Episode 5: Stacks & Queues rev.

17/02/2009 15:49

Reading: GT 2.0, p. 56, GT 2.1.1–2.1.2, pp. 57–64, GT 1.3.3, pp.24–27

The most important concepts of this lecture are the abstract data types of stack
and queue (i.e. what operations are expected of stacks and queues and what is
the semantics of these operations).
In 1.3.3 focus on the concept of loop invariant.
Ingredients: Stacks, Stacks in PLs, Queues

The Stack Data Structure

Operations:
P USH(o) – insert object o at the top
P OP() – remove and return the top element (error if empty)
Stack semantics is LIFO = last in first out

An Array-based Implementation

An n-element array S and an integer index t that points to the top of the stack.

t
a b c
0 1 2 3 4 5 6 7 n−3 n−2 n−1

P USH(o)
P OP()
t
a b c o
0 1 2 3 4 5 6 7 n−3 n−2 n−1

Figure 1: An array-based stack, before and after P USHing an object o.

1
2 Algorithms and Data Structures

P USH(o) P OP()
1 if t = n − 1 1 if t = −1
2 then error 2 then error
3 t←t+1 3 e ← S[t]
4 S[t] ← o 4 S[t] ← N ULL
5 t←t−1
6 return e

Stacks in Implementation of Programming Languages

F (x : int) : int
1 y ← 2x
2 if 2x < 10
3 then return F(y)
4 else return y − 1

The (abstract) stack during the evaluation of F(1):

y x y x y x y x
return 15;

··· 16 8 8 4 4 2 2 1

increasing stack depth SP addresses growing

There is a lot more on the actual stack. Here is a more detailed view:

stack pointer (SP) frame pointer (FP)

higher addresses
lower addresses

return address
argument m

temporaries

argument n
argument 1
argument 2

argument 1
argument 2
variables
registers
saved

local
···

···

next
current frame previous frame
frame
 Algorithms and Data Structures 3

In short: the same constant space for all calls of a specific function, only de-
pends on the parameters, local variables and the return type.
Remember: Function calls (including recursion) cost time and space!

The Queue Data Structure

Operations:
E NQUEUE(o) – insert object o at the rear
D EQUEUE() – remove and return the object at the front (error if empty)
Queue semantics is FIFO = first in first out

An Array-based Implementation

An array Q of n cells and two integer counters f (front) and r (rear).

f r
a b c ···
0 1 2 3 4 5 6 7 n−3 n−2 n−1

E NQUEUE(o)
f r
a b c o ···
0 1 2 3 4 5 6 7 n−3 n−2 n−1

D EQUEUE()
f r
b c o ···
0 1 2 3 4 5 6 7 n−3 n−2 n−1

Figure 2: Invariant: the array is empty iff f = r.

Incrementation: f ← (f +1) mod n on dequeue, r ← (g+1) mod n on enqueue

Also called a ring buffer, an implementation often used in low-level program-
ming (for example embedded systems). Fast, requires no dynamic memory
management, but note a built-in limitation for the queue size.
4 Algorithms and Data Structures

D EQUEUE() E NQUEUE(o)
1 if I S E MPTY() 1 if S IZE() = N − 1
2 then error 2 then error
3 temp ← Q[f ] 3 Q[r] ← o
4 Q[f ] ← null 4 r ← (r + 1) mod N
5 f ← (f + 1) mod N 5
6 return temp 6
The size of the queue is computed as (N − f + r). To avoid amibguity between
the empty and full queue (both happen when f = r) we only allow the queue
to hold N − 1 elements (line 1) in E NQUEUE.