[/ 
  Copyright 2006-2007 John Maddock.
  Distributed under the Boost Software License, Version 1.0.
  (See accompanying file LICENSE_1_0.txt or copy at
  http://www.boost.org/LICENSE_1_0.txt).
]


[section:icu Working With Unicode and ICU String Types]

[section:intro Introduction to using Regex with ICU]

The header:

   <boost/regex/icu.hpp>

contains the data types and algorithms necessary for working with regular 
expressions in a Unicode aware environment. 

In order to use this header you will need the 
[@http://www.ibm.com/software/globalization/icu/ ICU library], and you will need 
to have built the Boost.Regex library with 
[link boost_regex.install.building_with_unicode_and_icu_support ICU support enabled].

The header will enable you to:

* Create regular expressions that treat Unicode strings as sequences of UTF-32 code points.
* Create regular expressions that support various Unicode data properties, including character classification.
* Transparently search Unicode strings that are encoded as either UTF-8, UTF-16 or UTF-32.

[endsect]

[section:unicode_types Unicode regular expression types]

Header `<boost/regex/icu.hpp>` provides a regular expression traits class that 
handles UTF-32 characters:

   class icu_regex_traits;

and a regular expression type based upon that:

   typedef basic_regex<UChar32,icu_regex_traits> u32regex;

The type `u32regex` is regular expression type to use for all Unicode 
regular expressions; internally it uses UTF-32 code points, but can be 
created from, and used to search, either UTF-8, or UTF-16 encoded strings 
as well as UTF-32 ones.

The constructors, and assign member functions of `u32regex`, require UTF-32 
encoded strings, but there are a series of overloaded algorithms called 
`make_u32regex` which allow regular expressions to be created from 
UTF-8, UTF-16, or UTF-32 encoded strings:

   template <class InputIterator> 
   u32regex make_u32regex(InputIterator i, 
                          InputIterator j, 
                          boost::regex_constants::syntax_option_type opt);

[*Effects]: Creates a regular expression object from the iterator sequence \[i,j). 
The character encoding of the sequence is determined based upon sizeof(*i): 
1 implies UTF-8, 2 implies UTF-16, and 4 implies UTF-32.

   u32regex make_u32regex(const char* p, 
                          boost::regex_constants::syntax_option_type opt 
                              = boost::regex_constants::perl);

[*Effects]: Creates a regular expression object from the Null-terminated 
UTF-8 characater sequence /p/.

   u32regex make_u32regex(const unsigned char* p, 
                          boost::regex_constants::syntax_option_type opt 
                              = boost::regex_constants::perl);

[*Effects]: Creates a regular expression object from the Null-terminated UTF-8 characater sequence p.

   u32regex make_u32regex(const wchar_t* p, 
                          boost::regex_constants::syntax_option_type opt 
                              = boost::regex_constants::perl);

[*Effects]: Creates a regular expression object from the Null-terminated characater sequence p.  The character encoding of the sequence is determined based upon sizeof(wchar_t): 1 implies UTF-8, 2 implies UTF-16, and 4 implies UTF-32.

   u32regex make_u32regex(const UChar* p, 
                          boost::regex_constants::syntax_option_type opt 
                              = boost::regex_constants::perl);

[*Effects]: Creates a regular expression object from the Null-terminated UTF-16 characater sequence p.

   template<class C, class T, class A>
   u32regex make_u32regex(const std::basic_string<C, T, A>& s, 
                          boost::regex_constants::syntax_option_type opt 
                              = boost::regex_constants::perl);

[*Effects]: Creates a regular expression object from the string s.  The character encoding of the string is determined based upon sizeof(C): 1 implies UTF-8, 2 implies UTF-16, and 4 implies UTF-32.

   u32regex make_u32regex(const UnicodeString& s, 
                          boost::regex_constants::syntax_option_type opt 
                              = boost::regex_constants::perl);

[*Effects]: Creates a regular expression object from the UTF-16 encoding string s.

[endsect]

[section:unicode_algo Unicode Regular Expression Algorithms]

The regular expression algorithms [regex_match], [regex_search] and [regex_replace] 
all expect that the character sequence upon which they operate, 
is encoded in the same character encoding as the regular expression object 
with which they are used.  For Unicode regular expressions that behavior is 
undesirable: while we may want to process the data in UTF-32 "chunks", the 
actual data is much more likely to encoded as either UTF-8 or UTF-16.  
Therefore the header <boost/regex/icu.hpp> provides a series of thin wrappers 
around these algorithms, called `u32regex_match`, `u32regex_search`, and 
`u32regex_replace`.  These wrappers use iterator-adapters internally to 
make external UTF-8 or UTF-16 data look as though it's really a UTF-32 sequence, 
that can then be passed on to the "real" algorithm.

[h4 u32regex_match]

For each [regex_match] algorithm defined by `<boost/regex.hpp>`, then 
`<boost/regex/icu.hpp>` defines an overloaded algorithm that takes the 
same arguments, but which is called `u32regex_match`, and which will 
accept UTF-8, UTF-16 or UTF-32 encoded data, as well as an 
ICU UnicodeString as input.

Example: match a password, encoded in a UTF-16 UnicodeString:

   //
   // Find out if *password* meets our password requirements,
   // as defined by the regular expression *requirements*.
   //
   bool is_valid_password(const UnicodeString& password, const UnicodeString& requirements)
   {
      return boost::u32regex_match(password, boost::make_u32regex(requirements));
   }

Example: match a UTF-8 encoded filename:

   //
   // Extract filename part of a path from a UTF-8 encoded std::string and return the result
   // as another std::string:
   //
   std::string get_filename(const std::string& path)
   {
      boost::u32regex r = boost::make_u32regex("(?:\\A|.*\\\\)([^\\\\]+)");
      boost::smatch what;
      if(boost::u32regex_match(path, what, r))
      {
         // extract $1 as a std::string:
         return what.str(1);
      }
      else
      {
         throw std::runtime_error("Invalid pathname");
      }
   }

[h4 u32regex_search]

For each [regex_search] algorithm defined by `<boost/regex.hpp>`, then 
`<boost/regex/icu.hpp>` defines an overloaded algorithm that takes the 
same arguments, but which is called `u32regex_search`, and which will 
accept UTF-8, UTF-16 or UTF-32 encoded data, as well as an ICU 
UnicodeString as input.

Example: search for a character sequence in a specific language block:

   UnicodeString extract_greek(const UnicodeString& text)
   {
      // searches through some UTF-16 encoded text for a block encoded in Greek,
      // this expression is imperfect, but the best we can do for now - searching
      // for specific scripts is actually pretty hard to do right.
      //
      // Here we search for a character sequence that begins with a Greek letter,
      // and continues with characters that are either not-letters ( [^[:L*:]] )
      // or are characters in the Greek character block ( [\\x{370}-\\x{3FF}] ).
      //
      boost::u32regex r = boost::make_u32regex(
            L"[\\x{370}-\\x{3FF}](?:[^[:L*:]]|[\\x{370}-\\x{3FF}])*");
      boost::u16match what;
      if(boost::u32regex_search(text, what, r))
      {
         // extract $0 as a UnicodeString:
         return UnicodeString(what[0].first, what.length(0));
      }
      else
      {
         throw std::runtime_error("No Greek found!");
      }
   }

[h4 u32regex_replace]

For each [regex_replace] algorithm defined by `<boost/regex.hpp>`, then 
`<boost/regex/icu.hpp>` defines an overloaded algorithm that takes 
the same arguments, but which is called `u32regex_replace`, and which will 
accept UTF-8, UTF-16 or UTF-32 encoded data, as well as an ICU 
UnicodeString as input.  The input sequence and the format string specifier 
passed to the algorithm, can be encoded independently (for example one can 
be UTF-8, the other in UTF-16), but the result string / output iterator 
argument must use the same character encoding as the text being searched.

Example: Credit card number reformatting:

   //
   // Take a credit card number as a string of digits, 
   // and reformat it as a human readable string with "-"
   // separating each group of four digit;, 
   // note that we're mixing a UTF-32 regex, with a UTF-16
   // string and a UTF-8 format specifier, and it still all 
   // just works:
   //
   const boost::u32regex e = boost::make_u32regex(
         "\\A(\\d{3,4})[- ]?(\\d{4})[- ]?(\\d{4})[- ]?(\\d{4})\\z");
   const char* human_format = "$1-$2-$3-$4";

   UnicodeString human_readable_card_number(const UnicodeString& s)
   {
      return boost::u32regex_replace(s, e, human_format);
   }

[endsect]
[section:unicode_iter Unicode Aware Regex Iterators]

[h4 u32regex_iterator]

Type `u32regex_iterator` is in all respects the same as [regex_iterator] 
except that since the regular expression type is always `u32regex`
it only takes one template parameter (the iterator type). It also calls 
`u32regex_search` internally, allowing it to interface correctly with 
UTF-8, UTF-16, and UTF-32 data:

   template <class BidirectionalIterator>
   class u32regex_iterator
   {
      // for members see regex_iterator
   };

   typedef u32regex_iterator<const char*>     utf8regex_iterator;
   typedef u32regex_iterator<const UChar*>    utf16regex_iterator;
   typedef u32regex_iterator<const UChar32*>  utf32regex_iterator;

In order to simplify the construction of a `u32regex_iterator` from a string, 
there are a series of non-member helper functions called make_u32regex_iterator:

   u32regex_iterator<const char*> 
      make_u32regex_iterator(const char* s, 
                             const u32regex& e, 
                             regex_constants::match_flag_type m = regex_constants::match_default);
                             
   u32regex_iterator<const wchar_t*> 
      make_u32regex_iterator(const wchar_t* s, 
                             const u32regex& e, 
                             regex_constants::match_flag_type m = regex_constants::match_default);
                             
   u32regex_iterator<const UChar*> 
      make_u32regex_iterator(const UChar* s, 
                             const u32regex& e, 
                             regex_constants::match_flag_type m = regex_constants::match_default);
                             
   template <class charT, class Traits, class Alloc>
   u32regex_iterator<typename std::basic_string<charT, Traits, Alloc>::const_iterator> 
      make_u32regex_iterator(const std::basic_string<charT, Traits, Alloc>& s, 
                             const u32regex& e, 
                             regex_constants::match_flag_type m = regex_constants::match_default);
                             
   u32regex_iterator<const UChar*> 
      make_u32regex_iterator(const UnicodeString& s, 
                             const u32regex& e, 
                             regex_constants::match_flag_type m = regex_constants::match_default);

Each of these overloads returns an iterator that enumerates all occurrences 
of expression /e/, in text /s/, using match_flags /m/.

Example: search for international currency symbols, along with their associated numeric value:

   void enumerate_currencies(const std::string& text)
   {
      // enumerate and print all the currency symbols, along
      // with any associated numeric values:
      const char* re = 
         "([[:Sc:]][[:Cf:][:Cc:][:Z*:]]*)?"
         "([[:Nd:]]+(?:[[:Po:]][[:Nd:]]+)?)?"
         "(?(1)"
            "|(?(2)"
               "[[:Cf:][:Cc:][:Z*:]]*"
            ")"
            "[[:Sc:]]"
         ")";
      boost::u32regex r = boost::make_u32regex(re);
      boost::u32regex_iterator<std::string::const_iterator> 
            i(boost::make_u32regex_iterator(text, r)), j;
      while(i != j)
      {
         std::cout << (*i)[0] << std::endl;
         ++i;
      }
   }

Calling

[/this doesn't format correctly as code:]
[pre enumerate_currencies(" $100.23 or '''&#xA3;'''198.12 ");]

Yields the output:

[pre
$100.23
'''&#xA3;'''198.12
]

Provided of course that the input is encoded as UTF-8.

[h4 u32regex_token_iterator]

Type `u32regex_token_iterator` is in all respects the same as [regex_token_iterator] 
except that since the regular expression type is always `u32regex` it only 
takes one template parameter (the iterator type).  It also calls 
`u32regex_search` internally, allowing it to interface correctly with UTF-8, 
UTF-16, and UTF-32 data:

   template <class BidirectionalIterator>
   class u32regex_token_iterator
   {
      // for members see regex_token_iterator
   };

   typedef u32regex_token_iterator<const char*>     utf8regex_token_iterator;
   typedef u32regex_token_iterator<const UChar*>    utf16regex_token_iterator;
   typedef u32regex_token_iterator<const UChar32*>  utf32regex_token_iterator;

In order to simplify the construction of a `u32regex_token_iterator` from a string, 
there are a series of non-member helper functions called `make_u32regex_token_iterator`:

   u32regex_token_iterator<const char*> 
      make_u32regex_token_iterator(
            const char* s, 
            const u32regex& e, 
            int sub, 
            regex_constants::match_flag_type m = regex_constants::match_default);
                                  
   u32regex_token_iterator<const wchar_t*> 
      make_u32regex_token_iterator(
            const wchar_t* s, 
            const u32regex& e, 
            int sub, 
            regex_constants::match_flag_type m = regex_constants::match_default);
                                   
   u32regex_token_iterator<const UChar*> 
      make_u32regex_token_iterator(
            const UChar* s, 
            const u32regex& e, 
            int sub, 
            regex_constants::match_flag_type m = regex_constants::match_default);
                                   
   template <class charT, class Traits, class Alloc>
   u32regex_token_iterator<typename std::basic_string<charT, Traits, Alloc>::const_iterator> 
      make_u32regex_token_iterator(
            const std::basic_string<charT, Traits, Alloc>& s, 
            const u32regex& e, 
            int sub, 
            regex_constants::match_flag_type m = regex_constants::match_default);
                                   
   u32regex_token_iterator<const UChar*> 
      make_u32regex_token_iterator(
            const UnicodeString& s, 
            const u32regex& e, 
            int sub, 
            regex_constants::match_flag_type m = regex_constants::match_default);

Each of these overloads returns an iterator that enumerates all occurrences of 
marked sub-expression sub in regular expression /e/, found in text /s/, using 
match_flags /m/.

   template <std::size_t N>
   u32regex_token_iterator<const char*> 
      make_u32regex_token_iterator(
            const char* p, 
            const u32regex& e, 
            const int (&submatch)[N], 
            regex_constants::match_flag_type m = regex_constants::match_default);
                                   
   template <std::size_t N>
   u32regex_token_iterator<const wchar_t*> 
      make_u32regex_token_iterator(
            const wchar_t* p, 
            const u32regex& e, 
            const int (&submatch)[N], 
            regex_constants::match_flag_type m = regex_constants::match_default);
                                   
   template <std::size_t N>
   u32regex_token_iterator<const UChar*> 
      make_u32regex_token_iterator(
            const UChar* p, 
            const u32regex& e, 
            const int (&submatch)[N], 
            regex_constants::match_flag_type m = regex_constants::match_default);
                                   
   template <class charT, class Traits, class Alloc, std::size_t N>
   u32regex_token_iterator<typename std::basic_string<charT, Traits, Alloc>::const_iterator> 
      make_u32regex_token_iterator(
            const std::basic_string<charT, Traits, Alloc>& p, 
            const u32regex& e, 
            const int (&submatch)[N], 
            regex_constants::match_flag_type m = regex_constants::match_default);
                                   
   template <std::size_t N>
   u32regex_token_iterator<const UChar*> 
      make_u32regex_token_iterator(
            const UnicodeString& s, 
            const u32regex& e, 
            const int (&submatch)[N], 
            regex_constants::match_flag_type m = regex_constants::match_default);

Each of these overloads returns an iterator that enumerates one sub-expression 
for each submatch in regular expression /e/, found in text /s/, using match_flags /m/.

   u32regex_token_iterator<const char*> 
      make_u32regex_token_iterator(
            const char* p, 
            const u32regex& e, 
            const std::vector<int>& submatch, 
            regex_constants::match_flag_type m = regex_constants::match_default);
                                   
   u32regex_token_iterator<const wchar_t*> 
      make_u32regex_token_iterator(
            const wchar_t* p, 
            const u32regex& e, 
            const std::vector<int>& submatch, 
            regex_constants::match_flag_type m = regex_constants::match_default);
                                   
   u32regex_token_iterator<const UChar*> 
      make_u32regex_token_iterator(
            const UChar* p, 
            const u32regex& e, 
            const std::vector<int>& submatch, 
            regex_constants::match_flag_type m = regex_constants::match_default);
                                   
   template <class charT, class Traits, class Alloc>
   u32regex_token_iterator<typename std::basic_string<charT, Traits, Alloc>::const_iterator> 
      make_u32regex_token_iterator(
            const std::basic_string<charT, Traits, Alloc>& p, 
            const u32regex& e, 
            const std::vector<int>& submatch, 
            regex_constants::match_flag_type m = regex_constants::match_default);
                                   
   u32regex_token_iterator<const UChar*> 
      make_u32regex_token_iterator(
            const UnicodeString& s, 
            const u32regex& e, 
            const std::vector<int>& submatch, 
            regex_constants::match_flag_type m = regex_constants::match_default);

Each of these overloads returns an iterator that enumerates one sub-expression for 
each submatch in regular expression /e/, found in text /s/, using match_flags /m/.

Example: search for international currency symbols, along with their associated numeric value:

   void enumerate_currencies2(const std::string& text)
   {
      // enumerate and print all the currency symbols, along
      // with any associated numeric values:
      const char* re = 
         "([[:Sc:]][[:Cf:][:Cc:][:Z*:]]*)?"
         "([[:Nd:]]+(?:[[:Po:]][[:Nd:]]+)?)?"
         "(?(1)"
            "|(?(2)"
               "[[:Cf:][:Cc:][:Z*:]]*"
            ")"
            "[[:Sc:]]"
         ")";
      boost::u32regex r = boost::make_u32regex(re);
      boost::u32regex_token_iterator<std::string::const_iterator> 
         i(boost::make_u32regex_token_iterator(text, r, 1)), j;
      while(i != j)
      {
         std::cout << *i << std::endl;
         ++i;
      }
   }

[endsect]

[endsect]