Boost C++ Libraries

What is xpressive?

xpressive is a regular expression template library. Regular expressions (regexes) can be written as strings that are parsed dynamically at runtime (dynamic regexes), or as expression templates ^[4] that are parsed at compile-time (static regexes). Dynamic regexes have the advantage that they can be accepted from the user as input at runtime or read from an initialization file. Static regexes have several advantages. Since they are C++ expressions instead of strings, they can be syntax-checked at compile-time. Also, they can naturally refer to code and data elsewhere in your program, giving you the ability to call back into your code from within a regex match. Finally, since they are statically bound, the compiler can generate faster code for static regexes.

xpressive's dual nature is unique and powerful. Static xpressive is a bit like the Spirit Parser Framework. Like Spirit, you can build grammars with static regexes using expression templates. (Unlike Spirit, xpressive does exhaustive backtracking, trying every possibility to find a match for your pattern.) Dynamic xpressive is a bit like Boost.Regex. In fact, xpressive's interface should be familiar to anyone who has used Boost.Regex. xpressive's innovation comes from allowing you to mix and match static and dynamic regexes in the same program, and even in the same expression! You can embed a dynamic regex in a static regex, or vice versa, and the embedded regex will participate fully in the search, back-tracking as needed to make the match succeed.

Hello, world!

Enough theory. Let's have a look at Hello World, xpressive style:

#include <iostream>
#include <boost/xpressive/xpressive.hpp>

using namespace boost::xpressive;

int main()
{
    std::string hello( "hello world!" );

    sregex rex = sregex::compile( "(\\w+) (\\w+)!" );
    smatch what;

    if( regex_match( hello, what, rex ) )
    {
        std::cout << what[0] << '\n'; // whole match
        std::cout << what[1] << '\n'; // first capture
        std::cout << what[2] << '\n'; // second capture
    }

    return 0;
}

This program outputs the following:

hello world!
hello
world

The first thing you'll notice about the code is that all the types in xpressive live in the boost::xpressive namespace.

	Note
	Most of the rest of the examples in this document will leave off the using namespace boost::xpressive; directive. Just pretend it's there.

Next, you'll notice the type of the regular expression object is sregex. If you are familiar with Boost.Regex, this is different than what you are used to. The "s" in "sregex" stands for "string", indicating that this regex can be used to find patterns in std::string objects. I'll discuss this difference and its implications in detail later.

Notice how the regex object is initialized:

sregex rex = sregex::compile( "(\\w+) (\\w+)!" );

To create a regular expression object from a string, you must call a factory method such as basic_regex<>::compile(). This is another area in which xpressive differs from other object-oriented regular expression libraries. Other libraries encourage you to think of a regular expression as a kind of string on steroids. In xpressive, regular expressions are not strings; they are little programs in a domain-specific language. Strings are only one representation of that language. Another representation is an expression template. For example, the above line of code is equivalent to the following:

sregex rex = (s1= +_w) >> ' ' >> (s2= +_w) >> '!';

This describes the same regular expression, except it uses the domain-specific embedded language defined by static xpressive.

As you can see, static regexes have a syntax that is noticeably different than standard Perl syntax. That is because we are constrained by C++'s syntax. The biggest difference is the use of >> to mean "followed by". For instance, in Perl you can just put sub-expressions next to each other:

abc

But in C++, there must be an operator separating sub-expressions:

a >> b >> c

In Perl, parentheses () have special meaning. They group, but as a side-effect they also create back-references like $1 and $2. In C++, there is no way to overload parentheses to give them side-effects. To get the same effect, we use the special s1, s2, etc. tokens. Assign to one to create a back-reference (known as a sub-match in xpressive).

You'll also notice that the one-or-more repetition operator + has moved from postfix to prefix position. That's because C++ doesn't have a postfix + operator. So:

"\\w+"

is the same as:

+_w

We'll cover all the other differences later.

Getting xpressive

There are three ways to get xpressive. The first and simplest is to download the latest version of Boost. Just go to http://sf.net/projects/boost and follow the “Download” link.

The second way is by downloading xpressive.zip at the Boost File Vault in the “Strings - Text Processing” directory. In addition to the source code and the Boost license, this archive contains a copy of this documentation in PDF format. This version will always be stable and at least as current as the version in the latest Boost release. It may be more recent. The version in the File Vault is always guaranteed to work with the latest official Boost release.

The third way is by directly accessing the Boost Subversion repository. Just go to http://svn.boost.org/trac/boost/ and follow the instructions there for anonymous Subversion access. The version in Boost Subversion is unstable.

Building with xpressive

Xpressive is a header-only template library, which means you don't need to alter your build scripts or link to any separate lib file to use it. All you need to do is #include <boost/xpressive/xpressive.hpp>. If you are only using static regexes, you can improve compile times by only including xpressive_static.hpp. Likewise, you can include xpressive_dynamic.hpp if you only plan on using dynamic regexes.

If you would also like to use semantic actions or custom assertions with your static regexes, you will need to additionally include regex_actions.hpp.

Requirements

Xpressive requires Boost version 1.34.1 or higher.

Supported Compilers

Currently, Boost.Xpressive is known to work on the following compilers:

Check the latest tests results at Boost's Regression Results Page.

	Note
	Please send any questions, comments and bug reports to eric <at> boost-consulting <dot> com.

You don't need to know much to start being productive with xpressive. Let's begin with the nickel tour of the types and algorithms xpressive provides.

Now that you know a bit about the tools xpressive provides, you can pick the right tool for you by answering the following two questions:

Know Your Iterator Type

Most of the classes in xpressive are templates that are parameterized on the iterator type. xpressive defines some common typedefs to make the job of choosing the right types easier. You can use the table below to find the right types based on the type of your iterator.

You should notice the systematic naming convention. Many of these types are used together, so the naming convention helps you to use them consistently. For instance, if you have a sregex, you should also be using a smatch.

If you are not using one of those four iterator types, then you can use the templates directly and specify your iterator type.

Know Your Task

Do you want to find a pattern once? Many times? Search and replace? xpressive has tools for all that and more. Below is a quick reference:

These algorithms and classes are described in excruciating detail in the Reference section.

	Tip
	Try clicking on a task in the table above to see a complete example program that uses xpressive to solve that particular task.

When using xpressive, the first thing you'll do is create a basic_regex<> object. This section goes over the nuts and bolts of building a regular expression in the two dialects xpressive supports: static and dynamic.

Static Regexes

Overview

The feature that really sets xpressive apart from other C/C++ regular expression libraries is the ability to author a regular expression using C++ expressions. xpressive achieves this through operator overloading, using a technique called expression templates to embed a mini-language dedicated to pattern matching within C++. These "static regexes" have many advantages over their string-based brethren. In particular, static regexes:

• are syntax-checked at compile-time; they will never fail at run-time due to a syntax error.
• can naturally refer to other C++ data and code, including other regexes, making it simple to build grammars out of regular expressions and bind user-defined actions that execute when parts of your regex match.
• are statically bound for better inlining and optimization. Static regexes require no state tables, virtual functions, byte-code or calls through function pointers that cannot be resolved at compile time.
• are not limited to searching for patterns in strings. You can declare a static regex that finds patterns in an array of integers, for instance.

Since we compose static regexes using C++ expressions, we are constrained by the rules for legal C++ expressions. Unfortunately, that means that "classic" regular expression syntax cannot always be mapped cleanly into C++. Rather, we map the regex constructs, picking new syntax that is legal C++.

Construction and Assignment

You create a static regex by assigning one to an object of type basic_regex<>. For instance, the following defines a regex that can be used to find patterns in objects of type std::string:

sregex re = '$' >> +_d >> '.' >> _d >> _d;

Assignment works similarly.

Character and String Literals

In static regexes, character and string literals match themselves. For instance, in the regex above, '$' and '.' match the characters '$' and '.' respectively. Don't be confused by the fact that $ and . are meta-characters in Perl. In xpressive, literals always represent themselves.

When using literals in static regexes, you must take care that at least one operand is not a literal. For instance, the following are not valid regexes:

sregex re1 = 'a' >> 'b';         // ERROR!
sregex re2 = +'a';               // ERROR!

The two operands to the binary >> operator are both literals, and the operand of the unary + operator is also a literal, so these statements will call the native C++ binary right-shift and unary plus operators, respectively. That's not what we want. To get operator overloading to kick in, at least one operand must be a user-defined type. We can use xpressive's as_xpr() helper function to "taint" an expression with regex-ness, forcing operator overloading to find the correct operators. The two regexes above should be written as:

sregex re1 = as_xpr('a') >> 'b'; // OK
sregex re2 = +as_xpr('a');       // OK

Sequencing and Alternation

As you've probably already noticed, sub-expressions in static regexes must be separated by the sequencing operator, >>. You can read this operator as "followed by".

// Match an 'a' followed by a digit
sregex re = 'a' >> _d;

Alternation works just as it does in Perl with the | operator. You can read this operator as "or". For example:

// match a digit character or a word character one or more times
sregex re = +( _d | _w );

Grouping and Captures

In Perl, parentheses () have special meaning. They group, but as a side-effect they also create back-references like $1 and $2. In C++, parentheses only group -- there is no way to give them side-effects. To get the same effect, we use the special s1, s2, etc. tokens. Assigning to one creates a back-reference. You can then use the back-reference later in your expression, like using \1 and \2 in Perl. For example, consider the following regex, which finds matching HTML tags:

"<(\\w+)>.*?</\\1>"

In static xpressive, this would be:

'<' >> (s1= +_w) >> '>' >> -*_ >> "</" >> s1 >> '>'

Notice how you capture a back-reference by assigning to s1, and then you use s1 later in the pattern to find the matching end tag.

	Tip
	Grouping without capturing a back-reference In xpressive, if you just want grouping without capturing a back-reference, you can just use () without s1. That is the equivalent of Perl's (?:) non-capturing grouping construct.

Case-Insensitivity and Internationalization

Perl lets you make part of your regular expression case-insensitive by using the (?i:) pattern modifier. xpressive also has a case-insensitivity pattern modifier, called icase. You can use it as follows:

sregex re = "this" >> icase( "that" );

In this regular expression, "this" will be matched exactly, but "that" will be matched irrespective of case.

Case-insensitive regular expressions raise the issue of internationalization: how should case-insensitive character comparisons be evaluated? Also, many character classes are locale-specific. Which characters are matched by digit and which are matched by alpha? The answer depends on the std::locale object the regular expression object is using. By default, all regular expression objects use the global locale. You can override the default by using the imbue() pattern modifier, as follows:

std::locale my_locale = /* initialize a std::locale object */;
sregex re = imbue( my_locale )( +alpha >> +digit );

This regular expression will evaluate alpha and digit according to my_locale. See the section on Localization and Regex Traits for more information about how to customize the behavior of your regexes.

Static xpressive Syntax Cheat Sheet

The table below lists the familiar regex constructs and their equivalents in static xpressive.

Table 26.4. Perl syntax vs. Static xpressive syntax

Perl	Static xpressive	Meaning
.	_	any character (assuming Perl's /s modifier).
ab	a >> b	sequencing of a and b sub-expressions.
a|b	a | b	alternation of a and b sub-expressions.
(a)	(s1= a)	group and capture a back-reference.
(?:a)	(a)	group and do not capture a back-reference.
\1	s1	a previously captured back-reference.
a*	*a	zero or more times, greedy.
a+	+a	one or more times, greedy.
a?	!a	zero or one time, greedy.
a{n,m}	repeat<n,m>(a)	between n and m times, greedy.
a*?	-*a	zero or more times, non-greedy.
a+?	-+a	one or more times, non-greedy.
a??	-!a	zero or one time, non-greedy.
a{n,m}?	-repeat<n,m>(a)	between n and m times, non-greedy.
^	bos	beginning of sequence assertion.
$	eos	end of sequence assertion.
\b	_b	word boundary assertion.
\B	~_b	not word boundary assertion.
\n	_n	literal newline.
.	~_n	any character except a literal newline (without Perl's /s modifier).
\r?\n|\r	_ln	logical newline.
[^\r\n]	~_ln	any single character not a logical newline.
\w	_w	a word character, equivalent to set[alnum | '_'].
\W	~_w	not a word character, equivalent to ~set[alnum | '_'].
\d	_d	a digit character.
\D	~_d	not a digit character.
\s	_s	a space character.
\S	~_s	not a space character.
[:alnum:]	alnum	an alpha-numeric character.
[:alpha:]	alpha	an alphabetic character.
[:blank:]	blank	a horizontal white-space character.
[:cntrl:]	cntrl	a control character.
[:digit:]	digit	a digit character.
[:graph:]	graph	a graphable character.
[:lower:]	lower	a lower-case character.
[:print:]	print	a printing character.
[:punct:]	punct	a punctuation character.
[:space:]	space	a white-space character.
[:upper:]	upper	an upper-case character.
[:xdigit:]	xdigit	a hexadecimal digit character.
[0-9]	range('0','9')	characters in range '0' through '9'.
[abc]	as_xpr('a') | 'b' |'c'	characters 'a', 'b', or 'c'.
[abc]	(set= 'a','b','c')	same as above
[0-9abc]	set[ range('0','9') | 'a' | 'b' | 'c' ]	characters 'a', 'b', 'c' or in range '0' through '9'.
[0-9abc]	set[ range('0','9') | (set= 'a','b','c') ]	same as above
[^abc]	~(set= 'a','b','c')	not characters 'a', 'b', or 'c'.
(?i:stuff)	icase(stuff)	match stuff disregarding case.
(?>stuff)	keep(stuff)	independent sub-expression, match stuff and turn off backtracking.
(?=stuff)	before(stuff)	positive look-ahead assertion, match if before stuff but don't include stuff in the match.
(?!stuff)	~before(stuff)	negative look-ahead assertion, match if not before stuff.
(?<=stuff)	after(stuff)	positive look-behind assertion, match if after stuff but don't include stuff in the match. (stuff must be constant-width.)
(?<!stuff)	~after(stuff)	negative look-behind assertion, match if not after stuff. (stuff must be constant-width.)

sregex re = sregex::compile( "this|that", regex_constants::icase );

sregex_compiler compiler;
sregex re = compiler.compile( "this|that", regex_constants::icase );

std::locale my_locale = /* initialize your locale object here */;
sregex_compiler compiler;
compiler.imbue( my_locale );
sregex re = compiler.compile( "\\w+|\\d+" );

Overview

Once you have created a regex object, you can use the regex_match() and regex_search() algorithms to find patterns in strings. This page covers the basics of regex matching and searching. In all cases, if you are familiar with how regex_match() and regex_search() in the Boost.Regex library work, xpressive's versions work the same way.

Seeing if a String Matches a Regex

The regex_match() algorithm checks to see if a regex matches a given input.

	Warning
	The regex_match() algorithm will only report success if the regex matches the whole input, from beginning to end. If the regex matches only a part of the input, regex_match() will return false. If you want to search through the string looking for sub-strings that the regex matches, use the regex_search() algorithm.

The input can be a bidirectional range such as std::string, a C-style null-terminated string or a pair of iterators. In all cases, the type of the iterator used to traverse the input sequence must match the iterator type used to declare the regex object. (You can use the table in the Quick Start to find the correct regex type for your iterator.)

cregex cre = +_w;  // this regex can match C-style strings
sregex sre = +_w;  // this regex can match std::strings

if( regex_match( "hello", cre ) )              // OK
    { /*...*/ }

if( regex_match( std::string("hello"), sre ) ) // OK
    { /*...*/ }

if( regex_match( "hello", sre ) )              // ERROR! iterator mis-match!
    { /*...*/ }

The regex_match() algorithm optionally accepts a match_results<> struct as an out parameter. If given, the regex_match() algorithm fills in the match_results<> struct with information about which parts of the regex matched which parts of the input.

cmatch what;
cregex cre = +(s1= _w);

// store the results of the regex_match in "what"
if( regex_match( "hello", what, cre ) )
{
    std::cout << what[1] << '\n'; // prints "o"
}

The regex_match() algorithm also optionally accepts a match_flag_type bitmask. With match_flag_type, you can control certain aspects of how the match is evaluated. See the match_flag_type reference for a complete list of the flags and their meanings.

std::string str("hello");
sregex sre = bol >> +_w;

// match_not_bol means that "bol" should not match at [begin,begin)
if( regex_match( str.begin(), str.end(), sre, regex_constants::match_not_bol ) )
{
    // should never get here!!!
}

Click here to see a complete example program that shows how to use regex_match(). And check the regex_match() reference to see a complete list of the available overloads.

Searching for Matching Sub-Strings

Use regex_search() when you want to know if an input sequence contains a sub-sequence that a regex matches. regex_search() will try to match the regex at the beginning of the input sequence and scan forward in the sequence until it either finds a match or exhausts the sequence.

In all other regards, regex_search() behaves like regex_match() (see above). In particular, it can operate on a bidirectional range such as std::string, C-style null-terminated strings or iterator ranges. The same care must be taken to ensure that the iterator type of your regex matches the iterator type of your input sequence. As with regex_match(), you can optionally provide a match_results<> struct to receive the results of the search, and a match_flag_type bitmask to control how the match is evaluated.

Click here to see a complete example program that shows how to use regex_search(). And check the regex_search() reference to see a complete list of the available overloads.

Overview

Sometimes, it is not enough to know simply whether a regex_match() or regex_search() was successful or not. If you pass an object of type match_results<> to regex_match() or regex_search(), then after the algorithm has completed successfully the match_results<> will contain extra information about which parts of the regex matched which parts of the sequence. In Perl, these sub-sequences are called back-references, and they are stored in the variables $1, $2, etc. In xpressive, they are objects of type sub_match<>, and they are stored in the match_results<> structure, which acts as a vector of sub_match<> objects.

match_results

So, you've passed a match_results<> object to a regex algorithm, and the algorithm has succeeded. Now you want to examine the results. Most of what you'll be doing with the match_results<> object is indexing into it to access its internally stored sub_match<> objects, but there are a few other things you can do with a match_results<> object besides.

The table below shows how to access the information stored in a match_results<> object named what.

There is more you can do with the match_results<> object, but that will be covered when we talk about Grammars and Nested Matches.

sub_match

When you index into a match_results<> object, you get back a sub_match<> object. A sub_match<> is basically a pair of iterators. It is defined like this:

template< class BidirectionalIterator >
struct sub_match
    : std::pair< BidirectionalIterator, BidirectionalIterator >
{
    bool matched;
    // ...
};

Since it inherits publicaly from std::pair<>, sub_match<> has first and second data members of type BidirectionalIterator. These are the beginning and end of the sub-sequence this sub_match<> represents. sub_match<> also has a Boolean matched data member, which is true if this sub_match<> participated in the full match.

The following table shows how you might access the information stored in a sub_match<> object called sub.

Results Invalidation

Results are stored as iterators into the input sequence. Anything which invalidates the input sequence will invalidate the match results. For instance, if you match a std::string object, the results are only valid until your next call to a non-const member function of that std::string object. After that, the results held by the match_results<> object are invalid. Don't use them!

Regular expressions are not only good for searching text; they're good at manipulating it. And one of the most common text manipulation tasks is search-and-replace. xpressive provides the regex_replace() algorithm for searching and replacing.

regex_replace()

Performing search-and-replace using regex_replace() is simple. All you need is an input sequence, a regex object, and a format string or a formatter object. There are several versions of the regex_replace() algorithm. Some accept the input sequence as a bidirectional container such as std::string and returns the result in a new container of the same type. Others accept the input as a null terminated string and return a std::string. Still others accept the input sequence as a pair of iterators and writes the result into an output iterator. The substitution may be specified as a string with format sequences or as a formatter object. Below are some simple examples of using string-based substitutions.

std::string input("This is his face");
sregex re = as_xpr("his");                // find all occurrences of "his" ...
std::string format("her");                // ... and replace them with "her"

// use the version of regex_replace() that operates on strings
std::string output = regex_replace( input, re, format );
std::cout << output << '\n';

// use the version of regex_replace() that operates on iterators
std::ostream_iterator< char > out_iter( std::cout );
regex_replace( out_iter, input.begin(), input.end(), re, format );

The above program prints out the following:

Ther is her face
Ther is her face

Notice that all the occurrences of "his" have been replaced with "her".

Click here to see a complete example program that shows how to use regex_replace(). And check the regex_replace() reference to see a complete list of the available overloads.

Replace Options

The regex_replace() algorithm takes an optional bitmask parameter to control the formatting. The possible values of the bitmask are:

These flags live in the xpressive::regex_constants namespace. If the substitution parameter is a function object instead of a string, the flags format_literal, format_perl, format_sed, and format_all are ignored.

The ECMA-262 Format Sequences

When you haven't specified a substitution string dialect with one of the format flags above, you get the dialect defined by ECMA-262, the standard for ECMAScript. The table below shows the escape sequences recognized in ECMA-262 mode.

Any other sequence beginning with '$' simply represents itself. For example, if the format string were "$a" then "$a" would be inserted into the output sequence.

The Sed Format Sequences

When specifying the format_sed flag to regex_replace(), the following escape sequences are recognized:

The Perl Format Sequences

When specifying the format_perl flag to regex_replace(), the following escape sequences are recognized:

The Boost-Specific Format Sequences

When specifying the format_all flag to regex_replace(), the escape sequences recognized are the same as those above for format_perl. In addition, conditional expressions of the following form are recognized:

?Ntrue-expression:false-expression

where N is a decimal digit representing a sub-match. If the corresponding sub-match participated in the full match, then the substitution is true-expression. Otherwise, it is false-expression. In this mode, you can use parens () for grouping. If you want a literal paren, you must escape it as \(.

Formatter Objects

Format strings are not always expressive enough for all your text substitution needs. Consider the simple example of wanting to map input strings to output strings, as you may want to do with environment variables. Rather than a format string, for this you would use a formatter object. Consider the following code, which finds embedded environment variables of the form "$(XYZ)" and computes the substitution string by looking up the environment variable in a map.

#include <map>
#include <string>
#include <iostream>
#include <boost/xpressive/xpressive.hpp>
using namespace boost;
using namespace xpressive;

std::map<std::string, std::string> env;

std::string const &format_fun(smatch const &what)
{
    return env[what[1].str()];
}

int main()
{
    env["X"] = "this";
    env["Y"] = "that";

    std::string input("\"$(X)\" has the value \"$(Y)\"");

    // replace strings like "$(XYZ)" with the result of env["XYZ"]
    sregex envar = "$(" >> (s1 = +_w) >> ')';
    std::string output = regex_replace(input, envar, format_fun);
    std::cout << output << std::endl;
}

In this case, we use a function, format_fun() to compute the substitution string on the fly. It accepts a match_results<> object which contains the results of the current match. format_fun() uses the first submatch as a key into the global env map. The above code displays:

"this" has the value "that"

The formatter need not be an ordinary function. It may be an object of class type. And rather than return a string, it may accept an output iterator into which it writes the substitution. Consider the following, which is functionally equivalent to the above.

#include <map>
#include <string>
#include <iostream>
#include <boost/xpressive/xpressive.hpp>
using namespace boost;
using namespace xpressive;

struct formatter
{
    typedef std::map<std::string, std::string> env_map;
    env_map env;

    template<typename Out>
    Out operator()(smatch const &what, Out out) const
    {
        env_map::const_iterator where = env.find(what[1]);
        if(where != env.end())
        {
            std::string const &sub = where->second;
            out = std::copy(sub.begin(), sub.end(), out);
        }
        return out;
    }

};

int main()
{
    formatter fmt;
    fmt.env["X"] = "this";
    fmt.env["Y"] = "that";

    std::string input("\"$(X)\" has the value \"$(Y)\"");

    sregex envar = "$(" >> (s1 = +_w) >> ')';
    std::string output = regex_replace(input, envar, fmt);
    std::cout << output << std::endl;
}

The formatter must be a callable object -- a function or a function object -- that has one of three possible signatures, detailed in the table below. For the table, fmt is a function pointer or function object, what is a match_results<> object, out is an OutputIterator, and flags is a value of regex_constants::match_flag_type:

Formatter Expressions

In addition to format strings and formatter objects, regex_replace() also accepts formatter expressions. A formatter expression is a lambda expression that generates a string. It uses the same syntax as that for Semantic Actions, which are covered later. The above example, which uses regex_replace() to substitute strings for environment variables, is repeated here using a formatter expression.

#include <map>
#include <string>
#include <iostream>
#include <boost/xpressive/xpressive.hpp>
#include <boost/xpressive/regex_actions.hpp>
using namespace boost::xpressive;

int main()
{
    std::map<std::string, std::string> env;
    env["X"] = "this";
    env["Y"] = "that";

    std::string input("\"$(X)\" has the value \"$(Y)\"");

    sregex envar = "$(" >> (s1 = +_w) >> ')';
    std::string output = regex_replace(input, envar, ref(env)[s1]);
    std::cout << output << std::endl;
}

In the above, the formatter expression is ref(env)[s1]. This means to use the value of the first submatch, s1, as a key into the env map. The purpose of xpressive::ref() here is to make the reference to the env local variable lazy so that the index operation is deferred until we know what to replace s1 with.

regex_token_iterator<> is the Ginsu knife of the text manipulation world. It slices! It dices! This section describes how to use the highly-configurable regex_token_iterator<> to chop up input sequences.

Overview

You initialize a regex_token_iterator<> with an input sequence, a regex, and some optional configuration parameters. The regex_token_iterator<> will use regex_search() to find the first place in the sequence that the regex matches. When dereferenced, the regex_token_iterator<> returns a token in the form of a std::basic_string<>. Which string it returns depends on the configuration parameters. By default it returns a string corresponding to the full match, but it could also return a string corresponding to a particular marked sub-expression, or even the part of the sequence that didn't match. When you increment the regex_token_iterator<>, it will move to the next token. Which token is next depends on the configuration parameters. It could simply be a different marked sub-expression in the current match, or it could be part or all of the next match. Or it could be the part that didn't match.

As you can see, regex_token_iterator<> can do a lot. That makes it hard to describe, but some examples should make it clear.

Example 1: Simple Tokenization

This example uses regex_token_iterator<> to chop a sequence into a series of tokens consisting of words.

std::string input("This is his face");
sregex re = +_w;                      // find a word

// iterate over all the words in the input
sregex_token_iterator begin( input.begin(), input.end(), re ), end;

// write all the words to std::cout
std::ostream_iterator< std::string > out_iter( std::cout, "\n" );
std::copy( begin, end, out_iter );

This program displays the following:

This
is
his
face

Example 2: Simple Tokenization, Reloaded

This example also uses regex_token_iterator<> to chop a sequence into a series of tokens consisting of words, but it uses the regex as a delimiter. When we pass a -1 as the last parameter to the regex_token_iterator<> constructor, it instructs the token iterator to consider as tokens those parts of the input that didn't match the regex.

std::string input("This is his face");
sregex re = +_s;                      // find white space

// iterate over all non-white space in the input. Note the -1 below:
sregex_token_iterator begin( input.begin(), input.end(), re, -1 ), end;

// write all the words to std::cout
std::ostream_iterator< std::string > out_iter( std::cout, "\n" );
std::copy( begin, end, out_iter );

This program displays the following:

This
is
his
face

Example 3: Simple Tokenization, Revolutions

This example also uses regex_token_iterator<> to chop a sequence containing a bunch of dates into a series of tokens consisting of just the years. When we pass a positive integer N as the last parameter to the regex_token_iterator<> constructor, it instructs the token iterator to consider as tokens only the N-th marked sub-expression of each match.

std::string input("01/02/2003 blahblah 04/23/1999 blahblah 11/13/1981");
sregex re = sregex::compile("(\\d{2})/(\\d{2})/(\\d{4})"); // find a date

// iterate over all the years in the input. Note the 3 below, corresponding to the 3rd sub-expression:
sregex_token_iterator begin( input.begin(), input.end(), re, 3 ), end;

// write all the words to std::cout
std::ostream_iterator< std::string > out_iter( std::cout, "\n" );
std::copy( begin, end, out_iter );

This program displays the following:

2003
1999
1981

Example 4: Not-So-Simple Tokenization

This example is like the previous one, except that instead of tokenizing just the years, this program turns the days, months and years into tokens. When we pass an array of integers {I,J,...} as the last parameter to the regex_token_iterator<> constructor, it instructs the token iterator to consider as tokens the I-th, J-th, etc. marked sub-expression of each match.

std::string input("01/02/2003 blahblah 04/23/1999 blahblah 11/13/1981");
sregex re = sregex::compile("(\\d{2})/(\\d{2})/(\\d{4})"); // find a date

// iterate over the days, months and years in the input
int const sub_matches[] = { 2, 1, 3 }; // day, month, year
sregex_token_iterator begin( input.begin(), input.end(), re, sub_matches ), end;

// write all the words to std::cout
std::ostream_iterator< std::string > out_iter( std::cout, "\n" );
std::copy( begin, end, out_iter );

This program displays the following:

02
01
2003
23
04
1999
13
11
1981

The sub_matches array instructs the regex_token_iterator<> to first take the value of the 2nd sub-match, then the 1st sub-match, and finally the 3rd. Incrementing the iterator again instructs it to use regex_search() again to find the next match. At that point, the process repeats -- the token iterator takes the value of the 2nd sub-match, then the 1st, et cetera.

Overview

One of the key benefits of representing regexes as C++ expressions is the ability to easily refer to other C++ code and data from within the regex. This enables programming idioms that are not possible with other regular expression libraries. Of particular note is the ability for one regex to refer to another regex, allowing you to build grammars out of regular expressions. This section describes how to embed one regex in another by value and by reference, how regex objects behave when they refer to other regexes, and how to access the tree of results after a successful parse.

Embedding a Regex by Value

The basic_regex<> object has value semantics. When a regex object appears on the right-hand side in the definition of another regex, it is as if the regex were embedded by value; that is, a copy of the nested regex is stored by the enclosing regex. The inner regex is invoked by the outer regex during pattern matching. The inner regex participates fully in the match, back-tracking as needed to make the match succeed.

Consider a text editor that has a regex-find feature with a whole-word option. You can implement this with xpressive as follows:

find_dialog dlg;
if( dialog_ok == dlg.do_modal() )
{
    std::string pattern = dlg.get_text();          // the pattern the user entered
    bool whole_word = dlg.whole_word.is_checked(); // did the user select the whole-word option?

    sregex re = sregex::compile( pattern );        // try to compile the pattern

    if( whole_word )
    {
        // wrap the regex in begin-word / end-word assertions
        re = bow >> re >> eow;
    }

    // ... use re ...
}

Look closely at this line:

// wrap the regex in begin-word / end-word assertions
re = bow >> re >> eow;

This line creates a new regex that embeds the old regex by value. Then, the new regex is assigned back to the original regex. Since a copy of the old regex was made on the right-hand side, this works as you might expect: the new regex has the behavior of the old regex wrapped in begin- and end-word assertions.

	Note
	Note that re = bow >> re >> eow does not define a recursive regular expression, since regex objects embed by value by default. The next section shows how to define a recursive regular expression by embedding a regex by reference.

Embedding a Regex by Reference

If you want to be able to build recursive regular expressions and context-free grammars, embedding a regex by value is not enough. You need to be able to make your regular expressions self-referential. Most regular expression engines don't give you that power, but xpressive does.

Tip

The theoretical computer scientists out there will correctly point out that a self-referential regular expression is not "regular", so in the strict sense, xpressive isn't really a regular expression engine at all. But as Larry Wall once said, "the term [regular expression] has grown with the capabilities of our pattern matching engines, so I'm not going to try to fight linguistic necessity here."

Consider the following code, which uses the by_ref() helper to define a recursive regular expression that matches balanced, nested parentheses:

sregex parentheses;
parentheses                          // A balanced set of parentheses ...
    = '('                            // is an opening parenthesis ...
        >>                           // followed by ...
         *(                          // zero or more ...
            keep( +~(set='(',')') )  // of a bunch of things that are not parentheses ...
          |                          // or ...
            by_ref(parentheses)      // a balanced set of parentheses
          )                          //   (ooh, recursion!) ...
        >>                           // followed by ...
      ')'                            // a closing parenthesis
    ;

Matching balanced, nested tags is an important text processing task, and it is one that "classic" regular expressions cannot do. The by_ref() helper makes it possible. It allows one regex object to be embedded in another by reference. Since the right-hand side holds parentheses by reference, assigning the right-hand side back to parentheses creates a cycle, which will execute recursively.

Building a Grammar

Once we allow self-reference in our regular expressions, the genie is out of the bottle and all manner of fun things are possible. In particular, we can now build grammars out of regular expressions. Let's have a look at the text-book grammar example: the humble calculator.

sregex group, factor, term, expression;

group       = '(' >> by_ref(expression) >> ')';
factor      = +_d | group;
term        = factor >> *(('*' >> factor) | ('/' >> factor));
expression  = term >> *(('+' >> term) | ('-' >> term));

The regex expression defined above does something rather remarkable for a regular expression: it matches mathematical expressions. For example, if the input string were "foo 9*(10+3) bar", this pattern would match "9*(10+3)". It only matches well-formed mathematical expressions, where the parentheses are balanced and the infix operators have two arguments each. Don't try this with just any regular expression engine!

Let's take a closer look at this regular expression grammar. Notice that it is cyclic: expression is implemented in terms of term, which is implemented in terms of factor, which is implemented in terms of group, which is implemented in terms of expression, closing the loop. In general, the way to define a cyclic grammar is to forward-declare the regex objects and embed by reference those regular expressions that have not yet been initialized. In the above grammar, there is only one place where we need to reference a regex object that has not yet been initialized: the definition of group. In that place, we use by_ref() to embed expression by reference. In all other places, it is sufficient to embed the other regex objects by value, since they have already been initialized and their values will not change.

	Tip
	Embed by value if possible In general, prefer embedding regular expressions by value rather than by reference. It involves one less indirection, making your patterns match a little faster. Besides, value semantics are simpler and will make your grammars easier to reason about. Don't worry about the expense of "copying" a regex. Each regex object shares its implementation with all of its copies.

Dynamic Regex Grammars

Using regex_compiler<>, you can also build grammars out of dynamic regular expressions. You do that by creating named regexes, and referring to other regexes by name. Each regex_compiler<> instance keeps a mapping from names to regexes that have been created with it.

You can create a named dynamic regex by prefacing your regex with "(?$name=)", where name is the name of the regex. You can refer to a named regex from another regex with "(?$name)". The named regex does not need to exist yet at the time it is referenced in another regex, but it must exist by the time you use the regex.

Below is a code fragment that uses dynamic regex grammars to implement the calculator example from above.

using namespace boost::xpressive;
using namespace regex_constants;

sregex expr;

{
     sregex_compiler compiler;
     syntax_option_type x = ignore_white_space;

            compiler.compile("(? $group  = ) \\( (? $expr ) \\) ", x);
            compiler.compile("(? $factor = ) \\d+ | (? $group ) ", x);
            compiler.compile("(? $term   = ) (? $factor )"
                             " ( \\* (? $factor ) | / (? $factor ) )* ", x);
     expr = compiler.compile("(? $expr   = ) (? $term )"
                             "   ( \\+ (? $term ) | - (? $term )   )* ", x);
}

std::string str("foo 9*(10+3) bar");
smatch what;

if(regex_search(str, what, expr))
{
     // This prints "9*(10+3)":
     std::cout << what[0] << std::endl;
}

As with static regex grammars, nested regex invocations create nested match results (see Nested Results below). The result is a complete parse tree for string that matched. Unlike static regexes, dynamic regexes are always embedded by reference, not by value.

Cyclic Patterns, Copying and Memory Management, Oh My!

The calculator examples above raises a number of very complicated memory-management issues. Each of the four regex objects refer to each other, some directly and some indirectly, some by value and some by reference. What if we were to return one of them from a function and let the others go out of scope? What becomes of the references? The answer is that the regex objects are internally reference counted, such that they keep their referenced regex objects alive as long as they need them. So passing a regex object by value is never a problem, even if it refers to other regex objects that have gone out of scope.

Those of you who have dealt with reference counting are probably familiar with its Achilles Heel: cyclic references. If regex objects are reference counted, what happens to cycles like the one created in the calculator examples? Are they leaked? The answer is no, they are not leaked. The basic_regex<> object has some tricky reference tracking code that ensures that even cyclic regex grammars are cleaned up when the last external reference goes away. So don't worry about it. Create cyclic grammars, pass your regex objects around and copy them all you want. It is fast and efficient and guaranteed not to leak or result in dangling references.

Nested Regexes and Sub-Match Scoping

Nested regular expressions raise the issue of sub-match scoping. If both the inner and outer regex write to and read from the same sub-match vector, chaos would ensue. The inner regex would stomp on the sub-matches written by the outer regex. For example, what does this do?

sregex inner = sregex::compile( "(.)\\1" );
sregex outer = (s1= _) >> inner >> s1;

The author probably didn't intend for the inner regex to overwrite the sub-match written by the outer regex. The problem is particularly acute when the inner regex is accepted from the user as input. The author has no way of knowing whether the inner regex will stomp the sub-match vector or not. This is clearly not acceptable.

Instead, what actually happens is that each invocation of a nested regex gets its own scope. Sub-matches belong to that scope. That is, each nested regex invocation gets its own copy of the sub-match vector to play with, so there is no way for an inner regex to stomp on the sub-matches of an outer regex. So, for example, the regex outer defined above would match "ABBA", as it should.

Nested Results

If nested regexes have their own sub-matches, there should be a way to access them after a successful match. In fact, there is. After a regex_match() or regex_search(), the match_results<> struct behaves like the head of a tree of nested results. The match_results<> class provides a nested_results() member function that returns an ordered sequence of match_results<> structures, representing the results of the nested regexes. The order of the nested results is the same as the order in which the nested regex objects matched.

Take as an example the regex for balanced, nested parentheses we saw earlier:

sregex parentheses;
parentheses = '(' >> *( keep( +~(set='(',')') ) | by_ref(parentheses) ) >> ')';

smatch what;
std::string str( "blah blah( a(b)c (c(e)f (g)h )i (j)6 )blah" );

if( regex_search( str, what, parentheses ) )
{
    // display the whole match
    std::cout << what[0] << '\n';

    // display the nested results
    std::for_each(
        what.nested_results().begin(),
        what.nested_results().end(),
        output_nested_results() );
}

This program displays the following:

( a(b)c (c(e)f (g)h )i (j)6 )
    (b)
    (c(e)f (g)h )
        (e)
        (g)
    (j)

Here you can see how the results are nested and that they are stored in the order in which they are found.

	Tip
	See the definition of output_nested_results in the Examples section.

Filtering Nested Results

Sometimes a regex will have several nested regex objects, and you want to know which result corresponds to which regex object. That's where basic_regex<>::regex_id() and match_results<>::regex_id() come in handy. When iterating over the nested results, you can compare the regex id from the results to the id of the regex object you're interested in.

To make this a bit easier, xpressive provides a predicate to make it simple to iterate over just the results that correspond to a certain nested regex. It is called regex_id_filter_predicate, and it is intended to be used with Boost.Iterator. You can use it as follows:

sregex name = +alpha;
sregex integer = +_d;
sregex re = *( *_s >> ( name | integer ) );

smatch what;
std::string str( "marsha 123 jan 456 cindy 789" );

if( regex_match( str, what, re ) )
{
    smatch::nested_results_type::const_iterator begin = what.nested_results().begin();
    smatch::nested_results_type::const_iterator end   = what.nested_results().end();

    // declare filter predicates to select just the names or the integers
    sregex_id_filter_predicate name_id( name.regex_id() );
    sregex_id_filter_predicate integer_id( integer.regex_id() );

    // iterate over only the results from the name regex
    std::for_each(
        boost::make_filter_iterator( name_id, begin, end ),
        boost::make_filter_iterator( name_id, end, end ),
        output_result
        );

    std::cout << '\n';

    // iterate over only the results from the integer regex
    std::for_each(
        boost::make_filter_iterator( integer_id, begin, end ),
        boost::make_filter_iterator( integer_id, end, end ),
        output_result
        );
}

where output_results is a simple function that takes a smatch and displays the full match. Notice how we use the regex_id_filter_predicate together with basic_regex<>::regex_id() and boost::make_filter_iterator() from the Boost.Iterator to select only those results corresponding to a particular nested regex. This program displays the following:

marsha
jan
cindy
123
456
789

Overview

Imagine you want to parse an input string and build a std::map<> from it. For something like that, matching a regular expression isn't enough. You want to do something when parts of your regular expression match. Xpressive lets you attach semantic actions to parts of your static regular expressions. This section shows you how.

Semantic Actions

Consider the following code, which uses xpressive's semantic actions to parse a string of word/integer pairs and stuffs them into a std::map<>. It is described below.

#include <string>
#include <iostream>
#include <boost/xpressive/xpressive.hpp>
#include <boost/xpressive/regex_actions.hpp>
using namespace boost::xpressive;

int main()
{
    std::map<std::string, int> result;
    std::string str("aaa=>1 bbb=>23 ccc=>456");

    // Match a word and an integer, separated by =>,
    // and then stuff the result into a std::map<>
    sregex pair = ( (s1= +_w) >> "=>" >> (s2= +_d) )
        [ ref(result)[s1] = as<int>(s2) ];

    // Match one or more word/integer pairs, separated
    // by whitespace.
    sregex rx = pair >> *(+_s >> pair);

    if(regex_match(str, rx))
    {
        std::cout << result["aaa"] << '\n';
        std::cout << result["bbb"] << '\n';
        std::cout << result["ccc"] << '\n';
    }

    return 0;
}

This program prints the following:

1
23
456

The regular expression pair has two parts: the pattern and the action. The pattern says to match a word, capturing it in sub-match 1, and an integer, capturing it in sub-match 2, separated by "=>". The action is the part in square brackets: [ ref(result)[s1] = as<int>(s2) ]. It says to take sub-match one and use it to index into the results map, and assign to it the result of converting sub-match 2 to an integer.

	Note
	To use semantic actions with your static regexes, you must #include <boost/xpressive/regex_actions.hpp>

How does this work? Just as the rest of the static regular expression, the part between brackets is an expression template. It encodes the action and executes it later. The expression ref(result) creates a lazy reference to the result object. The larger expression ref(result)[s1] is a lazy map index operation. Later, when this action is getting executed, s1 gets replaced with the first sub_match<>. Likewise, when as<int>(s2) gets executed, s2 is replaced with the second sub_match<>. The as<> action converts its argument to the requested type using Boost.Lexical_cast. The effect of the whole action is to insert a new word/integer pair into the map.

Note

There is an important difference between the function boost::ref() in <boost/ref.hpp> and boost::xpressive::ref() in <boost/xpressive/regex_actions.hpp>. The first returns a plain reference_wrapper<> which behaves in many respects like an ordinary reference. By contrast, boost::xpressive::ref() returns a lazy reference that you can use in expressions that are executed lazily. That is why we can say ref(result)[s1], even though result doesn't have an operator[] that would accept s1.

In addition to the sub-match placeholders s1, s2, etc., you can also use the placeholder _ within an action to refer back to the string matched by the sub-expression to which the action is attached. For instance, you can use the following regex to match a bunch of digits, interpret them as an integer and assign the result to a local variable:

int i = 0;
// Here, _ refers back to all the
// characters matched by (+_d)
sregex rex = (+_d)[ ref(i) = as<int>(_) ];

Lazy Action Execution

What does it mean, exactly, to attach an action to part of a regular expression and perform a match? When does the action execute? If the action is part of a repeated sub-expression, does the action execute once or many times? And if the sub-expression initially matches, but ultimately fails because the rest of the regular expression fails to match, is the action executed at all?

The answer is that by default, actions are executed lazily. When a sub-expression matches a string, its action is placed on a queue, along with the current values of any sub-matches to which the action refers. If the match algorithm must backtrack, actions are popped off the queue as necessary. Only after the entire regex has matched successfully are the actions actually exeucted. They are executed all at once, in the order in which they were added to the queue, as the last step before regex_match() returns.

For example, consider the following regex that increments a counter whenever it finds a digit.

int i = 0;
std::string str("1!2!3?");
// count the exciting digits, but not the
// questionable ones.
sregex rex = +( _d [ ++ref(i) ] >> '!' );
regex_search(str, rex);
assert( i == 2 );

The action ++ref(i) is queued three times: once for each found digit. But it is only executed twice: once for each digit that precedes a '!' character. When the '?' character is encountered, the match algorithm backtracks, removing the final action from the queue.

Immediate Action Execution

When you want semantic actions to execute immediately, you can wrap the sub-expression containing the action in a keep(). keep() turns off back-tracking for its sub-expression, but it also causes any actions queued by the sub-expression to execute at the end of the keep(). It is as if the sub-expression in the keep() were compiled into an independent regex object, and matching the keep() is like a separate invocation of regex_search(). It matches characters and executes actions but never backtracks or unwinds. For example, imagine the above example had been written as follows:

int i = 0;
std::string str("1!2!3?");
// count all the digits.
sregex rex = +( keep( _d [ ++ref(i) ] ) >> '!' );
regex_search(str, rex);
assert( i == 3 );

We have wrapped the sub-expression _d [ ++ref(i) ] in keep(). Now, whenever this regex matches a digit, the action will be queued and then immediately executed before we try to match a '!' character. In this case, the action executes three times.

	Note
	Like keep(), actions within before() and after() are also executed early when their sub-expressions have matched.

Lazy Functions

So far, we've seen how to write semantic actions consisting of variables and operators. But what if you want to be able to call a function from a semantic action? Xpressive provides a mechanism to do this.

The first step is to define a function object type. Here, for instance, is a function object type that calls push() on its argument:

struct push_impl
{
    // Result type, needed for tr1::result_of
    typedef void result_type;

    template<typename Sequence, typename Value>
    void operator()(Sequence &seq, Value const &val) const
    {
        seq.push(val);
    }
};

The next step is to use xpressive's function<> template to define a function object named push:

// Global "push" function object.
function<push_impl>::type const push = {{}};

The initialization looks a bit odd, but this is because push is being statically initialized. That means it doesn't need to be constructed at runtime. We can use push in semantic actions as follows:

std::stack<int> ints;
// Match digits, cast them to an int
// and push it on the stack.
sregex rex = (+_d)[push(ref(ints), as<int>(_))];

You'll notice that doing it this way causes member function invocations to look like ordinary function invocations. You can choose to write your semantic action in a different way that makes it look a bit more like a member function call:

sregex rex = (+_d)[ref(ints)->*push(as<int>(_))];

Xpressive recognizes the use of the ->* and treats this expression exactly the same as the one above.

When your function object must return a type that depends on its arguments, you can use a result<> member template instead of the result_type typedef. Here, for example, is a first function object that returns the first member of a std::pair<> or sub_match<>:

// Function object that returns the
// first element of a pair.
struct first_impl
{
    template<typename Sig> struct result {};

    template<typename This, typename Pair>
    struct result<This(Pair)>
    {
        typedef typename remove_reference<Pair>
            ::type::first_type type;
    };

    template<typename Pair>
    typename Pair::first_type
    operator()(Pair const &p) const
    {
        return p.first;
    }
};

// OK, use as first(s1) to get the begin iterator
// of the sub-match referred to by s1.
function<first_impl>::type const first = {{}};

Referring to Local Variables

As we've seen in the examples above, we can refer to local variables within an actions using xpressive::ref(). Any such variables are held by reference by the regular expression, and care should be taken to avoid letting those references dangle. For instance, in the following code, the reference to i is left to dangle when bad_voodoo() returns:

sregex bad_voodoo()
{
    int i = 0;
    sregex rex = +( _d [ ++ref(i) ] >> '!' );
    // ERROR! rex refers by reference to a local
    // variable, which will dangle after bad_voodoo()
    // returns.
    return rex;
}

When writing semantic actions, it is your responsibility to make sure that all the references do not dangle. One way to do that would be to make the variables shared pointers that are held by the regex by value.

sregex good_voodoo(boost::shared_ptr<int> pi)
{
    // Use val() to hold the shared_ptr by value:
    sregex rex = +( _d [ ++*val(pi) ] >> '!' );
    // OK, rex holds a reference count to the integer.
    return rex;
}

In the above code, we use xpressive::val() to hold the shared pointer by value. That's not normally necessary because local variables appearing in actions are held by value by default, but in this case, it is necessary. Had we written the action as ++*pi, it would have executed immediately. That's because ++*pi is not an expression template, but ++*val(pi) is.

It can be tedious to wrap all your variables in ref() and val() in your semantic actions. Xpressive provides the reference<> and value<> templates to make things easier. The following table shows the equivalencies:

int i = 0;

sregex rex = +( _d [ ++ref(i) ] >> '!' );

int i = 0;
reference<int> ri(i);
sregex rex = +( _d [ ++ri ] >> '!' );

boost::shared_ptr<int> pi(new int(0));

sregex rex = +( _d [ ++*val(pi) ] >> '!' );

boost::shared_ptr<int> pi(new int(0));
value<boost::shared_ptr<int> > vpi(pi);
sregex rex = +( _d [ ++*vpi ] >> '!' );

As you can see, when using reference<>, you need to first declare a local variable and then declare a reference<> to it. These two steps can be combined into one using local<>.

local<int> i(0);

sregex rex = +( _d [ ++i ] >> '!' );

int i = 0;
reference<int> ri(i);
sregex rex = +( _d [ ++ri ] >> '!' );

We can use local<> to rewrite the above example as follows:

local<int> i(0);
std::string str("1!2!3?");
// count the exciting digits, but not the
// questionable ones.
sregex rex = +( _d [ ++i ] >> '!' );
regex_search(str, rex);
assert( i.get() == 2 );

Notice that we use local<>::get() to access the value of the local variable. Also, beware that local<> can be used to create a dangling reference, just as reference<> can.

Referring to Non-Local Variables

In the beginning of this section, we used a regex with a semantic action to parse a string of word/integer pairs and stuff them into a std::map<>. That required that the map and the regex be defined together and used before either could go out of scope. What if we wanted to define the regex once and use it to fill lots of different maps? We would rather pass the map into the regex_match() algorithm rather than embed a reference to it directly in the regex object. What we can do instead is define a placeholder and use that in the semantic action instead of the map itself. Later, when we call one of the regex algorithms, we can bind the reference to an actual map object. The following code shows how.

// Define a placeholder for a map object:
placeholder<std::map<std::string, int> > _map;

// Match a word and an integer, separated by =>,
// and then stuff the result into a std::map<>
sregex pair = ( (s1= +_w) >> "=>" >> (s2= +_d) )
    [ _map[s1] = as<int>(s2) ];

// Match one or more word/integer pairs, separated
// by whitespace.
sregex rx = pair >> *(+_s >> pair);

// The string to parse
std::string str("aaa=>1 bbb=>23 ccc=>456");

// Here is the actual map to fill in:
std::map<std::string, int> result;

// Bind the _map placeholder to the actual map
smatch what;
what.let( _map = result );

// Execute the match and fill in result map
if(regex_match(str, what, rx))
{
    std::cout << result["aaa"] << '\n';
    std::cout << result["bbb"] << '\n';
    std::cout << result["ccc"] << '\n';
}

This program displays:

1
23
456

We use placeholder<> here to define _map, which stands in for a std::map<> variable. We can use the placeholder in the semantic action as if it were a map. Then, we define a match_results<> struct and bind an actual map to the placeholder with "what.let( _map = result );". The regex_match() call behaves as if the placeholder in the semantic action had been replaced with a reference to result.

	Note
	Placeholders in semantic actions are not actually replaced at runtime with references to variables. The regex object is never mutated in any way during any of the regex algorithms, so they are safe to use in multiple threads.

The syntax for late-bound action arguments is a little different if you are using regex_iterator<> or regex_token_iterator<>. The regex iterators accept an extra constructor parameter for specifying the argument bindings. There is a let() function that you can use to bind variables to their placeholders. The following code demonstrates how.

// Define a placeholder for a map object:
placeholder<std::map<std::string, int> > _map;

// Match a word and an integer, separated by =>,
// and then stuff the result into a std::map<>
sregex pair = ( (s1= +_w) >> "=>" >> (s2= +_d) )
    [ _map[s1] = as<int>(s2) ];

// The string to parse
std::string str("aaa=>1 bbb=>23 ccc=>456");

// Here is the actual map to fill in:
std::map<std::string, int> result;

// Create a regex_iterator to find all the matches
sregex_iterator it(str.begin(), str.end(), pair, let(_map=result));
sregex_iterator end;

// step through all the matches, and fill in
// the result map
while(it != end)
    ++it;

std::cout << result["aaa"] << '\n';
std::cout << result["bbb"] << '\n';
std::cout << result["ccc"] << '\n';

This program displays:

1
23
456

User-Defined Assertions

You are probably already familiar with regular expression assertions. In Perl, some examples are the ^ and $ assertions, which you can use to match the beginning and end of a string, respectively. Xpressive lets you define your own assertions. A custom assertion is a contition which must be true at a point in the match in order for the match to succeed. You can check a custom assertion with xpressive's check() function.

There are a couple of ways to define a custom assertion. The simplest is to use a function object. Let's say that you want to ensure that a sub-expression matches a sub-string that is either 3 or 6 characters long. The following struct defines such a predicate:

// A predicate that is true IFF a sub-match is
// either 3 or 6 characters long.
struct three_or_six
{
    bool operator()(ssub_match const &sub) const
    {
        return sub.length() == 3 || sub.length() == 6;
    }
};

You can use this predicate within a regular expression as follows:

// match words of 3 characters or 6 characters.
sregex rx = (bow >> +_w >> eow)[ check(three_or_six()) ] ;

The above regular expression will find whole words that are either 3 or 6 characters long. The three_or_six predicate accepts a sub_match<> that refers back to the part of the string matched by the sub-expression to which the custom assertion is attached.

	Note
	The custom assertion participates in determining whether the match succeeds or fails. Unlike actions, which execute lazily, custom assertions execute immediately while the regex engine is searching for a match.

Custom assertions can also be defined inline using the same syntax as for semantic actions. Below is the same custom assertion written inline:

// match words of 3 characters or 6 characters.
sregex rx = (bow >> +_w >> eow)[ check(length(_)==3 || length(_)==6) ] ;

In the above, length() is a lazy function that calls the length() member function of its argument, and _ is a placeholder that receives the sub_match.

Once you get the hang of writing custom assertions inline, they can be very powerful. For example, you can write a regular expression that only matches valid dates (for some suitably liberal definition of the term “valid”).

int const days_per_month[] =
    {31, 29, 31, 30, 31, 30, 31, 31, 30, 31, 31, 31};

mark_tag month(1), day(2);
// find a valid date of the form month/day/year.
sregex date =
    (
        // Month must be between 1 and 12 inclusive
        (month= _d >> !_d)     [ check(as<int>(_) >= 1
                                    && as<int>(_) <= 12) ]
    >>  '/'
        // Day must be between 1 and 31 inclusive
    >>  (day=   _d >> !_d)     [ check(as<int>(_) >= 1
                                    && as<int>(_) <= 31) ]
    >>  '/'
        // Only consider years between 1970 and 2038
    >>  (_d >> _d >> _d >> _d) [ check(as<int>(_) >= 1970
                                    && as<int>(_) <= 2038) ]
    )
    // Ensure the month actually has that many days!
    [ check( ref(days_per_month)[as<int>(month)-1] >= as<int>(day) ) ]
;

smatch what;
std::string str("99/99/9999 2/30/2006 2/28/2006");

if(regex_search(str, what, date))
{
    std::cout << what[0] << std::endl;
}

The above program prints out the following:

2/28/2006

Notice how the inline custom assertions are used to range-check the values for the month, day and year. The regular expression doesn't match "99/99/9999" or "2/30/2006" because they are not valid dates. (There is no 99th month, and February doesn't have 30 days.)

Overview

Symbol tables can be built into xpressive regular expressions with just a std::map<>. The map keys are the strings to be matched and the map values are the data to be returned to your semantic action. Xpressive attributes, named a1, a2, through a9, hold the value corresponding to a matching key so that it can be used in a semantic action. A default value can be specified for an attribute if a symbol is not found.

Symbol Tables

An xpressive symbol table is just a std::map<>, where the key is a string type and the value can be anything. For example, the following regular expression matches a key from map1 and assigns the corresponding value to the attribute a1. Then, in the semantic action, it assigns the value stored in attribute a1 to an integer result.

int result;
std::map<std::string, int> map1;
// ... (fill the map)
sregex rx = ( a1 = map1 ) [ ref(result) = a1 ];

Consider the following example code, which translates number names into integers. It is described below.

#include <string>
#include <iostream>
#include <boost/xpressive/xpressive.hpp>
#include <boost/xpressive/regex_actions.hpp>
using namespace boost::xpressive;

int main()
{
    std::map<std::string, int> number_map;
    number_map["one"] = 1;
    number_map["two"] = 2;
    number_map["three"] = 3;
    // Match a string from number_map
    // and store the integer value in 'result'
    // if not found, store -1 in 'result'
    int result = 0;
    cregex rx = ((a1 = number_map ) | *_)
        [ ref(result) = (a1 | -1)];

    regex_match("three", rx);
    std::cout << result << '\n';
    regex_match("two", rx);
    std::cout << result << '\n';
    regex_match("stuff", rx);
    std::cout << result << '\n';
    return 0;
}

This program prints the following:

3
2
-1

First the program builds a number map, with number names as string keys and the corresponding integers as values. Then it constructs a static regular expression using an attribute a1 to represent the result of the symbol table lookup. In the semantic action, the attribute is assigned to an integer variable result. If the symbol was not found, a default value of -1 is assigned to result. A wildcard, *_, makes sure the regex matches even if the symbol is not found.

A more complete version of this example can be found in libs/xpressive/example/numbers.cpp ^[5] . It translates number names up to "nine hundred ninety nine million nine hundred ninety nine thousand nine hundred ninety nine" along with some special number names like "dozen".

Symbol table matches are case sensitive by default, but they can be made case-insensitive by enclosing the expression in icase().

Attributes

Up to nine attributes can be used in a regular expression. They are named a1, a2, ..., a9 in the boost::xpressive namespace. The attribute type is the same as the second component of the map that is assigned to it. A default value for an attribute can be specified in a semantic action with the syntax (a1 | default-value).

Attributes are properly scoped, so you can do crazy things like: ( (a1=sym1) >> (a1=sym2)[ref(x)=a1] )[ref(y)=a1]. The inner semantic action sees the inner a1, and the outer semantic action sees the outer one. They can even have different types.

	Note
	Xpressive builds a hidden ternary search trie from the map so it can search quickly. If BOOST_DISABLE_THREADS is defined, the hidden ternary search trie "self adjusts", so after each search it restructures itself to improve the efficiency of future searches based on the frequency of previous searches.

Overview

Matching a regular expression against a string often requires locale-dependent information. For example, how are case-insensitive comparisons performed? The locale-sensitive behavior is captured in a traits class. xpressive provides three traits class templates: cpp_regex_traits<>, c_regex_traits<> and null_regex_traits<>. The first wraps a std::locale, the second wraps the global C locale, and the third is a stub traits type for use when searching non-character data. All traits templates conform to the Regex Traits Concept.

Setting the Default Regex Trait

By default, xpressive uses cpp_regex_traits<> for all patterns. This causes all regex objects to use the global std::locale. If you compile with BOOST_XPRESSIVE_USE_C_TRAITS defined, then xpressive will use c_regex_traits<> by default.

Using Custom Traits with Dynamic Regexes

To create a dynamic regex that uses a custom traits object, you must use regex_compiler<>. The basic steps are shown in the following example:

// Declare a regex_compiler that uses the global C locale
regex_compiler<char const *, c_regex_traits<char> > crxcomp;
cregex crx = crxcomp.compile( "\\w+" );

// Declare a regex_compiler that uses a custom std::locale
std::locale loc = /* ... create a locale here ... */;
regex_compiler<char const *, cpp_regex_traits<char> > cpprxcomp(loc);
cregex cpprx = cpprxcomp.compile( "\\w+" );

The regex_compiler objects act as regex factories. Once they have been imbued with a locale, every regex object they create will use that locale.

Using Custom Traits with Static Regexes

If you want a particular static regex to use a different set of traits, you can use the special imbue() pattern modifier. For instance:

// Define a regex that uses the global C locale
c_regex_traits<char> ctraits;
sregex crx = imbue(ctraits)( +_w );

// Define a regex that uses a customized std::locale
std::locale loc = /* ... create a locale here ... */;
cpp_regex_traits<char> cpptraits(loc);
sregex cpprx1 = imbue(cpptraits)( +_w );

// A shorthand for above
sregex cpprx2 = imbue(loc)( +_w );

The imbue() pattern modifier must wrap the entire pattern. It is an error to imbue only part of a static regex. For example:

// ERROR! Cannot imbue() only part of a regex
sregex error = _w >> imbue(loc)( _w );

Searching Non-Character Data With null_regex_traits

With xpressive static regexes, you are not limitted to searching for patterns in character sequences. You can search for patterns in raw bytes, integers, or anything that conforms to the Char Concept. The null_regex_traits<> makes it simple. It is a stub implementation of the Regex Traits Concept. It recognizes no character classes and does no case-sensitive mappings.

For example, with null_regex_traits<>, you can write a static regex to find a pattern in a sequence of integers as follows:

// some integral data to search
int const data[] = {0, 1, 2, 3, 4, 5, 6};

// create a null_regex_traits<> object for searching integers ...
null_regex_traits<int> nul;

// imbue a regex object with the null_regex_traits ...
basic_regex<int const *> rex = imbue(nul)(1 >> +((set= 2,3) | 4) >> 5);
match_results<int const *> what;

// search for the pattern in the array of integers ...
regex_search(data, data + 7, what, rex);

assert(what[0].matched);
assert(*what[0].first == 1);
assert(*what[0].second == 6);

Squeeze the most performance out of xpressive with these tips and tricks.

Use Static Regexes

On average, static regexes execute about 10 to 15% faster than their dynamic counterparts. It's worth familiarizing yourself with the static regex dialect.

Reuse match_results<> Objects

The match_results<> object caches dynamically allocated memory. For this reason, it is far better to reuse the same match_results<> object if you have to do many regex searches.

Caveat: match_results<> objects are not thread-safe, so don't go wild reusing them across threads.

Prefer Algorithms That Take A match_results<> Object

This is a corollary to the previous tip. If you are doing multiple searches, you should prefer the regex algorithms that accept a match_results<> object over the ones that don't, and you should reuse the same match_results<> object each time. If you don't provide a match_results<> object, a temporary one will be created for you and discarded when the algorithm returns. Any memory cached in the object will be deallocated and will have to be reallocated the next time.

Prefer Algorithms That Accept Iterator Ranges Over Null-Terminated Strings

xpressive provides overloads of the regex_match() and regex_search() algorithms that operate on C-style null-terminated strings. You should prefer the overloads that take iterator ranges. When you pass a null-terminated string to a regex algorithm, the end iterator is calculated immediately by calling strlen. If you already know the length of the string, you can avoid this overhead by calling the regex algorithms with a [begin, end) pair.

Compile Patterns Once And Reuse Them

Compiling a regex (dynamic or static) is more expensive than executing a match or search. If you have the option, prefer to compile a pattern into a basic_regex<> object once and reuse it rather than recreating it over and over.

Understand syntax_option_type::optimize

The optimize flag tells the regex compiler to spend some extra time analyzing the pattern. It can cause some patterns to execute faster, but it increases the time to compile the pattern, and often increases the amount of memory consumed by the pattern. If you plan to reuse your pattern, optimize is usually a win. If you will only use the pattern once, don't use optimize.

Common Pitfalls

Keep the following tips in mind to avoid stepping in potholes with xpressive.

Create Grammars On A Single Thread

With static regexes, you can create grammars by nesting regexes inside one another. When compiling the outer regex, both the outer and inner regex objects, and all the regex objects to which they refer either directly or indirectly, are modified. For this reason, it's dangerous for global regex objects to participate in grammars. It's best to build regex grammars from a single thread. Once built, the resulting regex grammar can be executed from multiple threads without problems.

Beware Nested Quantifiers

This is a pitfall common to many regular expression engines. Some patterns can cause exponentially bad performance. Often these patterns involve one quantified term nested withing another quantifier, such as "(a*)*", although in many cases, the problem is harder to spot. Beware of patterns that have nested quantifiers.

CharT requirements

If type BidiIterT is used as a template argument to basic_regex<>, then CharT is iterator_traits<BidiIterT>::value_type. Type CharT must have a trivial default constructor, copy constructor, assignment operator, and destructor. In addition the following requirements must be met for objects; c of type CharT, c1 and c2 of type CharT const, and i of type int:

Traits Requirements

In the following table X denotes a traits class defining types and functions for the character container type CharT; u is an object of type X; v is an object of type const X; p is a value of type const CharT*; I1 and I2 are Input Iterators; c is a value of type const CharT; s is an object of type X::string_type; cs is an object of type const X::string_type; b is a value of type bool; i is a value of type int; F1 and F2 are values of type const CharT*; loc is an object of type X::locale_type; and ch is an object of const char.