Pattern Matching

exact pattern matching

Knuth-Morris-Pratt

RE pattern matching

grep

References:
Algorithms in C (2nd edition), Chapter 19
http://www.cs.princeton.edu/introalgsds/63long
http://www.cs.princeton.edu/introalgsds/72regular
1

exact pattern matching

Knuth-Morris-Pratt

RE pattern matching

grep

2
Exact pattern matching

Problem:
Find first match of a pattern of length M in a text stream of length N.

pattern typically N >> M

n e e d l e M=6

text
i n a h a y s t a c k a n e e d l e i n a N = 21

Applications.
• parsers.
• spam filters.
• digital libraries.
• screen scrapers.
• word processors.
• web search engines.
• natural language processing.
• computational molecular biology.
• feature detection in digitized images.
... 3
Brute-force exact pattern match

Check for pattern starting at each text position.

h a y n e e d s a n n e e d l e x
n e e d l e
n e e d l e
n e e d l e
n e e d l e
n e e d l e
n e e d l e
n e e d l e
n e e d l e
public static int search(String pattern, String text) n e e d l e
{
int M = pattern.length(); n e e d l e
int N = text.length(); n e e d l e
for (int i = 0; i < N - M; i++)
{
int j;
for (j = 0; j < M; j++)
if (text.charAt(i+j) != pattern.charAt(j))
break;
if (j == M) return i; pattern start index in text
}
return -1; not found
}
4
Brute-force exact pattern match: worst case

Brute-force algorithm can be slow if text and pattern are repetitive

a a a a a a a a a a a a a a a a b text length N
a a a a a b
a a a a a b
a a a a a b
a a a a a b
a a a a a b MN char compares
a a a a a b
a a a a a b
a a a a a b
a a a a a b
a a a a a b
a a a a a b
a a a a a b pattern length M

but this situation is rare in typical applications

Hence, the indexOf() method in Java’s String class uses brute-force

5
Exact pattern matching in Java

Exact pattern matching is implemented in Java’s String class

s.indexOf(t, i): index of first occurrence of pattern t
in string s, starting at offset i.

Ex: Screen scraping. Exact match to extract info from website

public class StockQuote http://finance.yahoo.com/q?s=goog

{ ...
public static void main(String[] args) <tr>
{ <td class= "yfnc_tablehead1"
width= "48%">
String name = "http://finance.yahoo.com/q?s="; Last Trade:
In in = new In(name + args[0]); </td>
String input = in.readAll(); <td class= "yfnc_tabledata1">
int start = input.indexOf("Last Trade:", 0); <big><b>688.04</b></big>
int from = input.indexOf("<b>", start); </td></tr>
<td class= "yfnc_tablehead1"
int to = input.indexOf("</b>", from); width= "48%">
String price = input.substring(from + 3, to); Trade Time:
System.out.println(price); </td>
} <td class= "yfnc_tabledata1">
}

6
Algorithmic challenges in pattern matching

Brute-force is not good enough for all applications

Theoretical challenge: Linear-time guarantee. fundamental algorithmic problem

Practical challenge: Avoid backup in text stream. often no room or time to save text

Now is the time for all people to come to the aid of their party. Now is the time for all
good people to come to the aid of their party. Now is the time for many good people to
come to the aid of their party. Now is the time for all good people to come to the aid of
their party. Now is the time for a lot of good people to come to the aid of their party.
Now is the time for all of the good people to come to the aid of their party. Now is the
time for all good people to come to the aid of their party. Now is the time for each good
person to come to the aid of their party. Now is the time for all good people to come to
the aid of their party. Now is the time for all good Republicans to come to the aid of
their party. Now is the time for all good people to come to the aid of their party. Now
is the time for many or all good people to come to the aid of their party. Now is the
time for all good people to come to the aid of their party. Now is the time for all good
Democrats to come to the aid of their party. Now is the time for all people to come to
the aid of their party. Now is the time for all good people to come to the aid of their
party. Now is the time for many good people to come to the aid of their party. Now is the
time for all good people to come to the aid of their party. Now is the time for a lot of
good people to come to the aid of their party. Now is the time for all of the good people
to come to the aid of their party. Now is the time for all good people to come to the aid
of their attack at dawn party. Now is the time for each person to come to the aid of
their party. Now is the time for all good people to come to the aid of their party. Now
is the time for all good Republicans to come to the aid of their party. Now is the time
for all good people to come to the aid of their party. Now is the time for many or all
good people to come to the aid of their party. Now is the time for all good people to
come to the aid of their party. Now is the time for all good Democrats to come to the aid
of their party.
7

exact pattern matching

Knuth-Morris-Pratt

RE pattern matching

grep

8
Knuth-Morris-Pratt (KMP) exact pattern-matching algorithm

Classic algorithm that meets both challenges

• linear-time guarantee
• no backup in text stream

Don Knuth Jim Morris Vaughan Pratt

Basic plan (for binary alphabet)
• build DFA from pattern
• simulate DFA with text as input

t pattern in text
ep
text DFA acc

a a a b a a b a a a b for
pattern reje
ct
pattern NOT
a a b a a a in text

No backup in a DFA
Linear-time because each step is just a state change
9
Knuth-Morris-Pratt DFA example

One state for each pattern character

• Match input character: move from i to i+1
• Mismatch: move to previous state

DFA
for
pattern
a a b a a a

b b b
a a b a a a
0 1 2 3 4 5
b a
b

How to construct? Stay tuned

10
Knuth-Morris-Pratt DFA simulation

b b b
a a b a a a
0 1 2 3 4 5
b a b

0 a a a b a a b a a a b

b b b
a a b a a a
0 1 2 3 4 5
b a b

1 a a a b a a b a a a b

b b b
a a b a a a
0 1 2 3 4 5
b a b

2 a a a b a a b a a a b

b b b
a a b a a a
0 1 2 3 4 5
b a b

2 a a a b a a b a a a b

b b b
a a b a a a
0 1 2 3 4 5
b a b

3 a a a b a a b a a a b
b b b
a a b a a a
0 1 2 3 4 5
b a b
11
Knuth-Morris-Pratt DFA simulation

b b b
a a b a a a
0 1 2 3 4 5
b a b

4 a a a b a a b a a a b

b b b
a a b a a a
0 1 2 3 4 5
b a b

5 a a a b a a b a a a b

b b b
a a b a a a
0 1 2 3 4 5
b a b

3 a a a b a a b a a a b

b b b
a a b a a a
0 1 2 3 4 5
b a b

4 a a a b a a b a a a b

b b b
a a b a a a
0 1 2 3 4 5
b a b

5 a a a b a a b a a a b
b b b
a a b a a a
0 1 2 3 4 5
accept! b a b
12
Knuth-Morris-Pratt DFA simulation

When in state i:
•have found match in i previous input chars
•that is the longest such match

Ex. End in state 4 iff text ends in aaba.

Ex. End in state 2 iff text ends in aa (but not aabaa or aabaaa).

0 a a a b a a b a a a b
1 a a a b a a b a a a b
2 a a a b a a b a a a b
2 a a a b a a b a a a b
3 a a a b a a b a a a b
4 a a a b a a b a a a b
5 a a a b a a b a a a b
3 a a a b a a b a a a b
4 a a a b a a b a a a b
5 a a a b a a b a a a b
a a a b a a b a a a b

b b b
a a b a a a
0 1 2 3 4 5
b a
b
13
KMP implementation

DFA representation: a single state-indexed array next[]

• Upon character match in state j, go forward to state j+1.
• Upon character mismatch in state j, go back to state next[j].

0 1 2 3 4 5

a 1 2 2 4 5 6
DFA b 0 0 3 0 0 3
for
only need to
pattern next 0 0 2 0 0 3
store mismatches

a a b a a a

b b b
a a b a a a
0 1 2 3 4 5
b a
b

14
KMP implementation

Two key differences from brute-force implementation:

• Text pointer i never decrements
• Need to precompute next[] table (DFA) from pattern.

int j = 0;
for (int i = 0; i < N; i++)
{
if (t.charAt(i) == p.charAt(j)) j++; // match
else j = next[j]; // mismatch
if (j == M) return i - M + 1; // found
}
return -1; // not found

Simulation of KMP DFA

15
Knuth-Morris-Pratt: Iterative DFA construction

DFA for first i states contains the information needed to build state i+1

Ex: given DFA for pattern aabaaa.

how to compute DFA for pattern aabaaab ?

Key idea
• on mismatch at 7th char, need to simulate 6-char backup
• previous 6 chars are known (abaaaa in example)
• 6-state DFA (known) determines next state!
0 a b a a a a
Keep track of DFA state for start at 2nd char of pattern 1 a b a a a a

•
0 a b a a a a
compare char at that position with next pattern char 1 a b a a a a
• match/mismatch provides all needed info 2
2
a
a
b
b
a
a
a
a
a
a
a
a
2 a b a a a a

b b b
a a b a a a
0 1 2 3 4 5
b a
b
16
KMP iterative DFA construction: two cases

Let X be the next state in the simulation and j the next state to build.

If p[X] and p[j] match, copy and increment 0 1 2 3 4 5 6

next[j] = next[X]; p[] a a b a a a b
X = X+1 next[] 0 0 2 0 0 3 2

X j

state for a b a a a b

b
DFA for b b
b
a a b a a a 6
a a b a a a b 0 1 2 3 4 5
b a
b a

If p[X] and p[j] mismatch, do the opposite 0 1 2 3 4 5 6

next[j] = X+1; p[] a a b a a a a
next[] 0 0 2 0 0 3 3
X = next[X];
state for a b a a a a X j

b b b
DFA for a a b a a a a
a a b a a a a 0 1 2 3 4 5 6
b a
b b
17
Knuth-Morris-Pratt DFA construction

0
DFA X: current state in simulation
compare p[j] with p[X]
0 a
a 0 1
0 b match: copy and increment
next[j] = next[X];
X = X + 1;
0 1
p[1..j-1] X b mismatch: do the opposite
a a match a a next[j] = X + 1;
0 0 0 1 2
0 b X = next[X];
X j
0 1 2
a a b mismatch b
a a b
0 0 2 a 1 0 1 2 3
b a

0 1 2 3
a a b a b b
match a a b a
0 0 2 0 a b 0 0 1 2 3 4
b a

0 1 2 3 4
a a b a a match b b b
a a b a a
0 0 2 0 0 a b a 1 0 1 2 3 4 5
b a

0 1 2 3 4 5
a a b a a a mismatch b b b
0 0 2 0 0 3 a b a a 2 a a b a a a
0 1 2 3 4 5
b a b
18
Knuth-Morris-Pratt DFA construction examples
ex: a a b a a a b ex: a b b a b b b
0 0 X: current state in simulation
a a compare p[j] with p[X]
0 0

0 1 0 1 match: copy and increment

a a a b mismatch next[j] = next[X];
next[] 0 0 match 0 1 X = X + 1;
mismatch: do the opposite
X j X j next[j] = X + 1;
X = next[X];
0 1 2 0 1 2
a a b mismatch a b b mismatch
0 0 2 0 1 1

0 1 2 3 0 1 2 3
a a b a match a b b a match
0 0 2 0 0 1 1 0

0 1 2 3 4 0 1 2 3 4
a a b a a match a b b a b match
0 0 2 0 0 0 1 1 0 1

0 1 2 3 4 5 0 1 2 3 4 5
a a b a a a mismatch a b b a b b match
0 0 2 0 0 3 0 1 1 0 1 1

0 1 2 3 4 5 6 0 1 2 3 4 5 6
a a b a a a b match a b b a b b b mismatch
0 0 2 0 0 3 2 0 1 1 0 1 1 4
19
DFA construction for KMP: Java implementation

Takes time and space proportional to pattern length.

int X = 0;
int[] next = new int[M];
for (int j = 1; j < M; j++)
{
if (p.charAt(X) == p.charAt(j))
{ // match
next[j] = next[X];
X = X + 1;
}
else
{ // mismatch
next[j] = X + 1;
X = next[X];
}
}

DFA Construction for KMP (assumes binary alphabet)

20
Optimized KMP implementation

Ultimate search program for any given pattern:

• one statement comparing each pattern character to next
• match: proceed to next statement
• mismatch: go back as dictated by DFA
• translates to machine language (three instructions per pattern char)

int kmpsearch(char t[])

{
int i = 0;
s0: if (t[i++] != 'a') goto s0;
s1: if (t[i++] != 'a') goto s0;
s2: if (t[i++] != 'b') goto s2;
s3: if (t[i++] != 'a') goto s0; assumes pattern is in text
(o/w use sentinel)
s4: if (t[i++] != 'a') goto s0;
s5: if (t[i++] != 'a') goto s3;
s6: if (t[i++] != 'b') goto s2;
s7: if (t[i++] != 'b') goto s4;
return i - 8;
} pattern[] next[]

Lesson: Your computer is a DFA! 21

KMP summary

General alphabet
• more difficult
• easy with next[][] indexed by mismatch position, character
• KMP paper has ingenious solution that is not difficult to implement
[ build NFA, then prove that it finishes in 2N steps ]

Bottom line: linear-time pattern matching is possible (and practical)

Short history:
• inspired by esoteric theorem of Cook
[ linear time 2-way pushdown automata simulation is possible ]
• discovered in 1976 independently by two theoreticians and a hacker
Knuth: discovered linear time algorithm
Pratt: made running time independent of alphabet
Morris: trying to build a text editor.
• theory meets practice 22
Exact pattern matching: other approaches

Rabin-Karp: make a digital signature of the pattern

• hashing without the table
• linear-time probabilistic guarantee
• plus: extends to 2D patterns
• minus: arithmetic ops much slower than char comparisons

Boyer-Moore: scan from right to left in pattern

• main idea: can skip M text chars when finding one not in the pattern
• needs additional KMP-like heuristic
• plus: possibility of sublinear-time performance (~ N/M )
• used in Unix, emacs

pattern s y z y g y

text a a a b b a a b a b a a a b b a a a b a a
s y z y g y
s y z y g y
s y z y g y
23
Exact pattern match cost summary

Cost of searching for M-character pattern in N-character text

algorithm typical worst-case

brute-force 1.1 N char compares † M N char compares

Karp-Rabin 3N arithmetic ops 3N arithmetic ops ‡

KMP 1.1 N char compares † 2N char compares

Boyer-Moore ~ N/M char compares † 3N char compares

† assumes appropriate model
‡ randomized

24

exact pattern matching

Knuth-Morris-Pratt

RE pattern matching

grep

25
Regular-expression pattern matching

Exact pattern matching:

Search for occurrences of a single pattern in a text file.

Regular expression (RE) pattern matching:

Search for occurrences of one of multiple patterns in a text file.

Ex. (genomics)
• Fragile X syndrome is a common cause of mental retardation.
• human genome contains triplet repeats of cgg or agg
bracketed by gcg at the beginning and ctg at the end
• number of repeats is variable, and correlated with syndrome
• use regular expression to specify pattern: gcg(cgg|agg)*ctg
• do RE pattern match on person’s genome to detect Fragile X

pattern (RE) gcg(cgg|agg)*ctg

text gcggcgtgtgtgcgagagagtgggtttaaagctggcgcggaggcggctggcgcggaggctg

26
RE pattern matching: applications

Test if a string matches some pattern.

• Process natural language.
• Scan for virus signatures.
• Search for information using Google.
• Access information in digital libraries.
• Retrieve information from Lexis/Nexis.
• Search-and-replace in a word processors.
• Filter text (spam, NetNanny, Carnivore, malware).
• Validate data-entry fields (dates, email, URL, credit card).
• Search for markers in human genome using PROSITE patterns.

Parse text files.

• Compile a Java program.
• Crawl and index the Web.
• Read in data stored in ad hoc input file format.
• Automatically create Java documentation from Javadoc comments.

27
Regular expression examples

A regular expression is a notation to specify a set of strings.

operation example RE in set not in set

concatenation aabaab aabaab every other string

cumulus succubus
wildcard .u.u.u.
jugulum tumultuous

aa
union aa | baab
baab every other string

aa ab
closure ab*a
abbba ababa

aaaab
a(a|b)aab
abaab every other string
parentheses
a aa
(ab)*a
ababababa abbba

28
Regular expression examples (continued)

Notation is surprisingly expressive

regular expression in set not in set

.*spb.* raspberry subspace

contains the trigraph spb crispbread subspecies

a* | (a*ba*ba*ba*)* bbb b
aaa bb
number of b’s is a multiple of 3 bbbaababbaa baabbbaa

.*0.... 1000234 111111111

fifth to last digit is 0 98701234 403982772

gcg(cgg|agg)*ctg gcgctg gcgcgg

gcgcggctg cggcggcggctg
fragile X syndrome indicator gcgcggaggctg gcgcaggctg

and plays a well-understood role in the theory of computation

29
Generalized regular expressions

Additional operations are often added

• Ex: [a-e]+ is shorthand for (a|b|c|d|e)(a|b|c|d|e)*
• for convenience only
• need to be alert for non-regular additions (Ex: Java /)

operation example in set not in set

abcde ade
one or more a(bc)+de
abcbcde bcde

word camelCase
character classes [A-Za-z][a-z]*
Capitalized 4illegal

08540-1321 111111111
exactly k [0-9]{5}-[0-9]{4}
19072-5541 166-54-111

negations [^aeiou]{6} rhythm decade

30
Regular expressions in Java

RE pattern matching is implemented in Java’s String class

• basic: match() method
• various other methods also available (stay tuned)

Ex: Validity checking. Is input in the set described by the re?

public class Validate
{
public static void main(String[] args)
{
String re = args[0];
String input = args[1];
System.out.println(input.matches(re));
}
}

need help solving

% java Validate "..oo..oo." bloodroot
true crosswords?

% java Validate "[$_A-Za-z][$_A-Za-z0-9]*" ident123 legal Java identifier

true
% java Validate "[a-z]+@([a-z]+\.)+(edu|com)" [email protected] valid email address
true (simplified)
% java Validate "[0-9]{3}-[0-9]{2}-[0-9]{4}" 166-11-4433 Social Security number
true
31
Regular expressions in other languages

Broadly applicable programmer's tool.

• originated in UNIX in the 1970s
• many languages support extended regular expressions
• built into grep, awk, emacs, Perl, PHP, Python, JavaScript

print all lines containing NEWLINE which

grep NEWLINE */*.java
occurs in any file with a .java extension

egrep '^[qwertyuiop]*[zxcvbnm]*$' dict.txt | egrep '...........'

PERL. Practical Extraction and Report Language.

perl -p -i -e 's|from|to|g' input.txt replace all occurrences of from

with to in the file input.txt

perl -n -e 'print if /^[A-Za-z][a-z]*$/' dict.txt

do for each line

32
Regular expression caveat

Writing a RE is like writing a program.

•need to understand programming model
•can be easier to write than read
•can be difficult to debug

"Sometimes you have a programming problem

and it seems like the best solution is to use
regular expressions; now you have two problems."

33
Can the average web surfer learn to use REs?

Google. Supports * for full word wildcard and | for union.

34
Can the average TV viewer learn to use REs?

TiVo. WishList has very limited pattern matching.

Reference: page 76, Hughes DirectTV TiVo manual

35
Can the average programmer learn to use REs?

Perl RE for Valid RFC822 Email Addresses Reference: http://www.ex-parrot.com/~pdw/Mail-RFC822-Address.html

(?:(?:\r\n)?[ \t])*(?:(?:(?:[^()<>@,;:\\".\[\] \000-\031]+(?:(?:(?:\r\n)?[ \t]

)+|\Z|(?=[\["()<>@,;:\\".\[\]]))|"(?:[^\"\r\\]|\\.|(?:(?:\r\n)?[ \t]))*"(?:(?:
\r\n)?[ \t])*)(?:\.(?:(?:\r\n)?[ \t])*(?:[^()<>@,;:\\".\[\] \000-\031]+(?:(?:(
?:\r\n)?[ \t])+|\Z|(?=[\["()<>@,;:\\".\[\]]))|"(?:[^\"\r\\]|\\.|(?:(?:\r\n)?[
\t]))*"(?:(?:\r\n)?[ \t])*))*@(?:(?:\r\n)?[ \t])*(?:[^()<>@,;:\\".\[\] \000-\0
31]+(?:(?:(?:\r\n)?[ \t])+|\Z|(?=[\["()<>@,;:\\".\[\]]))|\[([^\[\]\r\\]|\\.)*\
](?:(?:\r\n)?[ \t])*)(?:\.(?:(?:\r\n)?[ \t])*(?:[^()<>@,;:\\".\[\] \000-\031]+
(?:(?:(?:\r\n)?[ \t])+|\Z|(?=[\["()<>@,;:\\".\[\]]))|\[([^\[\]\r\\]|\\.)*\](?:
(?:\r\n)?[ \t])*))*|(?:[^()<>@,;:\\".\[\] \000-\031]+(?:(?:(?:\r\n)?[ \t])+|\Z
|(?=[\["()<>@,;:\\".\[\]]))|"(?:[^\"\r\\]|\\.|(?:(?:\r\n)?[ \t]))*"(?:(?:\r\n)
?[ \t])*)*\<(?:(?:\r\n)?[ \t])*(?:@(?:[^()<>@,;:\\".\[\] \000-\031]+(?:(?:(?:\
r\n)?[ \t])+|\Z|(?=[\["()<>@,;:\\".\[\]]))|\[([^\[\]\r\\]|\\.)*\](?:(?:\r\n)?[
\t])*)(?:\.(?:(?:\r\n)?[ \t])*(?:[^()<>@,;:\\".\[\] \000-\031]+(?:(?:(?:\r\n) “Implementing validation
?[ \t])+|\Z|(?=[\["()<>@,;:\\".\[\]]))|\[([^\[\]\r\\]|\\.)*\](?:(?:\r\n)?[ \t]
)*))*(?:,@(?:(?:\r\n)?[ \t])*(?:[^()<>@,;:\\".\[\] \000-\031]+(?:(?:(?:\r\n)?[ with regular expressions
\t])+|\Z|(?=[\["()<>@,;:\\".\[\]]))|\[([^\[\]\r\\]|\\.)*\](?:(?:\r\n)?[ \t])*
)(?:\.(?:(?:\r\n)?[ \t])*(?:[^()<>@,;:\\".\[\] \000-\031]+(?:(?:(?:\r\n)?[ \t]
somewhat pushes the limits
)+|\Z|(?=[\["()<>@,;:\\".\[\]]))|\[([^\[\]\r\\]|\\.)*\](?:(?:\r\n)?[ \t])*))*) of what it is sensible to do
*:(?:(?:\r\n)?[ \t])*)?(?:[^()<>@,;:\\".\[\] \000-\031]+(?:(?:(?:\r\n)?[ \t])+
|\Z|(?=[\["()<>@,;:\\".\[\]]))|"(?:[^\"\r\\]|\\.|(?:(?:\r\n)?[ \t]))*"(?:(?:\r with regular expressions,
\n)?[ \t])*)(?:\.(?:(?:\r\n)?[ \t])*(?:[^()<>@,;:\\".\[\] \000-\031]+(?:(?:(?: although Perl copes well.”
\r\n)?[ \t])+|\Z|(?=[\["()<>@,;:\\".\[\]]))|"(?:[^\"\r\\]|\\.|(?:(?:\r\n)?[ \t
]))*"(?:(?:\r\n)?[ \t])*))*@(?:(?:\r\n)?[ \t])*(?:[^()<>@,;:\\".\[\] \000-\031
]+(?:(?:(?:\r\n)?[ \t])+|\Z|(?=[\["()<>@,;:\\".\[\]]))|\[([^\[\]\r\\]|\\.)*\](
?:(?:\r\n)?[ \t])*)(?:\.(?:(?:\r\n)?[ \t])*(?:[^()<>@,;:\\".\[\] \000-\031]+(?
:(?:(?:\r\n)?[ \t])+|\Z|(?=[\["()<>@,;:\\".\[\]]))|\[([^\[\]\r\\]|\\.)*\](?:(?
:\r\n)?[ \t])*))*\>(?:(?:\r\n)?[ \t])*)|(?:[^()<>@,;:\\".\[\] \000-\031]+(?:(?
:(?:\r\n)?[ \t])+|\Z|(?=[\["()<>@,;:\\".\[\]]))|"(?:[^\"\r\\]|\\.|(?:(?:\r\n)?
[ \t]))*"(?:(?:\r\n)?[ \t])*)*:(?:(?:\r\n)?[ \t])*(?:(?:(?:[^()<>@,;:\\".\[\]
\000-\031]+(?:(?:(?:\r\n)?[ \t])+|\Z|(?=[\["()<>@,;:\\".\[\]]))|"(?:[^\"\r\\]|
\\.|(?:(?:\r\n)?[ \t]))*"(?:(?:\r\n)?[ \t])*)(?:\.(?:(?:\r\n)?[ \t])*(?:[^()<>
@,;:\\".\[\] \000-\031]+(?:(?:(?:\r\n)?[ \t])+|\Z|(?=[\["()<>@,;:\\".\[\]]))|"
(?:[^\"\r\\]|\\.|(?:(?:\r\n)?[ \t]))*"(?:(?:\r\n)?[ \t])*))*@(?:(?:\r\n)?[ \t]
)*(?:[^()<>@,;:\\".\[\] \000-\031]+(?:(?:(?:\r\n)?[ \t])+|\Z|(?=[\["()<>@,;:\\
".\[\]]))|\[([^\[\]\r\\]|\\.)*\](?:(?:\r\n)?[ \t])*)(?:\.(?:(?:\r\n)?[ \t])*(?
:[^()<>@,;:\\".\[\] \000-\031]+(?:(?:(?:\r\n)?[ \t])+|\Z|(?=[\["()<>@,;:\\".\[
\]]))|\[([^\[\]\r\\]|\\.)*\](?:(?:\r\n)?[ \t])*))*|(?:[^()<>@,;:\\".\[\] \000-
\031]+(?:(?:(?:\r\n)?[ \t])+|\Z|(?=[\["()<>@,;:\\".\[\]]))|"(?:[^\"\r\\]|\\.|(
?:(?:\r\n)?[ \t]))*"(?:(?:\r\n)?[ \t])*)*\<(?:(?:\r\n)?[ \t])*(?:@(?:[^()<>@,;
:\\".\[\] \000-\031]+(?:(?:(?:\r\n)?[ \t])+|\Z|(?=[\["()<>@,;:\\".\[\]]))|\[([
^\[\]\r\\]|\\.)*\](?:(?:\r\n)?[ \t])*)(?:\.(?:(?:\r\n)?[ \t])*(?:[^()<>@,;:\\"
.\[\] \000-\031]+(?:(?:(?:\r\n)?[ \t])+|\Z|(?=[\["()<>@,;:\\".\[\]]))|\[([^\[\ 37 more lines
]\r\\]|\\.)*\](?:(?:\r\n)?[ \t])*))*(?:,@(?:(?:\r\n)?[ \t])*(?:[^()<>@,;:\\".\
[\] \000-\031]+(?:(?:(?:\r\n)?[ \t])+|\Z|(?=[\["()<>@,;:\\".\[\]]))|\[([^\[\]\
36
r\\]|\\.)*\](?:(?:\r\n)?[ \t])*)(?:\.(?:(?:\r\n)?[ \t])*(?:[^()<>@,;:\\".\[\]

exact pattern matching

Knuth-Morris-Pratt

RE pattern matching

grep

37
GREP implementation: basic plan

Overview is the same as for KMP !

• linear-time guarantee
• no backup in text stream

Basic plan for GREP Ken Thompson

• build DFA from RE

• simulate DFA with text as input

t pattern in text
DFA acc
ep
text
for
pattern
actgtgcaggaggcggcgcggcggaggaggctggcga
reje
ct
pattern NOT
gcg(cgg|agg)*ctg in text

No backup in a DFA
Linear-time because each step is just a state change
38
Deterministic finite-state automata

DFA review.

int pc = 0;
while (!tape.isEmpty())
{
boolean bit = tape.read();
if (pc == 0) { if (!bit) pc = 0; else pc = 1; }
else if (pc == 1) { if (!bit) pc = 1; else pc = 2; }
else if (pc == 2) { if (!bit) pc = 2; else pc = 0; }
}
if (pc == 0) System.out.println("accepted");
else System.out.println("rejected");

39
Duality

RE. Concise way to describe a set of strings.

DFA. Machine to recognize whether a given string is in a given set.

Kleene's theorem.
• for any DFA, there exists a RE that describes the same set of strings
• for any RE, there exists a DFA that recognizes the same set of strings

Ex: set of strings whose number of 1's is a multiple of 3

DFA

RE
0* | (0*10*10*10*)*

Good news: The basic plan works

(build DFA from RE and run with text as input)
Bad news : The DFA can be exponentially large (can’t afford to build it).
Consequence: We need a smaller abstract machine. 40
Nondeterministic finite-state automata

NFA.
• may have 0, 1, or more transitions for each input symbol
• may have
-transitions (move to another state without reading input)
• accept if any sequence of transitions leads to accept state

Ex: set of strings that do not contain 110

convention:
unlabelled arrows in set: 111, 00011, 101001011
areε- transitions
not in set: 110, 00011011, 00110

Implication of proof of Kleene’s theorem: RE -> NFA -> DFA

Basic plan for GREP (revised)

• build NFA from RE
• simulate NFA with text as input
• give up on linear-time guarantee 41
Simulating an NFA

How to simulate an NFA? Maintain set of all possible states that NFA
could be in after reading in the first i symbols.

42
NFA Simulation

43
NFA Representation

NFA representation. Maintain several digraphs, one for each symbol in

the alphabet, plus one for
.

0-graph 1-graph

-graph

44
NFA: Java Implementation

public class NFA

{
private int START = 0; // start state
private int ACCEPT = 1; // accept state
private int N = 2; // number of states
private String ALPHABET = "01"; // RE alphabet
private int EPS = ALPHABET.length(); // symbols in alphabet
private Digraph[] G;

public NFA(String re)

{
G = new Digraph[EPS + 1];
for (int i = 0; i <= EPS; i++)
G[i] = new Digraph();
build(0, 1, re);
}

private void build(int from, int to, String re) { }

public boolean simulate(Tape tape) { }
}

45
NFA Simulation

How to simulate an NFA?

• Maintain a SET of all possible states that NFA could be in after
reading in the first i symbols.
• Use Digraph adjacency and reachability ops to update.

next = neighbors states reachable

pc
of pc in G[c] updated pc
from next in G[
]

46
NFA Simulation: Java Implementation

public boolean simulate(Tape tape)

{
SET<Integer> pc = G[EPS].reachable(START); states reachable from
start by
-transitions

while (!tape.isEmpty())
{ // Simulate NFA taking input from tape.
char c = tape.read();
int i = ALPHABET.indexOf(c); all possible states after
SET<Integer> next = G[i].neighbors(pc); reading character c from tape

pc = G[EPS].reachable(next); follow
-transitions

}

for (int state : pc)

if (state == ACCEPT) return true; check whether
return false; in accept state at end

47
Converting from an RE to an NFA: basic transformations

Use generalized NFA with full RE on trasitions arrows

• start with one transition having given RE
• remove operators with transformations given below
• goal: standard NFA (all single-character or epsilon-transitions)

start
concatenation closure
from R to

from from
union
from from
c
c
from from
c R c* S

R S
R | S R S to to

to
to
to to

48
Converting from an RE to an NFA example: ab* | ab*

0 0
0
a
ab* | a*b ab* a*b 2 a*b

b*
1 1
1

0 0 0

a a a
a*
a
2 2 2 5

a*b 4

3 3 3 4
b b b
b

1 1 1
49
NFA Construction: Java Implementation

private void build(int from, int to, String re)

{
int or = re.indexOf('|');
if (re.length() == 0) G[EPSILON].addEdge(from, to); union

else if (re.length() == 1)
{ single char from from
char c = re.charAt(0);
for (int i = 0; i < EPSILON; i++) R | S R S
if (c == ALPHABET.charAt(i) || c == '.')
G[i].addEdge(from, to);
to to
}
else if (or != -1)
{ union
build(from, to, re.substring(0, or)); R | S
build(from, to, re.substring(or + 1));
} re.charAt(or)

else if (re.charAt(1) == '*')

{ closure
G[EPSILON].addEdge(from, N);
build(N, N, re.substring(0, 1));
build(N++, to, re.substring(2));
}
else
{ concatenation
build(from, N, re.substring(0, 1));
build(N++, to, re.substring(1));
c* S
}
}
re.charAt(1)
50
Grep running time

Input. Text with N characters, RE with M characters.

Claim. The number of edges in the NFA is at most 2M.

• Single character: consumes 1 symbol, creates 1 edge.
• Wildcard character: consumes 1 symbol, creates 2 edges.
• Concatenation: consumes 1 symbols, creates 0 edges.
• Union: consumes 1 symbol, creates 1 edges.
• Closure: consumes one symbol, creates 2 edges.

NFA simulation. O(MN) since NFA has 2M transitions

• bottleneck: 1 graph reachability per input character
• can be substantially faster in practice if few
-transitions
NFA construction. Ours is O(M2) but not hard to make O(M).

Surprising bottom line:

Worst-case cost for grep is the same as for elementary exact match!
51
Industrial-strength grep implementation

To complete the implementation,

• Deal with parentheses.
• Extend the alphabet.
• Add character classes.
• Add capturing capabilities.
• Deal with meta characters.
• Extend the closure operator.
• Error checking and recovery.
• Greedy vs. reluctant matching.

52
Regular expressions in Java (revisited)

RE pattern matching is implemented in Java’s Pattern and Matcher classes

Ex: Harvesting. Print substrings of input that match re

import java.util.regex.Pattern;
import java.util.regex.Matcher; compile() creates a
Pattern (NFA) from RE
public class Harvester
{ matcher() creates a
public static void main(String[] args) Matcher (NFA simulator)
{
from NFA and text
String re = args[0];
In in = new In(args[1]);
String input = in.readAll();
Pattern pattern = Pattern.compile(re); find() looks for
Matcher matcher = pattern.matcher(input); the next match
while (matcher.find())
System.out.println(matcher.group()); group() returns
} the substring most
} recently found by find()

% java Harvester "gcg(cgg|agg)*ctg" chromosomeX.txt

gcgcggcggcggcggcggctg harvest patterns
gcgctg from DNA
gcgctg
gcgcggcggcggaggcggaggcggctg

% java Harvester "http://(\\w+\\.)*(\\w+)" http://www.cs.princeton.edu harvest links

http://www.princeton.edu from website
http://www.google.com 53
Typical application: Parsing a data file

Example. NCBI genome file, …

LOCUS AC146846 128142 bp DNA linear HTG 13-NOV-2003

DEFINITION Ornithorhynchus anatinus clone CLM1-393H9,
ACCESSION AC146846
KEYWORDS HTG; HTGS_PHASE2; HTGS_DRAFT.
SOURCE Ornithorhynchus anatinus (platypus)
ORIGIN
1 tgtatttcat ttgaccgtgc tgttttttcc cggtttttca gtacggtgtt agggagccac
61 gtgattctgt ttgttttatg ctgccgaata gctgctcgat gaatctctgc atagacagct // a comment
121 gccgcaggga gaaatgacca gtttgtgatg acaaaatgta ggaaagctgt ttcttcataa
...
128101 ggaaatgcga cccccacgct aatgtacagc ttctttagat tg

String regexp = "[ ]*[0-9]+([actg ]*).*";

Pattern pattern = Pattern.compile(regexp);
In in = new In(filename);
while (!in.isEmpty())
{
String line = in.readLine();
Matcher matcher = pattern.matcher(line);
if (matcher.find())
{
String s = matcher.group(1).replaceAll(" ", "");
// Do something with s.
}
} replace this RE with this string
the part of the match delimited
by the first group of parentheses 54
Algorithmic complexity attacks

Warning. Typical implementations do not guarantee performance!

grep, Java, Perl

java Validate "(a|aa)*b" aaaaaaaaaaaaaaaaaaaaaaaaaaaaaac 1.6 seconds

java Validate "(a|aa)*b" aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaac 3.7 seconds
java Validate "(a|aa)*b" aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaac 9.7 seconds
java Validate "(a|aa)*b" aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaac 23.2 seconds
java Validate "(a|aa)*b" aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaac 62.2 seconds
java Validate "(a|aa)*b" aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaac 161.6 seconds

SpamAssassin regular expression.

java RE "[a-z]+@[a-z]+([a-z\.]+\.)+[a-z]+" spammer@x......................

• Takes exponential time.

• Spammer can use a pathological email address to DOS a mail server.

55
Not-so-regular expressions

Back-references.
• \1 notation matches sub-expression that was matched earlier.
• Supported by typical RE implementations.

java Harvester "\b(.+)\1\b" dictionary.txt

beriberi
couscous word boundary

Some non-regular languages.

• set of strings of the form ww for some string w: beriberi.
• set of bitstrings with an equal number of 0s and 1s: 01110100.
• set of Watson-Crick complemented palindromes: atttcggaaat.

Remark. Pattern matching with back-references is intractable.

56
Context

Abstract machines, languages, and nondeterminism.

• basis of the theory of computation
• intensively studied since the 1930s
• basis of programming languages

Compiler. A program that translates a program to machine code.

• KMP string
DFA.
• grep RE
NFA.
• javac Java language
Java byte code.

KMP grep Java

pattern string RE program

parser unnecessary check if legal check if legal

compiler output DFA NFA byte code

simulator DFA simulator NFA simulator JVM

57
Summary of pattern-matching algorithms

Programmer:
• Implement exact pattern matching by DFA simulation (KMP).
• REs are a powerful pattern matching tool.
• Implement RE pattern matching by NFA simulation (grep).

Theoretician:
• RE is a compact description of a set of strings.
• NFA is an abstract machine equivalent in power to RE.
• DFAs and REs have limitations.

You: Practical application of core CS principles.

Example of essential paradigm in computer science.

• Build intermediate abstractions.
• Pick the right ones!
• Solve important practical problems.

58