Matchmaking With Regular Expressions

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 999-999999 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 writing Java code using the regular expression library. Using the Jakarta-ORO library To begin using the Jakarta-ORO library, first create the regular

Figure 10. Matches: The IP address and timestamp by combining two regular expressions. Click on thumbnail to view full-size 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}\\.[09]{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:

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 size color

Arial, Serif +2 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: etc.

to etc.

The regular expression to perform the search is shown in Figure 13.

If this regular expression is found, you can make your substitution for the link in Figure 13 with the following expression:

Figure 13. Matches: The link "http://widgets.acme.com/interface.html#(any anchor). Click on thumbnail to view full-size image. (30 KB)



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=""; 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( ""), 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.