Matchmaking with regular expressions

Use the power of regular expressions to ease text parsing and processing

By Benedict Chng, JavaWorld.com, 07/13/01

If you've programmed in Perl or any other language with built-in regular-expression capabilities,
then you probably know how much easier regular expressions make text processing and pattern
matching. If you're unfamiliar with the term, a regular expression is simply a string of characters
that defines a pattern used to search for a matching string.

Many languages, including Perl, PHP, Python, JavaScript, and JScript, now support regular
expressions for text processing, and some text editors use regular expressions for powerful
search-and-replace functionality. What about Java? At the time of this writing, a Java
Specification Request that includes a regular expression library for text processing has been
approved; you can expect to see it in a future version of the JDK.

But what if you need a regular expression library now? Luckily, you can download the open
source Jakarta ORO library from Apache.org. In this article, I'll first give you a short primer on
regular expressions, and then I'll show you how to use regular expressions with the open source
Jakarta-ORO API.

Regular expressions 101

Let's start simple. Suppose you want to search for a string with the word "cat" in it; your regular
expression would simply be "cat". If your search is case-insensitive, the words "catalog",
"Catherine", or "sophisticated" would also match:

Regular expression: cat

Matches: cat, catalog, Catherine, sophisticated

The period notation

Imagine you are playing Scrabble and need a three-letter word starting with the letter "t" and
ending with the letter "n". Imagine also that you have an English dictionary and will search
through its entire contents for a match using a regular expression. To form such a regular
expression, you would use a wildcard notation -- the period (.) character. The regular expression
would then be "t.n" and would match "tan", "Ten", "tin", and "ton"; it would also match "t#n",
"tpn", and even "t n", as well as many other nonsensical words. This is because the period
character matches everything, including the space, the tab character, and even line breaks:

Regular expression: t.n

Matches: tan, Ten, tin, ton, t n, t#n, tpn, etc.
The bracket notation

To solve the problem of the period's indiscriminate matches, you can specify characters you
consider meaningful with the bracket ("[]") expression, so that only those characters would
match the regular expression. Thus, "t[aeio]n" would just match "tan", "Ten", "tin", and "ton".
"Toon" would not match because you can only match a single character within the bracket
notation:

Regular expression: t[aeio]n

Matches: tan, Ten, tin, ton

The OR operator

If you want to match "toon" in addition to all the words matched in the previous section, you can
use the "|" notation, which is basically an OR operator. To match "toon", use the regular
expression "t(a|e|i|o|oo)n". You cannot use the bracket notation here because it will only match a
single character. Instead, use parentheses -- "()". You can also use parentheses for groupings
(more on that later):

Regular expression: t(a|e|i|o|oo)n

Matches: tan, Ten, tin, ton, toon

The quantifier notations

Table 1 shows the quantifier notations used to determine how many times a given notation to the
immediate left of the quantifier notation should repeat itself:

Notation Number of Times

* 0 or more times
+ 1 or more times
? 0 or 1 time
{n} Exactly n number of times
{n,m} n to m number of times
Table 1. Quantifier notations

Let's say you want to search for a social security number in a text file. The format for US social
security numbers is 999-99-9999. The regular expression you would use to match this is shown
in Figure 1. In regular expressions, the hyphen ("-") notation has special meaning; it indicates a
range that would match any number from 0 to 9. As a result, you must escape the "-" character
with a forward slash ("\") when matching the literal hyphens in a social security number.

Figure 1. Matches: All social security numbers of the form 123-12-1234

If, in your search, you wish to make the hyphen optional -- if, say, you consider both 999-99-
9999 and 999999999 acceptable formats -- you can use the "?" quantifier notation. Figure 2
shows that regular expression:

Figure 2. Matches: All social security numbers of the forms 123-12-1234 and 123121234

Let's take a look at another example. One format for US car plate numbers consists of four
numeric characters followed by two letters. The regular expression first comprises the numeric
part, "[0-9]{4}", followed by the textual part, "[A-Z]{2}". Figure 3 shows the complete regular
expression:

Figure 3. Matches: Typical US car plate numbers, such as 8836KV

The NOT notation

The "^" notation is also called the NOT notation. If used in brackets, "^" indicates the character
you don't want to match. For example, the expression in Figure 4 matches all words except those
starting with the letter X.
 Figure 4. Matches: All words except those that start with the letter X

The parentheses and space notations

Say you're trying to extract the birth month from a person's birthdate. The typical birthdate is in
the following format: June 26, 1951. The regular expression to match the string would be like the
one in Figure 5:

Figure 5. Matches: All dates with the format of Month DD, YYYY

The new "\s" notation is the space notation and matches all blank spaces, including tabs. If the
string matches perfectly, how do you extract the month field? You simply put parentheses
around the month field, creating a group, and later retrieve the value using the ORO API
(discussed in a following section). The appropriate regular expression is in Figure 6:

Figure 6. Matches: All dates with the format Month DD, YYYY, and extracts Month field
as Group 1
Other miscellaneous notations

To make life easier, some shorthand notations for commonly used regular expressions have been
created, as shown in Table 2:

Notation Equivalent Notation

\d [0-9]
\D [^0-9]
\w [A-Z0-9]
\W [^A-Z0-9]
\s [ \t\n\r\f]
\S [^ \t\n\r\f]
Table 2. Commonly used notations

To illustrate, we can use "\d" for all instances of "[0-9]" we used before, as was the case with our
social security number expressions. The revised regular expression is in Figure 7:

Figure 7. Matches: All social security numbers of the form 123-12-1234

Jakarta-ORO library

Many open source regular expression libraries are available for Java programmers, and many
support the Perl 5-compatible regular expression syntax. I use the Jakarta-ORO regular
expression library because it is one of the most comprehensive APIs available and is fully
compatible with Perl 5 regular expressions. It is also one of the most optimized APIs around.

The Jakarta-ORO library was formerly known as OROMatcher and has been kindly donated to
the Jakarta Project by Daniel Savarese. You can download the package from a link in the
Resources section below.

The Jakarta-ORO objects

I'll start by briefly describing the objects you need to create and access in order to use this
library, and then I will show how you use the Jakarta-ORO API.
The PatternCompiler object

First, create an instance of the Perl5Compiler class and assign it to the PatternCompiler
interface object. Perl5Compiler is an implementation of the PatternCompiler interface and
lets you compile a regular expression string into a Pattern object used for matching:

PatternCompiler compiler=new Perl5Compiler();

The Pattern object

To compile a regular expression into a Pattern object, call the compile() method of the
compiler object, passing in the regular expression. For example, you can compile the regular
expression "t[aeio]n" like so:
Pattern pattern=null;
try {
pattern=compiler.compile("t[aeio]n");
} catch (MalformedPatternException e) {
e.printStackTrace();
}

By default, the compiler creates a case-sensitive pattern, so that the above setup only matches
"tin", "tan", "ten", and "ton", but not "Tin" or "taN". To create a case-insensitive pattern, you
would call a compiler with an additional mask:

pattern=compiler.compile("t[aeio]n",Perl5Compiler.CASE_INSENSITIVE_MASK);

Once you've created the Pattern object, you can use it for pattern matching with the
PatternMatcher class.

The PatternMatcher object

The PatternMatcher object tests for a match based on the Pattern object and a string. You
instantiate a Perl5Matcher class and assign it to the PatternMatcher interface. The
Perl5Matcher class is an implementation of the PatternMatcher interface and matches
patterns based on the Perl 5 regular expression syntax:

PatternMatcher matcher=new Perl5Matcher();

You can obtain a match using the PatternMatcher object in one of several ways, with the string
to be matched against the regular expression passed in as the first parameter:
  boolean matches(String input, Pattern pattern): Used if the input string and the
regular expression should match exactly; in other words, the regular expression should
totally describe the string input
 boolean matchesPrefix(String input, Pattern pattern): Used if the regular
expression should match the beginning of the input string
 boolean contains(String input, Pattern pattern): Used if the regular
expression should match part of the input string (i.e., should be a substring)

You could also pass in a PatternMatcherInput object instead of a String object to the above
three method calls; if you did so, you could continue matching from the point at which the last
match was found in the string. This is useful when you have many substrings that are likely to be
matched by a given regular expression. The method signatures with the PatternMatcherInput
object instead of String are as follows:

 boolean matches(PatternMatcherInput input, Pattern pattern)

 boolean matchesPrefix(PatternMatcherInput input, Pattern pattern)
 boolean contains(PatternMatcherInput input, Pattern pattern)

Scenarios for using the API

Now let's discuss some example uses of the Jakarta-ORO library.

Log file processing

Your job: analyze a Web server log file and determine how long each user spends on the
Website. An entry from a typical BEA WebLogic log file looks like this:

172.26.155.241 - - [26/Feb/2001:10:56:03 -0500] "GET /IsAlive.htm HTTP/1.0"

200 15

After analyzing this entry, you'll realize that you need to extract two things from the log file: the
IP address and a page's access time. You can use the grouping notation (parentheses) to extract
the IP address field and the timestamp field from the log entry.

Let's first discuss the IP address. It consists of 4 bytes, each with values between 0 and 255; each
byte is separated from the others by a period. Thus, in each individual byte in the IP address, you
have at least one and at most three digits. You can see the regular expression for this field in
Figure 8:
Figure 8. Matches: IP addresses that consist of 4 bytes, each with values between 0 and 255

You need to escape the period character because you literally want it to be there; you do not want
it read in terms of its special meaning in regular expression syntax, which I explained earlier.

The log entry's timestamp part is surrounded by square brackets. You can extract whatever is
within these brackets by first searching for the opening square bracket character ("[") and
extracting whatever is not within the closing square bracket character ("]"), continuing until you
reach the closing square bracket. Figure 9 shows the regular expression for this:

Figure 9. Matches: At least one character until "]" is found

Now you combine these two regular expressions into a single expression with grouping notation
(parentheses) for extraction of your IP address and timestamp. Notice that "\s-\s-\s" is added in
the middle so that matching occurs, although you won't extract that. You can see the complete
regular expression in Figure 10.

Now that you've formulated this regular expression, you can begin Figure 10. Matches: The
writing Java code using the regular expression library. IP address and timestamp
by combining two regular
Using the Jakarta-ORO library expressions. Click on
thumbnail to view full-size
To begin using the Jakarta-ORO library, first create the regular image. (4 KB)
expression string and the sample string to parse:
String logEntry="172.26.155.241 - - [26/Feb/2001:10:56:03 -0500] \"GET
/IsAlive.htm HTTP/1.0\" 200 15 ";
String regexp="([0-9]{1,3}\\.[0-9]{1,3}\\.[0-9]{1,3}\\.[0-
9]{1,3})\\s-\\s-\\s\\[([^\\]]+)\\]";

The regular expression used here is nearly identical to the one found in Figure 10, with only one
difference: in Java, you need to escape every forward slash ("\"). Figure 10 is not in Java, so we
need to escape the forward-slash character so as not to cause a compilation error. Unfortunately,
this process is prone to error and you must do it carefully. You can type in the regular expression
first without escaping the forward slashes, and then visually scan the string from left to right and
replace every occurrence of the "\" character with "\\". To double check, print out the resulting
string to the console.

After initializing the strings, instantiate the PatternCompiler object and create a Pattern
object by using the PatternCompiler to compile the regular expression:

PatternCompiler compiler=new Perl5Compiler();

Pattern pattern=compiler.compile(regexp);

Now, create the PatternMatcher object and call the contain() method in the PatternMatcher
interface to see if you have a match:

PatternMatcher matcher=new Perl5Matcher();

if (matcher.contains(logEntry,pattern)) {
MatchResult result=matcher.getMatch();
System.out.println("IP: "+result.group(1));
System.out.println("Timestamp: "+result.group(2));
}

Next, print out the matched groups using the MatchResult object returned from the
PatternMatcher interface. Since the logEntry string contains the pattern to be matched, you
could expect the following output:

IP: 172.26.155.241
Timestamp: 26/Feb/2001:10:56:03 -0500

HTML processing

Your next task is to churn through your company's HTML pages and perform an analysis of all
of a font tag's attributes. The typical font tag in your HTML looks like this:

<font face="Arial, Serif" size="+2" color="red">

Your program will print out the attributes for every font tag encountered in the following format:

face=Arial, Serif
size=+2
color=red

In this case, I would suggest that you use two regular expressions. The first, shown in Figure 11,
extracts "face="Arial, Serif" size="+2" color="red" from the font tag:

Figure 11. Matches: The all-attribute part of the font tag

The second regular expression, shown in Figure 12, breaks down each individual attribute into a
name-value pair:

Figure 12. Matches: Each individual attribute, broken down into a name-value pair

Figure 12 breaks into:

font Arial, Serif

size +2
color red

Let's now discuss the code to achieve this. First, create the two regular expression strings and
compile them into a Pattern object using the Perl5Compiler. Use the
Perl5Compiler.CASE_INSENSITIVE_MASK option here when compiling the regular expression
for a case-insensitive match.
Next, create a Perl5Matcher object to perform matching:

String regexpForFontTag="<\\s*font\\s+([^>]*)\\s*>";
String regexpForFontAttrib="([a-z]+)\\s*=\\s*\"([^\"]+)\"";
PatternCompiler compiler=new Perl5Compiler();
Pattern
patternForFontTag=compiler.compile(regexpForFontTag,Perl5Compiler.CASE_INSENS
ITIVE_MASK);
Pattern
patternForFontAttrib=compiler.compile(regexpForFontAttrib,Perl5Compiler.CASE_
INSENSITIVE_MASK);
PatternMatcher matcher=new Perl5Matcher();

Assume you have a variable called html of type String that represents a line in the HTML file.
If the content of the html string contains the font tag, the matcher will return true, and you'll use
the MatchResult object returned from the matcher object to get your first group, which includes
all of your font attributes:

if (matcher.contains(html,patternForFontTag)) {
MatchResult result=matcher.getMatch();
String attribs=result.group(1);
PatternMatcherInput input=new PatternMatcherInput(attribs);
while (matcher.contains(input,patternForFontAttrib)) {
result=matcher.getMatch();
System.out.println(result.group(1)+": "+result.group(2));
}
}

Next, create a PatternMatcherInput object. As previously mentioned, this object lets you
continue matching from where the last match was found in the string; thus, it's perfect for
extracting the font tag's name-value pair. Create a PatternMatcherInput object by passing in
the string to be matched. Then, use the matcher instance to extract each font attribute as it is
encountered. This is done by repeatedly calling the contains() method of the PatternMatcher
object with the PatternMatcherInput object instead of a string. Every iteration through the
PatternMatcherInput object will advance a pointer within it, so the next test will start where
the previous one left off.

The output of the example is as follows:

face: Arial, Serif

size: +1
color: red

More HTML processing

Let's continue with another HTML example. This time, imagine that your Web server has moved
from widgets.acme.com to newserver.acme.com. You'll need to change the links on some of
your Webpages from:

<a href="http://widgets.acme.com/interface.html#How_To_Buy">
<a href="http://widgets.acme.com/interface.html#How_To_Sell">
etc.

to

<a href="http://newserver.acme.com/interface.html#How_To_Buy">
<a href="http://newserver.acme.com/interface.html#How_To_Sell">
etc.

The regular expression to perform the search

is shown in Figure 13.

Figure 13. Matches: The link

If this regular expression is found, you can "http://widgets.acme.com/interface.html#(any
make your substitution for the link in Figure anchor). Click on thumbnail to view full-size
13 with the following expression: image. (30 KB)

<a href="http://newserver.acme.com/interface.html#">

Notice that you use after the # character. Perl regular expression syntax uses , , and so forth to
represent groups that have been matched and extracted. The expression shown in Figure 13
appends whatever text has been matched and extracted as Group 1 to the link.

Now, back to Java. As usual, you must create your testing strings, the necessary object for
compiling the regular expression into a Pattern object, and a PatternMatcher object:

String link="<a
href=\"http://widgets.acme.com/interface.html#How_To_Trade\">";
String
regexpForLink="<\\s*a\\s+href\\s*=\\s*\"http://widgets.acme.com/interface.htm
l#([^\"]+)\">";
PatternCompiler compiler=new Perl5Compiler();
Pattern
patternForLink=compiler.compile(regexpForLink,Perl5Compiler.CASE_INSENSITIVE_
MASK);
PatternMatcher matcher=new Perl5Matcher();

Next, use the static method substitute() from the Util class in the com.oroinc.text.regex
package for performing a substitution, and print out the resulting string:
 String result=Util.substitute(matcher,
patternForLink,
new Perl5Substitution(
"<a
href=\"http://newserver.acme.com/interface.html#\">"),
link,
Util.SUBSTITUTE_ALL);
System.out.println(result);

The syntax of the Util.substitute() method is as follows:

public static String substitute(PatternMatcher matcher,

Pattern pattern,
Substitution sub,
String input,
int numSubs)

The first two parameters for this call are the PatternMatcher and Pattern objects created
earlier. The input for the third parameter is a Substitution object that determines how the
substitution is to be performed. In this case, use the Perl5Substitution object, which lets you
use a Perl 5-style substitution. The fourth parameter is the actual string on which you wish to
perform the substitution, and the last parameter lets you specify whether you wish to substitute
on every occurrence of the pattern found (Util.SUBSTITUTE_ALL) or only substitute a specified
number of times.