CS2040 Tutorial 6

Week 8, AY 23/24 Sem 1

Lee Yong Ler

[email protected]

Wang Zhi Jian

[email protected]
Disjoint Set ADT

Disjoint Set ADT

● Find: Given an element, find which set it is in. O(α(n)).

● Merge: Given two sets, merge them into one set. O(α(n)).
Union-Find Disjoint Sets (UFDS)

Representative

Each set is identified by a representative.

1 4 5 7

2 6

3
Union-Find Disjoint Sets (UFDS)

Find

Given an element, we can find the representative of the set it is in.

1 4 5 7

2 6

3
Union-Find Disjoint Sets (UFDS)

Find

Given an element, we can find the representative of the set it is in.

e.g. find(3) → 1

1 4 5
7

2 6
3
Union-Find Disjoint Sets (UFDS)

Merge

When we need to merge two sets, we do the following two steps.

1 4 5 7

2 6

3
Union-Find Disjoint Sets (UFDS)

Merge

1. Find the representatives of the two sets that we want to merge.

e.g. merge(3, 6)

1 4 5
7

2 6
3
Union-Find Disjoint Sets (UFDS)

Merge

2. Make one of them the representative of both sets.

e.g. merge(3, 6)

1 5 7
4

2 6
3
Union-Find Disjoint Sets (UFDS)

Find: Path Compression

After a series of merges, a set might be very long, such that it looks like a chain. This is not
good as find operations may now run in O(n) for a set of n elements.

1 2 3 4 5
Union-Find Disjoint Sets (UFDS)

Find: Path Compression

We can try to shorten the chains by linking elements directly to their representative during a find,
so that future finds will be faster.

1 2 3 4 5
Union-Find Disjoint Sets (UFDS)

Find: Path Compression

We can try to shorten the chains by linking elements directly to their representative during a find,
so that future finds will be faster. e.g. find(4)

1 2 3 4 5
Union-Find Disjoint Sets (UFDS)

Find: Path Compression

We can try to shorten the chains by linking elements directly to their representative during a find,
so that future finds will be faster. e.g. find(4)

1 2 3 4 5
Union-Find Disjoint Sets (UFDS)

Find: Path Compression

We can try to shorten the chains by linking elements directly to their representative during a find,
so that future finds will be faster. e.g. find(4)

1 2 3 4 5
Union-Find Disjoint Sets (UFDS)

Find: Path Compression

We can try to shorten the chains by linking elements directly to their representative during a find,
so that future finds will be faster. e.g. find(4)

1 2 3 4 5
Union-Find Disjoint Sets (UFDS)

Find: Path Compression

We can try to shorten the chains by linking elements directly to their representative during a find,
so that future finds will be faster. e.g. find(4)

1 2 3 4 5
Union-Find Disjoint Sets (UFDS)

Merge: Union By Rank

Initially, every representative has rank 0.

0 0 0 0

1 2 3 4
Union-Find Disjoint Sets (UFDS)

Merge: Union By Rank

When merging two representatives of the same rank, merge in either direction, and increase
the rank of the new representative by 1.
0 0 0 0

1 2 3 4
Union-Find Disjoint Sets (UFDS)

Merge: Union By Rank

When merging two representatives of the same rank, merge in either direction, and increase
the rank of the new representative by 1.
0 1 0 0

1 2 3 4
Union-Find Disjoint Sets (UFDS)

Merge: Union By Rank

When merging two representatives of the different ranks, merge the one with the smaller rank to
that with the greater rank. Rank of new representative is unchanged.
0 1 0 0

1 2 3 4
Union-Find Disjoint Sets (UFDS)

Merge: Union By Rank

When merging two representatives of the different ranks, merge the one with the smaller rank to
that with the greater rank. Rank of new representative is unchanged.
0 1 0 0

1 2 3 4
Union-Find Disjoint Sets (UFDS)

Merge: Analysis of Union By Rank

If we are using union by rank only (without path compression), notice that the rank actually
represents the height of tree rooted at the representative.
Union-Find Disjoint Sets (UFDS)

Merge: Analysis of Union By Rank

If we are using union by rank only (without path compression), notice that the rank actually
represents the height of tree rooted at the representative.

For a representative of rank n, the minimum number of nodes in its tree is n + 1.

Union-Find Disjoint Sets (UFDS)

Merge: Analysis of Union By Rank

If we are using union by rank only (without path compression), notice that the rank actually
represents the height of tree rooted at the representative.

For a representative of rank n, the minimum number of nodes in its tree is n + 1.

Since we increase the rank only when merging trees of equal rank, notice that every time the rank
of a representative increases, the number of elements in the tree doubles.
Union-Find Disjoint Sets (UFDS)

Merge: Analysis of Union By Rank

If we are using union by rank only (without path compression), notice that the rank actually
represents the height of tree rooted at the representative.

For a representative of rank n, the minimum number of nodes in its tree is n + 1.

Since we increase the rank only when merging trees of equal rank, notice that every time the rank
of a representative increases, the number of elements in the tree doubles.

The maximum attainable rank would thus be bounded by the number of time we can double the
size of a tree, which would be O(log n) given n elements.
Union-Find Disjoint Sets (UFDS)

Merge: Analysis of Union By Rank

Thus the maximum height of a tree would be O(log n) given a tree with n elements. As such, if

we use union by rank only, both find and merge will run in O(log n).
Problem 2: Intergalactic Wars

Given n warlords and n worlds, where initially, each warlord has one unique world under its
domain, support the following two types of queries in sub O(log n) time:

● Warlord x conquers warlord y. Add all worlds of warlord y under warlord x.

● Answer the query of whether a world z is under a given warlord x.
Problem 2: Intergalactic Wars

Given n warlords and n worlds, where initially, each warlord has one unique world under its
domain, support the following two types of queries in sub O(log n) time:

● Warlord x conquers warlord y. Add all worlds of warlord y under warlord x.

● Answer the query of whether a world z is under a given warlord x.

The first type of query looks like a union operation, and the second type of query looks like a
findset operation. This hints us towards a UFDS solution.
Problem 2: Intergalactic Wars

Construct a UFDS on the worlds.

Add an additional parameter to every set: the warlord that is in charge of the worlds in the set.

When we merge two sets of worlds together, ensure that the representative of the merged
set is updated with the warlord in charge of the worlds in the entire set.

The second type of queries can then be answered by first doing a findset to find the representative
of the set, then checking the warlord in charge of all the worlds in the set.
Problem 2: Intergalactic Wars

Warlord 2 Warlord 4 Warlord 5 Warlord 3

1 4 5 3

2 6
Problem 2: Intergalactic Wars

Warlord 4 conquers Warlord 2.

Warlord 2 Warlord 4 Warlord 5 Warlord 3

1 4 5 3

2 6
Problem 2: Intergalactic Wars

Warlord 4 conquers Warlord 2.

Merge the two sets together using the merge operation in UFDS.
Warlord 2 Warlord 4 Warlord 5
Warlord 3

1 4 5
3

2 6
Problem 2: Intergalactic Wars

Warlord 4 conquers Warlord 2.

Update the new representative with the new ruler.

Warlord 4 Warlord 3

Warlord 5
1 4 5 3

2 6
Problem 2: Intergalactic Wars

To check if a world is under a certain warlord, simply find the representative of the set that the world
is in, and check the warlord that rules the worlds in that set.

Warlord 4 Warlord 5 Warlord 3

1 4 5 3

2 6
Problem 2: Intergalactic Wars

Make modifications / additional operations to support an additional type of query:

● Are all n worlds under warlord x?

Problem 2: Intergalactic Wars

Make modifications / additional operations to support an additional type of query:

● Are all n worlds under warlord x?

Note that since merges are performed, after every merge, the number of different sets of worlds
decreases by 1. After n - 1 merges, no more merges can be performed, since all worlds will be
under one warlord.

Thus, we can simply check if n - 1 merges have been performed, and then finding the warlord in
the representative of the whole set of worlds.
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

e.g. If you have the following sweets, the longest unbroken sequence has length 4.

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Observation: Each unbroken sequence of candies is like a group of candies.

one group another group

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Observation: When a new candy is bought, if it’s next to a candy that was bought before, it
joins the unbroken sequence / group of candies.
e.g. Buys candy 5
one group another group

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Observation: When a new candy is bought, if it’s next to a candy that was bought before, it
joins the unbroken sequence / group of candies.
e.g. Buys candy 5 ⇒ 5 is next to 4 ⇒ 5 joins the group of candies with 4
one group another group

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Observation: When a new candy is bought, if it’s next to a candy that was bought before, it
joins the unbroken sequence / group of candies.
Joining a group is like the ‘merge’ operation in UFDS!
one group another group

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Observation: What is the longest unbroken sequence of candies that you have starting
from candy 1?

one group another group

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Observation: What is the longest unbroken sequence of candies that you have starting from
candy 1? The number of candies in the same group as candy 1. Augment the
UFDS to maintain the size of each set.
one group. size = 5 another group. size = 2

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Example

GotCandy?

UFDS

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Example

Buy(5).

GotCandy?

UFDS

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Example

Buy(5). Don’t have 4 and 6, don’t merge. GotCandy(1) = false ⇒ longest = 0.

GotCandy?

UFDS

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Example

Buy(1).

GotCandy?

UFDS

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Example

Buy(1). Don’t have 2, don’t merge. Size of set of 1 = 1 ⇒ longest = 1.

GotCandy?

UFDS

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Example

Buy(3).

GotCandy?

UFDS

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Example

Buy(3). Don’t have 2 and 4, don’t merge. Size of set of 1 = 1 ⇒ longest = 1.

GotCandy?

UFDS

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Example

Buy(2).

GotCandy?

UFDS

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Example

Buy(3). Have 1 and 3, merge with both. Size of set of 1= 3 ⇒ longest = 3.

GotCandy?

UFDS

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Example

Buy(6).

GotCandy?

UFDS

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Example

Buy(6). Have 5, merge with 5. Size of set of 1= 3 ⇒ longest = 3.

GotCandy?

UFDS

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Example

Buy(4).

GotCandy?

UFDS

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Example

Buy(4). Have 3 and 5, merge with both. Size of set of 1= 6 ⇒ longest = 6.

GotCandy?

UFDS

1 2 3 4 5 6 7 8 9 10
Problem 3: Number Candies

There are m types of candies. Initially, you have no candies.

You will buy l candies. After buying each candy, output the longest unbroken sequence of
candies you have from candy 1.

Each merge can be done in O(α(n)) time. Checking the size of the set of 1 takes O(α (n)) time.
Overall, each query can be answered in O(α(n)) time.

GotCandy?

UFDS

1 2 3 4 5 6 7 8 9 10
Problem 4: Funfair Ride

There is a circular ring of n seats (seat 1 is next to seat 2 and n − 1, seat 2 is next to seat 1 and
3, and so on). m children will arrive one by one.

Each child has a desired seat. For each child, do the following:

● If the desired seat is not taken, assign that seat to the child.
● If the desired seat is taken, ‘linear probe’ for the next empty seat, and assigned that seat to the
child.

For each child, output the seat assigned to the child.

Problem 4: Funfair Ride

Similar ideas to Problem 3.

● A sequence of consecutive children form a ‘group’.

● If a child wants a seat in the middle of the group, the child gets the seat at the end of the
group.

Seats

1 2 3 4 5 6 7 8 9 10
Problem 4: Funfair Ride

Similar ideas to Problem 3.

● A sequence of consecutive children form a ‘group’.

● If a child wants a seat in the middle of the group, the child gets the seat at the end of the
group.
child wants this seat

Seats

1 2 3 4 5 6 7 8 9 10
Problem 4: Funfair Ride

Similar ideas to Problem 3.

● A sequence of consecutive children form a ‘group’.

● If a child wants a seat in the middle of the group, the child gets the seat at the end of the
group.
child wants this seat gets assigned this one

Seats

1 2 3 4 5 6 7 8 9 10
Problem 4: Funfair Ride

Similar ideas to Problem 3.

● A sequence of consecutive children form a ‘group’.

● If a child wants a seat in the middle of the group, the child gets the seat at the end of the
group.
child wants this seat

Seats

1 2 3 4 5 6 7 8 9 10
Problem 4: Funfair Ride

Similar ideas to Problem 3.

● A sequence of consecutive children form a ‘group’.

● If a child wants a seat in the middle of the group, the child gets the seat at the end of the
group.
gets assigned this one child wants this seat

Seats

1 2 3 4 5 6 7 8 9 10
Problem 4: Funfair Ride

Similar ideas to Problem 3.

● If a new seat is assigned and adjacent seats are occupied, merge the two groups of seats
together. Merge operation in UFDS.

Seats

1 2 3 4 5 6 7 8 9 10
Problem 4: Funfair Ride

Similar ideas to Problem 3.

● If a new seat is assigned and adjacent seats are occupied, merge the two groups of seats
together. Merge operation in UFDS.

Seats

1 2 3 4 5 6 7 8 9 10
Problem 4: Funfair Ride

Similar ideas to Problem 3.

● How to find seat assigned if child wants a seat in a group?

child wants this seat gets assigned this one

Seats

1 2 3 4 5 6 7 8 9 10
Problem 4: Funfair Ride

Similar ideas to Problem 3.

● How to find seat assigned if child wants a seat in a group?

Maximum seat number in that group + 1

child wants this seat gets assigned this one

Seats

1 2 3 4 5 6 7 8 9 10
Problem 4: Funfair Ride

Similar ideas to Problem 3.

● How to find seat assigned if child wants a seat in a group?

If that exceeds n: if seat 1 is not assigned, then seat 1.

gets assigned this one child wants this seat

Seats

1 2 3 4 5 6 7 8 9 10
Problem 4: Funfair Ride

Similar ideas to Problem 3.

● How to find seat assigned if child wants a seat in a group? If that

exceeds n: if seat 1 is not assigned, then seat 1.
Otherwise, maximum seat number of group with seat 1 +
1.
gets assigned this one

child wants this seat

1 2 3 4 5 6 7 8 9 10
Seats
Problem 4: Funfair Ride

Similar ideas to Problem 3.

● How to find seat assigned if child wants a seat in a group?

Augment UFDS to maintain the maximum seat number in each
group.
gets assigned this one child wants this seat

Seats

1 2 3 4 5 6 7 8 9 10
Problem 4: Funfair Ride

function merge(x, y)
x = find(x)
y = find(y)
if rank[x] <
rank[y]
parent[x] = y

else
parent[y] = x
if rank[x] == rank[y]
rank[x] ← rank[x] + 1
Problem 4: Funfair Ride

function merge(x, y)
x = find(x)
y = find(y)
if rank[x] <
rank[y]
parent[x] = y
max[y] =
max(max[x],
max[y])
else if rank[x] == rank[y]
parent[y] = x
rank[x] ← rank[x] + 1
max[x] =
max(max[x],
max[y])
 Thank you.

Have a good weekend and see you next week.