Queues and Stacks

Atul Prakash Downey: Chapter 15 and 16

Queues
Queues occur in real life a lot. Queues at checkout Queues in banks In software systems: Queue of requests at a web servers

Properties of Queues
Queue is a FIFO data structure. First-inrst-out. Basic operations: end of the queue

enqueue(element): add element to the dequeue() ->

returns the element from the front of the queue, removing it

isEmpty() -> returns true if queue empty Some queues provide additional
operations, such as peeking at the front

Wrapping Lists to Create Queues

Can use lists as the underlying structure.

Create empty queue boolean Q.isEmpty(): true if queue is empty Q.enqueue(T elem): add element at the end of the list T Q.dequeue(): remove element from the front of the list and returns it.

Design Outline
List a; isEmpty() { if a.size() == 0 return true; else return false; } enqueue(T element) { a.add(element); } dequeue() { T result = a.get(0); a.remove(result); return result; } Queue() { a = new List() }

What we just did

List a; T data[]; isEmpty() { if a.size() == 0 return true; else return false; } enqueue(T element) { a.add(element); } dequeue() { T result = a.get(0); a.remove(result); return result; } Queue() { a = new List() } size() { .... } add(T element) { .... } remove() { .... T get(int i) { }

Queue.java

List.java

Discussion
Lists can do everything that queues can So, why implement a queue class? Why not
directly use lists?

Wrapping
We created lists by using an array and
starting point providing functions such as add(), remove(), get(), size(), etc.

We created queues by using lists as the We are sometimes restricting functionality

by doing this, but also making the data structure safer and easier to use for the intended task

Queue using ArrayList a (pseudo-code)

enqueue(o): same as a.add(o) a.dequeue(): three steps using ArrayList API result = a.get(0) a.remove(0) or a.remove(result),
depending on the interface of arraylist a

return result

Performance using Array-based Lists

enqueue operation:

a[size] = element; size++. Performance: O(1) But, if array lls up, expensive. all n elements have to be copied to a new, larger array. Cost: O(n) in copying. Think about if you double every time array lls up, what is the average cost per insertion? Suppose array size is 100.

First 100 insertions are cheap (constant time). Next one costs 100 operations. Then, the next 100 will be cheap. Next one costs 200. Then, the next 200 will be cheap, next one costs 400, etc.

One expensive operation makes subsequent ones cheap. Average performance per insert is at most one element copy + one element update. It is still constant time.

deque Operation
save A[0], shift all elements left, and
decrement size

Have to shift everything to the left. Cost: O(size)

Overall Cost
With array-based lists, adding elements at
the end and removing from the front:

Enqueue: O(n) worst-case. O(1): average Dequeue: O(n)

Mini-exercise
Create a stack by wrapping a list. Three
operations:

isEmpty() push(T element): push an element on top

of the stack

T pop(): remove and return the top

element from the stack

Just need to give pseudo-code of Stack.java

Discussion

Stacks: What is the cost of push and pop()

in terms of big-Oh?

Queues: The cost of dequeue is O(n) and

the cost of enqueue is O(1). Can we make them both O(1)?

Making Queues Efcient

Yes, we can make both operations O(1) Using arrays with a data structure called
circular queues

Using linked lists (more on that later)

Circular Queues: Basic Idea

Right now, our implementation is as
follows:

It turns out that we can implement queues

directly on top of arrays more efciently.

Arrays used to implement lists Lists used to implement queues

Circular queue

Keep a head pointer to the the front of the queue. Tail pointer to the next empty slot. Best to imagine the array to be CIRCULAR. Position 0 follows 17.

Insertion

We added 15 to the queue. tail advances. Basic code (without error checking for full queue) A[tail] = elem; tail = (tail + 1) % size.

Deletion

head advances on deletion. Basic code without error checking for empty queue: result = A[head]; head = (head + 1) mod size; return result

Wrap around

After some deletions and insertions, we can get the above situation. Queue: 4, 5, 8, ., 12. Arithmetic on head/tail must be mod 18

Inserts after wraparound

Queue: 4, 5, 8, 4 11, 12, 13

Empty Queue
What happens when we dequeue the last element from the queue?

Empty queue situation

Rule: (head == tail) means the queue is empty. Initialization of a circular queue: head = 0; tail = 0 would be a reasonable choice.

Full Queue
What happens on adding to the following
queue?

Use the last slot?

head will become same as tail. How do we know whether the queue is full
or empty?

Note
If the queue contains values, we can't really
use the values to determine whether the queue is empty or full.

Equivalent queues

10

30

40

There is always some value in "empty" slots in the array.

Avoiding the trap

Always leave one slot empty. This is a full
circular queue

Rule: if (tail + 1) mod size equals head, then the queue is full.

Cost of insertion/ deletion

Insertion cost: add at tail, advance tail.
head. O(1). O(1). (assuming no overow allowed).

Deletion cost: remove from head, advance

Discussion
Circular queues, as described, are of
bounded size, limited by array size. queues?

What if we wanted unbounded-length