Scheme Regular Expressions


Alex Shinn


This SRFI is currently in final status. Here is an explanation of each status that a SRFI can hold. To provide input on this SRFI, please send email to srfi-115@nospamsrfi.schemers.org. To subscribe to the list, follow these instructions. You can access previous messages via the mailing list archive.

Table of Contents


This SRFI provides a library for matching strings with regular expressions described using the SRE "Scheme Regular Expression" notation first introduced by SCSH, and extended heavily by IrRegex.


Regular expressions, coming from a long history of formal language theory, are today the lingua franca of simple string matching. A regular expression is an expression describing a regular language, the simplest level in the Chomsky hierarchy. They have the nice property that they can match in linear time, whereas parsers for the next level in the hierarchy require cubic time. This combined with their conciseness led them to be a popular choice for searching in editors, tools and search interfaces. Other tools may be better suited to specific purposes, but it is assumed any modern language will provide regular expression support.

SREs were first introduced in SCSH as an s-expression based alternative to the more common string based description. This format offers many advantages, including being easier to read and write (notably with structured editors), easier to compose (with no escaping issues), and faster and simpler to compile. An efficient reference implementation of this SRFI can be written in under 1000 lines of code, whereas in IrRegex the full PCRE parser alone requires over 500 lines.

Procedure Index

regexp   rx     regexp->sre       char-set->sre
valid-sre?      regexp?           regexp-match?
regexp-matches  regexp-matches?   regexp-search
regexp-fold     regexp-extract    regexp-split   regexp-partition
regexp-replace          regexp-replace-all
regexp-match-submatch   regexp-match-submatch-start   regexp-match-submatch-end
regexp-match-count      regexp-match->list

Sre Syntax Index

Basic Patterns
<string> seq : or |

w/nocase w/case w/ascii w/unicode

optional ?     zero-or-more *
one-or-more +  at-least >=
exactly =      repeated **

submatch $ submatch-named ->

Character Sets
<char> (<string>) char-set
char-range / or | and &
complement ~ difference -

Named Character Sets
any nonl ascii   lower-case lower
upper-case upper title-case title
alphabetic       alpha numeric num
alphanumeric     alphanum alnum
punctuation      punct symbol
graphic graph    printing print
whitespace       white space
control cntrl    hex-digit xdigit

Boundary Checks
bos  eos   bol eol
bow  eow   nwb
word word+ word
bog  eog   grapheme

Non-greedy Patterns
non-greedy-optional     ??
non-greedy-zero-or-more *?
non-greedy-repeated     **?

Look Around Patterns
look-ahead look-behind neg-look-ahead neg-look-behind

Types and Naming Conventions

We introduce two new types, regexp and regexp-match, which are disjoint from all other types. We also introduce the concept of an "SRE," which is not a disjoint type but is a Scheme object following the specification described below.

SRFI 14 defines the char-set type, which can be used as part of an SRE.

In the prototypes below the following naming conventions imply type restrictions:

Compatibility Levels and Features

We specify a thorough, though not exhaustive, syntax with many extensions popular in modern regular expression libraries such as PCRE. This is because it is assumed in many cases said libraries will be used as the underlying implementation, the features will be desirable, and if left unspecified people will provide their own, often incompatible, extensions.

On the other hand it is acknowledged that not all implementations will be able to support all extensions. Some are difficult to implement for DFA implementations, and some, like backref, are prohibitively expensive for any implementation. Furthermore, even if an implementation has Unicode support, its regexp library may not.

To resolve these differences we divide the syntax into a minimal core which all implementations are required to support, and additional extensions. In R7RS or other implementations which support SRFI 0 cond-expand with library level features, the availability can be tested with the following cond-expand features:

The first three simply refer to support for certain SRE patterns.

regexp-unicode indicates support for Unicode contexts. Toggling between Unicode and ASCII can be done with the w/unicode and w/ascii patterns. In a Unicode context, the named character sets have their full Unicode definition as described below and grapheme boundaries are "extended grapheme clusters" as defined in UAX #29 (Unicode Text Segmentation). Implementations which provide this feature may still support non-Unicode characters.

Library Procedures and Syntax

(regexp re) => regexp

Compiles a regexp if given an object whose structure matches the SRE syntax. This may be written as a literal or partial literal with quote or quasiquote, or may be generated entirely programmatically. Returns re unmodified if it is already a regexp. Raises an error if re is neither a regexp nor a valid representation of an SRE.

Mutating re may invalidate the resulting regexp, causing unspecified results if subsequently used for matching.

(rx sre ...) => regexp

Macro shorthand for (regexp `(: sre ...)). May be able to perform some or all computation at compile time if sre is not unquoted. Note because of this equivalence with the procedural constructor regexp, the semantics of unquote differs from the original SCSH implementation in that unquoted expressions can expand into any object matching the SRE syntax, but not a compiled regexp object. Further, unquote and unquote-splicing both expand all matches.

Rationale: Providing a procedural interface provides for greater flexibility, and without loss of potential compile-time optimizations by preserving the syntactic shorthand. The alternative is to rely on eval to dynamically generate regular expressions. However regexps in many cases come from untrusted sources, such as search parameters to a server, or from serialized sources such as config files or command-line arguments. Moreover many applications may want to keep many thousands of regexps in memory at once. Given the relatively heavy cost and insecurity of eval, and the frequency with which SREs are read and written as text, we prefer the procedural interface.

(regexp->sre re) => sre

Returns an SRE corresponding to the given regexp re. The SRE will be equivalent to (will match the same strings) but not necessarily equal? to the SRE originally used to compile re. Mutating the result may invalidate re, causing unspecified results if subsequently used for matching.

(char-set->sre char-set) => sre

Returns an SRE corresponding to the given SRFI 14 character set. The resulting SRE expands the character set into notation which does not make use of embedded SRFI 14 character sets, and so is suitable for writing portably.

(valid-sre? obj) => boolean

Returns true iff obj can be safely passed to regexp.

(regexp? obj) => boolean

Returns true iff obj is a regexp.

(regexp-matches re str [start [end]]) => regexp-match-or-false

Returns an regexp-match object if re successfully matches the entire string str from start (inclusive) to end (exclusive), or #f is the match fails. The regexp-match object will contain information needed to extract any submatches.

(regexp-matches? re str [start [end]]) => boolean?

Returns #t if re matches str as in regexp-matches, or #f otherwise. May be faster than regexp-matches since it doesn't need to return submatch data.

(regexp-search re str [start [end]]) => regexp-match-or-false

Returns a regexp-match object if re successfully matches a substring of str between start (inclusive) and end (exclusive), or #f if the match fails. The regexp-match object will contain information needed to extract any submatches.

(regexp-fold re kons knil str [finish [start [end]]]) => obj

The fundamental regexp matching iterator. Repeatedly searches str for the regexp re so long as a match can be found. On each successful match, applies (kons i regexp-match str acc) where i is the index since the last match (beginning with start), regexp-match is the resulting match, and acc is the result of the previous kons application, beginning with knil. When no more matches can be found, calls finish with the same arguments, except that regexp-match is #f.

By default finish just returns acc.

   (regexp-fold 'word
                (lambda (i m str acc)
                  (let ((s (regexp-match-submatch m 0)))
                   (cond ((assoc s acc)
                          => (lambda (x) (set-cdr! x (+ 1 (cdr x))) acc))
                         (else `((,s . 1) ,@acc)))))
                "to be or not to be")
   => '(("not" . 1) ("or" . 1) ("be" . 2) ("to" . 2))

(regexp-extract re str [start [end]]) => list

Extracts all non-empty substrings of str which match re between start and end as a list of strings.

   (regexp-extract '(+ numeric) "")
   => ("192" "168" "0" "1")
(regexp-split re str [start [end]]) => list

Splits str into a list of (possibly empty) strings separated by non-empty matches of re.

   (regexp-split '(+ space) " fee fi  fo\tfum\n")
   => ("" "fee" "fi" "fo" "fum" "")
   (regexp-split '(",;") "a,,b,")
   => ("a" "" "b" "")
   (regexp-split '(* digit) "abc123def456ghi789")
   => ("abc" "def" "ghi" "")
(regexp-partition re str [start [end]]) => list

Partitions str into a list of non-empty strings matching re, interspersed with the unmatched portions of the string. The first and every odd element is an unmatched substring, which will be the empty string if re matches at the beginning of the string or end of the previous match. The second and every even element will be a substring matching re. If the final match ends at the end of the string, no trailing empty string will be included. Thus, in the degenerate case where str is the empty string, the result is ("").

Note that regexp-partition is equivalent to interleaving the results of regexp-split and regexp-extract, starting with the former.

   (regexp-partition '(+ (or space punct)) "")
   => ("")
   (regexp-partition '(+ (or space punct)) "Hello, world!\n")
   => ("Hello" ", " "world" "!\n")
   (regexp-partition '(+ (or space punct)) "¿Dónde Estás?")
   => ("" "¿" "Dónde" " " "Estás" "?")
   (regexp-partition '(* digit) "abc123def456ghi789")
   => ("abc" "123" "def" "456" "ghi" "789")
(regexp-replace re str subst [start [end [count]]) => string

Returns a new string replacing the countth match of re in str the subst, where the zero-indexed count defaults to zero (i.e. the first match). If there are not count matches, returns the selected substring unmodified.

subst can be a string, an integer or symbol indicating the contents of a numbered or named submatch of re, 'pre for the substring to the left of the match, or 'post for the substring to the right of the match.

The optional parameters start and end restrict both the matching and the substitution, to the given indices, such that the result is equivalent to omitting these parameters and replacing on (substring str start end). As a convenience, a value of #f for end is equivalent to (string-length str).

   (regexp-replace '(+ space) "one two three" "_")
   => "one_two three"
   (regexp-replace '(+ space) "one two three" "_" 0 #f 0)
   => "one_two three"
   (regexp-replace '(+ space) "one two three" "_" 0 #f 1)
   => "one two_three"
   (regexp-replace '(+ space) "one two three" "_" 0 #f 2)
   => "one two three"
(regexp-replace-all re str subst [start [end]]) => string

Equivalent to regexp-replace, but replaces all occurrences of re in str.

   (regexp-replace-all '(+ space) "one two three" "_")
   => "one_two_three"
(regexp-match? obj) => boolean

Returns true iff obj is a successful match from regexp-matches or regexp-search.

   (regexp-match? (regexp-matches "x" "x"))  => #t
   (regexp-match? (regexp-matches "x" "y"))  => #f

(regexp-match-count regexp-match) => integer

Returns the number of submatches of regexp-match, regardless of whether they matched or not. Does not include the implicit zero full match in the count.

   (regexp-match-count (regexp-matches "x" "x"))  => 0
   (regexp-match-count (regexp-matches '($ "x") "x"))  => 1

(regexp-match-submatch regexp-match field) => string-or-false

Returns the substring matched in regexp-match corresponding to field, either an integer or a symbol for a named submatch. Index 0 refers to the entire match, index 1 to the first lexicographic submatch, and so on. If there are multiple submatches with the same name, the first which matched is returned. If passed an integer outside the range of matches, or a symbol which does not correspond to a named submatch of the pattern, it is an error. If the corresponding submatch did not match, returns false.

The result of extracting a submatch after the original matched string has been mutated is unspecified.

   (regexp-match-submatch (regexp-search 'word "**foo**") 0)  => "foo"
    (regexp-search '(: "*" ($ word) "*") "**foo**") 0)  => "*foo*"
    (regexp-search '(: "*" ($ word) "*") "**foo**") 1)  => "foo"

(regexp-match-submatch-start regexp-match field) => integer-or-false

Returns the start index regexp-match corresponding to field, as in regexp-match-submatch.

    (regexp-search 'word "**foo**") 0)  => 2
    (regexp-search '(: "*" ($ word) "*") "**foo**") 0)  => 1
    (regexp-search '(: "*" ($ word) "*") "**foo**") 1)  => 2

(regexp-match-submatch-end regexp-match field) => integer-or-false

Returns the end index in regexp-match corresponding to field, as in regexp-match-submatch.

    (regexp-search 'word "**foo**") 0)  => 5
    (regexp-search '(: "*" ($ word) "*") "**foo**") 0)  => 6
    (regexp-search '(: "*" ($ word) "*") "**foo**") 1)  => 5

(regexp-match->list regexp-match) => list

Returns a list of all submatches in regexp-match as string or false, beginning with the entire match 0.

    (regexp-search '(: ($ word) (+ (or space punct)) ($ word)) "cats & dogs"))
    => '("cats & dogs" "cats" "dogs")

SRE Syntax

The grammar for SREs is summarized below. Note that an SRE is a first-class object consisting of nested lists of strings, chars, char-sets, symbols and numbers. Where the syntax is described as (foo bar), this can be constructed equivalently as '(foo bar) or (list 'foo 'bar), etc. The following sections explain the semantics in greater detail.

    <sre> ::=
     | <string>                    ; A literal string match.
     | <cset-sre>                  ; A character set match.
     | (* <sre> ...)               ; 0 or more matches.
     | (zero-or-more <sre> ...)
     | (+ <sre> ...)               ; 1 or more matches.
     | (one-or-more <sre> ...)
     | (? <sre> ...)               ; 0 or 1 matches.
     | (optional <sre> ...)
     | (= <n> <sre> ...)           ; <n> matches.
     | (exactly <n> <sre> ...)
     | (>= <n> <sre> ...)          ; <n> or more matches.
     | (at-least <n> <sre> ...)
     | (** <n> <m> <sre> ...)      ; <n> to <m> matches.
     | (repeated <n> <m> <sre> ...)

     | (|  <sre> ...)              ; Alternation.
     | (or <sre> ...)

     | (:   <sre> ...)             ; Sequence.
     | (seq <sre> ...)     
     | ($ <sre> ...)               ; Numbered submatch.
     | (submatch <sre> ...)
     | (-> <name> <sre> ...)               ;  Named submatch.  <name> is
     | (submatch-named <name> <sre> ...)   ;  a symbol.

     | (w/case   <sre> ...)        ; Introduce a case-sensitive context.
     | (w/nocase <sre> ...)        ; Introduce a case-insensitive context.

     | (w/unicode   <sre> ...)     ; Introduce a unicode context.
     | (w/ascii <sre> ...)         ; Introduce an ascii context.

     | (w/nocapture <sre> ...)     ; Ignore all enclosed submatches.

     | bos                         ; Beginning of string.
     | eos                         ; End of string.

     | bol                         ; Beginning of line.
     | eol                         ; End of line.

     | bog                         ; Beginning of grapheme cluster.
     | eog                         ; End of grapheme cluster.
     | grapheme                    ; A single grapheme cluster.

     | bow                         ; Beginning of word.
     | eow                         ; End of word.
     | nwb                         ; A non-word boundary.
     | (word <sre> ...)            ; An SRE wrapped in word boundaries.
     | (word+ <cset-sre> ...)      ; A single word restricted to a cset.
     | word                        ; A single word.

     | (?? <sre> ...)              ; A non-greedy pattern, 0 or 1 match.
     | (non-greedy-optional <sre> ...)
     | (*? <sre> ...)              ; Non-greedy 0 or more matches.
     | (non-greedy-zero-or-more <sre> ...)
     | (**? <m> <n> <sre> ...)     ; Non-greedy <m> to <n> matches.
     | (non-greedy-repeated <sre> ...)
     | (look-ahead <sre> ...)      ; Zero-width look-ahead assertion.
     | (look-behind <sre> ...)     ; Zero-width look-behind assertion.
     | (neg-look-ahead <sre> ...)  ; Zero-width negative look-ahead assertion.
     | (neg-look-behind <sre> ...) ; Zero-width negative look-behind assertion.

     | (backref <n-or-name>)       ; Match a previous submatch.
The grammar for cset-sre is as follows.

    <cset-sre> ::=
     | <char>                      ; literal char
     | "<char>"                    ; string of one char
     | <char-set>                  ; embedded SRFI 14 char set
     | (<string>)                  ; literal char set
     | (char-set <string>)
     | (/ <range-spec> ...)        ; ranges
     | (char-range <range-spec> ...)
     | (or <cset-sre> ...)         ; union
     | (|\|| <cset-sre> ...)
     | (and <cset-sre> ...)        ; intersection
     | (& <cset-sre> ...)
     | (- <cset-sre> ...)          ; difference
     | (- <difference> ...)
     | (~ <cset-sre> ...)          ; complement of union
     | (complement <cset-sre> ...)
     | (w/case <cset-sre> ...)     ; case and unicode toggling
     | (w/nocase <cset-sre> ...)
     | (w/ascii <cset-sre> ...)
     | (w/unicode <cset-sre> ...)
     | any | nonl | ascii | lower-case | lower
     | upper-case | upper | title-case | title
     | alphabetic | alpha | alphanumeric | alphanum | alnum
     | numeric | num | punctuation | punct | symbol
     | graphic | graph | whitespace | white | space
     | printing | print | control | cntrl | hex-digit | xdigit
    <range-spec> ::= <string> | <char>

Basic Patterns


A literal string.

   (regexp-search "needle" "hayneedlehay") => #<regexp-match>
   (regexp-search "needle" "haynEEdlehay") => #f
(seq sre ...)
(: sre ...)

Sequencing. Matches if each of sre matches adjacently in order.

   (regexp-search '(: "one" space "two" space "three") "one two three") => #<regexp-match>
(or sre ...)
(|\|| sre ...)

Alternation. Matches if any of sre match.

   (regexp-search '(or "eeney" "meeney" "miney") "meeney") => #<regexp-match>
   (regexp-search '(or "eeney" "meeney" "miney") "moe") => #f
(w/nocase sre ...)

Enclosed sres are case-insensitive. In a Unicode context character and string literals match with the default simple Unicode case-insensitive matching. Implementations may, but are not required to, handle variable length case conversions, such as #\x00DF "ß" matching the two characters "SS".

Character sets match if any character in the set matches case-insensitively to the input. Conceptually each cset-sre is expanded to contain all case variants for all of its characters. In a compound cset-sre the expansion is applied at the terminals consisting of characters, strings, embedded SRFI 14 char-sets, and named character sets. For simple unions this would be equivalent to computing the full union first and then expanding case variants, but the semantics can differ when differences and intersections are applied. For example, (w/nocase (~ ("Aab"))) is equivalent to (~ ("AaBb")), for which "B" is clearly not a member. However if you were to compute (~ ("Aab")) first then you would have a char-set containing "B", and after expanding case variants both "B" and "b" would be members.

In an ASCII context only the 52 ASCII letters (/ "a-zA-Z") match case-insensitively to each other.

In a Unicode context the only named cset-sre which are affected by w/nocase are upper and lower. Note that the case insensitive versions of these are not equivalent to alpha as there are characters with the letter property but no case.

   (regexp-search "needle" "haynEEdlehay") => #f
   (regexp-search '(w/nocase "needle") "haynEEdlehay") => #<regexp-match>

   (regexp-search '(~ ("Aab")) "B") => #<regexp-match>
   (regexp-search '(~ ("Aab")) "b") => #f
   (regexp-search '(w/nocase (~ ("Aab"))) "B") => #f
   (regexp-search '(w/nocase (~ ("Aab"))) "b") => #f
   (regexp-search '(~ (w/nocase ("Aab"))) "B") => #f
   (regexp-search '(~ (w/nocase ("Aab"))) "b") => #f
(w/case sre ...)

Enclosed sres are case-sensitive. This is the default, and overrides any enclosing w/nocase setting.

   (regexp-search '(w/nocase "SMALL" (w/case "BIG")) "smallBIGsmall") => #<regexp-match>
   (regexp-search '(w/nocase (~ (w/case ("Aab")))) "b") => #f
(w/ascii sre ...)

Enclosed sres are interpreted in an ASCII context. In practice many regular expressions are used for simple parsing and only ASCII characters are relevant. Switching to ASCII mode can improve performance in some implementations.

   (regexp-search '(w/ascii bos (* alpha) eos) "English") => #<regexp-match>
   (regexp-search '(w/ascii bos (* alpha) eos) "Ελληνική") => #f
(w/unicode sre ...)

Enclosed sres are interpreted in a Unicode context - character sets with both an ASCII and Unicode definition take the latter. Has no effect if the regexp-unicode feature is not provided. This is the default.

   (regexp-search '(w/unicode bos (* alpha) eos) "English") => #<regexp-match>
   (regexp-search '(w/unicode bos (* alpha) eos) "Ελληνική") => #<regexp-match>
(w/nocapture sre ...)

Disables capturing for all submatches ($, submatch, -> and submatch-named) in the enclosed sres. The resulting SRE matches exactly the same strings, but without any associated submatch info. Useful for utility SREs which you want to incorporate without affecting your submatch positions.

   (let ((number '($ (+ digit))))
       (regexp-search `(: ,number "-" ,number "-" ,number)
                      "555-867-5309")))  ; => '("555" "867" "5309")
       (regexp-search `(: ,number "-" (w/nocapture ,number) "-" ,number)
                      "555-867-5309"))))   => '("555" "5309")

Repeating patterns

(optional sre ...)
(? sre ...)

An optional pattern - matches 1 or 0 times.

   (regexp-search '(: "match" (? "es") "!") "matches!") => #<regexp-match>
   (regexp-search '(: "match" (? "es") "!") "match!") => #<regexp-match>
   (regexp-search '(: "match" (? "es") "!") "matche!") => #f
(zero-or-more sre ...)
(* sre ...)

Kleene star, matches 0 or more times.

   (regexp-search '(: "<" (* (~ #\>)) ">") "<html>") => #<regexp-match>
   (regexp-search '(: "<" (* (~ #\>)) ">") "<>") => #<regexp-match>
   (regexp-search '(: "<" (* (~ #\>)) ">") "<html") => #f
(one-or-more sre ...)
(+ sre ...)

1 or more matches. Like * but requires at least a single match.

   (regexp-search '(: "<" (+ (~ #\>)) ">") "<html>") => #<regexp-match>
   (regexp-search '(: "<" (+ (~ #\>)) ">") "<a>") => #<regexp-match>
   (regexp-search '(: "<" (+ (~ #\>)) ">") "<>") => #f
(at-least n sre ...)
(>= n sre ...)

More generally, n or more matches.

   (regexp-search '(: "<" (>= 3 (~ #\>)) ">") "<table>") => #<regexp-match>
   (regexp-search '(: "<" (>= 3 (~ #\>)) ">") "<pre>") => #<regexp-match>
   (regexp-search '(: "<" (>= 3 (~ #\>)) ">") "<tr>") => #f
(exactly n sre ...)
(= n sre ...)

Exactly n matches.

   (regexp-search '(: "<" (= 4 (~ #\>)) ">") "<html>") => #<regexp-match>
   (regexp-search '(: "<" (= 4 (~ #\>)) ">") "<table>") => #f
(repeated from to sre ...)
(** from to sre ...)

The most general form, from n to m matches, inclusive.

   (regexp-search '(: (= 3 (** 1 3 numeric) ".") (** 1 3 numeric)) "") => #<regexp-match>
   (regexp-search '(: (= 3 (** 1 3 numeric) ".") (** 1 3 numeric)) "192.0168.1.10") => #f

Submatch Patterns

(submatch sre ...)
($ sre ...)

A numbered submatch. The contents matching the pattern will be available in the resulting regexp-match.

(submatch-named name sre ...)
(-> name sre ...)

A named submatch. Behaves just like submatch, but the field may also be referred to by name.

(backref n-or-name)

Optional: Match a previously matched submatch. The feature regexp-backrefs will be provided if this pattern is supported. Backreferences are expensive, and can trivially be shown to be NP-hard, so one should avoid their use even in implementations which support them.

Character Sets

A character set pattern matches a single character.


A singleton char set.

   (regexp-matches '(* #\-) "---") => #<regexp-match>
   (regexp-matches '(* #\-) "-_-") => #f


A singleton char set written as a string of length one rather than a character. Equivalent to its interpretation as a literal string match, but included to clarify it can be composed in cset-sres.


A SRFI 14 character set, which matches any character in the set. Note that currently there is no portable written representation of SRFI 14 character sets, which means that this pattern is typically generated programmatically, such as with a quasiquoted expression.

   (regexp-partition `(+ ,char-set:vowels) "vowels")
   => ("v" "o" "w" "e" "ls")
Rationale: Many useful character sets are likely to be available as SRFI 14 char-sets, so it is desirable to reuse them in regular expressions. Since many Unicode character sets are extremely large, converting back and forth between an internal and external representation can be expensive, so the option of direct embedding is necessary. When a readable external representation is needed, char-set->sre can be used.

(char-set <string>)

The set of chars as formed by SRFI 14 (string->char-set <string>).

Note that char-sets contain code points, not grapheme clusters, so any combining characters in <string> will be inserted separately from any preceding base characters by string->char-set.

   (regexp-matches '(* ("aeiou")) "oui") => #<regexp-match>
   (regexp-matches '(* ("aeiou")) "ouais") => #f
   (regexp-matches '(* ("e\x0301")) "e\x0301") => #<regexp-match>
   (regexp-matches '("e\x0301") "e\x0301") => #f
   (regexp-matches '("e\x0301") "e") => #<regexp-match>
   (regexp-matches '("e\x0301") "\x0301") => #<regexp-match>
   (regexp-matches '("e\x0301") "\x00E9") => #f
(char-range <range-spec> ...)
(/ <range-spec> ...)

Ranged char set. The <range-spec> is a list of strings and characters. These are flattened and grouped into pairs of characters, and all ranges formed by the pairs are included in the char set.

   (regexp-matches '(* (/ "AZ09")) "R2D2") => #<regexp-match>
   (regexp-matches '(* (/ "AZ09")) "C-3PO") => #f
(or <cset-sre> ...)
(|\|| <cset-sre> ...)

Char set union. The single vertical bar form is provided for consistency and compatibility with SCSH, although it needs to be escaped in R7RS.

(complement <cset-sre> ...)
(~ <cset-sre> ...)

Char set complement (i.e. [^...] in PCRE notation). (~ x) is equivalent to (- any x), thus in an ASCII context the complement is always ASCII.

(difference <cset-sre> ...)
(- <cset-sre> ...)

Char set difference.

   (regexp-matches '(* (- (/ "az") ("aeiou"))) "xyzzy") => #<regexp-match>
   (regexp-matches '(* (- (/ "az") ("aeiou"))) "vowels") => #f
(and <cset-sre> ...)
(& <cset-sre> ...)

Char set intersection.

   (regexp-matches '(* (& (/ "az") (~ ("aeiou")))) "xyzzy") => #<regexp-match>
   (regexp-matches '(* (& (/ "az") (~ ("aeiou")))) "vowels") => #f

Named Character Sets


Match any character. Equivalent to ascii in an ASCII context.


Match any character other than #\return or #\newline.


Match any ASCII character [0..127].


Matches any character for which char-lower-case? returns true. In a Unicode context this corresponds to the Lowercase (Ll + Other_Lowercase) property. In an ASCII context corresponds to (/ "az").


Matches any character for which char-upper-case? returns true. In a Unicode context this corresponds to the Uppercase (Lu + Other_Uppercase) property. In an ASCII context corresponds to (/ "AZ").


Matches any character with the Unicode Titlecase (Lt) property. This property only exists for the sake of ligature characters, of which only 31 exist at time of writing. In an ASCII context this is empty.


Matches any character for which char-alphabetic? returns true. In a Unicode context this corresponds to the Alphabetic (L + Nl + Other_Alphabetic) property. In an ASCII context corresponds to (w/nocase (/ "az")).


Matches any character for which char-numeric? returns true. For In a Unicode context this corresponds to the Numeric_Digit (Nd) property. In an ASCII context corresponds to (/ "09").


Matches any character which is either a letter or number. Equivalent to:

   (or alphabetic numeric)

Matches any punctuation character. In a Unicode context this corresponds to the Punctuation property (P). In an ASCII context this corresponds to "!\"#%&'()*,-./:;?@[\]_{}".


Matches any symbol character. In a Unicode context this corresponds to the Symbol property (Sm, Sc, Sk, or So). In an ASCII this corresponds to "$+<=>^`|~".


Matches any graphic character. Equivalent to:

   (or alphanumeric punctuation symbol)

Matches any whitespace character. In a Unicode context this corresponds to the Separator property (Zs, Zl or Zp). In an ASCII context this corresponds to space, tab, line feed, form feed, and carriage return.


Matches any printing character. Equivalent to:

   (or graphic whitespace)

Matches any control or other character. In a Unicode context this corresponds to the Other property (Cc, Cf, Co, Cs or Cn). In an ASCII context this corresponds to:

   `(/ ,(integer->char 0) ,(integer-char 31))

Matches any valid digit in hexadecimal notation. Always ASCII-only. Equivalent to:

   (w/ascii (w/nocase (or numeric "abcdef")))

Boundary Assertions


Matches at the beginning/end of string without consuming any characters (a zero-width assertion). If the search was initiated with start/end parameters, these are considered the end points, rather than the full string.


Matches at the beginning/end of a line without consuming any characters (a zero-width assertion). A line is a possibly empty sequence of characters followed by an end of line sequence as understood by the R7RS read-line procedure, specifically any of a linefeed character, carriage return character, or a carriage return followed by a linefeed character. The string is assumed to contain end of line sequences before the start and after the end of the string, even if the search was made on a substring and the actual surrounding characters differ.


Matches at the beginning/end of a word without consuming any characters (a zero-width assertion). A word is a contiguous sequence of characters that are either alphanumeric or the underscore character, i.e. (or alphanumeric "_"), with the definition of alphanumeric depending on the Unicode or ASCII context. The string is assumed to contain non-word characters immediately before the start and after the end, even if the search was made on a substring and word constituent characters appear immediately before the beginning or after the end.

  (regexp-search '(: bow "foo") "foo") => #<regexp-match>
  (regexp-search '(: bow "foo") "<foo>>") => #<regexp-match>
  (regexp-search '(: bow "foo") "snafoo") => #f
  (regexp-search '(: "foo" eow) "foo") => #<regexp-match>
  (regexp-search '(: "foo" eow) "foo!") => #<regexp-match>
  (regexp-search '(: "foo" eow) "foobar") => #f


Matches a non-word-boundary (i.e. \B in PCRE). Equivalent to (neg-look-ahead (or bow eow)).

(word sre ...)

Anchors a sequence to word boundaries. Equivalent to:

   (: bow sre ... eow)
(word+ cset-sre ...)

Matches a single word composed of characters in the intersection of the given cset-sre and the word constituent characters. Equivalent to:

   (word (+ (and (or alphanumeric "_") (or cset-sre ...))))

A shorthand for (word+ any).


Matches at the beginning/end of a single extended grapheme cluster without consuming any characters (a zero-width assertion). Grapheme cluster boundaries are defined in Unicode TR29. The string is assumed to contain non-combining codepoints immediately before the start and after the end. These always succeed in an ASCII context.


Matches a single grapheme cluster (i.e. \X in PCRE). This is what the end-user typically thinks of as a single character, comprised of a base non-combining codepoint followed by zero or more combining marks. In an ASCII context this is equivalent to any.

Assuming char-set:mark contains all characters with the Extend or SpacingMark properties defined in TR29, and char-set:control, char-set:regional-indicator and char-set:hangul-* are defined similarly, then the following SRE can be used with regexp-extract to define grapheme:

   `(or (: (* ,char-set:hangul-l) (+ ,char-set:hangul-v)
           (* ,char-set:hangul-t))
        (: (* ,char-set:hangul-l) ,char-set:hangul-v
           (* ,char-set:hangul-v) (* ,char-set:hangul-t))
        (: (* ,char-set:hangul-l) ,char-set:hangul-lvt
           (* ,char-set:hangul-t))
        (+ ,char-set:hangul-l)
        (+ ,char-set:hangul-t)
        (+ ,char-set:regional-indicator)
        (: "\r\n")
        (: (~ control ("\r\n"))
           (* ,char-set:mark))

Non-Greedy Patterns

The following patterns are only supported if the feature regexp-non-greedy is provided.

(non-greedy-optional sre ...)
(?? sre ...)

Non-greedy pattern, matches 0 or 1 times, preferring the shorter match.

(non-greedy-zero-or-more< sre ...)
(*? sre ...)

Non-greedy Kleene star, matches 0 or more times, preferring the shorter match.

(non-greedy-repeated m n sre ...)
(**? m n sre ...)

Non-greedy Kleene star, matches m to n times, preferring the shorter match.

Look Around Patterns

The following patterns are only supported if the feature regexp-look-around is provided.

(look-ahead sre ...)

Zero-width look-ahead assertion. Asserts the sequence matches from the current position, without advancing the position.

   (regexp-matches '(: "regular" (look-ahead " expression") " expression") "regular expression")=> #<regexp-match>
   (regexp-matches '(: "regular" (look-ahead " ") "expression") "regular expression")=> #f

(look-behind sre ...)

Zero-width look-behind assertion. Asserts the sequence matches behind the current position, without advancing the position. It is an error if the sequence does not have a fixed length.

(neg-look-ahead sre ...)

Zero-width negative look-ahead assertion.

(neg-look-behind sre ...)

Zero-width negative look-behind assertion.


A sample implementation in portable R7RS is available at

depending only on SRFI 14, SRFI 33 and SRFI 69. This is implemented as a Thompson-style non-backtracking NFA, a discussion of which can be found at Russ Cox's Implementing Regexps. At time of writing the implementation supports full unicode, but does not yet support look-around patterns. See IrRegex for an alternate implementation supporting all features, providing both a backtracking algorithm and DFA compilation.


      Alex Shinn, John Cowan, Arthur Gleckler, Revised7 Report on the Algorithmic Language Scheme
      Olin Shivers, A Scheme Shell
      Massachusetts Institute of Technology Cambridge, MA, USA, 1994
      Alex Shinn, IrRegex - IrRegular Expressions
      Mark Davis, Andy Heninger, UTR #18: Unicode Regular Expressions
      Mark Davis, UAX #29: Unicode Text Segmentation
      Marc Feeley, Feature-based conditional expansion construct
      Olin Shivers, Character-set Library
      Russ Cox, Implementing Regular Expressions
      Russ Cox, Henry Spencer's Tcl Regex Library
      Shiro Kawai, Gauche Scheme - Regular Expressions
      Damian Conway, Perl6 Exegesis 5 - Regular Expressions
      Philip Hazel, PCRE - Perl Compatible Regular Expressions


Copyright (C) Alex Shinn 2014. All Rights Reserved.

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.


Editor: Mike Sperber