https://github.com/xieqing/avl-tree

An AVL Tree Implementation In C

There are several choices when implementing AVL trees:

  • store height or balance factor
  • store parent reference or not
  • recursive or non-recursive (iterative)

This implementation's choice:

  • store balance factor
  • store parent reference
  • non-recursive (iterative)

Files:

  • avl_bf.h - AVL tree header
  • avl_bf.c - AVL tree library
  • avl_data.h - data header
  • avl_data.c - data library
  • avl_example.c - example code for AVL tree application
  • avl_test.c - unit test program
  • avl_test.sh - unit test shell script
  • README.md - implementation note

If you have suggestions, corrections, or comments, please get in touch with xieqing.

DEFINITION

The AVL tree is named after its two Soviet inventors, Georgy Adelson-Velsky and Evgenii Landis, who published it in their 1962 paper "An algorithm for the organization of information". It was the first such data structure to be invented.

In an AVL tree, the heights of the two child subtrees of any node differ by at most one; each node stores its height (alternatively, can just store difference in heights), if at any time they differ by more than one, rebalancing is done to restore this property.

In a binary search tree the balance factor of a node N is defined to be the height difference of its two child subtrees.

BalanceFactor(N) = Height(RightSubtree(N)) – Height(LeftSubtree(N))

A binary search tree is defined to be an AVL tree if the invariant holds for every node N in the tree.

BalanceFactor(N) ∈ {–1,0,+1}

A node N with BalanceFactor(N) < 0 is called "left-heavy", one with BalanceFactor(N) > 0 is called "right-heavy", and one with BalanceFactor(N) = 0 is sometimes simply called "balanced".

Balance factors can be kept up-to-date by knowning the previous balance factors and the change in height - it is not necesary to know the absolute height.

Two main properties of AVL trees:

  • Binary Search Property: in-order sequence of the keys, ensures that we can search for any value in O(height);
  • Balance Factor Property: the heights of two child subtrees of any node differ by at most one, ensures that the height of an AVL tree is always O(log N).

ROTATION

It is easy to check that a single rotation preserves the ordering requirement for a binary search tree. The keys in subtree A are less than or equal to x, the keys in tree C are greater than or equal to y, and the keys in B are between x and y.

Before rotation

      x
/ \
A y
/ \
B C After rotation y
/ \
x C
/ \
A B

SEARCH

Searching for a specific key in an AVL tree can be done the same way as that of a normal binary search tree.

MODIFICATION

After a modifying operation (e.g. insertion, deletion) it is necessary to update the balance factors of all nodes, a little thought should convince you that all nodes requiring correction must be on the path from the root to the modified node, and these nodes are ancestors of the modified node.

If a temporary height difference of more than one arises between two child subtrees, the parent subtree has to be rebalanced by rotations.

Rotations never violate the binary search property and the balance factor property, insertions and deletions never violate the binary search property, and violations of the balance factor property can be restored by rotations.

INSERTION

The effective insertion of the new node increases the height of the corresponding child tree from 0 to 1. Starting at this subtree, it is necessary to check each of the ancestors for consistency with the invariants of AVL trees.

  1. insert as in simple binary search tree.
  2. backtrack the top-down path from the root to the new node: update the balance factor of parent node; rebalance if the balance factor of parent node temporarily becomes +2 or -2 (parent subtree has the same height as before, thus backtracking terminate immediately); terminate if the height of that parent subtree remains unchanged (has the same height as before insertion).

Inserting

Replace the termination NIL pointer with the new node

Before insertion

    parent
|
NIL (current) After insertion parent (height increased)
|
new_node (current)
/ \
NIL NIL

Rebalancing

Let x be the lowest node that violates the AVL property and let h be the height of its shorter subtree.

The first case: insert under x.left

1. insert under x.left.left

    Before insertion

                x (h+2)
/ \
(h+1) y C (h)
/ \
(h) A B (h)
^ insert (A may be NIL) bf(y) = 0; bf(x) = -1; height = h+2 After insertion (y's balance factor has been updated) x (h+3)
/ \
(h+2) y C (h)
/ \
(h+1) A B (h) bf(y) = -1; bf(x) = -1; height = h+2 After right rotation y (h+2)
/ \
(h+1) A x (h+1)
/ \
(h) B C (h) bf(x) = 0; bf(y) = 0; height = h+2 (height unchanged) 2. insert under x.left.right Before insertion x (h+2)
/ \
(h+1) y C (h)
/ \
(h) A B (h)
^ insert (B may be NIL) bf(y) = 0; bf(x) = -1; height = h+2 After insertion (y's balance factor has been updated) x (h+3)
/ \
(h+2) y C (h)
/ \
(h) A B (h+1) bf(y) = 1; bf(x) = -1; height = h+2 Let's expand B one more level (since B has height h+1, it cannot be empty) x (h+3)
/ \
(h+2) y C (h)
/ \
(h) A z' (h+1)
/ \
(h/h-1/h=0) U V (h-1/h/h=0) After left rotation x
/ \
z C
/ \
y V
/ \
A U After right rotation z (h+2)
/ \
/ \
(h+1) y x (h+1)
/ \ / \
(h) A U V C (h)
(h/h-1/h=0)(h-1/h/h=0) bf(z') = -1; bf(y) = 0; bf(x) = 1; bf(z) = 0; height = h+2 (height unchanged)
bf(z') = 1; bf(y) = -1; bf(x) = 0; bf(z) = 0; height = h+2 (height unchanged)
bf(z') = 0; bf(y) = 0; bf(x) = 0; bf(z) = 0; height = h+2 (height unchanged) The second case: insert under x.right 1. insert under x.right.right Before insertion x (h+2)
/ \
(h) A y (h+1)
/ \
(h) B C (h)
^ insert (C may be NIL) bf(y) = 0; bf(x) = 1; height = h+2 After insertion (y's balance factor has been updated) x
/ \
(h) A y (h+2)
/ \
(h) B C (h+1) bf(y) = 1; bf(x) = 1; height = h+2 After left rotation y (h+2)
/ \
(h+1) x C (h+1)
/ \
(h) A B (h) bf(x) = 0; bf(y) = 0; height = h+2 (height unchanged) 2. insert under x.right.left Before insertion x (h+2)
/ \
(h) A y (h+1)
/ \
(h) B C (h)
^ insert (B may be NIL) bf(y) = 0; bf(x) = 1; height = h+2 After insertion (y's balance factor has been updated) x
/ \
(h) A y (h+2)
/ \
(h+1) B C (h) bf(y) = -1; bf(x) = 1; height = h+2 Let's expand it one more level (since B has height h+1, it cannot be empty) x (h+3)
/ \
(h) A y (h+2)
/ \
(h+1) z' C (h)
/ \
(h/h-1/h=0) U V (h-1/h/h=0) After right rotation x
/ \
A z
/ \
U y
/ \
V C After left rotation z (h+2)
/ \
/ \
(h+1) y x (h+1)
/ \ / \
(h) A U V C (h)
(h/h-1/h=0)(h-1/h/h=0) bf(z') = -1; bf(y) = 0; bf(x) = 1; bf(z) = 0; height = h+2 (height unchanged)
bf(z') = 1; bf(y) = -1; bf(x) = 0; bf(z) = 0; height = h+2 (height unchanged)
bf(z') = 0; bf(y) = 0; bf(x) = 0; bf(z) = 0; height = h+2 (height unchanged)

DELETION

The effective deletion of the subject node or the replacement node decreases the height of the corresponding child tree either from 1 to 0 or from 2 to 1, if that node had a child. Starting at this subtree, it is necessary to check each of the ancestors for consistency with the invariants of AVL trees.

  1. find the subject node or its replacement node (in-order successor) if the subject node has two children, not to remove it for the time being.
  2. backtrack the top-down path from the root to the subject node or the replacement node: update the balance factor of parent node; rebalance if the balance factor of parent node temporarily becomes +2 or -2; terminate if the height of that parent subtree remains unchanged (has the same height as before deletion).
  3. remove the subject node or the replacement node.

Rebalancing

Let x be the lowest node that violates the AVL property and let h+1 be the height of its shorter subtree.

The first case: delete under x.right

1. x.left is left-heavy or balanced

    Before deletion

                      x (h+3)
/ \
(h+2) y C (h+1)
/ \ ^ delete
(h+1/h+1) A B (h/h+1) bf(y) = -1; bf(x) = -1; height = h+3
bf(y) = 0; bf(x) = -1; height = h+3 After deletion x
/ \
(h+2) y' C (h)
/ \
(h+1/h+1) A B (h/h+1) After right rotation y (h+2/h+3)
/ \
(h+1/h+1) A x (h+1/h+2)
/ \
(h/h+1) B C (h) bf(y') = -1; bf(x) = 0; bf(y) = 0; height = h+2 (height decreased)
bf(y') = 0; bf(x) = -1; bf(y) = 1; height = h+3 (height unchanged) 2. x.left is right-heavy Before deletion x (h+3)
/ \
(h+2) y C (h+1)
/ \ ^ delete
(h) A B (h+1) bf(y) = 1; bf(x) = -1; height = h+3 After deletion x (h+1)
/ \
(h+2) y C (h)
/ \
(h) A B (h+1) Let's expand B one more level (since B has height h+1, it cannot be empty) (h+3) x
/ \
(h+2) y C (h)
/ \
(h) A z' (h+1)
/ \
(h/h-1/h) U V (h-1/h/h) After left rotation x
/ \
z C
/ \
y V
/ \
A U After right rotation z (h+2)
/ \
/ \
(h+1) y x (h+1)
/ \ / \
(h) A U V C (h)
(h/h-1/h)(h-1/h/h) bf(z') = -1; bf(y) = 0; bf(x) = 1; bf(z) = 0; height = h+2 (height decreased)
bf(z') = 1; bf(y) = -1; bf(x) = 0; bf(z) = 0; height = h+2 (height decreased)
bf(z') = 0; bf(y) = 0; bf(x) = 0; bf(z) = 0; height = h+2 (height decreased) The sencond case: delete under x.left 1. x.right is right-heavy or balanced Before deletion x (h+3)
/ \
(h+1) A y (h+2)
delete ^ / \
(h/h+1) B C (h+1/h+1) bf(y) = 1; bf(x) = 1; height = h+3
bf(y) = 0; bf(x) = 1; height = h+3 After deletion x (h+3)
/ \
(h) A y' (h+2)
/ \
(h/h+1) B C (h+1/h+1) After left rotation y (h+2/h+3)
/ \
(h+1/h+2) x C (h+1/h+1)
/ \
(h) A B (h/h+1) bf(y') = 1; bf(x) = 0; bf(y) = 0; height = h+2 (height decreased)
bf(y') = 0; bf(x) = 1; bf(y) = -1; height = h+3 (height unchanged) 2. x.right is left-heavy Before deletion x (h+3)
/ \
(h+1) A y (h+2)
delete ^ / \
(h+1) B C (h) bf(y) = -1; bf(x) = 1; height = h+3 After deletion x (h+3)
/ \
(h) A y (h+2)
/ \
(h+1) B C (h) Let's expand B one more level (since B has height h+1, it cannot be empty) x (h+3)
/ \
(h) A y (h+2)
/ \
(h+1) z' C (h)
/ \
(h/h-1/h) U V (h-1/h/h) After right rotation x
/ \
A z
/ \
U y
/ \
V C After left rotation z (h+2)
/ \
/ \
(h+1) y x (h+1)
/ \ / \
(h) A U V C (h)
(h/h-1/h)(h-1/h/h) bf(z') = -1; bf(y) = 0; bf(x) = 1; bf(z) = 0; height = h+2 (height decreased)
bf(z') = 1; bf(y) = -1; bf(x) = 0; bf(z) = 0; height = h+2 (height decreased)
bf(z') = 0; bf(y) = 0; bf(x) = 0; bf(z) = 0; height = h+2 (height decreased)

Removing

Replace the subject node or the replacement node with its child (which may be NIL)

            parent
| parent
node -> |
/ \ child/NIL
child/NIL/NIL NIL/child/NIL

References

  1. https://en.wikipedia.org/wiki/AVL_tree
  2. https://www.cs.usfca.edu/~galles/visualization/AVLtree.html

License

Copyright (c) 2019 xieqing. https://github.com/xieqing

May be freely redistributed, but copyright notice must be retained.

AVL树实现记录的更多相关文章

  1. 算法与数据结构(十一) 平衡二叉树(AVL树)

    今天的博客是在上一篇博客的基础上进行的延伸.上一篇博客我们主要聊了二叉排序树,详情请戳<二叉排序树的查找.插入与删除>.本篇博客我们就在二叉排序树的基础上来聊聊平衡二叉树,也叫AVL树,A ...

  2. AVL树

    AVL树 在二叉查找树(BST)中,频繁的插入操作可能会让树的性能发生退化,因此,需要加入一些平衡操作,使树的高度达到理想的O(logn),这就是AVL树出现的背景.注意,AVL树的起名来源于两个发明 ...

  3. PAT树_层序遍历叶节点、中序建树后序输出、AVL树的根、二叉树路径存在性判定、奇妙的完全二叉搜索树、最小堆路径、文件路由

    03-树1. List Leaves (25) Given a tree, you are supposed to list all the leaves in the order of top do ...

  4. AVL树插入操作实现

    为了提高二插排序树的性能,规定树中的每个节点的左子树和右子树高度差的绝对值不能大于1.为了满足上面的要求需要在插入完成后对树进行调整.下面介绍各个调整方式. 右单旋转 如下图所示,节点A的平衡因子(左 ...

  5. 数据结构——二叉查找树、AVL树

    二叉查找树:由于二叉查找树建树的过程即为插入的过程,所以其中序遍历一定为升序排列! 插入:直接插入,插入后一定为根节点 查找:直接查找 删除:叶子节点直接删除,有一个孩子的节点删除后将孩子节点接入到父 ...

  6. 平衡二叉树,AVL树之图解篇

    学习过了二叉查找树,想必大家有遇到一个问题.例如,将一个数组{1,2,3,4}依次插入树的时候,形成了图1的情况.有建立树与没建立树对于数据的增删查改已经没有了任何帮助,反而增添了维护的成本.而只有建 ...

  7. My集合框架第三弹 AVL树

    旋转操作: 由于任意一个结点最多只有两个儿子,所以当高度不平衡时,只可能是以下四种情况造成的: 1. 对该结点的左儿子的左子树进行了一次插入. 2. 对该结点的左儿子的右子树进行了一次插入. 3. 对 ...

  8. 红黑树和AVL树的实现与比较-----算法导论

    一.问题描述 实现3种树中的两种:红黑树,AVL树,Treap树 二.算法原理 (1)红黑树 红黑树是一种二叉查找树,但在每个结点上增加一个存储位表示结点的颜色,可以是red或black.红黑树满足以 ...

  9. AVL树的插入删除查找算法实现和分析-1

    至于什么是AVL树和AVL树的一些概念问题在这里就不多说了,下面是我写的代码,里面的注释非常详细地说明了实现的思想和方法. 因为在操作时真正需要的是子树高度的差,所以这里采用-1,0,1来表示左子树和 ...

随机推荐

  1. git 本地修改、撤消操作

    // 撤消本地文件的修改,还原到最近版本 git checkout -- * 是撤销从上次提交之后所做的所有修改 git checkout -- <filaname> 是撤销从上次提交之后 ...

  2. python用字符串调用当前模块内的函数

    eval(字符串)() vars()[字符串]() 例如:

  3. Sql更新

    //要传一个DataSet,和若干个表 /// <summary> /// 更新数据库 /// </summary> /// <param name="data ...

  4. 02_编写Table的CRUD

    1.使用EF的Code First模式生成DbContext和表对应的实体类 2.编写CRUD接口: 3.集成Swagger接口生成工具,方便测试使用: https://www.cnblogs.com ...

  5. xshell完美开源替代方案(Kitty+MTPuTTY并设置全局字体)

    xshell是收费的,过了30天就不能用了.我们应该找一个开源的替代品.说实话windows平台没有什么可选的,就是putty.但是原生的putty不好用,记不住密码,又不支持多标签. Kitty是基 ...

  6. Ubuntu系统配置

    0.基本配置 0.1初始设置 (1)开户root账号并重启系统: 打开gedit /usr/share/lightdm/lightdm.conf.d/50-ubuntu.conf 添加greeter- ...

  7. 给查询出的SQL记录添加序号列,解决方法有以下两种

    第一: select ROW_NUMBER() OVER (ORDER BY a.字段 ASC) AS XUHAO,a.* from table a (table 为表名,字段为表a中的字段名) 第二 ...

  8. HashTable源码简单介绍

    首先是继承了字典类Dictionary, 这说明HashTable的一些找位置的实现与Dictionary有关, 看一看数据结构,是一个entry数组, Entry,这个不陌生,先看一下它的结构吧,这 ...

  9. 0008 合并K个排序链表

    合并 k 个排序链表,返回合并后的排序链表.请分析和描述算法的复杂度. 示例: 输入: [   1->4->5,   1->3->4,   2->6 ] 输出: 1-&g ...

  10. n2n网络穿透内网

    目录 前言 配置 网络拓扑: 公网服务器的配置 公司电脑的配置 家里笔记本的配置 注意事项 使用n2n网络 n2n的各edge之间传输数据 补充:NAT类型 后记 前言 在家里的时候比较经常需要对公司 ...