path: root/doc/regex.texi

@node Overview
@chapter Overview

A @dfn{regular expression} (or @dfn{regexp}, or @dfn{pattern}) is a text
string that describes some (mathematical) set of strings.  A regexp
@var{r} @dfn{matches} a string @var{s} if @var{s} is in the set of
strings described by @var{r}.

Using the Regex library, you can:

@itemize @bullet

@item
see if a string matches a specified pattern as a whole, and

@item
search within a string for a substring matching a specified pattern.

@end itemize

Some regular expressions match only one string, i.e., the set they
describe has only one member.  For example, the regular expression
@samp{foo} matches the string @samp{foo} and no others.  Other regular
expressions match more than one string, i.e., the set they describe has
more than one member.  For example, the regular expression @samp{f*}
matches the set of strings made up of any number (including zero) of
@samp{f}s.  As you can see, some characters in regular expressions match
themselves (such as @samp{f}) and some don't (such as @samp{*}); the
ones that don't match themselves instead let you specify patterns that
describe many different strings.

To either match or search for a regular expression with the Regex
library functions, you must first compile it with a Regex pattern
compiling function.  A @dfn{compiled pattern} is a regular expression
converted to the internal format used by the library functions.  Once
you've compiled a pattern, you can use it for matching or searching any
number of times.

The Regex library is used by including @file{regex.h}.
@pindex regex.h
Regex provides three groups of functions with which you can operate on
regular expressions.  One group---the GNU group---is more
powerful but not completely compatible with the other two, namely the
POSIX and Berkeley Unix groups; its interface was designed
specifically for GNU.

We wrote this chapter with programmers in mind, not users of
programs---such as Emacs---that use Regex.  We describe the Regex
library in its entirety, not how to write regular expressions that a
particular program understands.


@node Regular Expression Syntax
@chapter Regular Expression Syntax

@cindex regular expressions, syntax of
@cindex syntax of regular expressions

@dfn{Characters} are things you can type.  @dfn{Operators} are things in
a regular expression that match one or more characters.  You compose
regular expressions from operators, which in turn you specify using one
or more characters.

Most characters represent what we call the match-self operator, i.e.,
they match themselves; we call these characters @dfn{ordinary}.  Other
characters represent either all or parts of fancier operators; e.g.,
@samp{.} represents what we call the match-any-character operator
(which, no surprise, matches (almost) any character); we call these
characters @dfn{special}.  Two different things determine what
characters represent what operators:

@enumerate
@item
the regular expression syntax your program has told the Regex library to
recognize, and

@item
the context of the character in the regular expression.
@end enumerate

In the following sections, we describe these things in more detail.

@menu
* Syntax Bits::
* Predefined Syntaxes::
* Collating Elements vs. Characters::
* The Backslash Character::
@end menu


@node Syntax Bits
@section Syntax Bits

@cindex syntax bits

In any particular syntax for regular expressions, some characters are
always special, others are sometimes special, and others are never
special.  The particular syntax that Regex recognizes for a given
regular expression depends on the current syntax (as set by
@code{re_set_syntax}) when the pattern buffer of that regular expression
was compiled.

You get a pattern buffer by compiling a regular expression.  @xref{GNU
Pattern Buffers}, for more information on pattern buffers.  @xref{GNU
Regular Expression Compiling}, and @ref{BSD Regular Expression
Compiling}, for more information on compiling.

Regex considers the current syntax to be a collection of bits; we refer
to these bits as @dfn{syntax bits}.  In most cases, they affect what
characters represent what operators.  We describe the meanings of the
operators to which we refer in @ref{Common Operators} and @ref{GNU
Operators}.

For reference, here is the complete list of syntax bits, in alphabetical
order:

@table @code

@cnindex RE_BACKSLASH_ESCAPE_IN_LIST
@item RE_BACKSLASH_ESCAPE_IN_LISTS
If this bit is set, then @samp{\} inside a list (@pxref{List Operators})
quotes (makes ordinary, if it's special) the following character; if
this bit isn't set, then @samp{\} is an ordinary character inside lists.
(@xref{The Backslash Character}, for what @samp{\} does outside of lists.)

@cnindex RE_BK_PLUS_QM
@item RE_BK_PLUS_QM
If this bit is set, then @samp{\+} represents the match-one-or-more
operator and @samp{\?} represents the match-zero-or-more operator; if
this bit isn't set, then @samp{+} represents the match-one-or-more
operator and @samp{?} represents the match-zero-or-one operator.  This
bit is irrelevant if @code{RE_LIMITED_OPS} is set.

@cnindex RE_CHAR_CLASSES
@item RE_CHAR_CLASSES
If this bit is set, then you can use character classes in lists; if this
bit isn't set, then you can't.

@cnindex RE_CONTEXT_INDEP_ANCHORS
@item RE_CONTEXT_INDEP_ANCHORS
If this bit is set, then @samp{^} and @samp{$} are special anywhere outside
a list; if this bit isn't set, then these characters are special only in
certain contexts.  @xref{Match-beginning-of-line Operator}, and
@ref{Match-end-of-line Operator}.

@cnindex RE_CONTEXT_INDEP_OPS
@item RE_CONTEXT_INDEP_OPS
If this bit is set, then certain characters are special anywhere outside
a list; if this bit isn't set, then those characters are special only in
some contexts and are ordinary elsewhere.  Specifically, if this bit
isn't set then @samp{*}, and (if the syntax bit @code{RE_LIMITED_OPS}
isn't set) @samp{+} and @samp{?} (or @samp{\+} and @samp{\?}, depending
on the syntax bit @code{RE_BK_PLUS_QM}) represent repetition operators
only if they're not first in a regular expression or just after an
open-group or alternation operator.  The same holds for @samp{@{} (or
@samp{\@{}, depending on the syntax bit @code{RE_NO_BK_BRACES}) if
it is the beginning of a valid interval and the syntax bit
@code{RE_INTERVALS} is set.

@cnindex RE_CONTEXT_INVALID_DUP
@item RE_CONTEXT_INVALID_DUP
If this bit is set, then an open-interval operator cannot occur at the
start of a regular expression, or immediately after an alternation,
open-group or close-interval operator.

@cnindex RE_CONTEXT_INVALID_OPS
@item RE_CONTEXT_INVALID_OPS
If this bit is set, then repetition and alternation operators can't be
in certain positions within a regular expression.  Specifically, the
regular expression is invalid if it has:

@itemize @bullet

@item
a repetition operator first in the regular expression or just after a
match-beginning-of-line, open-group, or alternation operator; or

@item
an alternation operator first or last in the regular expression, just
before a match-end-of-line operator, or just after an alternation or
open-group operator.

@end itemize

If this bit isn't set, then you can put the characters representing the
repetition and alternation characters anywhere in a regular expression.
Whether or not they will in fact be operators in certain positions
depends on other syntax bits.

@cnindex RE_DEBUG
@item RE_DEBUG
If this bit is set, and the regex library was compiled with
@code{-DDEBUG}, then internal debugging is turned on; if unset, then
it is turned off.

@cnindex RE_DOT_NEWLINE
@item RE_DOT_NEWLINE
If this bit is set, then the match-any-character operator matches
a newline; if this bit isn't set, then it doesn't.

@cnindex RE_DOT_NOT_NULL
@item RE_DOT_NOT_NULL
If this bit is set, then the match-any-character operator doesn't match
a null character; if this bit isn't set, then it does.

@cnindex RE_HAT_LISTS_NOT_NEWLINE
@item RE_HAT_LISTS_NOT_NEWLINE
If this bit is set, nonmatching lists @samp{[^...]} do not match
newline; if not set, they do.

@cnindex RE_ICASE
@item RE_ICASE
If this bit is set, then ignore case when matching; otherwise, case is
significant.

@cnindex RE_INTERVALS
@item RE_INTERVALS
If this bit is set, then Regex recognizes interval operators; if this bit
isn't set, then it doesn't.

@cnindex RE_INVALID_INTERVAL_ORD
@item RE_INVALID_INTERVAL_ORD
If this bit is set, a syntactically invalid interval is treated as a
string of ordinary characters.  For example, the extended regular
expression @samp{a@{1} is treated as @samp{a\@{1}.

@cnindex RE_LIMITED_OPS
@item RE_LIMITED_OPS
If this bit is set, then Regex doesn't recognize the match-one-or-more,
match-zero-or-one or alternation operators; if this bit isn't set, then
it does.

@cnindex RE_NEWLINE_ALT
@item RE_NEWLINE_ALT
If this bit is set, then newline represents the alternation operator; if
this bit isn't set, then newline is ordinary.

@cnindex RE_NO_BK_BRACES
@item RE_NO_BK_BRACES
If this bit is set, then @samp{@{} represents the open-interval operator
and @samp{@}} represents the close-interval operator; if this bit isn't
set, then @samp{\@{} represents the open-interval operator and
@samp{\@}} represents the close-interval operator.  This bit is relevant
only if @code{RE_INTERVALS} is set.

@cnindex RE_NO_BK_PARENS
@item RE_NO_BK_PARENS
If this bit is set, then @samp{(} represents the open-group operator and
@samp{)} represents the close-group operator; if this bit isn't set, then
@samp{\(} represents the open-group operator and @samp{\)} represents
the close-group operator.

@cnindex RE_NO_BK_REFS
@item RE_NO_BK_REFS
If this bit is set, then Regex doesn't recognize @samp{\}@var{digit} as
the back-reference operator; if this bit isn't set, then it does.

@cnindex RE_NO_BK_VBAR
@item RE_NO_BK_VBAR
If this bit is set, then @samp{|} represents the alternation operator;
if this bit isn't set, then @samp{\|} represents the alternation
operator.  This bit is irrelevant if @code{RE_LIMITED_OPS} is set.

@cnindex RE_NO_EMPTY_RANGES
@item RE_NO_EMPTY_RANGES
If this bit is set, then a regular expression with a range whose ending
point collates lower than its starting point is invalid; if this bit
isn't set, then Regex considers such a range to be empty.

@cnindex RE_NO_GNU_OPS
@item RE_NO_GNU_OPS
If this bit is set, GNU regex operators are not recognized; otherwise,
they are.

@cnindex RE_NO_POSIX_BACKTRACKING
@item RE_NO_POSIX_BACKTRACKING
If this bit is set, succeed as soon as we match the whole pattern,
without further backtracking.  This means that a match may not be
the leftmost longest; @pxref{What Gets Matched?} for what this means.

@cnindex RE_NO_SUB
@item RE_NO_SUB
If this bit is set, then @code{no_sub} will be set to one during
@code{re_compile_pattern}.  This causes matching and searching routines
not to record substring match information.

@cnindex RE_UNMATCHED_RIGHT_PAREN_ORD
@item RE_UNMATCHED_RIGHT_PAREN_ORD
If this bit is set and the regular expression has no matching open-group
operator, then Regex considers what would otherwise be a close-group
operator (based on how @code{RE_NO_BK_PARENS} is set) to match @samp{)}.

@end table


@node Predefined Syntaxes
@section Predefined Syntaxes

If you're programming with Regex, you can set a pattern buffer's
(@pxref{GNU Pattern Buffers})
syntax either to an arbitrary combination of syntax bits
(@pxref{Syntax Bits}) or else to the configurations defined by Regex.
These configurations define the syntaxes used by certain
programs---GNU Emacs,
@cindex Emacs
POSIX Awk,
@cindex POSIX Awk
traditional Awk,
@cindex Awk
Grep,
@cindex Grep
@cindex Egrep
Egrep---in addition to syntaxes for POSIX basic and extended
regular expressions.

The predefined syntaxes---taken directly from @file{regex.h}---are:

@smallexample
#define RE_SYNTAX_EMACS 0

#define RE_SYNTAX_AWK                                                   \
  (RE_BACKSLASH_ESCAPE_IN_LISTS | RE_DOT_NOT_NULL                       \
   | RE_NO_BK_PARENS            | RE_NO_BK_REFS                         \
   | RE_NO_BK_VBAR               | RE_NO_EMPTY_RANGES                   \
   | RE_UNMATCHED_RIGHT_PAREN_ORD)

#define RE_SYNTAX_POSIX_AWK                                             \
  (RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS)

#define RE_SYNTAX_GREP                                                  \
  (RE_BK_PLUS_QM              | RE_CHAR_CLASSES                         \
   | RE_HAT_LISTS_NOT_NEWLINE | RE_INTERVALS                            \
   | RE_NEWLINE_ALT)

#define RE_SYNTAX_EGREP                                                 \
  (RE_CHAR_CLASSES        | RE_CONTEXT_INDEP_ANCHORS                    \
   | RE_CONTEXT_INDEP_OPS | RE_HAT_LISTS_NOT_NEWLINE                    \
   | RE_NEWLINE_ALT       | RE_NO_BK_PARENS                             \
   | RE_NO_BK_VBAR)

#define RE_SYNTAX_POSIX_EGREP                                           \
  (RE_SYNTAX_EGREP | RE_INTERVALS | RE_NO_BK_BRACES)

/* P1003.2/D11.2, section 4.20.7.1, lines 5078ff.  */
#define RE_SYNTAX_ED RE_SYNTAX_POSIX_BASIC

#define RE_SYNTAX_SED RE_SYNTAX_POSIX_BASIC

/* Syntax bits common to both basic and extended POSIX regex syntax.  */
#define _RE_SYNTAX_POSIX_COMMON                                         \
  (RE_CHAR_CLASSES | RE_DOT_NEWLINE      | RE_DOT_NOT_NULL              \
   | RE_INTERVALS  | RE_NO_EMPTY_RANGES)

#define RE_SYNTAX_POSIX_BASIC                                           \
  (_RE_SYNTAX_POSIX_COMMON | RE_BK_PLUS_QM)

/* Differs from ..._POSIX_BASIC only in that RE_BK_PLUS_QM becomes
   RE_LIMITED_OPS, i.e., \? \+ \| are not recognized.  Actually, this
   isn't minimal, since other operators, such as \`, aren't disabled.  */
#define RE_SYNTAX_POSIX_MINIMAL_BASIC                                   \
  (_RE_SYNTAX_POSIX_COMMON | RE_LIMITED_OPS)

#define RE_SYNTAX_POSIX_EXTENDED                                        \
  (_RE_SYNTAX_POSIX_COMMON | RE_CONTEXT_INDEP_ANCHORS                   \
   | RE_CONTEXT_INDEP_OPS  | RE_NO_BK_BRACES                            \
   | RE_NO_BK_PARENS       | RE_NO_BK_VBAR                              \
   | RE_UNMATCHED_RIGHT_PAREN_ORD)

/* Differs from ..._POSIX_EXTENDED in that RE_CONTEXT_INVALID_OPS
   replaces RE_CONTEXT_INDEP_OPS and RE_NO_BK_REFS is added.  */
#define RE_SYNTAX_POSIX_MINIMAL_EXTENDED                                \
  (_RE_SYNTAX_POSIX_COMMON  | RE_CONTEXT_INDEP_ANCHORS                  \
   | RE_CONTEXT_INVALID_OPS | RE_NO_BK_BRACES                           \
   | RE_NO_BK_PARENS        | RE_NO_BK_REFS                             \
   | RE_NO_BK_VBAR          | RE_UNMATCHED_RIGHT_PAREN_ORD)
@end smallexample

@node Collating Elements vs. Characters
@section Collating Elements vs.@: Characters

POSIX generalizes the notion of a character to that of a
collating element.  It defines a @dfn{collating element} to be ``a
sequence of one or more bytes defined in the current collating sequence
as a unit of collation.''

This generalizes the notion of a character in
two ways.  First, a single character can map into two or more collating
elements.  For example, the German ``ß''
collates as the collating element @samp{s} followed by another collating
element @samp{s}.  Second, two or more characters can map into one
collating element.  For example, the Czech @samp{ch} collates after
@samp{h} and before @samp{i}.

Since POSIX's ``collating element'' preserves the essential idea of
a ``character,'' we use the latter, more familiar, term in this document.

@node The Backslash Character
@section The Backslash Character

@cindex \
The @samp{\} character has one of four different meanings, depending on
the context in which you use it and what syntax bits are set
(@pxref{Syntax Bits}).  It can: 1) stand for itself, 2) quote the next
character, 3) introduce an operator, or 4) do nothing.

@enumerate
@item
It stands for itself inside a list
(@pxref{List Operators}) if the syntax bit
@code{RE_BACKSLASH_ESCAPE_IN_LISTS} is not set.  For example, @samp{[\]}
would match @samp{\}.

@item
It quotes (makes ordinary, if it's special) the next character when you
use it either:

@itemize @bullet
@item
outside a list,@footnote{Sometimes
you don't have to explicitly quote special characters to make
them ordinary.  For instance, most characters lose any special meaning
inside a list (@pxref{List Operators}).  In addition, if the syntax bits
@code{RE_CONTEXT_INVALID_OPS} and @code{RE_CONTEXT_INDEP_OPS}
aren't set, then (for historical reasons) the matcher considers special
characters ordinary if they are in contexts where the operations they
represent make no sense; for example, then the match-zero-or-more
operator (represented by @samp{*}) matches itself in the regular
expression @samp{*foo} because there is no preceding expression on which
it can operate.  It is poor practice, however, to depend on this
behavior; if you want a special character to be ordinary outside a list,
it's better to always quote it, regardless.} or

@item
inside a list and the syntax bit @code{RE_BACKSLASH_ESCAPE_IN_LISTS} is set.

@end itemize

@item
It introduces an operator when followed by certain ordinary
characters---sometimes only when certain syntax bits are set.  See the
cases @code{RE_BK_PLUS_QM}, @code{RE_NO_BK_BRACES}, @code{RE_NO_BK_VAR},
@code{RE_NO_BK_PARENS}, @code{RE_NO_BK_REF} in @ref{Syntax Bits}.  Also:

@itemize @bullet
@item
@samp{\b} represents the match-word-boundary operator
(@pxref{Match-word-boundary Operator}).

@item
@samp{\B} represents the match-within-word operator
(@pxref{Match-within-word Operator}).

@item
@samp{\<} represents the match-beginning-of-word operator @*
(@pxref{Match-beginning-of-word Operator}).

@item
@samp{\>} represents the match-end-of-word operator
(@pxref{Match-end-of-word Operator}).

@item
@samp{\w} represents the match-word-constituent operator
(@pxref{Match-word-constituent Operator}).

@item
@samp{\W} represents the match-non-word-constituent operator
(@pxref{Match-non-word-constituent Operator}).

@item
@samp{\s@var{class}} is equivalent to @code{[[:space:]]}
(@pxref{Match-space Operator}).

@item
@samp{\S@var{class}} is equivalent to @code{[^[:space]]}
(@pxref{Match-non-space Operator}).

@item
@samp{\`} represents the match-beginning-of-string
operator and @samp{\'} represents the match-end-of-string operator
(@pxref{Whole-string Operators}).

@end itemize

@item
In all other cases, Regex ignores @samp{\}.  For example,
@samp{\n} matches @samp{n}.

@end enumerate

@node Common Operators
@chapter Common Operators

You compose regular expressions from operators.  In the following
sections, we describe the regular expression operators specified by
POSIX; GNU also uses these.  Most operators have more than one
representation as characters.  @xref{Regular Expression Syntax}, for
what characters represent what operators under what circumstances.

For most operators that can be represented in two ways, one
representation is a single character and the other is that character
preceded by @samp{\}.  For example, either @samp{(} or @samp{\(}
represents the open-group operator.  Which one does depends on the
setting of a syntax bit, in this case @code{RE_NO_BK_PARENS}.  Why is
this so?  Historical reasons dictate some of the varying
representations, while POSIX dictates others.

Finally, almost all characters lose any special meaning inside a list
(@pxref{List Operators}).

@menu
* Match-self Operator::                 Ordinary characters.
* Match-any-character Operator::        .
* Concatenation Operator::              Juxtaposition.
* Repetition Operators::                *  +  ? @{@}
* Alternation Operator::                |
* List Operators::                      [...]  [^...]
* Grouping Operators::                  (...)
* Back-reference Operator::             \digit
* Anchoring Operators::                 ^  $
@end menu

@node Match-self Operator
@section The Match-self Operator (@var{ordinary character})

This operator matches the character itself.  All ordinary characters
(@pxref{Regular Expression Syntax}) represent this operator.  For
example, @samp{f} is always an ordinary character, so the regular
expression @samp{f} matches only the string @samp{f}.  In
particular, it does @emph{not} match the string @samp{ff}.

@node Match-any-character Operator
@section The Match-any-character Operator (@code{.})

@cindex @samp{.}

This operator matches any single printing or nonprinting character
except it won't match a:

@table @asis
@item newline
if the syntax bit @code{RE_DOT_NEWLINE} isn't set.

@item null
if the syntax bit @code{RE_DOT_NOT_NULL} is set.

@end table

The @samp{.} (period) character represents this operator.  For example,
@samp{a.b} matches any three-character string beginning with @samp{a}
and ending with @samp{b}.

@node Concatenation Operator
@section The Concatenation Operator

This operator concatenates two regular expressions @var{a} and @var{b}.
No character represents this operator; you simply put @var{b} after
@var{a}.  The result is a regular expression that will match a string if
@var{a} matches its first part and @var{b} matches the rest.  For
example, @samp{xy} (two match-self operators) matches @samp{xy}.

@node Repetition Operators
@section Repetition Operators

Repetition operators repeat the preceding regular expression a specified
number of times.

@menu
* Match-zero-or-more Operator::  *
* Match-one-or-more Operator::   +
* Match-zero-or-one Operator::   ?
* Interval Operators::           @{@}
@end menu

@node Match-zero-or-more Operator
@subsection The Match-zero-or-more Operator (@code{*})

@cindex @samp{*}

This operator repeats the smallest possible preceding regular expression
as many times as necessary (including zero) to match the pattern.
@samp{*} represents this operator.  For example, @samp{o*}
matches any string made up of zero or more @samp{o}s.  Since this
operator operates on the smallest preceding regular expression,
@samp{fo*} has a repeating @samp{o}, not a repeating @samp{fo}.  So,
@samp{fo*} matches @samp{f}, @samp{fo}, @samp{foo}, and so on.

Since the match-zero-or-more operator is a suffix operator, it may be
useless as such when no regular expression precedes it.  This is the
case when it:

@itemize @bullet
@item
is first in a regular expression, or

@item
follows a match-beginning-of-line, open-group, or alternation
operator.

@end itemize

@noindent
Three different things can happen in these cases:

@enumerate
@item
If the syntax bit @code{RE_CONTEXT_INVALID_OPS} is set, then the
regular expression is invalid.

@item
If @code{RE_CONTEXT_INVALID_OPS} isn't set, but
@code{RE_CONTEXT_INDEP_OPS} is, then @samp{*} represents the
match-zero-or-more operator (which then operates on the empty string).

@item
Otherwise, @samp{*} is ordinary.

@end enumerate

@cindex backtracking
The matcher processes a match-zero-or-more operator by first matching as
many repetitions of the smallest preceding regular expression as it can.
Then it continues to match the rest of the pattern.

If it can't match the rest of the pattern, it backtracks (as many times
as necessary), each time discarding one of the matches until it can
either match the entire pattern or be certain that it cannot get a
match.  For example, when matching @samp{ca*ar} against @samp{caaar},
the matcher first matches all three @samp{a}s of the string with the
@samp{a*} of the regular expression.  However, it cannot then match the
final @samp{ar} of the regular expression against the final @samp{r} of
the string.  So it backtracks, discarding the match of the last @samp{a}
in the string.  It can then match the remaining @samp{ar}.


@node Match-one-or-more Operator
@subsection The Match-one-or-more Operator (@code{+} or @code{\+})

@cindex @samp{+}

If the syntax bit @code{RE_LIMITED_OPS} is set, then Regex doesn't recognize
this operator.  Otherwise, if the syntax bit @code{RE_BK_PLUS_QM} isn't
set, then @samp{+} represents this operator; if it is, then @samp{\+}
does.

This operator is similar to the match-zero-or-more operator except that
it repeats the preceding regular expression at least once;
@pxref{Match-zero-or-more Operator}, for what it operates on, how some
syntax bits affect it, and how Regex backtracks to match it.

For example, supposing that @samp{+} represents the match-one-or-more
operator; then @samp{ca+r} matches, e.g., @samp{car} and
@samp{caaaar}, but not @samp{cr}.

@node Match-zero-or-one Operator
@subsection The Match-zero-or-one Operator (@code{?} or @code{\?})
@cindex @samp{?}

If the syntax bit @code{RE_LIMITED_OPS} is set, then Regex doesn't
recognize this operator.  Otherwise, if the syntax bit
@code{RE_BK_PLUS_QM} isn't set, then @samp{?} represents this operator;
if it is, then @samp{\?} does.

This operator is similar to the match-zero-or-more operator except that
it repeats the preceding regular expression once or not at all;
@pxref{Match-zero-or-more Operator}, to see what it operates on, how
some syntax bits affect it, and how Regex backtracks to match it.

For example, supposing that @samp{?} represents the match-zero-or-one
operator; then @samp{ca?r} matches both @samp{car} and @samp{cr}, but
nothing else.

@node Interval Operators
@subsection Interval Operators (@code{@{} @dots{} @code{@}} or @code{\@{} @dots{} @code{\@}})

@cindex interval expression
@cindex @samp{@{}
@cindex @samp{@}}
@cindex @samp{\@{}
@cindex @samp{\@}}

If the syntax bit @code{RE_INTERVALS} is set, then Regex recognizes
@dfn{interval expressions}.  They repeat the smallest possible preceding
regular expression a specified number of times.

If the syntax bit @code{RE_NO_BK_BRACES} is set, @samp{@{} represents
the @dfn{open-interval operator} and @samp{@}} represents the
@dfn{close-interval operator} ; otherwise, @samp{\@{} and @samp{\@}} do.

Specifically, supposing that @samp{@{} and @samp{@}} represent the
open-interval and close-interval operators; then:

@table @code
@item  @{@var{count}@}
matches exactly @var{count} occurrences of the preceding regular
expression.

@item @{@var{min},@}
matches @var{min} or more occurrences of the preceding regular
expression.

@item  @{@var{min}, @var{max}@}
matches at least @var{min} but no more than @var{max} occurrences of
the preceding regular expression.

@end table

The interval expression (but not necessarily the regular expression that
contains it) is invalid if:

@itemize @bullet
@item
@var{min} is greater than @var{max}, or

@item
any of @var{count}, @var{min}, or @var{max} are outside the range
zero to @code{RE_DUP_MAX} (which symbol @file{regex.h}
defines).

@end itemize

If the interval expression is invalid and the syntax bit
@code{RE_NO_BK_BRACES} is set, then Regex considers all the
characters in the would-be interval to be ordinary.  If that bit
isn't set, then the regular expression is invalid.

If the interval expression is valid but there is no preceding regular
expression on which to operate, then if the syntax bit
@code{RE_CONTEXT_INVALID_OPS} is set, the regular expression is invalid.
If that bit isn't set, then Regex considers all the characters---other
than backslashes, which it ignores---in the would-be interval to be
ordinary.


@node Alternation Operator
@section The Alternation Operator (@code{|} or @code{\|})

@kindex |
@kindex \|
@cindex alternation operator
@cindex or operator

If the syntax bit @code{RE_LIMITED_OPS} is set, then Regex doesn't
recognize this operator.  Otherwise, if the syntax bit
@code{RE_NO_BK_VBAR} is set, then @samp{|} represents this operator;
otherwise, @samp{\|} does.

Alternatives match one of a choice of regular expressions:
if you put the character(s) representing the alternation operator between
any two regular expressions @var{a} and @var{b}, the result matches
the union of the strings that @var{a} and @var{b} match.  For
example, supposing that @samp{|} is the alternation operator, then
@samp{foo|bar|quux} would match any of @samp{foo}, @samp{bar} or
@samp{quux}.

The alternation operator operates on the @emph{largest} possible
surrounding regular expressions.  (Put another way, it has the lowest
precedence of any regular expression operator.)
Thus, the only way you can
delimit its arguments is to use grouping.  For example, if @samp{(} and
@samp{)} are the open and close-group operators, then @samp{fo(o|b)ar}
would match either @samp{fooar} or @samp{fobar}.  (@samp{foo|bar} would
match @samp{foo} or @samp{bar}.)

@cindex backtracking
The matcher usually tries all combinations of alternatives so as to
match the longest possible string.  For example, when matching
@samp{(fooq|foo)*(qbarquux|bar)} against @samp{fooqbarquux}, it cannot
take, say, the first (``depth-first'') combination it could match, since
then it would be content to match just @samp{fooqbar}.

Note that since the default behavior is to return the leftmost longest
match, when more than one of a series of alternatives matches the actual
match will be the longest matching alternative, not necessarily the
first in the list.


@node List Operators
@section List Operators (@code{[} @dots{} @code{]} and @code{[^} @dots{} @code{]})

@cindex matching list
@cindex @samp{[}
@cindex @samp{]}
@cindex @samp{^}
@cindex @samp{-}
@cindex @samp{\}
@cindex @samp{[^}
@cindex nonmatching list
@cindex matching newline
@cindex bracket expression

@dfn{Lists}, also called @dfn{bracket expressions}, are a set of one or
more items.  An @dfn{item} is a character,
a collating symbol, an equivalence class expression,
a character class expression, or a range expression.  The syntax bits
affect which kinds of items you can put in a list.  We explain the last
four items in subsections below.  Empty lists are invalid.

A @dfn{matching list} matches a single character represented by one of
the list items.  You form a matching list by enclosing one or more items
within an @dfn{open-matching-list operator} (represented by @samp{[})
and a @dfn{close-list operator} (represented by @samp{]}).

For example, @samp{[ab]} matches either @samp{a} or @samp{b}.
@samp{[ad]*} matches the empty string and any string composed of just
@samp{a}s and @samp{d}s in any order.  Regex considers invalid a regular
expression with a @samp{[} but no matching
@samp{]}.

@dfn{Nonmatching lists} are similar to matching lists except that they
match a single character @emph{not} represented by one of the list
items.  You use an @dfn{open-nonmatching-list operator} (represented by
@samp{[^}@footnote{Regex therefore doesn't consider the @samp{^} to be
the first character in the list.  If you put a @samp{^} character first
in (what you think is) a matching list, you'll turn it into a
nonmatching list.}) instead of an open-matching-list operator to start a
nonmatching list.

For example, @samp{[^ab]} matches any character except @samp{a} or
@samp{b}.

If the syntax bit @code{RE_HAT_LISTS_NOT_NEWLINE} is set, then
nonmatching lists do not match a newline.

Most characters lose any special meaning inside a list.  The special
characters inside a list follow.

@table @samp
@item ]
ends the list if it's not the first list item.  So, if you want to make
the @samp{]} character a list item, you must put it first.

@item \
quotes the next character if the syntax bit @code{RE_BACKSLASH_ESCAPE_IN_LISTS} is
set.

@item [.
represents the open-collating-symbol operator (@pxref{Collating Symbol
Operators}).

@item .]
represents the close-collating-symbol operator.

@item [=
represents the open-equivalence-class operator (@pxref{Equivalence Class
Operators}).

@item =]
represents the close-equivalence-class operator.

@item [:
represents the open-character-class operator (@pxref{Character Class
Operators}) if the syntax bit @code{RE_CHAR_CLASSES} is set and what
follows is a valid character class expression.

@item :]
represents the close-character-class operator if the syntax bit
@code{RE_CHAR_CLASSES} is set and what precedes it is an
open-character-class operator followed by a valid character class name.

@item -
represents the range operator (@pxref{Range Operator}) if it's
not first or last in a list or the ending point of a range.

@end table

@noindent
All other characters are ordinary.  For example, @samp{[.*]} matches
@samp{.} and @samp{*}.

@menu
* Collating Symbol Operators::  [.elem.]
* Equivalence Class Operators:: [=class=]
* Character Class Operators::   [:class:]
* Range Operator::          start-end
@end menu


@node Collating Symbol Operators
@subsection Collating Symbol Operators (@code{[.} @dots{} @code{.]})

Collating symbols can be represented inside lists.
You form a @dfn{collating symbol} by
putting a collating element between an @dfn{open-collating-symbol
operator} and a @dfn{close-collating-symbol operator}.  @samp{[.}
represents the open-collating-symbol operator and @samp{.]} represents
the close-collating-symbol operator.  For example, if @samp{ll} is a
collating element, then @samp{[[.ll.]]} would match @samp{ll}.

@node Equivalence Class Operators
@subsection Equivalence Class Operators (@code{[=} @dots{} @code{=]})
@cindex equivalence class expression in regex
@cindex @samp{[=} in regex
@cindex @samp{=]} in regex

Regex recognizes equivalence class
expressions inside lists.  A @dfn{equivalence class expression} is a set
of collating elements which all belong to the same equivalence class.
You form an equivalence class expression by putting a collating
element between an @dfn{open-equivalence-class operator} and a
@dfn{close-equivalence-class operator}.  @samp{[=} represents the
open-equivalence-class operator and @samp{=]} represents the
close-equivalence-class operator.  For example, if @samp{a} and @samp{A}
were an equivalence class, then both @samp{[[=a=]]} and @samp{[[=A=]]}
would match both @samp{a} and @samp{A}.  If the collating element in an
equivalence class expression isn't part of an equivalence class, then
the matcher considers the equivalence class expression to be a collating
symbol.

@node Character Class Operators
@subsection Character Class Operators (@code{[:} @dots{} @code{:]})

@cindex character classes
@cindex @samp{[colon} in regex
@cindex @samp{colon]} in regex

If the syntax bit @code{RE_CHAR_CLASSES} is set, then Regex recognizes
character class expressions inside lists.  A @dfn{character class
expression} matches one character from a given class.  You form a
character class expression by putting a character class name between
an @dfn{open-character-class operator} (represented by @samp{[:}) and
a @dfn{close-character-class operator} (represented by @samp{:]}).
The character class names and their meanings are:

@table @code

@item alnum
letters and digits

@item alpha
letters

@item blank
system-dependent; for GNU, a space or tab

@item cntrl
control characters (in the ASCII encoding, code 0177 and codes
less than 040)

@item digit
digits

@item graph
same as @code{print} except omits space

@item lower
lowercase letters

@item print
printable characters (in the ASCII encoding, space
tilde---codes 040 through 0176)

@item punct
neither control nor alphanumeric characters

@item space
space, carriage return, newline, vertical tab, and form feed

@item upper
uppercase letters

@item xdigit
hexadecimal digits: @code{0}--@code{9}, @code{a}--@code{f}, @code{A}--@code{F}

@end table

@noindent
These correspond to the definitions in the C library's @file{<ctype.h>}
facility.  For example, @samp{[:alpha:]} corresponds to the standard
facility @code{isalpha}.  Regex recognizes character class expressions
only inside of lists; so @samp{[[:alpha:]]} matches any letter, but
@samp{[:alpha:]} outside of a bracket expression and not followed by a
repetition operator matches just itself.

@node Range Operator
@subsection The Range Operator (@code{-})

Regex recognizes @dfn{range expressions} inside a list. They represent
those characters
that fall between two elements in the current collating sequence.  You
form a range expression by putting a @dfn{range operator} between two
of any of the following: characters, collating elements, collating symbols,
and equivalence class expressions.  The starting point of the range and
the ending point of the range don't have to be the same kind of item,
e.g., the starting point could be a collating element and the ending
point could be an equivalence class expression.  If a range's ending
point is an equivalence class, then all the collating elements in that
class will be in the range.@footnote{You can't use a character class for the starting
or ending point of a range, since a character class is not a single
character.} @samp{-} represents the range operator.  For example,
@samp{a-f} within a list represents all the characters from @samp{a}
through @samp{f}
inclusively.

If the syntax bit @code{RE_NO_EMPTY_RANGES} is set, then if the range's
ending point collates less than its starting point, the range (and the
regular expression containing it) is invalid.  For example, the regular
expression @samp{[z-a]} would be invalid.  If this bit isn't set, then
Regex considers such a range to be empty.

Since @samp{-} represents the range operator, if you want to make a
@samp{-} character itself
a list item, you must do one of the following:

@itemize @bullet
@item
Put the @samp{-} either first or last in the list.

@item
Include a range whose starting point collates strictly lower than
@samp{-} and whose ending point collates equal or higher.  Unless a
range is the first item in a list, a @samp{-} can't be its starting
point, but @emph{can} be its ending point.  That is because Regex
considers @samp{-} to be the range operator unless it is preceded by
another @samp{-}.  For example, in the ASCII encoding, @samp{)},
@samp{*}, @samp{+}, @samp{,}, @samp{-}, @samp{.}, and @samp{/} are
contiguous characters in the collating sequence.  You might think that
@samp{[)-+--/]} has two ranges: @samp{)-+} and @samp{--/}.  Rather, it
has the ranges @samp{)-+} and @samp{+--}, plus the character @samp{/}, so
it matches, e.g., @samp{,}, not @samp{.}.

@item
Put a range whose starting point is @samp{-} first in the list.

@end itemize

For example, @samp{[-a-z]} matches a lowercase letter or a hyphen (in
English, in ASCII).


@node Grouping Operators
@section Grouping Operators (@code{(} @dots{} @code{)} or @code{\(} @dots{} @code{\)})

@kindex (
@kindex )
@kindex \(
@kindex \)
@cindex grouping
@cindex subexpressions
@cindex parenthesizing

A @dfn{group}, also known as a @dfn{subexpression}, consists of an
@dfn{open-group operator}, any number of other operators, and a
@dfn{close-group operator}.  Regex treats this sequence as a unit, just
as mathematics and programming languages treat a parenthesized
expression as a unit.

Therefore, using @dfn{groups}, you can:

@itemize @bullet
@item
delimit the argument(s) to an alternation operator (@pxref{Alternation
Operator}) or a repetition operator (@pxref{Repetition
Operators}).

@item
keep track of the indices of the substring that matched a given group.
@xref{Using Registers}, for a precise explanation.
This lets you:

@itemize @bullet
@item
use the back-reference operator (@pxref{Back-reference Operator}).

@item
use registers (@pxref{Using Registers}).

@end itemize

@end itemize

If the syntax bit @code{RE_NO_BK_PARENS} is set, then @samp{(} represents
the open-group operator and @samp{)} represents the
close-group operator; otherwise, @samp{\(} and @samp{\)} do.

If the syntax bit @code{RE_UNMATCHED_RIGHT_PAREN_ORD} is set and a
close-group operator has no matching open-group operator, then Regex
considers it to match @samp{)}.


@node Back-reference Operator
@section The Back-reference Operator (@dfn{\}@var{digit})

@cindex back-references

If the syntax bit @code{RE_NO_BK_REF} isn't set, then Regex recognizes
back-references.  A back-reference matches a specified preceding group.
The back-reference operator is represented by @samp{\@var{digit}}
anywhere after the end of a regular expression's @w{@var{digit}-th}
group (@pxref{Grouping Operators}).

@var{digit} must be between @samp{1} and @samp{9}.  The matcher assigns
numbers 1 through 9 to the first nine groups it encounters.  By using
one of @samp{\1} through @samp{\9} after the corresponding group's
close-group operator, you can match a substring identical to the
one that the group does.

Back-references match according to the following (in all examples below,
@samp{(} represents the open-group, @samp{)} the close-group, @samp{@{}
the open-interval and @samp{@}} the close-interval operator):

@itemize @bullet
@item
If the group matches a substring, the back-reference matches an
identical substring.  For example, @samp{(a)\1} matches @samp{aa} and
@samp{(bana)na\1bo\1} matches @samp{bananabanabobana}.  Likewise,
@samp{(.*)\1} matches any (newline-free if the syntax bit
@code{RE_DOT_NEWLINE} isn't set) string that is composed of two
identical halves; the @samp{(.*)} matches the first half and the
@samp{\1} matches the second half.

@item
If the group matches more than once (as it might if followed
by, e.g., a repetition operator), then the back-reference matches the
substring the group @emph{last} matched.  For example,
@samp{((a*)b)*\1\2} matches @samp{aabababa}; first @w{group 1} (the
outer one) matches @samp{aab} and @w{group 2} (the inner one) matches
@samp{aa}.  Then @w{group 1} matches @samp{ab} and @w{group 2} matches
@samp{a}.  So, @samp{\1} matches @samp{ab} and @samp{\2} matches
@samp{a}.

@item
If the group doesn't participate in a match, i.e., it is part of an
alternative not taken or a repetition operator allows zero repetitions
of it, then the back-reference makes the whole match fail.  For example,
@samp{(one()|two())-and-(three\2|four\3)} matches @samp{one-and-three}
and @samp{two-and-four}, but not @samp{one-and-four} or
@samp{two-and-three}.  For example, if the pattern matches
@samp{one-and-}, then its @w{group 2} matches the empty string and its
@w{group 3} doesn't participate in the match.  So, if it then matches
@samp{four}, then when it tries to back-reference @w{group 3}---which it
will attempt to do because @samp{\3} follows the @samp{four}---the match
will fail because @w{group 3} didn't participate in the match.

@end itemize

You can use a back-reference as an argument to a repetition operator.  For
example, @samp{(a(b))\2*} matches @samp{a} followed by two or more
@samp{b}s.  Similarly, @samp{(a(b))\2@{3@}} matches @samp{abbbb}.

If there is no preceding @w{@var{digit}-th} subexpression, the regular
expression is invalid.

Back-references can greatly slow down matching, as they can generate
exponentially many matching possibilities that can consume both time
and memory to explore.  Also, the POSIX specification for
back-references is at times unclear.  Furthermore, many regular
expression implementations have back-reference bugs that can cause
programs to return incorrect answers or even crash, and fixing these
bugs has often been low-priority: for example, as of 2020 the
@url{https://sourceware.org/bugzilla/,GNU C library bug database}
contained back-reference bugs
@url{https://sourceware.org/bugzilla/show_bug.cgi?id=52,,52},
@url{https://sourceware.org/bugzilla/show_bug.cgi?id=10844,,10844},
@url{https://sourceware.org/bugzilla/show_bug.cgi?id=11053,,11053},
@url{https://sourceware.org/bugzilla/show_bug.cgi?id=24269,,24269}
and @url{https://sourceware.org/bugzilla/show_bug.cgi?id=25322,,25322},
with little sign of forthcoming fixes.  Luckily,
back-references are rarely useful and it should be little trouble to
avoid them in practical applications.


@node Anchoring Operators
@section Anchoring Operators

@cindex anchoring
@cindex regexp anchoring

These operators can constrain a pattern to match only at the beginning or
end of the entire string or at the beginning or end of a line.

@menu
* Match-beginning-of-line Operator::  ^
* Match-end-of-line Operator::        $
@end menu


@node Match-beginning-of-line Operator
@subsection The Match-beginning-of-line Operator (@code{^})

@kindex ^
@cindex beginning-of-line operator
@cindex anchors

This operator can match the empty string either at the beginning of the
string or after a newline character.  Thus, it is said to @dfn{anchor}
the pattern to the beginning of a line.

In the cases following, @samp{^} represents this operator.  (Otherwise,
@samp{^} is ordinary.)

@itemize @bullet

@item
It (the @samp{^}) is first in the pattern, as in @samp{^foo}.

@cnindex RE_CONTEXT_INDEP_ANCHORS @r{(and @samp{^})}
@item
The syntax bit @code{RE_CONTEXT_INDEP_ANCHORS} is set, and it is outside
a bracket expression.

@cindex open-group operator and @samp{^}
@cindex alternation operator and @samp{^}
@item
It follows an open-group or alternation operator, as in @samp{a\(^b\)}
and @samp{a\|^b}.  @xref{Grouping Operators}, and @ref{Alternation
Operator}.

@end itemize

These rules imply that some valid patterns containing @samp{^} cannot be
matched; for example, @samp{foo^bar} if @code{RE_CONTEXT_INDEP_ANCHORS}
is set.

@vindex not_bol @r{field in pattern buffer}
If the @code{not_bol} field is set in the pattern buffer (@pxref{GNU
Pattern Buffers}), then @samp{^} fails to match at the beginning of the
string.  This lets you match against pieces of a line, as you would need to if,
say, searching for repeated instances of a given pattern in a line; it
would work correctly for patterns both with and without
match-beginning-of-line operators.


@node Match-end-of-line Operator
@subsection The Match-end-of-line Operator (@code{$})

@kindex $
@cindex end-of-line operator
@cindex anchors

This operator can match the empty string either at the end of
the string or before a newline character in the string.  Thus, it is
said to @dfn{anchor} the pattern to the end of a line.

It is always represented by @samp{$}.  For example, @samp{foo$} usually
matches, e.g., @samp{foo} and, e.g., the first three characters of
@samp{foo\nbar}.

Its interaction with the syntax bits and pattern buffer fields is
exactly the dual of @samp{^}'s; see the previous section.  (That is,
``@samp{^}'' becomes ``@samp{$}'', ``beginning'' becomes ``end'',
``next'' becomes ``previous'', ``after'' becomes ``before'', and
``@code{not_bol}'' becomes ``@code{not_eol}''.)


@node GNU Operators
@chapter GNU Operators

The following are operators that GNU defines (and POSIX doesn't) that
you can use unless the syntax bit @code{RE_NO_GNU_OPS} is set.

@menu
* Word Operators::
* Space Operators::
* Whole-string Operators::
@end menu

@node Word Operators
@section Word Operators

The operators in this section require Regex to recognize parts of words.
Characters that are part of words, which are called
@dfn{word-constituent}, are letters, digits, and the underscore
(@samp{_}); more precisely, any character in the POSIX class
@code{alnum} in the current locale, or underscore.

@menu
* Match-word-boundary Operator::        \b
* Match-within-word Operator::          \B
* Match-beginning-of-word Operator::    \<
* Match-end-of-word Operator::          \>
* Match-word-constituent Operator::     \w
* Match-non-word-constituent Operator:: \W
@end menu

@node Match-word-boundary Operator
@subsection The Match-word-boundary Operator (@code{\b})

@cindex @samp{\b}
@cindex word boundaries, matching

This operator (represented by @samp{\b}) matches the empty string at
either the beginning or the end of a word.  For example, @samp{\brat\b}
matches the separate word @samp{rat}.

@node Match-within-word Operator
@subsection The Match-within-word Operator (@code{\B})

@cindex @samp{\B}

This operator (represented by @samp{\B}) matches the empty string within
a word. For example, @samp{c\Brat\Be} matches @samp{crate}, but
@samp{dirty \Brat} doesn't match @samp{dirty rat}.

@node Match-beginning-of-word Operator
@subsection The Match-beginning-of-word Operator (@code{\<})

@cindex @samp{\<}

This operator (represented by @samp{\<}) matches the empty string at the
beginning of a word.

@node Match-end-of-word Operator
@subsection The Match-end-of-word Operator (@code{\>})

@cindex @samp{\>}

This operator (represented by @samp{\>}) matches the empty string at the
end of a word.

@node Match-word-constituent Operator
@subsection The Match-word-constituent Operator (@code{\w})

@cindex @samp{\w}

This operator (represented by @samp{\w}) matches any word-constituent
character.

@node Match-non-word-constituent Operator
@subsection The Match-non-word-constituent Operator (@code{\W})

@cindex @samp{\W}

This operator (represented by @samp{\W}) matches any character that is
not word-constituent.


@node Space Operators
@section Space Operators

@node Match-space Operator
@subsection The Match-space Operator (@code{\s})

@cindex @samp{\s}

This operator (represented by @samp{\s}) matches any space
character (that is, in the POSIX class @code{[:space:]}).

@node Match-non-space Operator
@subsection The Match-non-space Operator (@code{\S})

@cindex @samp{\S}

This operator (represented by @samp{\S}) matches any character
that is not a space (that is, in the POSIX class @code{[:space:]}).


@node Whole-string Operators
@section Whole-string Operators

Following are operators which work on the whole string.

@menu
* Match-beginning-of-string Operator::  \`
* Match-end-of-string Operator::        \'
@end menu


@node Match-beginning-of-string Operator
@subsection The Match-beginning-of-string Operator (@code{\`})

@cindex @samp{\`}

This operator (represented by @samp{\`}) matches the empty string at the
beginning of the string.

@node Match-end-of-string Operator
@subsection The Match-end-of-string Operator (@code{\'})

@cindex @samp{\'}

This operator (represented by @samp{\'}) matches the empty string at the
end of the string.


@node What Gets Matched?
@chapter What Gets Matched?

Regex usually matches strings according to the ``leftmost longest''
rule; that is, it chooses the longest of the leftmost matches.  This
does not mean that for a regular expression containing subexpressions
that it simply chooses the longest match for each subexpression, left to
right; the overall match must also be the longest possible one.

For example, @samp{(ac*)(c*d[ac]*)\1} matches @samp{acdacaaa}, not
@samp{acdac}, as it would if it were to choose the longest match for the
first subexpression.


@node Programming with Regex
@chapter Programming with Regex

Here we describe how you use the Regex data structures and functions in
C programs.  Regex has three interfaces: one designed for GNU, one
compatible with POSIX (as specified by POSIX, draft
1003.2/D11.2), and one compatible with Berkeley Unix.  The
POSIX interface is not documented here; see the documentation of
GNU libc, or the POSIX man pages.  The Berkeley Unix interface is
documented here for convenience, since its documentation is not
otherwise readily available on GNU systems.

@menu
* GNU Regex Functions::
* BSD Regex Functions::
@end menu


@node GNU Regex Functions
@section GNU Regex Functions

If you're writing code that doesn't need to be compatible with either
POSIX or Berkeley Unix, you can use these functions.  They
provide more options than the other interfaces.

@menu
* GNU Pattern Buffers::         The re_pattern_buffer type.
* GNU Regular Expression Compiling::  re_compile_pattern ()
* GNU Matching::                re_match ()
* GNU Searching::               re_search ()
* Matching/Searching with Split Data::  re_match_2 (), re_search_2 ()
* Searching with Fastmaps::     re_compile_fastmap ()
* GNU Translate Tables::        The @code{translate} field.
* Using Registers::             The re_registers type and related fns.
* Freeing GNU Pattern Buffers::  regfree ()
@end menu


@node GNU Pattern Buffers
@subsection GNU Pattern Buffers

@cindex pattern buffer, definition of
@tindex re_pattern_buffer @r{definition}
@tindex struct re_pattern_buffer @r{definition}

To compile, match, or search for a given regular expression, you must
supply a pattern buffer.  A @dfn{pattern buffer} holds one compiled
regular expression.@footnote{Regular expressions are also referred to as
``patterns,'' hence the name ``pattern buffer.''}

You can have several different pattern buffers simultaneously, each
holding a compiled pattern for a different regular expression.

@file{regex.h} defines the pattern buffer @code{struct} with the
following public fields:

@example
  unsigned char *buffer;
  unsigned long allocated;
  char *fastmap;
  char *translate;
  size_t re_nsub;
  unsigned no_sub : 1;
  unsigned not_bol : 1;
  unsigned not_eol : 1;
@end example


@node GNU Regular Expression Compiling
@subsection GNU Regular Expression Compiling

In GNU, you can both match and search for a given regular
expression.  To do either, you must first compile it in a pattern buffer
(@pxref{GNU Pattern Buffers}).

@cindex syntax initialization
@vindex re_syntax_options @r{initialization}
Regular expressions match according to the syntax with which they were
compiled; with GNU, you indicate what syntax you want by setting
the variable @code{re_syntax_options} (declared in @file{regex.h})
before calling the compiling function, @code{re_compile_pattern} (see
below).  @xref{Syntax Bits}, and @ref{Predefined Syntaxes}.

You can change the value of @code{re_syntax_options} at any time.
Usually, however, you set its value once and then never change it.

@cindex pattern buffer initialization
@code{re_compile_pattern} takes a pattern buffer as an argument.  You
must initialize the following fields:

@table @code

@item translate @r{initialization}

@item translate
@vindex translate @r{initialization}
Initialize this to point to a translate table if you want one, or to
zero if you don't.  We explain translate tables in @ref{GNU Translate
Tables}.

@item fastmap
@vindex fastmap @r{initialization}
Initialize this to nonzero if you want a fastmap, or to zero if you
don't.

@item buffer
@itemx allocated
@vindex buffer @r{initialization}
@vindex allocated @r{initialization}
@findex malloc
If you want @code{re_compile_pattern} to allocate memory for the
compiled pattern, set both of these to zero.  If you have an existing
block of memory (allocated with @code{malloc}) you want Regex to use,
set @code{buffer} to its address and @code{allocated} to its size (in
bytes).

@code{re_compile_pattern} uses @code{realloc} to extend the space for
the compiled pattern as necessary.

@end table

To compile a pattern buffer, use:

@findex re_compile_pattern
@example
char *
re_compile_pattern (const char *@var{regex}, const int @var{regex_size},
                    struct re_pattern_buffer *@var{pattern_buffer})
@end example

@noindent
@var{regex} is the regular expression's address, @var{regex_size} is its
length, and @var{pattern_buffer} is the pattern buffer's address.

If @code{re_compile_pattern} successfully compiles the regular
expression, it returns zero and sets @code{*@var{pattern_buffer}} to the
compiled pattern.  It sets the pattern buffer's fields as follows:

@table @code
@item buffer
@vindex buffer @r{field, set by @code{re_compile_pattern}}
to the compiled pattern.

@item syntax
@vindex syntax @r{field, set by @code{re_compile_pattern}}
to the current value of @code{re_syntax_options}.

@item re_nsub
@vindex re_nsub @r{field, set by @code{re_compile_pattern}}
to the number of subexpressions in @var{regex}.

@end table

If @code{re_compile_pattern} can't compile @var{regex}, it returns an
error string corresponding to a POSIX error code.


@node GNU Matching
@subsection GNU Matching

@cindex matching with GNU functions

Matching the GNU way means trying to match as much of a string as
possible starting at a position within it you specify.  Once you've compiled
a pattern into a pattern buffer (@pxref{GNU Regular Expression
Compiling}), you can ask the matcher to match that pattern against a
string using:

@findex re_match
@example
int
re_match (struct re_pattern_buffer *@var{pattern_buffer},
          const char *@var{string}, const int @var{size},
          const int @var{start}, struct re_registers *@var{regs})
@end example

@noindent
@var{pattern_buffer} is the address of a pattern buffer containing a
compiled pattern.  @var{string} is the string you want to match; it can
contain newline and null characters.  @var{size} is the length of that
string.  @var{start} is the string index at which you want to
begin matching; the first character of @var{string} is at index zero.
@xref{Using Registers}, for an explanation of @var{regs}; you can safely
pass zero.

@code{re_match} matches the regular expression in @var{pattern_buffer}
against the string @var{string} according to the syntax of
@var{pattern_buffer}.  (@xref{GNU Regular Expression Compiling}, for how
to set it.)  The function returns @math{-1} if the compiled pattern does
not match any part of @var{string} and @math{-2} if an internal error
happens; otherwise, it returns how many (possibly zero) characters of
@var{string} the pattern matched.

An example: suppose @var{pattern_buffer} points to a pattern buffer
containing the compiled pattern for @samp{a*}, and @var{string} points
to @samp{aaaaab} (whereupon @var{size} should be 6). Then if @var{start}
is 2, @code{re_match} returns 3, i.e., @samp{a*} would have matched the
last three @samp{a}s in @var{string}.  If @var{start} is 0,
@code{re_match} returns 5, i.e., @samp{a*} would have matched all the
@samp{a}s in @var{string}.  If @var{start} is either 5 or 6, it returns
zero.

If @var{start} is not between zero and @var{size}, then
@code{re_match} returns @math{-1}.


@node GNU Searching
@subsection GNU Searching

@cindex searching with GNU functions

@dfn{Searching} means trying to match starting at successive positions
within a string.  The function @code{re_search} does this.

Before calling @code{re_search}, you must compile your regular
expression.  @xref{GNU Regular Expression Compiling}.

Here is the function declaration:

@findex re_search
@example
int
re_search (struct re_pattern_buffer *@var{pattern_buffer},
           const char *@var{string}, const int @var{size},
           const int @var{start}, const int @var{range},
           struct re_registers *@var{regs})
@end example

@noindent
@vindex start @r{argument to @code{re_search}}
@vindex range @r{argument to @code{re_search}}
whose arguments are the same as those to @code{re_match} (@pxref{GNU
Matching}) except that the two arguments @var{start} and @var{range}
replace @code{re_match}'s argument @var{start}.

If @var{range} is positive, then @code{re_search} attempts a match
starting first at index @var{start}, then at @math{@var{start} + 1} if
that fails, and so on, up to @math{@var{start} + @var{range}}; if
@var{range} is negative, then it attempts a match starting first at
index @var{start}, then at @math{@var{start} -1} if that fails, and so
on.

If @var{start} is not between zero and @var{size}, then @code{re_search}
returns @math{-1}.  When @var{range} is positive, @code{re_search}
adjusts @var{range} so that @math{@var{start} + @var{range} - 1} is
between zero and @var{size}, if necessary; that way it won't search
outside of @var{string}.  Similarly, when @var{range} is negative,
@code{re_search} adjusts @var{range} so that @math{@var{start} +
@var{range} + 1} is between zero and @var{size}, if necessary.

If the @code{fastmap} field of @var{pattern_buffer} is zero,
@code{re_search} matches starting at consecutive positions; otherwise,
it uses @code{fastmap} to make the search more efficient.
@xref{Searching with Fastmaps}.

If no match is found, @code{re_search} returns @math{-1}.  If
a match is found, it returns the index where the match began.  If an
internal error happens, it returns @math{-2}.


@node Matching/Searching with Split Data
@subsection Matching and Searching with Split Data

Using the functions @code{re_match_2} and @code{re_search_2}, you can
match or search in data that is divided into two strings.

The function:

@findex re_match_2
@example
int
re_match_2 (struct re_pattern_buffer *@var{buffer},
            const char *@var{string1}, const int @var{size1},
            const char *@var{string2}, const int @var{size2},
            const int @var{start},
            struct re_registers *@var{regs},
            const int @var{stop})
@end example

@noindent
is similar to @code{re_match} (@pxref{GNU Matching}) except that you
pass @emph{two} data strings and sizes, and an index @var{stop} beyond
which you don't want the matcher to try matching.  As with
@code{re_match}, if it succeeds, @code{re_match_2} returns how many
characters of @var{string} it matched.  Regard @var{string1} and
@var{string2} as concatenated when you set the arguments @var{start} and
@var{stop} and use the contents of @var{regs}; @code{re_match_2} never
returns a value larger than @math{@var{size1} + @var{size2}}.

The function:

@findex re_search_2
@example
int
re_search_2 (struct re_pattern_buffer *@var{buffer},
             const char *@var{string1}, const int @var{size1},
             const char *@var{string2}, const int @var{size2},
             const int @var{start}, const int @var{range},
             struct re_registers *@var{regs},
             const int @var{stop})
@end example

@noindent
is similarly related to @code{re_search}.


@node Searching with Fastmaps
@subsection Searching with Fastmaps

@cindex fastmaps
If you're searching through a long string, you should use a fastmap.
Without one, the searcher tries to match at consecutive positions in the
string.  Generally, most of the characters in the string could not start
a match.  It takes much longer to try matching at a given position in the
string than it does to check in a table whether or not the character at
that position could start a match.  A @dfn{fastmap} is such a table.

More specifically, a fastmap is an array indexed by the characters in
your character set.  Under the ASCII encoding, therefore, a fastmap
has 256 elements.  If you want the searcher to use a fastmap with a
given pattern buffer, you must allocate the array and assign the array's
address to the pattern buffer's @code{fastmap} field.  You either can
compile the fastmap yourself or have @code{re_search} do it for you;
when @code{fastmap} is nonzero, it automatically compiles a fastmap the
first time you search using a particular compiled pattern.

By setting the buffer's @code{fastmap} field before calling
@code{re_compile_pattern}, you can reuse a buffer data structure across
multiple searches with different patterns, and allocate the fastmap only
once.  Nonetheless, the fastmap must be recompiled each time the buffer
has a new pattern compiled into it.

To compile a fastmap yourself, use:

@findex re_compile_fastmap
@example
int
re_compile_fastmap (struct re_pattern_buffer *@var{pattern_buffer})
@end example

@noindent
@var{pattern_buffer} is the address of a pattern buffer.  If the
character @var{c} could start a match for the pattern,
@code{re_compile_fastmap} makes
@code{@var{pattern_buffer}->fastmap[@var{c}]} nonzero.  It returns
@math{0} if it can compile a fastmap and @math{-2} if there is an
internal error.  For example, if @samp{|} is the alternation operator
and @var{pattern_buffer} holds the compiled pattern for @samp{a|b}, then
@code{re_compile_fastmap} sets @code{fastmap['a']} and
@code{fastmap['b']} (and no others).

@code{re_search} uses a fastmap as it moves along in the string: it
checks the string's characters until it finds one that's in the fastmap.
Then it tries matching at that character.  If the match fails, it
repeats the process.  So, by using a fastmap, @code{re_search} doesn't
waste time trying to match at positions in the string that couldn't
start a match.

If you don't want @code{re_search} to use a fastmap,
store zero in the @code{fastmap} field of the pattern buffer before
calling @code{re_search}.

Once you've initialized a pattern buffer's @code{fastmap} field, you
need never do so again---even if you compile a new pattern in
it---provided the way the field is set still reflects whether or not you
want a fastmap.  @code{re_search} will still either do nothing if
@code{fastmap} is null or, if it isn't, compile a new fastmap for the
new pattern.

@node GNU Translate Tables
@subsection GNU Translate Tables

If you set the @code{translate} field of a pattern buffer to a translate
table, then the GNU Regex functions to which you've passed that
pattern buffer use it to apply a simple transformation
to all the regular expression and string characters at which they look.

A @dfn{translate table} is an array indexed by the characters in your
character set.  Under the ASCII encoding, therefore, a translate
table has 256 elements.  The array's elements are also characters in
your character set.  When the Regex functions see a character @var{c},
they use @code{translate[@var{c}]} in its place, with one exception: the
character after a @samp{\} is not translated.  (This ensures that, the
operators, e.g., @samp{\B} and @samp{\b}, are always distinguishable.)

For example, a table that maps all lowercase letters to the
corresponding uppercase ones would cause the matcher to ignore
differences in case.@footnote{A table that maps all uppercase letters to
the corresponding lowercase ones would work just as well for this
purpose.}  Such a table would map all characters except lowercase letters
to themselves, and lowercase letters to the corresponding uppercase
ones.  Under the ASCII encoding, here's how you could initialize
such a table (we'll call it @code{case_fold}):

@example
for (i = 0; i < 256; i++)
  case_fold[i] = i;
for (i = 'a'; i <= 'z'; i++)
  case_fold[i] = i - ('a' - 'A');
@end example

You tell Regex to use a translate table on a given pattern buffer by
assigning that table's address to the @code{translate} field of that
buffer.  If you don't want Regex to do any translation, put zero into
this field.  You'll get weird results if you change the table's contents
anytime between compiling the pattern buffer, compiling its fastmap, and
matching or searching with the pattern buffer.

@node Using Registers
@subsection Using Registers

A group in a regular expression can match a (possibly empty) substring
of the string that regular expression as a whole matched.  The matcher
remembers the beginning and end of the substring matched by
each group.

To find out what they matched, pass a nonzero @var{regs} argument to a
GNU matching or searching function (@pxref{GNU Matching} and
@ref{GNU Searching}), i.e., the address of a structure of this type, as
defined in @file{regex.h}:

@c We don't bother to include this directly from regex.h,
@c since it changes so rarely.
@example
@tindex re_registers
@vindex num_regs @r{in @code{struct re_registers}}
@vindex start @r{in @code{struct re_registers}}
@vindex end @r{in @code{struct re_registers}}
struct re_registers
@{
  unsigned num_regs;
  regoff_t *start;
  regoff_t *end;
@};
@end example

Except for (possibly) the @var{num_regs}'th element (see below), the
@var{i}th element of the @code{start} and @code{end} arrays records
information about the @var{i}th group in the pattern.  (They're declared
as C pointers, but this is only because not all C compilers accept
zero-length arrays; conceptually, it is simplest to think of them as
arrays.)

The @code{start} and @code{end} arrays are allocated in one of two ways.
The simplest and perhaps most useful is to let the matcher (re)allocate
enough space to record information for all the groups in the regular
expression.  If @code{re_set_registers} is not called before searching
or matching, then the matcher allocates two arrays each of @math{1 +
@var{re_nsub}} elements (@var{re_nsub} is another field in the pattern
buffer; @pxref{GNU Pattern Buffers}).  The extra element is set to
@math{-1}.  Then on subsequent calls with the same pattern buffer and
@var{regs} arguments, the matcher reallocates more space if necessary.

The function:

@findex re_set_registers
@example
void
re_set_registers (struct re_pattern_buffer *@var{buffer},
                  struct re_registers *@var{regs},
                  size_t @var{num_regs},
                  regoff_t *@var{starts}, regoff_t *@var{ends})
@end example

@noindent sets @var{regs} to hold @var{num_regs} registers, storing
them in @var{starts} and @var{ends}.  Subsequent matches using
@var{buffer} and @var{regs} will use this memory for recording
register information.  @var{starts} and @var{ends} must be allocated
with malloc, and must each be at least @math{@var{num_regs} *
@code{sizeof (regoff_t)}} bytes long.

If @var{num_regs} is zero, then subsequent matches should allocate
their own register data.

Unless this function is called, the first search or match using
@var{buffer} will allocate its own register data, without freeing the
old data.

The following examples illustrate the information recorded in the
@code{re_registers} structure.  (In all of them, @samp{(} represents the
open-group and @samp{)} the close-group operator.  The first character
in the string @var{string} is at index 0.)

@itemize @bullet

@item
If the regular expression has an @w{@var{i}-th}
group that matches a
substring of @var{string}, then the function sets
@code{@w{@var{regs}->}start[@var{i}]} to the index in @var{string} where
the substring matched by the @w{@var{i}-th} group begins, and
@code{@w{@var{regs}->}end[@var{i}]} to the index just beyond that
substring's end.  The function sets @code{@w{@var{regs}->}start[0]} and
@code{@w{@var{regs}->}end[0]} to analogous information about the entire
pattern.

For example, when you match @samp{((a)(b))} against @samp{ab}, you get:

@itemize
@item
0 in @code{@w{@var{regs}->}start[0]} and 2 in @code{@w{@var{regs}->}end[0]}

@item
0 in @code{@w{@var{regs}->}start[1]} and 2 in @code{@w{@var{regs}->}end[1]}

@item
0 in @code{@w{@var{regs}->}start[2]} and 1 in @code{@w{@var{regs}->}end[2]}

@item
1 in @code{@w{@var{regs}->}start[3]} and 2 in @code{@w{@var{regs}->}end[3]}
@end itemize

@item
If a group matches more than once (as it might if followed by,
e.g., a repetition operator), then the function reports the information
about what the group @emph{last} matched.

For example, when you match the pattern @samp{(a)*} against the string
@samp{aa}, you get:

@itemize
@item
0 in @code{@w{@var{regs}->}start[0]} and 2 in @code{@w{@var{regs}->}end[0]}

@item
1 in @code{@w{@var{regs}->}start[1]} and 2 in @code{@w{@var{regs}->}end[1]}
@end itemize

@item
If the @w{@var{i}-th} group does not participate in a
successful match, e.g., it is an alternative not taken or a
repetition operator allows zero repetitions of it, then the function
sets @code{@w{@var{regs}->}start[@var{i}]} and
@code{@w{@var{regs}->}end[@var{i}]} to @math{-1}.

For example, when you match the pattern @samp{(a)*b} against
the string @samp{b}, you get:

@itemize
@item
0 in @code{@w{@var{regs}->}start[0]} and 1 in @code{@w{@var{regs}->}end[0]}

@item
@math{-1} in @code{@w{@var{regs}->}start[1]} and @math{-1} in @code{@w{@var{regs}->}end[1]}
@end itemize

@item
If the @w{@var{i}-th} group matches a zero-length string, then the
function sets @code{@w{@var{regs}->}start[@var{i}]} and
@code{@w{@var{regs}->}end[@var{i}]} to the index just beyond that
zero-length string.

For example, when you match the pattern @samp{(a*)b} against the string
@samp{b}, you get:

@itemize
@item
0 in @code{@w{@var{regs}->}start[0]} and 1 in @code{@w{@var{regs}->}end[0]}

@item
0 in @code{@w{@var{regs}->}start[1]} and 0 in @code{@w{@var{regs}->}end[1]}
@end itemize

@item
If an @w{@var{i}-th} group contains a @w{@var{j}-th} group
in turn not contained within any other group within group @var{i} and
the function reports a match of the @w{@var{i}-th} group, then it
records in @code{@w{@var{regs}->}start[@var{j}]} and
@code{@w{@var{regs}->}end[@var{j}]} the last match (if it matched) of
the @w{@var{j}-th} group.

For example, when you match the pattern @samp{((a*)b)*} against the
string @samp{abb}, @w{group 2} last matches the empty string, so you
get what it previously matched:

@itemize
@item
0 in @code{@w{@var{regs}->}start[0]} and 3 in @code{@w{@var{regs}->}end[0]}

@item
2 in @code{@w{@var{regs}->}start[1]} and 3 in @code{@w{@var{regs}->}end[1]}

@item
2 in @code{@w{@var{regs}->}start[2]} and 2 in @code{@w{@var{regs}->}end[2]}
@end itemize

When you match the pattern @samp{((a)*b)*} against the string
@samp{abb}, @w{group 2} doesn't participate in the last match, so you
get:

@itemize
@item
0 in @code{@w{@var{regs}->}start[0]} and 3 in @code{@w{@var{regs}->}end[0]}

@item
2 in @code{@w{@var{regs}->}start[1]} and 3 in @code{@w{@var{regs}->}end[1]}

@item
0 in @code{@w{@var{regs}->}start[2]} and 1 in @code{@w{@var{regs}->}end[2]}
@end itemize

@item
If an @w{@var{i}-th} group contains a @w{@var{j}-th} group
in turn not contained within any other group within group @var{i}
and the function sets
@code{@w{@var{regs}->}start[@var{i}]} and
@code{@w{@var{regs}->}end[@var{i}]} to @math{-1}, then it also sets
@code{@w{@var{regs}->}start[@var{j}]} and
@code{@w{@var{regs}->}end[@var{j}]} to @math{-1}.

For example, when you match the pattern @samp{((a)*b)*c} against the
string @samp{c}, you get:

@itemize
@item
0 in @code{@w{@var{regs}->}start[0]} and 1 in @code{@w{@var{regs}->}end[0]}

@item
@math{-1} in @code{@w{@var{regs}->}start[1]} and @math{-1} in @code{@w{@var{regs}->}end[1]}

@item
@math{-1} in @code{@w{@var{regs}->}start[2]} and @math{-1} in @code{@w{@var{regs}->}end[2]}
@end itemize

@end itemize

@node Freeing GNU Pattern Buffers
@subsection Freeing GNU Pattern Buffers

To free any allocated fields of a pattern buffer, use the POSIX
function @code{regfree}:

@findex regfree
@example
void
regfree (regex_t *@var{preg})
@end example

@noindent
@var{preg} is the pattern buffer whose allocated fields you want freed;
this works because since the type @code{regex_t}---the type for
POSIX pattern buffers---is equivalent to the type
@code{re_pattern_buffer}.

@code{regfree} also sets @var{preg}'s @code{allocated} field to zero.
After a buffer has been freed, it must have a regular expression
compiled in it before passing it to a matching or searching function.


@node BSD Regex Functions
@section BSD Regex Functions

If you're writing code that has to be Berkeley Unix compatible,
you'll need to use these functions whose interfaces are the same as those
in Berkeley Unix.

@menu
* BSD Regular Expression Compiling::    re_comp ()
* BSD Searching::                       re_exec ()
@end menu

@node BSD Regular Expression Compiling
@subsection  BSD Regular Expression Compiling

With Berkeley Unix, you can only search for a given regular
expression; you can't match one.  To search for it, you must first
compile it.  Before you compile it, you must indicate the regular
expression syntax you want it compiled according to by setting the
variable @code{re_syntax_options} (declared in @file{regex.h}) to some
syntax (@pxref{Regular Expression Syntax}).

To compile a regular expression use:

@findex re_comp
@example
char *
re_comp (char *@var{regex})
@end example

@noindent
@var{regex} is the address of a null-terminated regular expression.
@code{re_comp} uses an internal pattern buffer, so you can use only the
most recently compiled pattern buffer.  This means that if you want to
use a given regular expression that you've already compiled---but it
isn't the latest one you've compiled---you'll have to recompile it.  If
you call @code{re_comp} with the null string (@emph{not} the empty
string) as the argument, it doesn't change the contents of the pattern
buffer.

If @code{re_comp} successfully compiles the regular expression, it
returns zero.  If it can't compile the regular expression, it returns
an error string.  @code{re_comp}'s error messages are identical to those
of @code{re_compile_pattern} (@pxref{GNU Regular Expression
Compiling}).

@node BSD Searching
@subsection BSD Searching

Searching the Berkeley Unix way means searching in a string
starting at its first character and trying successive positions within
it to find a match.  Once you've compiled a pattern using @code{re_comp}
(@pxref{BSD Regular Expression Compiling}), you can ask Regex
to search for that pattern in a string using:

@findex re_exec
@example
int
re_exec (char *@var{string})
@end example

@noindent
@var{string} is the address of the null-terminated string in which you
want to search.

@code{re_exec} returns either 1 for success or 0 for failure.  It
automatically uses a GNU fastmap (@pxref{Searching with Fastmaps}).