parent;
while(Parent != Nil && node == Parent->Right)
{
node = Parent;
Parent = node->Parent;
}
return Parent;
}

You can use the `Successor`

function with the `Min`

function to iterate tree nodes in ascending order.

### Delete Function

`Delete`

function has three cases, I found it too complicated, and I hope I can describe it, The three cases are:

- The node has no child, so we just remove it. As you can see in figure 2, node K can be deleted by just resetting the pointer from node
`J`

. - The node has one child, so we need to splice it as node
`H`

in figure 2, we should make node `D`

point to node `F`

as a right child, and set node `F`

parent to point to node `D`

. - The node has two children, so we choose its successor to take its position, and splice its successor (join successor parent and child, like the previous case). For example, if we delete node
`C`

at figure 2, we get node `C`

successor (`D`

), so we will splice node `D`

so that node `I`

and `H`

will be connected directly and node `C`

will take node `C`

place as in Figure 3.

Code in the `Delete`

is optimized to handle the three cases altogether, so I find it hard to describe it here, but you can try to apply the code in each case and read comments carefully, and you will find it working well.

void Delete(TREENODE* node)
{
TREENODE *pSplice = (node->Left == Nil ||
node->Right == Nil)?node:Successor(node);
TREENODE *pChild = pSplice->Childs[pSplice->Left == Nil];
if(pChild != Nil)
pChild->Parent = pSplice->Parent;
if(pSplice->Parent == Nil)
Root = pChild;
else
pSplice->Parent->Childs[pSplice !=
pSplice->Parent->Childs[0]] = pChild;
if(pSplice != node)
{
*node = *pSplice;
delete pSplice;
}
else
delete node;
Count--;
}

### Iterate Tree Nodes

You can call `Min(Tree.Root)`

and then `Successor`

in a loop to retrieve all nodes in order (ascending):

TREENODE* node = Min(Root);
while(node)
{
node = Successor(node);
}

And in the same way, you can call `Max(Tree.Root)`

and then `Predecessor`

in a loop to retrieve all nodes in a reversed order (descending):

TREENODE* node = Max(Root);
while(node)
{
node = Predecessor(node);
}

## Class Usage

`CBinaryTree`

usage is very simple, we just need to decide the `KEY`

and `DATA`

data types, then you can define the class, but you must choose a `KEY`

that supports the `compare`

function as it is used in the `Insert`

and `Search`

functions. For example:

CBiaryTree<string, LPCSTR, int, int> tree;
tree.NoRepeat = true;
for(...)
{
tree.Insert(str);
}

`class`

`String`

of the STL supports the `compare`

function, but in some cases you have to add it yourself, as in case of integers' sort:

class CInt
{
public:
int m_nKey;
int compare(int nKey)
{ return nKey == m_nKey ? 0 : (nKey >= m_nKey ? -1 : 1); }
void operator=(int nKey) { m_nKey = nKey; }
bool operator==(int nKey) { return m_nKey == nKey; }
};
CBiaryTree<CInt, int, int, int> tree;
tree.NoRepeat = true;
for(...)
{
tree.Insert(n);
}

## Tree Load Balancing

Tree balancing can be achieved using many techniques. In my class here, I am using Red-Black Tree functions for the insertion in the tree. `Red-Black Tree`

simply depends on keep tree height as short as possible, as the search and the insertion operations time depend on the tree height. As in the following figure, if the sequence (`C, A, B`

) is added to the tree, the height will be like case 1, but if we have an operation to change to the valid tree in case 3, it will be good as the tree height reduced.

So, I have included the functions `LeftRotate`

, `RightRotate`

, and `RBInsert`

to balance the tree functions.

## Sample Demo

The attached source zip file with article contains a sample that parses any folder in a recursive function and parses all its files with the extension specified in the extensions editor (any text files format), and it adds all files tokens to a binary tree. Then you can navigate all tree tokens through the tree control. You can test it in VC6 or 7.1, or just run the attached EXE, and don't hesitate to mail me for any help.

## Source Code Files

*BinaryTree.h*: Binary Tree code *RBTree.h*: Red-Black Tree code

## References

- [1] Thomas H.Cormen, Charles E.Leiserson, Ronald L.Rivest,
*Algorithms* 1990

## Thanks to...

I owe a lot to my colleagues for helping me in implementing and testing this code. (JAK)

## License

This article has no explicit license attached to it, but may contain usage terms in the article text or the download files themselves. If in doubt, please contact the author via the discussion board below.

A list of licenses authors might use can be found here.