Binary Search Trees

‣ basic implementations
‣ randomized BSTs
‣ deletion in BSTs

References:
Algorithms in Java, Chapter 12
Intro to Programming, Section 4.4 Except as otherwise noted, the content of this presentation
http://www.cs.princeton.edu/algs4 is licensed under the Creative Commons Attribution 2.5 License.

Algorithms in Java, 4th Edition · Robert Sedgewick and Kevin Wayne · Copyright © 2008 · May 2, 2008 10:48:13 AM
Elementary implementations: summary

worst case average case ordered operations

implementation
iteration? on keys
search insert search hit insert

unordered array N N N/2 N no equals()

unordered list N N N/2 N no equals()

ordered array lg N N lg N N/2 yes compareTo()

ordered list N N N/2 N/2 yes compareTo()

Challenge. Efficient implementations of search and insert.

2
‣ binary search tree
‣ randomized BSTs
‣ deletion in BSTs

3
Binary search trees

Def. A BST is a binary tree in symmetric order.

it

A binary tree is either: best was

• Empty. the

• A key-value pair and two disjoint binary trees. of times

Symmetric order. Every node’s key is:

• Larger than all keys in its left subtree. BST

• Smaller than all keys in its right subtree. Node key val

left right

BST with smaller keys BST with larger keys

Binary search tree

4
BST representation

A BST is a reference to a root node.

A Node is comprised of four fields: key val

• A Key and a Value. left right

• A reference to the left and right subtree.

smaller keys larger keys

root

private class Node it 2

{
private Key key;
private Value val; best 1 was 2

private Node left; null null null

private Node right;

} the 1

Key and Value are generic types;

Key is Comparable
of 1 times 1
null null null null

5
BST implementation (skeleton)

public class BST<Key extends Comparable<Key>, Value>

{
private Node root; instance variable

private class Node

{
private Key key;
private Value val;
private Node left, right;
instance variable
public Node(Key key, Value val)
{
this.key = key;
this.val = val;
}
}

public void put(Key key, Value val)

{ /* see next slides */ }

public Val get(Key key)

{ /* see next slides */ }
}

6
BST search

Get. Return value corresponding to given key, or null if no such key.

successful
successful searchsearch unsuccessful
unsuccessful searchsearch
for a node
for a node withthe
with key key the for a node with key times
for a node with key times
times
times is it
is after after it
it it it it
so go so
to go
thetoright
the right
best best was was best best was was
the isthe is after
after it it
so go so
to go
thetoright
the right
the the the the

of of of of

it it it it

best best was was best best was was

times
times is before
is before was was
the the the the
so go so
to go
thetoleft
the left
the
the is is before
before was was
of of of of
to go
so go so the left
thetoleft
it it it it

best best was was best best was was

the the the the times

times is after
is after the the
butright
but the the right
link islink
nullis null
of of of of so thesoBST
the has
BSTnohasnodeno node
success!
success! havinghaving that key
that key
Searching
Searching in a in a BST
BST
7
BST search: Java implementation

Get. Return value corresponding to given key, or null if no such key.

public Value get(Key key)

{
Node x = root;
while (x != null)
{
int cmp = key.compareTo(x.key);
if (cmp < 0) x = x.left;
else if (cmp > 0) x = x.right;
else if (cmp == 0) return x.val;
}
return null;
}

Running time. Proportional to depth of node.

8
BST insert

Put. Associate value with key. insert times

times is after it
it so go to the right
best was

the

of

it

best was
times is before was
the
so go to the left
of

it

best was

the

of times is after the

so it goes on the right

it

best was

the

of times

Inserting a new node into a BST 9

BST insert: Java implementation

Put. Associate value with key.

public void put(Key key, Value val)

{ root = put(root, key, val); }

private Node put(Node x, Key key, Value val)

{
if (x == null) return new Node(key, val);
int cmp = key.compareTo(x.key);
if (cmp < 0) x.left = put(x.left, key, val); concise, but tricky,
else if (cmp > 0) x.right = put(x.right, key, val); recursive code;
read carefully!
else if (cmp == 0) x.val = val;
return x;
}

Running time. Proportional to depth of node.

10
 best was

the

BST construction example of it

best was

the

key of
inserted
it it times it

was it best was

was the

the it
of times

was
worst it
the
best was
best it
the worst
best was
of times
the
Constructing a BST
of it

best was

the

of

times it
11
best was

the
BST insertion: visualization

Ex. Insert keys in random order.

12
Correspondence between BSTs and quicksort partitioning

K
E S

C I Q T

A E M R U

L P U

Remark. Correspondence is 1-1 if no duplicate keys.

13
Tree shape

• Many BSTs correspond to same input data.

• Cost of search/insert is proportional to depth of node.

best case typical case worst case

the of best

it was it times it

best of times worst best the was of

worst the

times

was

worst

Typical BSTs constructed from randomly ordered keys

Typical BSTs constructed from randomly ordered keys worst

was

Remark. Tree shape depends on order of insertion. times

the

of

14
it

best
BSTs: mathematical analysis

Proposition. If keys are inserted in random order, the expected number of

compares for a search/insert is ~ 2 ln N.

Pf. 1-1 correspondence with quicksort partitioning.

Proposition. [Reed, 2003] If keys are inserted in random order,

expected height of tree is ~ 4.311 ln N – 1.953 ln ln N.

But… Worst-case for search/insert/height is N (but occurs with

exponentially small chance when keys are inserted in random order).

15
ST implementations: summary

guarantee average case

ordered operations
implementation
iteration? on keys
search insert search hit insert

unordered array N N N/2 N no equals()

unordered list N N N/2 N no equals()

ordered array lg N N lg N N/2 yes compareTo()

ordered list N N N/2 N/2 yes compareTo()

BST N N 1.38 lg N 1.38 lg N ? compareTo()

Next challenge. Ordered iteration.

16
Inorder traversal

Traversing the tree inorder yields keys in ascending order.

public void show()

{ return show(root); } BST

key val

private void show(Node x)

{ left right
if (x == null) return;
show(x.left); BST with smaller keys BST with larger keys

StdOut.println(x.key + " " + x.val);

smaller keys, in order key larger keys, in order
show(x.right);
} all keys, in order

Recursive inorder traversal of a binary search tree

To implement an iterator: need a non-recursive version.

17
Non-recursive inorder traversal

To process a node:
• Follow left links until empty (pushing onto stack).
• Pop and process.
• Follow right link (push onto stack).

visit(E) E
visit(B) E B
visit(A) E B A E
print A A E B
print B B E
B S
visit(C) E C
print C C E
print E E - A C I
visit(S) S
visit(I) S I
visit(H) S I H H N
print H H S I
print I I S
visit(N) S N
print N N S
print S S -

recursive calls output stack contents

18
Inorder iteartor: Java implementation

public Iterator<Key> iterator()

{ return new Inorder(); }

private class Inorder implements Iterator<Key>

{
private Stack<Node> stack = new Stack<Node>();

private void pushLeft(Node x)

{ go down left
while (x != null) spine and push
{ stack.push(x); x = x.left; } all keys onto stack
}

BSTIterator()
{ pushLeft(root); }

public boolean hasNext()

{ return !stack.isEmpty(); }

public Key next()

{
Node x = stack.pop();
pushLeft(x.right);
return x.key;
}
}

19
ST implementations: summary

guarantee average case

ordered operations
implementation
iteration? on keys
search insert search hit insert

unordered array N N N/2 N no equals()

unordered list N N N/2 N no equals()

ordered array lg N N lg N N/2 yes compareTo()

ordered list N N N/2 N/2 yes compareTo()

BST N N 1.38 lg N 1.38 lg N yes compareTo()

Next challenge. Guaranteed efficiency for search and insert.

20
Searching challenge 3 (revisited):

Problem. Frequency counts in “Tale of Two Cities”

Assumptions. Book has 135,000+ words; about 10,000 distinct words.

Which searching method to use?

1) Unordered array.
2) Unordered linked list
3) Ordered array with binary search.
4) Need better method, all too slow.
5) Doesn’t matter much, all fast enough.
6) BSTs.

insertion cost < 10000 * 1.38 * lg 10000 < .2 million

lookup cost < 135000 * 1.38 * lg 10000 < 2.5 million

21
‣ basic implementations
‣ randomized BSTs
‣ deletion in BSTs

22
Rotation in BSTs

Two fundamental operations to rearrange nodes in a tree.

• Maintain symmetric order.
• Local transformations (change just 3 pointers).
• Basis for advanced BST algorithms.

Strategy. Use rotations on insert to adjust tree shape to be more balanced.

h
u h
v

h = rotL(u)
v
u

A h = rotR(v) C

B C A B

Key point. No change to BST search code (!)

23
 Rotation in BSTs

Two fundamental operations to rearrange nodes in a tree.

• Easier done than said.
• Raise some nodes, lowers some others.

root = rotL(A) a.right = rotR(S)

private Node rotL(Node h)
{
Node v = h.right;
h.right = v.left;
v.left = h;
return v;
}

private Node rotR(Node h)

{
Node u = h.left;
h.left = u.right;
u.right = h;
return u;
}

24
BST root insertion

insert G
Insert a node and make it the new root.
• Insert node at bottom, as in standard BST.
• Rotate inserted node to the root.
• Compact recursive implementation.

private Node putRoot(Node x, Key key, Val val)

{
if (x == null) return new Node(key, val);
int cmp = key.compareTo(x.key);
if (cmp < 0)
{
x.left = putRoot(x.left, key, val);
x = rotR(x);
} tricky recursive
else if (cmp > 0) code; read very
{ carefully!
x.right = putRoot(x.right, key, val);
x = rotL(x);
}
else if (cmp == 0) x.val = val;
return x;
}

25
BST root insertion: construction

Ex. A S E R C H I N G X M P L

Why bother?
• Recently inserted keys are near the top (better for some clients).
• Basis for randomized BST.

26
Randomized BSTs (Roura, 1996)

Intuition. If tree is random, height is logarithmic.

Fact. Each node in a random tree is equally likely to be the root.

Idea. Since new node should be the root with probability 1/(N+1),
make it the root (via root insertion) with probability 1/(N+1).

and apply idea recursively

private Node put(Node x, Key key, Value val)
{
if (x == null) return new Node(key, val);
int cmp = key.compareTo(x.key);

if (cmp == 0) { x.val = val; return x; } no rotations if key in table

if (StdRandom.bernoulli(1.0 / (x.N + 1.0)) root insert with the right

return putRoot(h, key, val); probability

if (cmp < 0) x.left = put(x.left, key, val);

else if (cmp > 0) x.right = put(x.right, key, val);

maintain count of nodes in

x.N++;
subtree rooted at x
return x;
}

27
Randomized BST: construction

Ex. Insert 15 keys in ascending order.

28
Randomized BST construction: visualization

Ex. Insert 500 keys in random order.

29
Randomized BST: analysis

Proposition. Randomized BSTs have

of
the same distribution as BSTs under
random insertion order, no matter
it in whattimes
order keys are inserted.
best the was

worst

Typical BSTs constructed from randomly ordered keys

• Expected height is ~ 4.31107 ln N.

• Average search cost is ~ 2 ln N.
• Exponentially small chance of bad balance.

Implementation cost. Need to maintain subtree size in each node.

30
ST implementations: summary

guarantee average case

ordered operations
implementation
iteration? on keys
search insert search hit insert

unordered array N N N/2 N no equals()

unordered list N N N/2 N no equals()

ordered array lg N N lg N N/2 yes compareTo()

ordered list N N N/2 N/2 yes compareTo()

BST N N 1.38 lg N 1.38 lg N yes compareTo()

randomized BST 3 lg N 3 lg N 1.38 lg N 1.38 lg N yes compareTo()

Bottom line. Randomized BSTs provide the desired guarantee.

probabilistic, with exponentially

Bonus. Randomized BSTs also support delete (!)
small chance of linear time
31
‣ basic implementations
‣ randomized BSTs
‣ deletion in BSTs

32
BST deletion: lazy approach

To remove a node with a given key:

• Set its value to null.
• Leave key in tree to guide searches (but don't consider it equal to search key).

E E

A S A S

delete I
C I C I tombstone

H R H R

N N

Cost. O(log N') per insert, search, and delete, where N' is the number of
elements ever inserted in the BST.

Unsatisfactory solution. Tombstone overload.

33
BST deletion: Hibbard deletion

To remove a node from a BST:

• Zero children: just remove it.
• One child: pass the child up.
• Two children: find the next largest node using right-left*, swap with
next largest, remove as above.

zero children one child two children

34
BST deletion: Hibbard deletion

To remove a node from a BST:

• Zero children: just remove it.
• One child: pass the child up.
• Two children: find the next largest node using right-left*, swap with
next largest, remove as above.

zero children one child two children

Unsatisfactory solution. Not symmetric, code is clumsy.

Surprising consequence. Trees not random (!) ⇒ sqrt(N) per op.
Longstanding open problem. Simple and efficient delete for BSTs.
35
Randomized BST deletion

To delete a node containing a given key:

• Find the node containing the key.
• Remove the node.
• Join its two subtrees to make a tree.

Ex. Delete S.

A S

C I X

H R

36
Randomized BST deletion

To delete a node containing a given key: private Node remove(Node x, Key key)

• Find the node containing the key. {

if (x == null) return null;
• Remove the node. int cmp = key.compareTo(x.key);

•
if (cmp < 0)
Join its two subtrees to make a tree. x.left = remove(x.left, key);
else if (cmp > 0)
x.right = remove(x.right, key);
else if (cmp == 0)
return join(x.left, x.right);
Ex. Delete S. return x;
}

E
join these two subtrees
A

C I X

H R

37
Randomized BST join

To join two subtrees with all keys in one less than all keys in the other:
• Maintain counts of nodes in subtrees a and b.
• With probability |a|/(|a|+|b|):
- root = root of a
- left subtree = left subtree of a
- right subtree = join b and right subtree of a
• With probability |a|/(|a|+|b|) do the symmetric operations.

make I the root and recursively join these

to join these two subtrees with probability 4/5 two subtrees to form right
subtree of I

I XX I X

H R H R

N N

38
Randomized BST join

To join two subtrees with all keys in one less than all keys in the other:
• Maintain counts of nodes in subtrees a and b.
• With probability |a|/(|a|+|b|):
- root = root of a
- left subtree = left subtree of a
- right subtree = join b and right subtree of a
• With probability |a|/(|a|+|b|) do the symmetric operations.

to join these two

subtrees

private Node join(Node a, Node b)

{
if (a == null) return b; X
if (b == null) return a;
if (StdRandom.bernoulli((double) a.N / (a.N + b.N))
R
{ a.right = join(a.right, b); return a; }
else N make R the root
{ b.left = join(a, b.left ); return b; } with probability 2/3
}
R

N X
39
Randomized BST deletion

To delete a node containing a given key:

• Find the node containing the key.
• Remove the node.
• Join its two subtrees to make a tree.

E E

A S A I

C I X C H R

H R
N X
N

Proposition. Tree still random after delete (!).

Bottom line. Logarithmic guarantee for search/insert/delete.
40
 ST implementations: summary

guarantee average case

ordered operations
implementation
iteration? on keys
search insert delete search hit insert delete

unordered array N N N N/2 N N/2 no equals()

unordered list N N N N/2 N N/2 no equals()

ordered array lg N N N lg N N/2 N/2 yes compareTo()

ordered list N N N N/2 N/2 N/2 yes compareTo()

BST N N N 1.38 lg N 1.38 lg N ? yes compareTo()

randomized BST 3 lg N 3 lg N 3 lg N 1.38 lg N 1.38 lg N 1.38 lg N yes compareTo()

Bottom line. Randomized BSTs provide the desired guarantee.

Next lecture. Can we do better?
probabilistic, with exponentially
small chance of linear time
41