regex : Java Glossary

* 0-9 A B C D E F G H I J K L M N O P Q R S T U V W X Y Z (all)

regular expression: a system of pattern masks to describe strings to search for, sort of a more generalised wildcarding. You can use them to scan text to extract data embedded in boilerplate. It is usually pronounced reg (with a hard g as is regular), ex (as in expression). Jan Goyvaerts, the author of RegexBuddy, pronounces it reejecs. You can use them to replace boilerplate patterns.

Regexes are notoriously difficult to proofread and debug. You must test them in isolation (unit tests) with every pathological data string and every corner case you can think of.

menu
Introduction	Negative Regex
Other Regex Engines	Excluding Characters
Quoting, why you need \\\\	Matching vs Finding vs LookingAt
Recipes for Quoting	Start and End
Regex Variations Table	Splitting
Multiple Characters	Replacing
Awkward Characters	Tips
Terminology	String
Pattern Flags	Books
Named Fragments	Learning More
Examples	RegexBuddy
Finding Quoted Strings	Links

Introduction

Java version 1.4 introduces the java.util.regex package. If they don’t work, use Wassup to check out the version of Java you are using. You may be inadvertently using an old one. Perl-like Regex expressions are compiled into a Pattern (parsed into an internal state machine format, not byte code). You don’t use a constructor to create a Pattern; you use the static method Pattern. compile( String). Then you create a Matcher object with Pattern. matcher( String) feeding it the String you wonder if matches the pattern. Finally, you call Matcher. matches to see if the xfString fits the pattern. There are many other things you can do, for example, to find multiple matches in your String.

Regex cannot do tasks like look for balanced ( ) or deal with a simple precedence grammar. For that you need a parser. Regexes are also very awkward if your fields are not in some standard order. They drive you nuts analysing HTML (Hypertext Markup Language) form parameters for example where the parms can come in any order. They are great when data come is some standard order, with some missing with alternate forms and variable separators.

Regexes will drive you insane like no other kind of computer programming will. You can stare at them for hours and have no clue why they fail to match. If you change them the tiniest bit, they will refuse to work. The problem is they are black boxes. You can’t watch them work to figure out why and where they are failing. Failures are often subtle reluctant/greedy issues, or a failure in a totally different part of the regex that you presumed. Escaping requires great precision since there are two escaping/quoting mechanisms interacting, one for regex and one for Java String literals. If you are having trouble, write a one-shot Java program or write a simple parser.

Other Regex Engines

Daniel Savarese has written a second Regex package based on Perl regexes. Look at the Apache Jakarta Regexp project. IBM Alphaworks has one. Search for regex. Jakarta-ORO (née OROMatcher), lets you add regex ability to your own Java programs. Funduc Search and Replace is a utility for doing global search and replace on files using regular expressions. The Quoter Amanuensis helps you compose regex expressions for Funduc Search and Replace. SlickEdit® is a text editor that has supports several kinds of regular expressions for global search and replace. Forté Agent newsreader has a regular expression scheme for describing junk filters. However, it is completely unlike Java regex. It is more like Google search expressions.

Quoting

Reserved characters, aka meta characters are command characters that have special meaning in regexes must be quoted when you mean them literally, as just characters. This does not mean you must enclose them in quotation marks, but rather you must specially mark them as meant literally by preceding them with a \. e.g. \- \+ \?. If you are unsure, quote. It won’t hurt to quote punctuation that does not need it. However, Don’t quote : in Vslick since \:… has special meaning.

Unfortunately, the regex people used the same quoting character \ as the designers of Java did for String literals. In a non-regex Java String literal, every literal \ must be doubled. In a regex every literal \ must be doubled. So when you express a regex as a Java String literal, every literal \ must be quadrupled! and written as \\\\.

When you compose a regex String on the fly, character by character, then Java String literal quoting is no longer at play. There you merely need double each \. Be especially careful with File.file SeparatorChar in composed on the fly regexes. If it is \ it must be doubled.

Java 1.4.1+ also offers \Q… \E quoting long passages without having to quote command characters individually. You still have to quote for String literals though.

The quoter amanuensis will let you compose your literal regex strings then convert them to deal with both regex and Java \\ quoting.

In Java version 1.5 or later, Pattern.quote( String ) will do the same thing the quoter amanuensis does to a String to give you the equivalent regex, properly quoted to match it literally. It just mindlessly sandwiches the string in \Q … \E, whether it needs it or not.

Again, it won’t hurt to quote punctuation that doesn’t need it. Note that " and ' don’t need regex quoting, though they need Java quoting.

Recipes for Quoting Awkward Characters in Java Regexes in Java source code String Literals

How to write various awkward characters literally in Java String literals
How to Write Awkward Characters Literally in Java Regex String Literals
Character name	Character	Java literal	Regex	Java literal + Regex
left bracket, acting as a regex command character	[	[	[	[
left bracket, reserved regex command character metacharacter acting as a literal [	[	[	\[	\\[
A literal newline character	???	\n	\n	\\n
A literal carriage return character	???	\r	\r	\\r
A literal double quote character, magic to Java, nothing special to regex.	"	\"	"	\"
A literal single quote character, magic to Java, nothing special to regex.	'	\'	'	\'
A literal backslash character, magic to both Java and regex.	\	\\	\\	\\\\

Regex Variations Table

I use three different regex engines many times a day. I have a heck of a time remembering which commands work with which one. So I composed this table. Lucky I don’t need Perl too.

Operators for three regex schemes: Java 1.4+, SlickEdit and Funduc Search and Replace
Regex Variations Table
Use	Java	SlickEdit® Unix	Funduc SR	Function
Use	Java version 1.4 or later	SlickEdit Unix	Funduc SR	Function
search reserved chars (quoted with \)	search metachars outside character class [] $()+.?[\]^{\|} Prune this string to get just the chars you want: !"#\$%&'\\+,\-\./0…9:;<=>\?@A…Z\[\\\]\^_`a…z\{\\|\} search metachars inside character class [] [-^[\]] but not dot. - does not need to be quoted if it is the first character after [. Prune this string to get just the chars you want for a character class: [ !"#$%&'()*+,\-./0-9:;<=>@A-Z\[\\\]\^_`a-z{\|}]	+.-?[\]{\|} Prune this string to get just the chars you want: !"#$%&'()\\+,\-\./0…9:;<=>\?@A…Z\[\\\]^_`a…z\{\\|\}	!$()+-?[\]^\| Prune this string to get just the chars you want: \!"#\$%&'\\+,\-./0…9:;<=>\?@A…Z\[\\\]\^_`a…z{\\|}	Reserved metacharacters in search strings must be \-quoted if used as data chars, e. g. \+ \* \\|. If in doubt, quote. It won’t hurt.
replace reserved chars (quoted with \)	\$	\	% < > \	Reserved metacharacters in replace strings must be \-quoted if used as data chars, e. g. \% \\ \< \> If in doubt, quote. It won’t hurt. In Java, you can abbreviate [a-z\.] as [a-z.] since . is clearly a character not a command inside [].
any	.	.	.	Matches anything. In Java . sometimes matches Cr and Lf and sometimes not.
0+	greedy: * reluctant: *?	*	*	Zero or More of the preceding thing. .* matches anything. .* is nearly always useless. You normally want .? so that the tail end of your regex will have effect. In Funduc, the comes before the thing repeated, e.g. [] to match anything even over multiple lines. In Java and SlickEdit, the comes after, e.g. [a-z]. Normally you want .?, the reluctant form instead of .* for wildcard matching. As a rule of thumb, if your regex is matching too long a string, try replacing a greedy quantifier with a reluctant one. The documentation mislead me. It made it sound as if reluctant would only every match a single character — pretty lame, but that is not so. It just finds the first match to your complete regex.
1+	greedy: + reluctant: +? possessive: ++	+	+	One or More of the preceding thing. In Funduc, the + comes before the thing repeated, e.g. +[0-9\,\.\+\-] to crudely match a number. In Java and SlickEdit, the + comes after, e.g. [0-9\,\.\+\-]+.
1	{1}	{1}	default	Exactly One of the preceding things, similarly for any {n}. Here is a cute trick to use this Java feature to count characters, inserting a dash between pairs of characters: // insert a dash between chars String cute = "AA54BG4G3G".replaceAll( "(\\w{2})(?!$)", "$1:" ); // cute is "AA-54-BG-4G-3G"
0 or 1	greedy: ? reluctant: ??	?		Zero or One of the preceding thing. e.g (abc)? will match "" or abc
group	capturing: ( — ) non-capturing: (?: — )	( — )	( — )	Delimits a group of characters or patterns. The characters matching the group will show up when you call group(i). However, they won’t if you make the group non-capturing. Java group( 0 ) gets you the entire pattern, including text outside (). group( 1 ) gets you the text inside the first (). groupCount() gets you the number of captured fields, not including the whole pattern. group(1) can sometimes return null, even when groupCount() returns 1. Funduc () works only on expressions. xref?!(=) finds the letters xref followed by anything but =. Normally you leave the () for replacement off, e.g. +[a-z] not (+a-z).
not char	^	~	?!()	Not character operator, e.g. In Java, [^abc] means anything but a, b or c. In other contexts ^ means start of line. In Vslick [~abc] means the same. In Funduc () works only on expressions. xref?!(=) finds the letters xref followed by anything but =. Normally you leave the () for replacement off, e.g. +[a-z] not (+a-z). In Java you can say [a-z&&[^m-p]] to get a through z, except m through p.
not exp	(?!X)	~	!	Not expression operator. In Java anything but X, via zero width negative lookahead. After the non-match, you continue where you left off, not at the end of the non-matching string. In Java, you might search for a word beginning with l but not a lion like this: "(?!lion)l[a-z]+ ". (?! looks ahead and aborts the match if it sees the undesirable pattern. In Funduc xref?!(=) finds the letters xref followed by anything but =. In Funduc you cannot use ! inside […] range operators. If you wanted the printable ASCII (American Standard Code for Information Interchange) chars except < and > for example, you would have to code it terms of the chars you wanted like this: [!-;=?-~].
or	\|	\|	\|	infix or Operator, (cat\|dog) matches cat or dog. Like any () group, the set gets its own dedicated group(i) slot. Funduc \| is quite limited since the or expressions must be simple strings. They may not contain operators. For example, in Funduc, <(td\|li)[a-z =]> is legit, but (<td[a-z =]>)\|(<li*[a-z ="]>) is not.
any	.	.	?	any char but newline. To make newline \n also match dot, in Java, embed (?s) early in the string. (?s) does not match anything, it just switches mode. in Java, you can also turn the scan-over-line-endings mode on with a Pattern.compile("xxx", Pattern.DOTALL) to control whether \n is considered any character. You can turn it off again with (?-s). Use plain . not [.] because inside square brackets dot just means a literal period, not any-character.
nl	\r\n	\n	\r\n	newline, given for Windows.
sol	^	^	^	Start of Line. In other contexts means not. See notes on $.
eol	$	$	$	End of Line. For Windows, matches a pair of characters \r\n. For Linux matches \n. For Mac matches \r. For Java, $ means end of string not end of line, unless you turn on multiline mode by embedding (?m) first. You can turn it off again with (?-m). You can also turn it on with Pattern. compile( xxx, Pattern.MULTILINE ). This is subtle and will drive you nuts. ^ and $ do not match any actual character. In multiline (?m) mode, they match an empty string right after or before a line terminator. Most of the time use (?s) to turn on multiline mode and allow \s*? to swallow newlines as if there were spaces.
sof			^^	Start of File
eof			$$	End of File
range	[]	[]	[]	Range Operator, list of chars,[ab] means match a or b. [a-z] matches any character in range a through z. [0-9] is a digit. [a-z] is lower case. [A-Z] is upper case. [ -_] (space dash underscore) is any printable ASCII char. In Funduc, you don’t need parenthesis around [a-z] in the search string. Keep strings of selection characters inside [] in alphabetical order. It will make proofreading easier and comparing regexes easier, e. g. [ a-z0-9\"%&'\$\$\\-./:;\\?=_] The quoter amanuensis will compute the span of any string, a canonical regex expression that will hop over the string. It will create tidy complex range expressions sorted in alphabetical order.
negation	[^, ]	[~, ]	n/a	any character except a comma or space
intersection	[a-z&&[^bc]]	n/a	n/a	a through z, except for b and c
sub	()	()	()	Sub-Expression. In Funduc, you don’t need parenthesis around *[a-z] in the search string. Further, you must not use them!
col			+n	Column Operator
replace	$1	\1 \2 etc.	%1 %2 etc. %1< (to lower case) %1> (to upper case) can also do math.	back reference to tagged expression #1, in () for replace. E.g. in SlickEdit to replace all occurrences of <span class=jmethod> used before an upper case name, converting them to <span class=jclass>.. Search string : <span class=jmethod>([A-Z]) Replace string : <span class=jclass>\1 Remember to turn exact case matching on for these to work. In Funduc, you don’t need parenthesis around [a-z] in the search string. [a-z]* in Funduc will put the first character in %1 and the rest of the match in %2, very confusing. Java regex has only very primitive replace ability. Every match must be replaced by the same string, with $1 $2 etc to bring over matched pieces from the original String. However, in Java you can also use \1 in the match string to insist on a match for some expression found earlier in the string, i.e. a repeated pattern, most commonly used to make sure single or double quotes balance. Use Matcher. replaceAll. IntelliJ editor uses standard Java regex, including $1 to mark a replacement parameter.
replace example	search: \(([a-zA-z\(\"]) replace: \( $1	search: \(([a-zA-z\(\"]) replace: \( \1	search: \([a-zA-z\(\"] replace: \( %1	Replace all (x with ( x but only if x is alphabetic or ( or "
single white space	\s = [ \t\n\x0B\f\r]	[ \t\n]	[ \t\r\n]	single white space
white spaces	\s+	\:b	+[ \t\r\n]	one or more white spaces, [ \t\n\x0B\f\r] Watch out, matches line end as well!
poss white spaces	\s*	[ \t\r\n]*	*[ \t\r\n]	zero or more white spaces, [ \t\n\x0B\f\r]* Watch out, matches line end as well!
black	\S	[^ \t\n]	[! \t\r\n]	single non white space (blank, tab)
blacks	\S+	[^ \n\t]+	+[! \r\n\t]	one or more non-white spaces
word	(\p{Alpha}+)	\:w	+[A-Za-z]	alphabetic word (string of A-Z a-z )
number	([0-9\,\.\+\-]+)	([0-9\,\.\+\-]+)	+[0-9\,\.\+\-]	number (string of digits, commas, decimal points and signs)
quoted	$\\\\|([ A-Za-z\'\[\]\+\=\!\@\# \$\%\^\&\\($ \<\>\:\;\?\\|\\]))\"	\:q	$\\\([ A-Za-z\'\[\]\+\=\!\@\# \$\%\^\&\\($ \<\>\:\;\?\\|\\]*))\"	quoted String. It easier just to quote all punctuation sometimes. It is easier to proofread. Don’t quote : in Vslick since \:… has special meaning.
special	The following work both inside and outside []. \d = digit = [0-9] \D = non digit = [^0-9] \s = single whitespace char = [ \t\n\x0B\f\r] \S = not whitespace = [^\s] \w = single alphanumeric char = [a-zA-Z_0-9] \W = not alphanumeric = [^\w] The following work both inside and outside []. The following are all case-sensitive. You must specify \p{Lower} not \P{lower} etc. \p{Lower} overrides CASE_INSENSITIVE. Even then it will not match upper case letters. \p{Lower} = [a-z] \p{Upper} = [A-Z] \p{ASCII } = [\x00-\x7F] \p{Alpha} = [A-za-z] \p{Digit} = [0-9] [\p{Digit}\.]+ = [0-9\.]+ decimal number \p{Alnum} = [[A-Za-z0-9] \p{Punct} = [!"#\$%&'\*\+,\-\./:;<=>\?@\[\\\]\^_`\{\\|\}~] \p{Graph} = [\p{Alnum}\p{Punct}] \p{Print} = [\p{Graph}\x20] \p{Blank} = [ \t] c.f. \s \p{Cntrl} = [\x00-\x1F\x7F] \p{XDigit} = [0-9a-fA-F] \p{Space} = [ \t\n\x0B\f\r] c.f. \s \p{IsAlphabetic} letter, possibly accented, possibly in some non-Latin alphabet. \p{javaLowerCase} lower case letter, possibly accented, possibly in some non-Latin alphabet. \p{javaUpperCase} upper case letter, possibly accented, possibly in some non-Latin alphabet. \p{Lu} = upper case letter \p{InGreek} = Greek letter \p{Sc} = a currency symbol [\p{L}&&[^\p{Lu}]] = anything but an upper case letter. (?i) = turn on case-insensitive mode (?-i) = turn on case-sensitive mode	\:a alphanumeric char = [A-Za-z0-9] \d0-\d27 ASCII codes 0…27 specified as 8-bit decimal. \:b blanks = ([ \t]+) \:c alpha char = [A-Za-z] \:d digit = [0-9] \:f filename part \:h hex = ([0-9A-Fa-f]+) \:i int = ([0-9]+) \:n float \:p path \:q quoted string \:v C language variable name = ([A-Za-z_$][A-Za-z0-9_$]*) \:w word = ([A-Za-z]+)		predefined match strings, e.g. \:w = ([A-Za-z]+) matches a word. Those are braces in \p{Alnum} not parentheses. It can be hard to tell in some typefaces. The strings are case-sensitive and when used in Java source code such strings must be coded as \\p{Alnum}. \d \D \s \S \w \W \p{Lower} etc. will also work inside […]. \p{Lower] is not quite identical to [a-z] If you have CASE_INSENSITIVE, \p{Lower} will only match lower case letters while [a-z] will also match upper case ones.
capture	X{n} X{n,m} capturing ( — ) non-capturing (?: — ) greedy: + reluctant: +? possessive: ++		%%srpath%% %%srfile%% %%srfiledate%% %%srfiletime%% %%srfilesize%% %%srdate%% %%srtime%% %%envvar=fruit%%	X{n,m} means X appears exactly n to m times. X{n} means X appears exactly n times. X{n,} means X appears at least n times

This table only covers the most common magic characters. See the documentation for each Regex package for details.

Multiple Characters

Multiples in Java Regex
Multiples in Java Regex
[A-Z]	A single upper-case letter
[A-Z]*	zero or more upper-case letters
[A-Z]+	one or more upper-case letters
[A-Z][A-Z]	Exactly two upper-case letters
[A-Z]{2}	Exactly two upper-case letters (same as above)
[A-Z]{2,}	Two or more upper-case letters
[A-Z]{2,10}	Between 2 and 10 (inclusive) upper-case letters
[a-zA-Z]	A single letter, upper- or lower-case

Awkward Characters

Here is how to represent various awkward characters. They represent the combined quoting needs for Java String literals and Regex Patterns.

How To Encode Awkward Characters
How	Desired
\\\\	\ The literal backslash character. You must double the \ twice since \ is the quoting character in both Java and Regex literals.
\\xhh	The character with hexadecimal value 0xhh, e.g. \\xff. Only works with two hex digits!
\uhhhh	The character with hexadecimal value 0xhhhh, e.g. \u20ac. Must always have exactly four hex digits. Don’t use for control characters e.g. 0..ff since \u expansion happens prior to compilation. In other words \u000a will start a new line in your program. Note there is only one lead \.
\\t	The tab character \u0009
\\n	The newline (line feed) character \u000a
\\r	The carriage-return character \u000d
\\x0c	The form-feed character \u000c.
\\a	The alert (bell) character \u0007 \a itself is illegal in Java Strings
\\e	The escape character \u001b
\\cx	control characters, e.g. \\cq for ctrl-q.
\\-	Literal -, not a regex range operator.
\\+	Literal +, not a regex operator.
\\*	Literal *, not a regex operator.
\\?	Literal ?, not a regex operator.
\\(	Literal (, not a regex expression bracketer.
\\)	Literal ), not a regex expression bracketer.
\\[	Literal [, not a regex expression bracketer.
\\]	Literal ], not a regex expression bracketer.
\\{	Literal {, not a regex expression bracketer.
\\}	Literal }, not a regex expression bracketer.
\\\|	Literal \|, not a regex operator.
\\$	Literal $, not a regex end of line.
\\^	Literal ^, not regex operator.
\\<	Literal <, not regex operator.
\\=	Literal =, not regex operator.

The Quoter utility will quote regexes for you, for Java, SlickEdit and Funduc. It will also work out regex patterns needed to span a given string of characters.

You can also use a sandwich to quote characters. \Q… \E

You can use the JetBrains IntelliJ Idea IDE which highlights characters that are improperly quoted or improperly nested.

Terminology

Pattern.CASE_INSENSITIVE is a flag you can feed to Pattern.compile to do case-insensitive searches. This is much easier than trying to do them directly in the regex strings.

Java 1.4.1+ regexes have assertions, extra conditions placed on the match. Colourful regex terminology includes:

capturing means characters are accumulated for Matcher.group.
greedy means find the longest possible match (consuming the most text).
lookahead means it looks ahead for X.
lookbehind means it looks behind for X.
negative means the match fails if it finds X.
positive means the match succeeds if it finds X.
possessive means greedily match as much as you can and do not back off, even when doing so would allow the overall match to succeed. For example, if you applied the greedy regex .+ to abc you get abc.
reluctant means find the shortest/first possible match, If you applied the reluctant regex .+? to abc, you would just get a.
zero-width means it doesn’t actually capture any characters, or prevent them from being used in further matching.

The easiest way to understand these terms is to experiment with the various regex operators on simple strings. You can make yourself a test program that reads strings from the console. That way, at least you can avoid having to deal with Java \ string quoting. You only need concern yourself with regex \ quoting. You can also use the Quoter Amanuensis to first apply regex quoting then Java string quoting and let you paste the result into your program.

Pattern Flags

You can specify flags to Pattern.compile( String regex, int flags) with:

By default regexes are case-sensitive.

Regex Pattern Flags
Possible Pattern flags
Flag	Alternate Embedded Code	Notes
CASE_INSENSITIVE	(?i)	Makes case does not matter on matching, s matches S. Even if you use it, \p{Lower} will not match upper case letters. [a] will match A though.
MULTILINE	(?m)	Make ^ and $ match embedded newlines. You might expect embedded newlines to match by default, but they don’t. For Java, $ means end of string not end of line, unless you turn on multiline mode by embedding (?m) first. You can turn it off again with (?-m). You can also turn it on with Pattern. compile( xxx, Pattern.MULTILINE ).
DOTALL	(?s)	Makes . match any character, including a line terminator. By default . does not match line terminators.
UNICODE_CASE	(?u)	Used in conjunction with CASE_INSENSITIVE to use the elaborate code-folding schemes to compare Unicode upper and lower case. By default, the presumption is all characters being matched are US-ASCII.
CANON_EQ		Treats canonically accented characters done with single char or with a pair as equivalent e.g. å : the pair a\u030A is the treated the same as the single character \u00E5.
UNIX_LINES	(?d)	\n is recognised in ^ and $ processing.
LITERAL	\Q… \E	Treat all characters as ordinary literals rather than as commands. You don’t then quote with \.
COMMENTS	?x	Makes whitespace ignored and allows embedded comments starting with # that are ignored until the end of a line.

Named Fragments

Naming fragments of regexes as String constants can make your code easier to proofread.

I name the fragment Strings beginning with A_ so the Rearranger or other code tidier will put them before my regex Patterns and will group the fragment Strings together.

Note how much easier the regex patterns are to proofread.
Note that if you have got a fragment pattern wrong, you need fix it in only one place.
Note that you can reuse your regex fragments from a previous program. You don’t have to work them out from first principles each time.
Note how you can implement the patterns in ever more refined ways without having to adjust all your patterns.
Note how you can debug more simply. Get your fragments debugged first, then your patterns will usually work first time.

Examples

The following examples use the Java conventions. For use on the command line, undouble the \\.

Finding Quoted Strings

There are all kinds of ways to write a regex that will only sometimes find quoted regex strings (characters enclosed in " or '). The sample code below will show you some ways not to do it and also some ways that work.

Here are the results of the program:

As the specifications get more and more complicated, regexes run out of steam. Instead you want to write a parser. The code will be special purpose, faster and easier to modify.

Negative Regex

(?!X) is the exclusion or negative regex operator, anything but X, via zero width negative lookahead. After the non-match, you continue where you left off, not at the end of the non-matching string. In Java, you might search for a word beginning with l but not a lion like this: (?!lion)l[a-z]+ . (?! looks ahead and aborts the match if it sees the undesirable pattern. I have not completely understood this operator. Sometimes exclusions don’t work and I have no idea why. It sometimes easier to let the regex collect too much stuff and then toss what you don’t need programmatically in Java.

Excluding Characters

Sometimes lists of characters in [ ] get so complicated, it would be easier to specify the characters you don’t want rather than the ones you do. Here is how to specify anything but a ". [^"] Here is how to specify anything but the letters wxyz [^w-z].

Matching vs Finding vs LookingAt

Matching means the pattern must match the entire String.
Finding means the pattern must appear somewhere in the String.
LookingAt mean the String must start with the pattern.

Matching

When you want the entire String to match your Pattern,

Finding

When you want to find fragments in your String that match the Pattern, use Matcher.find. If you only want to find the first occurrence of a regex in a String you can use this

Here is how you do a case-insensitive find.

The following example will help you understand how the or | operator works and the effects of using layers of capturing ().

LookingAt

When you want to see if you String starts with your Pattern, use Matcher.lookingAt.

Start and End

There are four Matcher methods you might confuse:

regionStart(): offset of start of entire String you are scanning for patterns.
regionEnd: offset of end of entire String you are scanning for patterns.
start(group): offset of start of particular group just found.
start(): offset of start of pattern just found.
end(group): offset of end of particular group just found.
end(): offset of end of pattern just found.

You probably do not want to use regionStart() and regionEnd(). You want start() and end().

Splitting

Regexes can be used to break phrases into individual words. Here is an example:

Beware, split treats leading, embedded and trailing separators differently. It ignores trailing separators unless you use split ( string, -1 /* limit */ ). It inherited this oddity from Perl. Normally you use trim() before split() to avoid getting empty elements.

Another oddity is when you split an empty String, you don’t get a 0-length array. You get an array with a 0-length String in the [0] position.

Here is an SSCCE (Simple Self Contained Compilable Example) to illustrate these gotchas.

Replacing

Here is how to search for instances of some pattern in a big string and replace them all with some computed modification of the pattern.

Tips

When proofreading, double check any ] not followed by a +.
Use the IntelliJ code inspector. It will warn you of any regex characters you quoted needlessly.
If you want to capture a variable number of fields, write a regex to extract all of them as a single lump then use Regex.split.
Test your regexes in isolation. Feed them pathological cases to make sure they behave as expected. Testing on live data will not test the corner cases.
Be careful with space, a pair of spaces and a newline. Regexes treat them as quite different. You may tend to treat them as equal by eye.
If you are having trouble composing a String to describe what you do want, try instead to compose one that describes what you don’t want and reverse the sense of the match.
When a regex does not work, give just as much attention to the right end as the left. It can be failing on the very last character. So often the problem is not in the complicated part where you expect, but in a trivial part of the regex.
Don’t try to put all your logic in one be-all-end-all DEBE (Does Everything But Eat) regex. Use several simpler ones in succession. For example, if there are four distinct patterns you are looking for, use four regexes rather than one giant complicated one.
Always check the your regex to make sure you did not use an unquoted magic character as if it were an ordinary one. It is so easy to forget that a character is a command that you rarely use.
Regex code often seems to work, but because you left out one letter from a Pattern, it will fail to catch all instances. Manually count instances and make sure all are accounted for. Rather than thinking of which characters to include, look at a Pattern that includes everything and decide letter by letter if you want to include it. In general include characters unless there is a specific reason to exclude them. Don’t exclude them just because they not commonly used.
You will need some code like this if you want to include the separator character in your regexes, since \ has to be quoted in regexes.
Regexes are a sledgehammer for complex pattern matching. For simple tasks you can do the job at least three times faster and more simply with String. substring, String. indexOf, String. lastIndexOf, String. startsWith, String.endsWith, or possibly StringTokenizer or StreamTokenizer. Free to mix regex logic with String method logic.
Regexes are not designed for complex language analysis like parsing XML (extensible Markup Language) or Java source code. Use a parser instead.
Regexes to extract information from HTML will drive you crazy unless you condition the HTML first with some sort of tidy program to use consistent " delimiters, consistent capitalisation and to order all parameters alphabetically and to collapse excess blank space down to one.
You would think a regex to find a number in the range 1 to 20 would be a simple task. You are better off to capture the number then check if it is in range outside the regex.
Compiling a Pattern is a non-trivial, time-consuming operation. If you are not careful, your Pattern will be recompiled on every use. For speed, use this idiom to compile the Pattern only once:
```
// ensuring the Pattern is compiled only once.
private static final Pattern p = Pattern.compile( "[a]*" );
```
Regexes rarely give an error message. They just fail to match anything. Start with just the first few chars of your regex and see what that matches. Then when that works and a few more characters at a time rather than trying to debug the whole thing at once. You can gradually add characters to the right end of your regex, or gradually chop them off until it starts matching.
IntelliJ and Eclipse both have a regex plug-in to help you compose and debug regexes.
Matcher.group( i ) is full of surprises. Print out everything from 0 to n to make sure you are grabbing the right thing. Groups that matched nothing will be null, not "".
Keep your Pattern characters in ASCII order. It makes them easier to proofread.
To search for numbers 1 to 20 try [1-9]|1[0-9]|20

String

The String class borrows some convenience regex methods, such as split, matches, replaceAll and replaceFirst. Normally you would use the more efficient java.util.regex methods such as Matcher. replaceFirst and Matcher.replaceAll where you precompile your Pattern and reuse it. The String versions are for one-shot use where efficiency is not a concern. Note that String. replace does not use regexes.

Books

recommend book⇒Regular Expressions Cookbook

Jan Goyvaerts, Steven Levithan

978-1-4493-1943-4

paperback

publisher

O’Reilly

978-1-4493-2748-4

eBook

published

2012-09-03

B008Y4OP1O

kindle

Shows regexes for C#, Java, JavaScript, Perl, PHP, Python, Ruby and VB.NET

Online bookstores carrying Regular Expressions Cookbook
	abe books anz	abe books.ca
	abe books.de	amazon.ca
	amazon.de	Chapters Indigo
	amazon.es	Chapters Indigo eBooks
	iberlibro.com	abe books.com
	abe books.fr	amazon.com
	amazon.fr	Barnes & Noble
	abe books.it	Nook at Barnes & Noble
	amazon.it	Kobo
	junglee.com	Google play
	abe books.co.uk	O’Reilly Safari
	amazon.co.uk	Powells
	other stores

Greyed out stores probably do not have the item in stock. Try looking for it with a bookfinder.

recommend book⇒Regular Expression Pocket Reference: Regular Expressions for Perl, Ruby, PHP, Python, C, Java and .NET, second edition

Tony Stubblebine

978-0-596-51427-3

paperback

birth

1978-04-30 age:40

978-1-4493-7886-8

eBook

publisher

O’Reilly

B0093SZ4QU

kindle

published

2007-07-25

The Owl Cheat Sheet for regexes. Pocket reference companion to Mastering Regular Expressions which also has an owl on the cover.

Online bookstores carrying Regular Expression Pocket Reference: Regular Expressions for Perl, Ruby, PHP, Python, C, Java and .NET, second edition
	abe books anz	abe books.ca
	abe books.de	amazon.ca
	amazon.de	Chapters Indigo
	amazon.es	Chapters Indigo eBooks
	iberlibro.com	abe books.com
	abe books.fr	amazon.com
	amazon.fr	Barnes & Noble
	abe books.it	Nook at Barnes & Noble
	amazon.it	Kobo
	junglee.com	Google play
	abe books.co.uk	O’Reilly Safari
	amazon.co.uk	Powells
	other stores

Greyed out stores probably do not have the item in stock. Try looking for it with a bookfinder.

recommend book⇒Mastering Regular Expressions, Powerful Techniques for Perl and Other Tools, third edition

Jeffrey E. Friedl andy Oram

978-0-596-52812-6

paperback

birth

1966 age:51

978-1-4493-3253-2

eBook

publisher

O’Reilly

B007I8S1X0

kindle

published

2006-08-08

The Owl Book. Includes scripting languages such as Perl, Tcl, auk and Python. Does not specifically cover Java, though Java regexes were modeled on Perl. More a book for regex experts to hone their skills than a newbie to learn regexes. It is a good place to find regex solutions to standard problems. While it isn’t made up in cookbook style, the examples are usually real-life problems that can be put into practical use.

Online bookstores carrying Mastering Regular Expressions, Powerful Techniques for Perl and Other Tools, third edition
	abe books anz	abe books.ca
	abe books.de	amazon.ca
	amazon.de	Chapters Indigo
	amazon.es	Chapters Indigo eBooks
	iberlibro.com	abe books.com
	abe books.fr	amazon.com
	amazon.fr	Barnes & Noble
	abe books.it	Nook at Barnes & Noble
	amazon.it	Kobo
	junglee.com	Google play
	abe books.co.uk	O’Reilly Safari
	amazon.co.uk	Powells
	other stores

Greyed out stores probably do not have the item in stock. Try looking for it with a bookfinder.

Learning More

Oracle’s Miscellaneous documentation on Regex

Oracle’s Miscellaneous documentation on Greedy, Reluctant Possessive Quantifiers

Oracle’s Javadoc on java.util.regex package : available:

on the web at Oracle.com
in the current JDK 1.8.0_131 on your local Windows J: drive.

Oracle’s Javadoc on Pattern class : Includes reference on meaning of regex command letters. : available:

on the web at Oracle.com
in the current JDK 1.8.0_131 on your local Windows J: drive.

Oracle’s Javadoc on Pattern.quote : available:

on the web at Oracle.com
in the current JDK 1.8.0_131 on your local Windows J: drive.

Oracle’s Javadoc on Matcher class : available:

on the web at Oracle.com
in the current JDK 1.8.0_131 on your local Windows J: drive.

Oracle’s Javadoc on Matcher.replaceAll : available:

on the web at Oracle.com
in the current JDK 1.8.0_131 on your local Windows J: drive.

Oracle’s Javadoc on String.matches : available:

on the web at Oracle.com
in the current JDK 1.8.0_131 on your local Windows J: drive.

Oracle’s Miscellaneous documentation on @Regex annotation to validate regexes

A common bug is to confuse String.replace (non regex replace all) with String.replaceAll (regex replace all) and String. replaceFirst (regex replace just one instance). It is probably too late now for Sun to assign the methods clearer names.

Oracle’s Javadoc on String.replace : available:

on the web at Oracle.com
in the current JDK 1.8.0_131 on your local Windows J: drive.

Oracle’s Javadoc on String.replaceAll : available:

on the web at Oracle.com
in the current JDK 1.8.0_131 on your local Windows J: drive.

Oracle’s Javadoc on String.replaceFirst : available:

on the web at Oracle.com
in the current JDK 1.8.0_131 on your local Windows J: drive.

Oracle’s Javadoc on String.split : available:

on the web at Oracle.com
in the current JDK 1.8.0_131 on your local Windows J: drive.

Slick Edit documentation available from Help | contents ⇒ Search and Replace ⇒ Regular Expressions ⇒ Unix Regular Expressions.

Funduc search and replace documentation is available from Help ⇒ contents ⇒ Regular Expressions | Search Operators.

tcc/TakeCommand documentation is available from help | contents ⇒ wildcards ⇒ advanced wildcards

Apache RegExp
Checker-Framework @Regex: unfortunately requires complex install
Eclipse Regex tester
Expresso Regex tester: for Windows .net style regex
finite state automaton
JavaRegex.com
JFlex
JRegex
KDE utilities
KRegExpEditor: part of KDE
list of regex parsers
literal
parser
PowerGrep
PowerGrep
PY regex tutorial
Quoter Regex Amanuensis
Regex Composer student project
Regex Debugger student project
Regex Proofreader student project
Regex testing tool: applies your regex to a set of test strings
RegexBuddy
regular-expressions.info
Savarese OroMatcher
Scanner
StreamTokenizer
String
StringTokenizer
TCC Regex

standard footer
	This page is posted on the web at:	http://mindprod.com/jgloss/regex.html
	Optional Replicator mirror of mindprod.com on local hard disk J:	J:\mindprod\jgloss\regex.html
	Please read the feedback from other visitors, or send your own feedback about the site. Contact Roedy. Please feel free to link to this page without explicit permission.
	Canadian Mind Products IP:[65.110.21.43] Your face IP:[3.149.234.50]
Feedback	You are visitor number