Regular Expressions (RegEx) - Quick Reference
Fundamentals
Match anywhere: By default, a regular expression matches a substring anywhere inside the string to be searched. For example, the regular expression abc matches abc123, 123abc, and 123abcxyz. To require the match to occur only at the beginning or end, use an anchor.
Escaped characters: Most characters like abc123 can be used literally inside a regular expression. However, the characters \.*?+[{|()^$ must be preceded by a backslash to be seen as literal. For example, \. is a literal period and \\ is a literal backslash. Escaping can be avoided by using \Q...\E. For example: \QLiteral Text\E.
Case-sensitive: By default, regular expressions are case-sensitive. This can be changed via the "i" option. For example, the pattern i)abc searches for "abc" without regard to case. See below for other modifiers.
Options (case sensitive)
At the very beginning of a regular expression, specify zero or more of the following options followed by a close-parenthesis. For example, the pattern "im)abc" would search for abc with the case-insensitive and multiline options (the parenthesis may be omitted when there are no options). Although this syntax breaks from tradition, it requires no special delimiters (such as forward-slash), and thus there is no need to escape such delimiters inside the pattern. In addition, performance is improved because the options are easier to parse.
| i | Case-insensitive matching, which treats the letters A through Z as identical to their lowercase counterparts. |
| m | Multiline. Views Haystack as a collection of individual lines (if it contains newlines) rather than as a single continuous line. Specifically, it changes the following: 1) Circumflex (^) matches immediately after all internal newlines -- as well as at the start of Haystack where it always matches (but it does not match after a newline at the very end of Haystack). 2) Dollar-sign ($) matches before any newlines in Haystack (as well as at the very end where it always matches). For example, the pattern "m)^abc$" would not match the Haystack"xyz`r`nabc" unless the "m" option is present. The "D" option is ignored when "m" is present. |
| s | DotAll. This causes a period (.) to match all characters including newlines (normally, it does not match newlines). However, when the newline character is at its default of CRLF (`r`n), two dots are required to match it (not one). Regardless of this option, a negative class such as [^a] always matches newlines. |
| x | Ignores whitespace characters in the pattern except when escaped or inside a character class. The characters `n and `t are among those ignored because by the time they get to PCRE, they are already raw/literal whitespace characters (by contrast, \n and \t are not ignored because they are PCRE escape sequences). The x option also ignores characters between a non-escaped # outside a character class and the next newline character, inclusive. This makes it possible to include comments inside complicated patterns. However, this applies only to data characters; whitespace may never appear within special character sequences such as (?(, which begins a conditional subpattern. |
| A | Forces the pattern to be anchored; that is, it can match only at the start of Haystack. Under most conditions, this is equivalent to explicitly anchoring the pattern by means such as "^". |
| D | Forces dollar-sign ($) to match at the very end of Haystack, even if Haystack's last item is a newline. Without this option, $ instead matches right before the final newline (if there is one). Note: This option is ignored when the "m" option is present. |
| J | Allows duplicate named subpatterns. This can be useful for patterns in which only one of a collection of identically-named subpatterns can match. Note: If more than one instance of a particular name matches something, only the leftmost one is stored. Also, variable names are not case-sensitive. |
| U | Ungreedy. Makes the quantifiers *+?{} consume only those characters absolutely necessary to form a match, leaving the remaining ones available for the next part of the pattern. When the "U" option is not in effect, an individual quantifier can be made non-greedy by following it with a question mark. Conversely, when "U" is in effect, the question mark makes an individual quantifier greedy. |
| X | PCRE_EXTRA. Enables PCRE features that are incompatible with Perl. Currently, the only such feature is that any backslash in a pattern that is followed by a letter that has no special meaning causes the match to fail and ErrorLevel to be set accordingly. This option helps reserve unused backslash sequences for future use. Without this option, a backslash followed by a letter with no special meaning is treated as a literal (e.g. \g and g are both recognized as a literal g). Regardless of this option, non-alphabetic backslash sequences that have no special meaning are always treated as literals (e.g. \/ and / are both recognized as forward-slash). |
| S | Studies the pattern to try improve its performance. This is useful when a particular pattern (especially a complex one) will be executed many times. If PCRE finds a way to improve performance, that discovery is stored alongside the pattern in the cache for use by subsequent executions of the same pattern (subsequent uses of that pattern should also specify the S option because finding a match in the cache requires that the option letters exactly match, including their order). |
| C | Enables the auto-callout mode. See Regular Expression Callouts for more info. |
| `a | Enables recognition of additional newline markers. By default, only `r`n, `n and `r are recognized. With this option enabled, `v/VT/vertical tab/chr(0xB), `f/FF/formfeed/chr(0xC), NEL/next-line/chr(0x85), LS/line separator/chr(0x2028) and PS/paragraph separator/chr(0x2029) are also recognized. The `a, `n and `r options affect the behavior of anchors (^ and $) and the dot/period pattern. `a also puts (*BSR_UNICODE) into effect, which causes \R to match any kind of newline. By default, \R matches `n, `r and `r`n; this behaviour can be restored by combining options as follows: `a)(*BSR_ANYCRLF) |
| `n | Causes a solitary linefeed (`n) to be the only recognized newline marker (see above). |
| `r | Causes a solitary carriage return (`r) to be the only recognized newline marker (see above). |
Note: Spaces and tabs may optionally be used to separate each option from the next.
Commonly Used Symbols and Syntax
| . | By default, a dot matches any single character which is not a newline character or part of a newline (`r`n) sequence, but this can be changed by using the DotAll (s), linefeed (`n), carriage return (`r), or `a options. For example, ab. matches abc and abz and ab_. |
| * | An asterisk matches zero or more of the preceding character, class, or subpattern. For example, a* matches ab and aaab. It also matches at the very beginning of any string that contains no "a" at all. Wildcard: The dot-star pattern .* is one of the most permissive because it matches zero or more occurrences of any character (except newline: `r and `n). For example, abc.*123 matches abcAnything123 as well as abc123. |
| ? | A question mark matches zero or one of the preceding character, class, or subpattern. Think of this as "the preceding item is optional". For example, colou?r matches both color and colour because the "u" is optional. |
| + | A plus sign matches one or more of the preceding character, class, or subpattern. For example a+ matches ab and aaab. But unlike a* and a?, the pattern a+ does not match at the beginning of strings that lack an "a" character. |
| {min,max} | Matches between min and max occurrences of the preceding character, class, or subpattern. For example, a{1,2} matches ab but only the first two a's in aaab. Also, {3} means exactly 3 occurrences, and {3,} means 3 or more occurrences. Note: The specified numbers must be less than 65536, and the first must be less than or equal to the second. |
| [...] | Classes of characters: The square brackets enclose a list or range of characters (or both). For example, [abc] means "any single character that is either a, b or c". Using a dash in between creates a range; for example, [a-z] means "any single character that is between lowercase a and z (inclusive)". Lists and ranges may be combined; for example [a-zA-Z0-9_] means "any single character that is alphanumeric or underscore". A character class may be followed by *, ?, +, or {min,max}. For example, [0-9]+ matches one or more occurrence of any digit; thus it matches xyz123 but not abcxyz. The following POSIX named sets are also supported via the form [[:xxx:]], where xxx is one of the following words: alnum, alpha, ascii (0-127), blank (space or tab), cntrl (control character), digit (0-9), xdigit (hex digit), print, graph (print excluding space), punct, lower, upper, space (whitespace), word (same as \w). Within a character class, characters do not need to be escaped except when they have special meaning inside a class; e.g. [\^a], [a\-b], [a\]], and [\\a]. |
| [^...] | Matches any single character that is not in the class. For example, [^/]* matches zero or more occurrences of any character that is not a forward-slash, such as http://. Similarly, [^0-9xyz] matches any single character that isn't a digit and isn't the letter x, y, or z. |
| \d | Matches any single digit (equivalent to the class [0-9]). Conversely, capital \D means "any non-digit". This and the other two below can also be used inside a class; for example, [\d.-] means "any single digit, period, or minus sign". |
| \s | Matches any single whitespace character, mainly space, tab, and newline (`r and `n). Conversely, capital \S means "any non-whitespace character". |
| \w | Matches any single "word" character, namely alphanumeric or underscore. This is equivalent to [a-zA-Z0-9_]. Conversely, capital \W means "any non-word character". |
| ^ $ |
Circumflex (^) and dollar sign ($) are called anchors because they don't consume any characters; instead, they tie the pattern to the beginning or end of the string being searched. ^ may appear at the beginning of a pattern to require the match to occur at the very beginning of a line. For example, ^abc matches abc123 but not 123abc. $ may appear at the end of a pattern to require the match to occur at the very end of a line. For example, abc$ matches 123abc but not abc123. The two anchors may be combined. For example, ^abc$ matches only abc (i.e. there must be no other characters before or after it). If the text being searched contains multiple lines, the anchors can be made to apply to each line rather than the text as a whole by means of the "m" option. For example, m)^abc$ matches 123`r`nabc`r`n789. But without the "m" option, it wouldn't match. |
| \b | \b means "word boundary", which is like an anchor because it doesn't consume any characters. It requires the current character's status as a word character (\w) to be the opposite of the previous character's. It is typically used to avoid accidentally matching a word that appears inside some other word. For example, \bcat\b doesn't match catfish, but it matches cat regardless of what punctuation and whitespace surrounds it. Capital \B is the opposite: it requires that the current character not be at a word boundary. |
| | | The vertical bar separates two or more alternatives. A match occurs if any of the alternatives is satisfied. For example, gray|grey matches both gray and grey. Similarly, the pattern gr(a|e)y does the same thing with the help of the parentheses described below. |
| (...) | Items enclosed in parentheses are most commonly used to:
|
| \t \r etc. |
These escape sequences stand for special characters. The most common ones are \t (tab), \r (carriage return), and \n (linefeed). In AutoHotkey, an accent (`) may optionally be used in place of the backslash in these cases. Escape sequences in the form \xhh are also supported, in which hh is the hex code of any ANSI character between 00 and FF. \R matches |
| \p{xx} \P{xx} \X |
Unicode character properties. \p{xx} matches a character with the xx property while \P{xx} matches any character without the xx property. For example, \pL matches any letter and \p{Lu} matches any upper-case letter. \X matches any number of characters that form an extended Unicode sequence. For a full list of supported property names and other details, search for "\p{xx}" at www.pcre.org/pcre.txt. |
| (*UCP) | For performance, \d, \D, \s, \S, \w, \W, \b and \B recognize only ASCII characters by default. If the pattern begins with (*UCP), Unicode properties will be used to determine which characters match. For example, \w becomes equivalent to [\p{L}\p{N}_] and \d becomes equivalent to \p{Nd}. |
Greed: By default, *, ?, +, and {min,max} are greedy because they consume all characters up through the last possible one that still satisfies the entire pattern. To instead have them stop at the first possible character, follow them with a question mark. For example, the pattern <.+> (which lacks a question mark) means: "search for a <, followed by one or more of any character, followed by a >". To stop this pattern from matching the entire string <em>text</em>, append a question mark to the plus sign: <.+?>. This causes the match to stop at the first '>' and thus it matches only the first tag <em>.
Look-ahead and look-behind assertions: The groups (?=...), (?!...), (?<=...), and (?<!...) are called assertions because they demand a condition to be met but don't consume any characters. For example, abc(?=.*xyz) is a look-ahead assertion that requires the string xyz to exist somewhere to the right of the string abc (if it doesn't, the entire pattern is not considered a match). (?=...) is called a positive look-ahead because it requires that the specified pattern exist. Conversely, (?!...) is a negative look-ahead because it requires that the specified pattern not exist. Similarly, (?<=...) and (?<!...) are positive and negative look-behinds (respectively) because they look to the left of the current position rather than the right. Look-behinds are more limited than look-aheads because they do not support quantifiers of varying size such as *, ?, and +. The escape sequence \K is similar to a look-behind assertion because it causes any previously-matched characters to be omitted from the final matched string. For example, foo\Kbar matches "foobar" but reports that it has matched "bar".
Related: Regular expressions are supported by RegExMatch(), RegExReplace(), and SetTitleMatchMode.
Final note: Although this page touches upon most of the commonly-used RegEx features, there are quite a few other features you may want to explore such as conditional subpatterns. The complete PCRE manual is at www.pcre.org/pcre.txt