Chris@16: //
Chris@16: //  Copyright (c) 2009-2011 Artyom Beilis (Tonkikh)
Chris@16: //
Chris@16: //  Distributed under the Boost Software License, Version 1.0. (See
Chris@16: //  accompanying file LICENSE_1_0.txt or copy at
Chris@16: //  http://www.boost.org/LICENSE_1_0.txt)
Chris@16: //
Chris@16: #ifndef BOOST_LOCALE_UTIL_HPP
Chris@16: #define BOOST_LOCALE_UTIL_HPP
Chris@16: #include <locale>
Chris@16: #include <typeinfo>
Chris@16: #include <boost/cstdint.hpp>
Chris@16: #include <boost/locale/utf.hpp>
Chris@16: #include <boost/locale/generator.hpp>
Chris@16: #include <boost/assert.hpp>
Chris@16: 
Chris@16: #include <vector>
Chris@16: namespace boost {
Chris@16: namespace locale {
Chris@16: ///
Chris@16: /// \brief This namespace provides various utility function useful for Boost.Locale backends
Chris@16: /// implementations
Chris@16: ///
Chris@16: namespace util {
Chris@16:     
Chris@16:     ///
Chris@16:     /// \brief Return default system locale name in POSIX format.
Chris@16:     ///
Chris@16:     /// This function tries to detect the locale using, LC_CTYPE, LC_ALL and LANG environment
Chris@16:     /// variables in this order and if all of them unset, in POSIX platforms it returns "C"
Chris@16:     /// 
Chris@16:     /// On Windows additionally to check the above environment variables, this function
Chris@16:     /// tries to creates locale name from ISO-339 and ISO-3199 country codes defined
Chris@16:     /// for user default locale.
Chris@16:     /// If \a use_utf8_on_windows is true it sets the encoding to UTF-8, otherwise, if system
Chris@16:     /// locale supports ANSI code-page it defines the ANSI encoding like windows-1252, otherwise it fall-backs
Chris@16:     /// to UTF-8 encoding if ANSI code-page is not available.
Chris@16:     ///
Chris@16:     BOOST_LOCALE_DECL
Chris@16:     std::string get_system_locale(bool use_utf8_on_windows = false);
Chris@16: 
Chris@16:     ///
Chris@16:     /// \brief Installs information facet to locale in based on locale name \a name
Chris@16:     ///
Chris@16:     /// This function installs boost::locale::info facet into the locale \a in and returns
Chris@16:     /// newly created locale.
Chris@16:     ///
Chris@16:     /// Note: all information is based only on parsing of string \a name;
Chris@16:     ///
Chris@16:     /// The name has following format: language[_COUNTRY][.encoding][\@variant]
Chris@16:     /// Where language is ISO-639 language code like "en" or "ru", COUNTRY is ISO-3166
Chris@16:     /// country identifier like "US" or "RU". the Encoding is a charracter set name
Chris@16:     /// like UTF-8 or ISO-8859-1. Variant is backend specific variant like \c euro or
Chris@16:     /// calendar=hebrew.
Chris@16:     ///
Chris@16:     /// If some parameters are missing they are specified as blanks, default encoding
Chris@16:     /// is assumed to be US-ASCII and missing language is assumed to be "C"
Chris@16:     ///
Chris@16:     BOOST_LOCALE_DECL
Chris@16:     std::locale create_info(std::locale const &in,std::string const &name); 
Chris@16: 
Chris@16: 
Chris@16:     ///
Chris@16:     /// \brief This class represent a simple stateless converter from UCS-4 and to UCS-4 for
Chris@16:     ///  each single code point
Chris@16:     ///
Chris@16:     /// This class is used for creation of std::codecvt facet for converting utf-16/utf-32 encoding
Chris@16:     /// to encoding supported by this converter
Chris@16:     ///
Chris@16:     /// Please note, this converter should be fully stateless. Fully stateless means it should
Chris@16:     /// never assume that it is called in any specific order on the text. Even if the
Chris@16:     /// encoding itself seems to be stateless like windows-1255 or shift-jis, some
Chris@16:     /// encoders (most notably iconv) can actually compose several code-point into one or
Chris@16:     /// decompose them in case composite characters are found. So be very careful when implementing
Chris@16:     /// these converters for certain character set.
Chris@16:     ///
Chris@16:     class base_converter {
Chris@16:     public:
Chris@16: 
Chris@16:         ///
Chris@16:         /// This value should be returned when an illegal input sequence or code-point is observed:
Chris@16:         /// For example if a UCS-32 code-point is in the range reserved for UTF-16 surrogates
Chris@16:         /// or an invalid UTF-8 sequence is found
Chris@16:         ///
Chris@16:         static const uint32_t illegal=utf::illegal;
Chris@16: 
Chris@16:         ///
Chris@16:         /// This value is returned in following cases: The of incomplete input sequence was found or 
Chris@16:         /// insufficient output buffer was provided so complete output could not be written.
Chris@16:         ///
Chris@16:         static const uint32_t incomplete=utf::incomplete;
Chris@16:         
Chris@16:         virtual ~base_converter() 
Chris@16:         {
Chris@16:         }
Chris@16:         ///
Chris@16:         /// Return the maximal length that one Unicode code-point can be converted to, for example
Chris@16:         /// for UTF-8 it is 4, for Shift-JIS it is 2 and ISO-8859-1 is 1
Chris@16:         ///
Chris@16:         virtual int max_len() const 
Chris@16:         {
Chris@16:             return 1;
Chris@16:         }
Chris@16:         ///
Chris@16:         /// Returns true if calling the functions from_unicode, to_unicode, and max_len is thread safe.
Chris@16:         ///
Chris@16:         /// Rule of thumb: if this class' implementation uses simple tables that are unchanged
Chris@16:         /// or is purely algorithmic like UTF-8 - so it does not share any mutable bit for
Chris@16:         /// independent to_unicode, from_unicode calls, you may set it to true, otherwise,
Chris@16:         /// for example if you use iconv_t descriptor or UConverter as conversion object return false,
Chris@16:         /// and this object will be cloned for each use.
Chris@16:         ///
Chris@16:         virtual bool is_thread_safe() const 
Chris@16:         {
Chris@16:             return false;
Chris@16:         }
Chris@16:         ///
Chris@16:         /// Create a polymorphic copy of this object, usually called only if is_thread_safe() return false
Chris@16:         ///
Chris@16:         virtual base_converter *clone() const 
Chris@16:         {
Chris@16:             BOOST_ASSERT(typeid(*this)==typeid(base_converter));
Chris@16:             return new base_converter();
Chris@16:         }
Chris@16: 
Chris@16:         ///
Chris@16:         /// Convert a single character starting at begin and ending at most at end to Unicode code-point.
Chris@16:         ///
Chris@16:         /// if valid input sequence found in [\a begin,\a code_point_end) such as \a begin < \a code_point_end && \a code_point_end <= \a end
Chris@16:         /// it is converted to its Unicode code point equivalent, \a begin is set to \a code_point_end
Chris@16:         ///
Chris@16:         /// if incomplete input sequence found in [\a begin,\a end), i.e. there my be such \a code_point_end that \a code_point_end > \a end
Chris@16:         /// and [\a begin, \a code_point_end) would be valid input sequence, then \a incomplete is returned begin stays unchanged, for example
Chris@16:         /// for UTF-8 conversion a *begin = 0xc2, \a begin +1 = \a end is such situation.
Chris@16:         ///
Chris@16:         /// if invalid input sequence found, i.e. there is a sequence [\a begin, \a code_point_end) such as \a code_point_end <= \a end
Chris@16:         /// that is illegal for this encoding, \a illegal is returned and begin stays unchanged. For example if *begin = 0xFF and begin < end
Chris@16:         /// for UTF-8, then \a illegal is returned.
Chris@16:         /// 
Chris@16:         ///
Chris@16:         virtual uint32_t to_unicode(char const *&begin,char const *end) 
Chris@16:         {
Chris@16:             if(begin == end)
Chris@16:                 return incomplete;
Chris@16:             unsigned char cp = *begin;
Chris@16:             if(cp <= 0x7F) {
Chris@16:                 begin++;
Chris@16:                 return cp;
Chris@16:             }
Chris@16:             return illegal;
Chris@16:         }
Chris@16:         ///
Chris@16:         /// Convert a single code-point \a u into encoding and store it in [begin,end) range.
Chris@16:         ///
Chris@16:         /// If u is invalid Unicode code-point, or it can not be mapped correctly to represented character set,
Chris@16:         /// \a illegal should be returned
Chris@16:         ///
Chris@16:         /// If u can be converted to a sequence of bytes c1, ... , cN (1<= N <= max_len() ) then
Chris@16:         /// 
Chris@16:         /// -# If end - begin >= N, c1, ... cN are written starting at begin and N is returned
Chris@16:         /// -# If end - begin < N, incomplete is returned, it is unspecified what would be
Chris@16:         ///    stored in bytes in range [begin,end)
Chris@16: 
Chris@16:         virtual uint32_t from_unicode(uint32_t u,char *begin,char const *end) 
Chris@16:         {
Chris@16:             if(begin==end)
Chris@16:                 return incomplete;
Chris@16:             if(u >= 0x80)
Chris@16:                 return illegal;
Chris@16:             *begin = static_cast<char>(u);
Chris@16:             return 1;
Chris@16:         }
Chris@16:     };
Chris@16: 
Chris@16:     ///
Chris@16:     /// This function creates a \a base_converter that can be used for conversion between UTF-8 and
Chris@16:     /// unicode code points
Chris@16:     ///
Chris@16:     BOOST_LOCALE_DECL std::auto_ptr<base_converter> create_utf8_converter();
Chris@16:     ///
Chris@16:     /// This function creates a \a base_converter that can be used for conversion between single byte
Chris@16:     /// character encodings like ISO-8859-1, koi8-r, windows-1255 and Unicode code points,
Chris@16:     /// 
Chris@16:     /// If \a encoding is not supported, empty pointer is returned. You should check if
Chris@16:     /// std::auto_ptr<base_converter>::get() != 0
Chris@16:     ///
Chris@16:     BOOST_LOCALE_DECL std::auto_ptr<base_converter> create_simple_converter(std::string const &encoding);
Chris@16: 
Chris@16: 
Chris@16:     ///
Chris@16:     /// Install codecvt facet into locale \a in and return new locale that is based on \a in and uses new
Chris@16:     /// facet.
Chris@16:     ///
Chris@16:     /// codecvt facet would convert between narrow and wide/char16_t/char32_t encodings using \a cvt converter.
Chris@16:     /// If \a cvt is null pointer, always failure conversion would be used that fails on every first input or output.
Chris@16:     /// 
Chris@16:     /// Note: the codecvt facet handles both UTF-16 and UTF-32 wide encodings, it knows to break and join
Chris@16:     /// Unicode code-points above 0xFFFF to and from surrogate pairs correctly. \a cvt should be unaware
Chris@16:     /// of wide encoding type
Chris@16:     ///
Chris@16:     BOOST_LOCALE_DECL
Chris@16:     std::locale create_codecvt(std::locale const &in,std::auto_ptr<base_converter> cvt,character_facet_type type);
Chris@16: 
Chris@16: } // util
Chris@16: } // locale 
Chris@16: } // boost
Chris@16: 
Chris@16: #endif
Chris@16: // vim: tabstop=4 expandtab shiftwidth=4 softtabstop=4