Heaps and Heapsort

15 August 2024 OSU CSE 1

 Heaps
• A heap is a binary tree of T that
satisfies two properties:
– Global shape property: it is a complete
binary tree
– Local ordering property: the label in each
node is “smaller than or equal to” the label in
each of its child nodes

15 August 2024 OSU CSE 2

 Heaps
• A heap is a binary tree of T that
satisfies two properties:
– Global shape property: it is a complete
binarybinary
A complete tree tree is one in which
all–levels
Localare “full” except
ordering possiblythe
property: thelabel in each
bottom level, with any nodes on the
node is “smaller than or equal
bottom level as far left as possible.
to” the label in
each of its child nodes

15 August 2024 OSU CSE 3

 Heaps
•Also
A heap is a binary
in the picture tree of T that
is (as with BSTs,
sorting, etc.) two
satisfies a totalproperties:
preorder that
makes this notion precise.
– Global shape property: it is a complete
binary tree
– Local ordering property: the label in each
node is “smaller than or equal to” the label in
each of its child nodes

15 August 2024 OSU CSE 4

 Simplification
• For simplicity in the following illustrations,
we use only one kind of example:
– T = integer
– The ordering is ≤
• Because heaps are used in sorting, where
duplicate values may be involved, we
allow that multiple nodes in a heap may
have the same labels (i.e., we will not
assume that the labels are unique)
15 August 2024 OSU CSE 5
 The Big Picture

15 August 2024 OSU CSE 6

 The Big Picture
This tree’s root
label y satisfies
x ≤ y

15 August 2024 OSU CSE 7

 The Big Picture
Observe: This tree is also
a heap; and for each
node in this tree with
label z, we have:
x ≤ y ≤ z.
x

15 August 2024 OSU CSE 8

 The Big PictureThis tree’s root
label y satisfies
x ≤ y

15 August 2024 OSU CSE 9

 The Big Picture
Observe: This tree is also
a heap; and for each
node in this tree with
label z, we have:
x ≤ y ≤ z.
x

15 August 2024 OSU CSE 10

 Examples of Heaps
7 1 1

2 8 1 8

5 8 2

15 August 2024 OSU CSE 11

 Non-Examples of Heaps
1 1 1

2 2 5 2 1

8 1 8

15 August 2024 OSU CSE 12

 Non-Examples of Heaps
1 1 1

2 2 5 2 1

8 1 8

Shape property is violated here: not 5

all nodes at the bottom level are as
far left as possible.

15 August 2024 OSU CSE 13

 Non-Examples of Heaps
1 1 1

2 2 5 2 1

8 1 8

Ordering property is violated here: 5

value is out of order with that of its
right child.

15 August 2024 OSU CSE 14

 Non-Examples of Heaps
1 1 1

2 2 5 2 1

8 1 8

Shape property is violated: two 5

levels are not “full”.

15 August 2024 OSU CSE 15

 Heapsort
• A heap can be used to represent the
values in a SortingMachine, as follows:
– In changeToExtractionMode, arrange all
the values into a heap
– In removeFirst, remove the root, and adjust
the slightly mutilated heap to make it a heap
again

15 August 2024 OSU CSE 16

 Heapsort
Why should this work?
• A heap can be used to represent the
values in a SortingMachine, as follows:
– In changeToExtractionMode, arrange all
the values into a heap
– In removeFirst, remove the root, and adjust
the slightly mutilated heap to make it a heap
again

15 August 2024 OSU CSE 17

 How removeFirst Can Work
• If the root is the only node in the heap,
then after removing it, what remains is
already a heap; nothing left to do
• If the root is not the only node, then
removing it leaves an “opening” that must
be filled by moving some other value in the
heap into the opening

15 August 2024 OSU CSE 18

 How removeFirst Can Work
The question is:
• If the root is the only node in
thewhich
heap,
one?
then after removing it, what remains is
already a heap; nothing left to do
• If the root is not the only node, then
removing it leaves an “opening” that must
be filled by moving some other value in the
heap into the opening

15 August 2024 OSU CSE 19

 Example: A First Attempt

2 5

4 3

15 August 2024 OSU CSE 20

 Example: A First Attempt
If we remove the
root, leaving this
opening ...

2 5

4 3

15 August 2024 OSU CSE 21

 Example: A First Attempt
... we might move
up the smaller
child ...
2

4 3

15 August 2024 OSU CSE 22

 Example: A First Attempt
... now creating
another opening
...
2

4 3

15 August 2024 OSU CSE 23

 Example: A First Attempt
... so, we might
move up the
smaller child.
2

3 5

15 August 2024 OSU CSE 24

 Example: A First Attempt
Is the result
necessarily a
heap?
2

3 5

15 August 2024 OSU CSE 25

 Example: A Second Attempt

2 5

4 3

15 August 2024 OSU CSE 26

 Example: A Second Attempt
This time, let’s
maintain the
shape property ...

2 5

4 3

15 August 2024 OSU CSE 27

 Example: A Second Attempt
... by promoting
the last node on
the bottom level.
3

2 5

15 August 2024 OSU CSE 28

 Example: A Second Attempt
Now, we can “sift
down” the root into
its proper place ...
3

2 5

15 August 2024 OSU CSE 29

 Example: A Second Attempt
... by swapping it
with its smaller
child ...
2

3 5

15 August 2024 OSU CSE 30

 Example: A Second Attempt
... and then “sifting
down” the root of
that subtree.
2

3 5

15 August 2024 OSU CSE 31

 Example: A Second Attempt
Is the result
necessarily a
heap?
2

3 5

15 August 2024 OSU CSE 32

 Pseudo-Contract
/**
* Restores a complete binary tree to be a heap
* if only the root might be out of place.
* @updates t
* @requires
* [t is a complete binary tree] and
* [both subtrees of the root of t are heaps]
* @ensures
* [t is a heap with the same values as #t]
*/
public static void siftDown (BinaryTree<T> t)
{...}

15 August 2024 OSU CSE 33

Building a Heap In the First Place
• Suppose we have n values in a complete
binary tree, but they are arranged without
regard to the heap ordering
• How can we “heapify” it?

15 August 2024 OSU CSE 34

 Pseudo-Contract
/**
* Makes a complete binary tree into a heap.
* @updates t
* @requires
* [t is a complete binary tree]
* @ensures
* [t is a heap with the same values as #t]
*/
public static void heapify (BinaryTree<T> t)
{...}

15 August 2024 OSU CSE 35

 Hint
• To see how you might implement
heapify, compare the contracts of
siftDown and heapify
• The only difference: before we can call
siftDown to make a heap, both subtrees
of the root must already be heaps
– Once they are heaps, just a call to siftDown
will finish the job

15 August 2024 OSU CSE 36

 Hint How do we make
the subtrees into
• To see how you might implementheaps?
heapify, compare the contracts of
siftDown and heapify
• The only difference: before we can call
siftDown to make a heap, both subtrees
of the root must already be heaps
– Once they are heaps, just a call to siftDown
will finish the job

15 August 2024 OSU CSE 37

 Example

5 2

1 4

15 August 2024 OSU CSE 38

 Example
First, recursively
“heapify” the left
subtree.
3

1 2

5 4

15 August 2024 OSU CSE 39

 Example
Then, recursively
“heapify” the right
subtree.
3

1 2

5 4

15 August 2024 OSU CSE 40

 Example
Then “sift down” the
root, because now
only the root might
3 be out of place.

1 2

5 4

15 August 2024 OSU CSE 41

 Embedding a Heap in an array
• While one could represent a heap using a
BinaryTree<T> (as suggested in the
pseudo-contracts above), it is generally
not done this way
• Instead, a heap is usually represented
“compactly” using an array of T

15 August 2024 OSU CSE 42

 Interpreting an array as a Heap

10

20 50

40 30

15 August 2024 OSU CSE 43

 Interpreting an array as a Heap
Because it’s a complete
binary tree, the nodes
can be numbered top-to-
0
10 bottom, left-to-right.

1 2
20 50

3 4
40 30

15 August 2024 OSU CSE 44

 0
10

1 2
20 50

3 4
40 30

10 20 50 40 30

0 1 2 3 4
15 August 2024 OSU CSE 45
 At what index in the
array is the left child of 0
the node at index i? 10

1 2
20 50

3 4
40 30

10 20 50 40 30

0 1 2 3 4
15 August 2024 OSU CSE 46
 At what index in the
array is the right child 0
of the node at index i? 10

1 2
20 50

3 4
40 30

10 20 50 40 30

0 1 2 3 4
15 August 2024 OSU CSE 47
 Resources
• Wikipedia: Heapsort
– https://en.wikipedia.org/wiki/Heapsort
• Wikipedia: Heap (data structure)
– https://en.wikipedia.org/wiki/Heap_(data_structure)
• Big Java (4th ed), Sections 16.8, 16.9
– https://library.ohio-state.edu/record=b8540788~S7

15 August 2024 OSU CSE 48