Chris@16: #ifndef DATE_TIME_TZ_DB_BASE_HPP__
Chris@16: #define DATE_TIME_TZ_DB_BASE_HPP__
Chris@16: 
Chris@16: /* Copyright (c) 2003-2005 CrystalClear Software, Inc.
Chris@16:  * Subject to the Boost Software License, Version 1.0. 
Chris@16:  * (See accompanying file LICENSE_1_0.txt or http://www.boost.org/LICENSE_1_0.txt)
Chris@16:  * Author: Jeff Garland, Bart Garst
Chris@101:  * $Date$
Chris@16:  */
Chris@16: 
Chris@16: #include <map>
Chris@16: #include <vector>
Chris@16: #include <string>
Chris@16: #include <sstream>
Chris@16: #include <fstream>
Chris@16: #include <stdexcept>
Chris@16: #include <boost/tokenizer.hpp>
Chris@16: #include <boost/shared_ptr.hpp>
Chris@16: #include <boost/throw_exception.hpp>
Chris@16: #include <boost/date_time/compiler_config.hpp>
Chris@16: #include <boost/date_time/time_zone_names.hpp>
Chris@16: #include <boost/date_time/time_zone_base.hpp>
Chris@16: #include <boost/date_time/time_parsing.hpp>
Chris@16: 
Chris@16: namespace boost {
Chris@16:   namespace date_time {
Chris@16: 
Chris@16:     //! Exception thrown when tz database cannot locate requested data file
Chris@16:     class data_not_accessible : public std::logic_error
Chris@16:     {
Chris@16:      public:
Chris@16:        data_not_accessible() : 
Chris@16:          std::logic_error(std::string("Unable to locate or access the required datafile.")) 
Chris@16:        {}
Chris@16:        data_not_accessible(const std::string& filespec) : 
Chris@16:          std::logic_error(std::string("Unable to locate or access the required datafile. Filespec: " + filespec)) 
Chris@16:        {}
Chris@16:     };
Chris@16:     
Chris@16:     //! Exception thrown when tz database locates incorrect field structure in data file
Chris@16:     class bad_field_count : public std::out_of_range
Chris@16:     {
Chris@16:      public:
Chris@16:        bad_field_count(const std::string& s) : 
Chris@16:          std::out_of_range(s) 
Chris@16:       {}
Chris@16:     };
Chris@16: 
Chris@16:     //! Creates a database of time_zones from csv datafile
Chris@16:     /*! The csv file containing the zone_specs used by the
Chris@16:      * tz_db_base is intended to be customized by the
Chris@16:      * library user. When customizing this file (or creating your own) the
Chris@16:      * file must follow a specific format.
Chris@16:      * 
Chris@16:      * This first line is expected to contain column headings and is therefore
Chris@16:      * not processed by the tz_db_base.
Chris@16:      *
Chris@16:      * Each record (line) must have eleven fields. Some of those fields can
Chris@16:      * be empty. Every field (even empty ones) must be enclosed in 
Chris@16:      * double-quotes.
Chris@16:      * Ex:
Chris@16:      * @code
Chris@16:      * "America/Phoenix" <- string enclosed in quotes
Chris@16:      * ""                <- empty field
Chris@16:      * @endcode
Chris@16:      * 
Chris@16:      * Some fields represent a length of time. The format of these fields 
Chris@16:      * must be:
Chris@16:      * @code
Chris@16:      * "{+|-}hh:mm[:ss]" <- length-of-time format
Chris@16:      * @endcode
Chris@16:      * Where the plus or minus is mandatory and the seconds are optional.
Chris@16:      * 
Chris@16:      * Since some time zones do not use daylight savings it is not always 
Chris@16:      * necessary for every field in a zone_spec to contain a value. All 
Chris@16:      * zone_specs must have at least ID and GMT offset. Zones that use 
Chris@16:      * daylight savings must have all fields filled except: 
Chris@16:      * STD ABBR, STD NAME, DST NAME. You should take note 
Chris@16:      * that DST ABBR is mandatory for zones that use daylight savings 
Chris@16:      * (see field descriptions for further details).
Chris@16:      *
Chris@16:      * ******* Fields and their description/details ********* 
Chris@16:      *     
Chris@16:      * ID: 
Chris@16:      * Contains the identifying string for the zone_spec. Any string will
Chris@16:      * do as long as it's unique. No two ID's can be the same. 
Chris@16:      *
Chris@16:      * STD ABBR:
Chris@16:      * STD NAME:
Chris@16:      * DST ABBR:
Chris@16:      * DST NAME:
Chris@16:      * These four are all the names and abbreviations used by the time 
Chris@16:      * zone being described. While any string will do in these fields, 
Chris@16:      * care should be taken. These fields hold the strings that will be 
Chris@16:      * used in the output of many of the local_time classes. 
Chris@16:      * Ex:
Chris@16:      * @code
Chris@16:      * time_zone nyc = tz_db.time_zone_from_region("America/New_York");
Chris@16:      * local_time ny_time(date(2004, Aug, 30), IS_DST, nyc);
Chris@16:      * cout << ny_time.to_long_string() << endl;
Chris@16:      * // 2004-Aug-30 00:00:00 Eastern Daylight Time
Chris@16:      * cout << ny_time.to_short_string() << endl;
Chris@16:      * // 2004-Aug-30 00:00:00 EDT
Chris@16:      * @endcode
Chris@16:      *
Chris@16:      * NOTE: The exact format/function names may vary - see local_time 
Chris@16:      * documentation for further details.
Chris@16:      *
Chris@16:      * GMT offset:
Chris@16:      * This is the number of hours added to utc to get the local time 
Chris@16:      * before any daylight savings adjustments are made. Some examples 
Chris@16:      * are: America/New_York offset -5 hours, & Africa/Cairo offset +2 hours.
Chris@16:      * The format must follow the length-of-time format described above.
Chris@16:      *
Chris@16:      * DST adjustment:
Chris@16:      * The amount of time added to gmt_offset when daylight savings is in 
Chris@16:      * effect. The format must follow the length-of-time format described
Chris@16:      * above.
Chris@16:      *
Chris@16:      * DST Start Date rule:
Chris@16:      * This is a specially formatted string that describes the day of year
Chris@16:      * in which the transition take place. It holds three fields of it's own,
Chris@16:      * separated by semicolons. 
Chris@16:      * The first field indicates the "nth" weekday of the month. The possible 
Chris@16:      * values are: 1 (first), 2 (second), 3 (third), 4 (fourth), 5 (fifth), 
Chris@16:      * and -1 (last).
Chris@16:      * The second field indicates the day-of-week from 0-6 (Sun=0).
Chris@16:      * The third field indicates the month from 1-12 (Jan=1).
Chris@16:      * 
Chris@16:      * Examples are: "-1;5;9"="Last Friday of September", 
Chris@16:      * "2;1;3"="Second Monday of March"
Chris@16:      *
Chris@16:      * Start time:
Chris@16:      * Start time is the number of hours past midnight, on the day of the
Chris@16:      * start transition, the transition takes place. More simply put, the 
Chris@16:      * time of day the transition is made (in 24 hours format). The format
Chris@16:      * must follow the length-of-time format described above with the 
Chris@16:      * exception that it must always be positive.
Chris@16:      *
Chris@16:      * DST End date rule:
Chris@16:      * See DST Start date rule. The difference here is this is the day 
Chris@16:      * daylight savings ends (transition to STD).
Chris@16:      *
Chris@16:      * End time:
Chris@16:      * Same as Start time.
Chris@16:      */
Chris@16:     template<class time_zone_type, class rule_type>
Chris@16:     class tz_db_base {
Chris@16:     public:
Chris@16:       /* Having CharT as a template parameter created problems 
Chris@16:        * with posix_time::duration_from_string. Templatizing 
Chris@16:        * duration_from_string was not possible at this time, however, 
Chris@16:        * it should be possible in the future (when poor compilers get 
Chris@16:        * fixed or stop being used). 
Chris@16:        * Since this class was designed to use CharT as a parameter it 
Chris@16:        * is simply typedef'd here to ease converting in back to a 
Chris@16:        * parameter the future */
Chris@16:       typedef char char_type;
Chris@16: 
Chris@16:       typedef typename time_zone_type::base_type time_zone_base_type;
Chris@16:       typedef typename time_zone_type::time_duration_type time_duration_type;
Chris@16:       typedef time_zone_names_base<char_type> time_zone_names;
Chris@16:       typedef boost::date_time::dst_adjustment_offsets<time_duration_type> dst_adjustment_offsets;
Chris@16:       typedef std::basic_string<char_type> string_type;
Chris@16: 
Chris@16:       //! Constructs an empty database
Chris@16:       tz_db_base() {}
Chris@16: 
Chris@16:       //! Process csv data file, may throw exceptions
Chris@16:       /*! May throw bad_field_count exceptions */
Chris@16:       void load_from_stream(std::istream &in)
Chris@16:       {
Chris@16:         std::string  buff;
Chris@16:         while( std::getline(in, buff)) {
Chris@16:           parse_string(buff);
Chris@16:         }
Chris@16:       }
Chris@16: 
Chris@16:       //! Process csv data file, may throw exceptions
Chris@16:       /*! May throw data_not_accessible, or bad_field_count exceptions */
Chris@16:       void load_from_file(const std::string& pathspec)
Chris@16:       {
Chris@16:         std::string  buff;
Chris@16:         
Chris@16:         std::ifstream ifs(pathspec.c_str());
Chris@16:         if(!ifs){
Chris@16:           boost::throw_exception(data_not_accessible(pathspec));
Chris@16:         }
Chris@16:         std::getline(ifs, buff); // first line is column headings
Chris@16:         this->load_from_stream(ifs);
Chris@16:       }
Chris@16: 
Chris@16:       //! returns true if record successfully added to map
Chris@16:       /*! Takes a region name in the form of "America/Phoenix", and a 
Chris@16:        * time_zone object for that region. The id string must be a unique 
Chris@16:        * name that does not already exist in the database. */
Chris@16:       bool add_record(const string_type& region, 
Chris@16:                       boost::shared_ptr<time_zone_base_type> tz)
Chris@16:       {
Chris@16:         typename map_type::value_type p(region, tz); 
Chris@16:         return (m_zone_map.insert(p)).second;
Chris@16:       }
Chris@16: 
Chris@16:       //! Returns a time_zone object built from the specs for the given region
Chris@16:       /*! Returns a time_zone object built from the specs for the given 
Chris@16:        * region. If region does not exist a local_time::record_not_found 
Chris@16:        * exception will be thrown */
Chris@16:       boost::shared_ptr<time_zone_base_type> 
Chris@16:       time_zone_from_region(const string_type& region) const 
Chris@16:       {
Chris@16:         // get the record
Chris@16:         typename map_type::const_iterator record = m_zone_map.find(region);
Chris@16:         if(record == m_zone_map.end()){
Chris@16:           return boost::shared_ptr<time_zone_base_type>(); //null pointer
Chris@16:         }
Chris@16:         return record->second;
Chris@16:       }
Chris@16: 
Chris@16:       //! Returns a vector of strings holding the time zone regions in the database
Chris@16:       std::vector<std::string> region_list() const
Chris@16:       {
Chris@16:         typedef std::vector<std::string> vector_type;
Chris@16:         vector_type regions;
Chris@16:         typename map_type::const_iterator itr = m_zone_map.begin();
Chris@16:         while(itr != m_zone_map.end()) {
Chris@16:           regions.push_back(itr->first);
Chris@16:           ++itr;
Chris@16:         }
Chris@16:         return regions;
Chris@16:       }
Chris@16:     
Chris@16:     private:
Chris@16:       typedef std::map<string_type, boost::shared_ptr<time_zone_base_type> > map_type;
Chris@16:       map_type m_zone_map;
Chris@16: 
Chris@16:       // start and end rule are of the same type
Chris@16:       typedef typename rule_type::start_rule::week_num week_num;
Chris@16: 
Chris@16:       /* TODO: mechanisms need to be put in place to handle different
Chris@16:        * types of rule specs. parse_rules() only handles nth_kday
Chris@16:        * rule types. */
Chris@16:       
Chris@16:       //! parses rule specs for transition day rules
Chris@16:       rule_type* parse_rules(const string_type& sr, const string_type& er) const
Chris@16:       {
Chris@16:         using namespace gregorian;
Chris@16:         // start and end rule are of the same type, 
Chris@16:         // both are included here for readability
Chris@16:         typedef typename rule_type::start_rule start_rule;
Chris@16:         typedef typename rule_type::end_rule end_rule;
Chris@16:        
Chris@16:         // these are: [start|end] nth, day, month
Chris@16:         int s_nth = 0, s_d = 0, s_m = 0;
Chris@16:         int e_nth = 0, e_d = 0, e_m = 0;
Chris@16:         split_rule_spec(s_nth, s_d, s_m, sr);
Chris@16:         split_rule_spec(e_nth, e_d, e_m, er);
Chris@16:         
Chris@16:         typename start_rule::week_num s_wn, e_wn;
Chris@16:         s_wn = get_week_num(s_nth);
Chris@16:         e_wn = get_week_num(e_nth);
Chris@16:         
Chris@16:         
Chris@101:         return new rule_type(start_rule(s_wn,
Chris@101:                                         static_cast<unsigned short>(s_d),
Chris@101:                                         static_cast<unsigned short>(s_m)),
Chris@101:                              end_rule(e_wn,
Chris@101:                                       static_cast<unsigned short>(e_d),
Chris@101:                                       static_cast<unsigned short>(e_m)));
Chris@16:       }
Chris@16:       //! helper function for parse_rules()
Chris@16:       week_num get_week_num(int nth) const
Chris@16:       {
Chris@16:         typedef typename rule_type::start_rule start_rule;
Chris@16:         switch(nth){
Chris@16:         case 1:
Chris@16:           return start_rule::first;
Chris@16:         case 2:
Chris@16:           return start_rule::second;
Chris@16:         case 3:
Chris@16:           return start_rule::third;
Chris@16:         case 4:
Chris@16:           return start_rule::fourth;
Chris@16:         case 5:
Chris@16:         case -1:
Chris@16:           return start_rule::fifth;
Chris@16:         default:
Chris@16:           // shouldn't get here - add error handling later
Chris@16:           break;
Chris@16:         }
Chris@16:         return start_rule::fifth; // silence warnings
Chris@16:       }
Chris@16:           
Chris@16:       //! splits the [start|end]_date_rule string into 3 ints
Chris@16:       void split_rule_spec(int& nth, int& d, int& m, string_type rule) const
Chris@16:       {
Chris@16:         typedef boost::char_separator<char_type, std::char_traits<char_type> > char_separator_type;
Chris@16:         typedef boost::tokenizer<char_separator_type,
Chris@16:                                  std::basic_string<char_type>::const_iterator,
Chris@16:                                  std::basic_string<char_type> > tokenizer;
Chris@16:         typedef boost::tokenizer<char_separator_type,
Chris@16:                                  std::basic_string<char_type>::const_iterator,
Chris@16:                                  std::basic_string<char_type> >::iterator tokenizer_iterator;
Chris@16:         
Chris@16:         const char_type sep_char[] = { ';', '\0'};
Chris@16:         char_separator_type sep(sep_char);
Chris@16:         tokenizer tokens(rule, sep); // 3 fields
Chris@16: 
Chris@16:         if ( std::distance ( tokens.begin(), tokens.end ()) != 3 ) {
Chris@16:           std::ostringstream msg;
Chris@16:           msg << "Expecting 3 fields, got " 
Chris@16:               << std::distance ( tokens.begin(), tokens.end ()) 
Chris@16:               << " fields in line: " << rule;
Chris@16:           boost::throw_exception(bad_field_count(msg.str()));
Chris@16:         }
Chris@16: 
Chris@16:         tokenizer_iterator tok_iter = tokens.begin(); 
Chris@16:         nth = std::atoi(tok_iter->c_str()); ++tok_iter;
Chris@16:         d   = std::atoi(tok_iter->c_str()); ++tok_iter;
Chris@16:         m   = std::atoi(tok_iter->c_str());
Chris@16:       }
Chris@16: 
Chris@16:      
Chris@16:       //! Take a line from the csv, turn it into a time_zone_type.
Chris@16:       /*! Take a line from the csv, turn it into a time_zone_type,
Chris@16:        * and add it to the map. Zone_specs in csv file are expected to 
Chris@16:        * have eleven fields that describe the time zone. Returns true if 
Chris@16:        * zone_spec successfully added to database */
Chris@16:       bool parse_string(string_type& s)
Chris@16:       {
Chris@16:         std::vector<string_type> result;
Chris@16:         typedef boost::token_iterator_generator<boost::escaped_list_separator<char_type>, string_type::const_iterator, string_type >::type token_iter_type;
Chris@16: 
Chris@16:         token_iter_type i = boost::make_token_iterator<string_type>(s.begin(), s.end(),boost::escaped_list_separator<char_type>());
Chris@16: 
Chris@16:         token_iter_type end;
Chris@16:         while (i != end) {
Chris@16:           result.push_back(*i);
Chris@16:           i++;
Chris@16:         }
Chris@16: 
Chris@16:         enum db_fields { ID, STDABBR, STDNAME, DSTABBR, DSTNAME, GMTOFFSET,
Chris@16:                          DSTADJUST, START_DATE_RULE, START_TIME, END_DATE_RULE,
Chris@16:                          END_TIME, FIELD_COUNT };
Chris@16: 
Chris@16:         //take a shot at fixing gcc 4.x error
Chris@16:         const unsigned int expected_fields = static_cast<unsigned int>(FIELD_COUNT);
Chris@16:         if (result.size() != expected_fields) { 
Chris@16:           std::ostringstream msg;
Chris@16:           msg << "Expecting " << FIELD_COUNT << " fields, got " 
Chris@16:             << result.size() << " fields in line: " << s;
Chris@16:           boost::throw_exception(bad_field_count(msg.str()));
Chris@16:           BOOST_DATE_TIME_UNREACHABLE_EXPRESSION(return false); // should never reach
Chris@16:         }
Chris@16: 
Chris@16:         // initializations
Chris@16:         bool has_dst = true; 
Chris@16:         if(result[DSTABBR] == std::string()){
Chris@16:           has_dst = false;
Chris@16:         }
Chris@16: 
Chris@16: 
Chris@16:         // start building components of a time_zone
Chris@16:         time_zone_names names(result[STDNAME], result[STDABBR],
Chris@16:                               result[DSTNAME], result[DSTABBR]);
Chris@16: 
Chris@16:         time_duration_type utc_offset = 
Chris@16:           str_from_delimited_time_duration<time_duration_type,char_type>(result[GMTOFFSET]);
Chris@16:         
Chris@16:         dst_adjustment_offsets adjust(time_duration_type(0,0,0),
Chris@16:                                       time_duration_type(0,0,0),
Chris@16:                                       time_duration_type(0,0,0));
Chris@16: 
Chris@16:         boost::shared_ptr<rule_type> rules;
Chris@16: 
Chris@16:         if(has_dst){
Chris@16:           adjust = dst_adjustment_offsets(
Chris@16:                                           str_from_delimited_time_duration<time_duration_type,char_type>(result[DSTADJUST]),
Chris@16:                                           str_from_delimited_time_duration<time_duration_type,char_type>(result[START_TIME]),
Chris@16:                                           str_from_delimited_time_duration<time_duration_type,char_type>(result[END_TIME])
Chris@16:                                           );
Chris@16: 
Chris@16:           rules = 
Chris@16:             boost::shared_ptr<rule_type>(parse_rules(result[START_DATE_RULE],
Chris@16:                                                      result[END_DATE_RULE]));
Chris@16:         }
Chris@16:         string_type id(result[ID]);
Chris@16:         boost::shared_ptr<time_zone_base_type> zone(new time_zone_type(names, utc_offset, adjust, rules));
Chris@16:         return (add_record(id, zone));
Chris@16:         
Chris@16:       } 
Chris@16:      
Chris@16:     };
Chris@16: 
Chris@16: } } // namespace
Chris@16: 
Chris@16: #endif // DATE_TIME_TZ_DB_BASE_HPP__