Aho-3 7

CHAPTER 3.
LEXICAL ANALYSIS
Figure 3.30: NFA for Exercise 3.6.4
Exercise 3.6.4 : Repeat Exercise 3.6.3 for the NFA of Fig. 3.30.
Exercise 3.6.5 : Give the transition tables for the NFA of:
a) Exercise 3.6.3.
b) Exercise 3.6.4.
c) Figure 3.26.
3.7 From Regular Expressions to Automata

The regular expression is the notation of choice for describing lexical analyzers
and other pattern-processing software, as was reflected in Section 3.5. How-
ever, implementation of that software requires the simulation of a DFA, as in
Algorithm 3.18, or perhaps simulation of an NFA. Because an NFA often has a
choice of move on an input symbol (as Fig. 3.24 does on input a from state 0)
or on e (as Fig. 3.26 does from state 0), or even a choice of making a transition
on E: or on a real input symbol, its simulation is less straightforward than for a
DFA. Thus often it is important to convert an NFA to a DFA that accepts the
same language.
In this section we shall first show how to convert NFA's to DFA's. Then, we
use this technique, known as "the subset construction," to give a useful algo-
rit hm for simulating NFA's directly, in situations (other than lexical analysis)
where the NFA-to-DFA conversion takes more time than the direct simulation.
Next, we show how to convert regular expressions to NFA's, from which a DFA
can be constructed if desired. We conclude with a discussion of the time-space
tradeoffs inherent in the various methods for implementing regular expressions,
and see how to choose the appropriate method for your application.
3.7.1 Conversion of an NFA to a DFA

The general idea behind the subset construction is that each state of the
constructed DFA corresponds to a set of NFA states. After reading input
3.7. FROM REGULAR EXPRESSIONS T O AUTOMATA
ala2 - - .a,, the DFA is in that state which corresponds to the set of states that
the NFA can reach, from its start state, following paths labeled ala2 . . an.
It is possible that the number of DFA states is exponential in the number
of NFA states, which could lead to difficulties when we try to implement this
DFA. However, part of the power of the automaton-based approach to lexical
analysis is that for real languages, the NFA and DFA have approximately the
same number of states, and the exponential behavior is not seen.
Algorithm 3.20 : The subset construction of a DFA from an NFA.
OUTPUT: A DFA D accepting the same language as N.

METHOD: Our algorithm constructs a transition table Dtran for D. Each
state of D is a set of NFA states, and we construct Dtran so D will simulate
"in parallel" all possible moves N can make on a given input string. Our first
problem is to deal with e-transitions of N properly. In Fig. 3.31 we see the
definitions of several functions that describe basic computations on the states
of N that are needed in the algorithm. Note that s is a single state of N , while
T is a set of states of N .
t-closure(s) I Set of NFA states reachable from NFA state s

on e-transitions alone.
e-closure(T) Set of NFA states reachable from some NFA state s
Iin set T on €-transitions alone; = Us in T e-closure(s).
move(T,a) ( Set of NFA states to which there is a transition on
I input symbol a from some state s in T.
Figure 3.31: Operations on NFA states
We must explore those sets of states that N can be in after seeing some input
string. As a basis, before reading the first input symbol, N can be in any of the
states of E-closure(so),where so is its start state. For the induction, suppose
that N can be in set of states T after reading input string x. If it next reads
input a, then N can immediately go to any of the states in move(T, a). However,
after reading a, it may also make several €-transitions; thus N could be in any
state of e-closure(move(T, a)) after reading input xu. Following these ideas, the
construction of the set of D's states, Dstates, and its transition function Dtran,
is shown in Fig. 3.32.
The start state of D is c-closure(so), and the accepting states of D are all
those sets of N's states that include at least one accepting state of N . To
complete our description of the subset construction, we need only to show how
CHAPTER 3.- LEXICAL ANALYSIS
initially, e-closure(so) is the only state in Dstates, and it is unmarked;

while ( there is an unmarked state T in Dstates ) {
mark T ;
for ( each input symbol a ) {
U = E- closure(moue(~,a));
if ( U is not in Dstates )
add U as an unmarked state to Dstates;
Dtran[T, a] = U;
3
Figure 3.32: The subset construction
E-closure(T) is computed for any set of NFA states T. This process, shown in
Fig. 3.33, is a straightforward search in a graph from a set of states. In this
case, imagine that only the €-labeled edges are available in the graph.
push all states of T onto stack;

initialize E- closure(T) to T;
while ( stack is not empty ) {
pop t, the top element, off stack;
for ( each state u with an edge from t to u labeled e )
if ( u is not in e-closure(T) ) {
add u to e-closure(T);
push u onto stack
1
Figure 3.33: Computing E- closure(T)
Example 3.21 : Figure 3.34 shows another NFA accepting (a1b ) *abb; it hap-
pens to be the one we shall construct directly from this regular expression in
Section 3.7. Let us apply Algorithm 3.20 to Fig. 3.29.
The start state A of the equivalent DFA is E-closure(O),or A = {0,1,2,4,7),
since these are exactly the states reachable from state 0 via a path all of whose
edges have label e. Note that a path can have zero edges, so state 0 is reachable
from itself by an €-labeled path.
The input alphabet is {a, b). Thus, our first step is to mark A and compute
Dtran[A, a] = E-closure(moue(A,a)) and Dtran[A, b] = t- closure(moue(A, b)) .
Among the states 0, 1, 2, 4, and 7, only 2 and 7 have transitions on a, to
3 and 8, respectively. Thus, move(A, a) = {3,8). Also, t-closure({3,8) =
{1,2,3,4,6,7,8),so we conclude
3.7. FROM REGULAR EXPRESSIONS T O AUTOMATA
Figure 3.34: NFA N for (alb)*abb
Dtran[A, a] = e-closure(rnoue(A,a ) ) = e-closure({3, 8)) = {I, 2,3,4,6,7,8)

Let us call this set B, so Dtran[A, a] = B.
Now, we must compute Dtran[A, b]. Among the states in A, only 4 has a
transition on b, and it goes to 5. Thus,
Let us call the above set C, so Dtran[A, b] = C.
Figure 3.35: Transition table Dtran for DFA D
If we continue this process with the unmarked sets B and C, we eventually

reach a point where all the states of the DFA are marked. This conclusion is
guaranteed, since there are "only" 2'' different subsets of a set of eleven NFA
states. The five different DFA states we actually construct, their correspond-
ing sets of NFA states, and the transition table for the DFA D are shown in
Fig. 3.35, and the transition graph for D is in Fig. 3.36. State A is the start
state, and state E, which contains state 10 of the NFA, is the only accepting
state.
Note that D has one more state than the DFA of Fig. 3.28 for the same lan-
guage. States A and C have the same move function, and so can be merged. We
discuss the matter of minimizing the number of states of a DFA in Section 3.9.6.
CHAPTER 3. LEXICAL ANALYSIS
Figure 3.36: Result of applying the subset construction to Fig. 3.34
3.7.2 Simulationofan NFA

A strategy that has been used in a number of text-editing programs is to con-
struct an NF'A from a regular expression and then simulate the NFA using
something like an on-the-fly subset construction. The simulation is outlined
below.
Algorit hrn 3.22 : Simulating an NFA.

INPUT: An input string x terminated by an end-of-file character eof. An NFA
N with start state so, accepting states F, and transition function moue.
OUTPUT: Answer "yes7' if M accepts x; "no" otherwise.
METHOD: The algorithm keeps a set of current states S, those that are reached
from so following a path labeled by the inputs read so far. If c is the next input
character, read by the function nextchar(), then we first compute move(S, c)
and then close that set using E-closure(). The algorithm is sketched in Fig. 3.37.
S = E-closure(so);
c = nextchar();
while ( c != eof ) {
S = E-closure (move(S, c)) ;
c = nextchar();
1
if ( S n F != 0 ) return Ityesll;
else return "no";
Figure 3.37: Simulating an NFA

Aho-3 7

Uploaded by

Copyright:

Available Formats

Aho-3 7

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Aho-3 7

Uploaded by

Copyright:

Available Formats

CHAPTER 3.

Figure 3.30: NFA for Exercise 3.6.4

3.7 From Regular Expressions to Automata

3.7.1 Conversion of an NFA to a DFA

Algorithm 3.20 : The subset construction of a DFA from an NFA.

OUTPUT: A DFA D accepting the same language as N.

t-closure(s) I Set of NFA states reachable from NFA state s

Figure 3.31: Operations on NFA states

initially, e-closure(so) is the only state in Dstates, and it is unmarked;

Figure 3.32: The subset construction

push all states of T onto stack;

Figure 3.33: Computing E- closure(T)

Figure 3.34: NFA N for (alb)*abb

Dtran[A, a] = e-closure(rnoue(A,a ) ) = e-closure({3, 8)) = {I, 2,3,4,6,7,8)

Let us call the above set C, so Dtran[A, b] = C.

Figure 3.35: Transition table Dtran for DFA D

If we continue this process with the unmarked sets B and C, we eventually

Figure 3.36: Result of applying the subset construction to Fig. 3.34

3.7.2 Simulationofan NFA

Algorit hrn 3.22 : Simulating an NFA.

Figure 3.37: Simulating an NFA

You might also like