UNIT-2

Disjoint Sets and Union/find operations

Dr.B.Srinu

Dept of Computer Science & Engineering

 Today’s Outline
 Disjoint Set Union/Find ADT
 Up-trees
 Weighted Union
 Path Compression(Collapsing Find)

Disjoint Sets 2
 Disjoint Set
 Sets and Disjoint Set :
 Disjoint Set Union:
 Considering a set S={1,2,3…10} (when n=10),
 then elements can be partitioned into three disjoint
sets
 s1={1,7,8,9}, s2={2,5,10} and s3={3,4,6}.

 Disjoint Set: if then Si and Sj

Si  S j  
are called as disjoint sets

Disjoint Sets 3
 Disjoint Set Union/Find

 1. Disjoint Set union:

 If Si and Sj are two disjoint sets, then their
union Si U Sj = all the elements x such that x
is in Si or Sj.
Thus S1 U S2 ={1,7,8,9,2,5,10}.
 2. Find(i):
 Given the element I, find the set containing i.
Thus, 4 is in set S3, 9 is in S1.

Disjoint Sets 4
 Disjoint Set Union/Find ADT
Union/Find ADT operations
 union find(4)
 find {1,4,8} {6}
 create
 destroy 8 {7} {2,3,6}
{5,9,10}
{2,3}
union(2,6)
Disjoint set equivalence property: every element of a DS U/F
structure belongs to exactly one set
Dynamic equivalence property: the set of an element can
change after execution of a union

Disjoint Sets CSE 326 Autumn 2001 5

 Example a 3 b 10 c
Construct the maze on the right
2 1 6
Initial state (set names underlined):
{a}{b}{c}{d}{e}{f}{g}{h}{i} d 4 e 7 f

Maze constructor (outside agent) 11 9 8

traverses edges in random order and
decides whether to union
g 12 h 5 i

Disjoint Sets CSE 326 Autumn 2001 7

 Example, First Step
{a}{b}{c}{d}{e}{f}{g}{h}{i} a 3 b 10 c

2 1 6
find(b)  b
find(e)  e
d 4 e 7 f
find(b)  find(e) so:
add 1 to E 11 9 8
union(b, e)
g 12 h 5 i
{a}{b,e}{c}{d}{f}{g}{h}{i}
Order of edges in blue

Disjoint Sets CSE 326 Autumn 2001 8

 Up-Tree Intuition
Finding the representative member of a set is
somewhat like the opposite of finding whether a
given item exists in a set.

So, instead of using trees with pointers from

each node to its children; let’s use trees with a
pointer from each node to its parent.

Disjoint Sets CSE 326 Autumn 2001 9

 Union-Find
Up-Tree Data Structure
 Each subset is an up-tree
with its root as its
representative member
 All members of a given a c g h
set are nodes in that
set’s up-tree d b f i
 Hash table maps input
data to the node
associated with that data e

Up-trees are not necessarily binary!

Disjoint Sets CSE 326 Autumn 2001 10
 Find
find(f)
find(e)
a b 10 c
a c g h

d e 7 f
d b f i 11 9 8
g 12 h i
e

Just traverse to the root!

runtime:
Disjoint Sets CSE 326 Autumn 2001 11
 Union
union(a,c)

a b 10 c
a c g h

d e f
d b f i 11 9 8
g 12 h i
e

runtime: Just hang one root from the other!

Disjoint Sets CSE 326 Autumn 2001 12

 a 3 b 10 c

Example (1/11)
2 1 6
d 4 e 7 f
11 9 8
union(b,e) g 12 h 5 i

a b c d e f g h i

a b c d f g h i

e
 a 3 b 10 c

Example (2/11)
2 6
d 4 e 7 f
11 9 8
union(a,d) g 12 h 5 i

a b c d f g h i
e

a b c f g h i
d e
 a 3 b 10 c

Example (3/11)
6
d 4 e 7 f
11 9 8
union(a,b) g 12 h 5 i

a b c f g h i
d e

a c f g h i
b d
e
 a b 10 c

Example (4/11)
6
d 4 e 7 f
11 9 8
find(d) = find(e) g 12 h 5 i
No union!

a c f g h i
b d
e

While we’re finding e,

could we do anything else?
 a b 10 c

Example (5/11)
6
d e 7 f
11 9 8
union(h,i) g 12 h 5 i

a c f g h i a c f g h

b d b d i

e e
 a b 10 c

Example (6/11)
6
d e 7 f
11 9 8
union(c,f) g 12 h i

a c f g h a c g h

b d i b d f i

e e
 a b 10 c

Example (7/11) d e 7 f
find(e)
find(f) 11 9 8
union(a,c) g 12 h i

a c g h c g h

b d f i a f i

e b d

Could we do a
better job on this union? e
 a b 10 c

Example (8/11) d e f
find(f)
find(i) 11 9 8
union(c,h) g 12 h i

c g h c g

a f i a f h

b d b d i

e e
 a b 10 c

Example (9/11) d e f
find(e) = find(h) and find(b) = find(c) 11 9
So, no unions for either of these. g 12 h i

c g

a f h

b d i

e
 a b c

Example (10/11) d e f
find(d)
find(g) 11
union(c, g) g 12 h i

c g g

c
a f h
a f h
b d i
b d i

e e
 a b c

Example (11/11) d e f
find(g) = find(h)
So, no union.
And, we’re done! g 12 h i

g a b c

c d e f

a f h
g h i
b d i
Ooh… scary!
e Such a hard maze!
 Disjoint set data structure
 A forest of up-trees
can easily be stored a c g h
in an array.
 Also, if the node b d f i
names are integers or
characters, we can
use a very simple, e
perfect hash.
Nifty storage trick!
0 (a) 1 (b) 2 (c) 3 (d) 4 (e) 5 (f) 6 (g) 7 (h) 8 (i)
up-index: -1 0 -1 0 1 2 -1 -1 7

Disjoint Sets CSE 326 Autumn 2001 24

 Implementation
 Simple Find Algorithm Simple Algorithm for Union:
 Algorithm Find(i) Algorithm Union(i,j)
 { {
//replace the disjoint sets
 j:=i; with roots i and j, I not equal
 while(p[j]>0) do to j by their union Integer i,j;
 j:=p[j];
 return j; P[j] :=i;
}
 }

runtime: O(depth) or … runtime: O(1)

Disjoint Sets CSE 326 Autumn 2001 25

 Degenerate Tree:

The time taken for n-1 unions is O(n).

runtime: O(depth) or … runtime: O(1)

Disjoint Sets CSE 326 Autumn 2001 26

 Room for Improvement:
Weighted Union
 Always makes the root of the larger tree the new root
 Often cuts down on height of the new up-tree

a c g h c g h a g h

b d f i a f i b d c i

e b d e f
Could we do a e
better job on this union?
Weighted union!

Disjoint Sets CSE 326 Autumn 2001 27

 Weighted Union Code
typedef ID int;
ID union(ID x, ID y) {
assert(up[x] == -1);
assert(up[y] == -1);

if (weight[x] > weight[y]) {

up[y] = x;
weight[x] += weight[y]; new runtime of union:
} else {
up[x] = y;
weight[y] += weight[x];
}
} new runtime of find:

Disjoint Sets CSE 326 Autumn 2001 28

 Weighted Union Find Analysis
 Finds with weighted union are O(max up-tree height)
 But, an up-tree of height h with weighted union must
have at least 2h nodes
Base case: h = 0, tree has 20 = 1 node
Induction hypothesis: assume true for h < h

A merge can only increase tree height by

one over the smaller tree. So, a tree of
  2max height  n and height h-1 was merged with a larger tree to
form the new tree. Each tree then has  2h-1
max height  log n nodes by the induction hypotheses for a
 So, find takes O(log n) total of at least 2h nodes. QED.

Disjoint Sets CSE 326 Autumn 2001 29

 Room for Improvement:
Path Compression(collapsing find)
 Points everything along the path of a find to the root
 Reduces the height of the entire access path to 1

a c f g h i a c f g h i
b d b d
e e
While we’re finding e,
could we do anything else?
Path compression!

Disjoint Sets CSE 326 Autumn 2001 30

 Path Compression Example
find(e)

c g c g

a f h a f h
b e
b d
d
i
i e

Disjoint Sets CSE 326 Autumn 2001 31

 Path Compression(collapsing find) Code
typedef ID int;
ID find(Object x) {
assert(hTable.contains(x));
ID xID = hTable[x];
ID hold = xID;
while(up[xID] != -1) {
xID = up[xID];
}
while(up[hold] != -1) {
temp = up[hold];
up[hold] = xID;
hold = temp; runtime:
}
return xID;
}

Disjoint Sets CSE 326 Autumn 2001 32

 Complexity of
Weighted Union + Path Compression
Ackermann created a function A(x, y) which grows very fast!
Inverse Ackermann function (x, y) grows very slooooowwwly …
Single-variable inverse Ackermann function is called log* n
How fast does log n grow? log n = 4 for n = 16
Let log(k) n = log (log (log … (log n)))

Then, let log* n = minimum k such that log(k) n  1

How fast does log* n grow? log* n = 4 for n = 65536
log* n = 5 for n = 265536 (20,000 digit number!)

How fast does (x, y) grow? Even sloooowwwer than log* n

(x, y) = 4 for n far larger than the number of atoms in the
universe (2300)

Disjoint Sets CSE 326 Autumn 2001 33

 Complexity of
Weighted Union + Path Compression
 Tarjan proved that m weighted union and find
operations on a set of n elements have worst case
complexity O(m(m, n))

 This is essentially amortized constant time

 In some practical cases, weighted union or path

compression or both are unnecessary because trees
do not naturally get very deep.

Disjoint Sets CSE 326 Autumn 2001 34