sl@0: // Boost token_functions.hpp  ------------------------------------------------//
sl@0: 
sl@0: // Copyright John R. Bandela 2001. 
sl@0: 
sl@0: // Distributed under the Boost Software License, Version 1.0. (See
sl@0: // accompanying file LICENSE_1_0.txt or copy at
sl@0: // http://www.boost.org/LICENSE_1_0.txt)
sl@0: 
sl@0: // See http://www.boost.org/libs/tokenizer/ for documentation.
sl@0: 
sl@0: // Revision History:
sl@0: // 01 Oct 2004   Joaquín M López Muñoz
sl@0: //      Workaround for a problem with string::assign in msvc-stlport
sl@0: // 06 Apr 2004   John Bandela
sl@0: //      Fixed a bug involving using char_delimiter with a true input iterator
sl@0: // 28 Nov 2003   Robert Zeh and John Bandela
sl@0: //      Converted into "fast" functions that avoid using += when
sl@0: //      the supplied iterator isn't an input_iterator; based on
sl@0: //      some work done at Archelon and a version that was checked into
sl@0: //      the boost CVS for a short period of time.
sl@0: // 20 Feb 2002   John Maddock
sl@0: //      Removed using namespace std declarations and added
sl@0: //      workaround for BOOST_NO_STDC_NAMESPACE (the library
sl@0: //      can be safely mixed with regex).
sl@0: // 06 Feb 2002   Jeremy Siek
sl@0: //      Added char_separator.
sl@0: // 02 Feb 2002   Jeremy Siek
sl@0: //      Removed tabs and a little cleanup.
sl@0: 
sl@0: 
sl@0: #ifndef BOOST_TOKEN_FUNCTIONS_JRB120303_HPP_
sl@0: #define BOOST_TOKEN_FUNCTIONS_JRB120303_HPP_
sl@0: 
sl@0: #include <vector>
sl@0: #include <stdexcept>
sl@0: #include <string>
sl@0: #include <cctype>
sl@0: #include <algorithm> // for find_if
sl@0: #include <boost/config.hpp>
sl@0: #include <boost/assert.hpp>
sl@0: #include <boost/detail/workaround.hpp>
sl@0: #include <boost/mpl/if.hpp>
sl@0: 
sl@0: //
sl@0: // the following must not be macros if we are to prefix them
sl@0: // with std:: (they shouldn't be macros anyway...)
sl@0: //
sl@0: #ifdef ispunct
sl@0: #  undef ispunct
sl@0: #endif
sl@0: #ifdef isspace
sl@0: #  undef isspace
sl@0: #endif
sl@0: //
sl@0: // fix namespace problems:
sl@0: //
sl@0: #ifdef BOOST_NO_STDC_NAMESPACE
sl@0: namespace std{
sl@0:  using ::ispunct;
sl@0:  using ::isspace;
sl@0: }
sl@0: #endif
sl@0: 
sl@0: namespace boost{
sl@0: 
sl@0:   //===========================================================================
sl@0:   // The escaped_list_separator class. Which is a model of TokenizerFunction
sl@0:   // An escaped list is a super-set of what is commonly known as a comma 
sl@0:   // separated value (csv) list.It is separated into fields by a comma or 
sl@0:   // other character. If the delimiting character is inside quotes, then it is
sl@0:   // counted as a regular character.To allow for embedded quotes in a field,
sl@0:   // there can be escape sequences using the \ much like C. 
sl@0:   // The role of the comma, the quotation mark, and the escape 
sl@0:   // character (backslash \), can be assigned to other characters.
sl@0: 
sl@0:   struct escaped_list_error : public std::runtime_error{
sl@0:     escaped_list_error(const std::string& what_arg):std::runtime_error(what_arg) { }
sl@0:   };
sl@0:   
sl@0: 
sl@0: // The out of the box GCC 2.95 on cygwin does not have a char_traits class.
sl@0: // MSVC does not like the following typename
sl@0: #if !defined(BOOST_MSVC) || BOOST_MSVC > 1300
sl@0:   template <class Char, 
sl@0:     class Traits = typename std::basic_string<Char>::traits_type >
sl@0: #else
sl@0:   template <class Char, 
sl@0:     class Traits = std::basic_string<Char>::traits_type >
sl@0: #endif
sl@0:   class escaped_list_separator {
sl@0: 
sl@0:   private:
sl@0:     typedef std::basic_string<Char,Traits> string_type;
sl@0:     struct char_eq {
sl@0:       Char e_;
sl@0:       char_eq(Char e):e_(e) { }
sl@0:       bool operator()(Char c) {
sl@0:         return Traits::eq(e_,c);
sl@0:       }
sl@0:     };
sl@0:     string_type  escape_;
sl@0:     string_type  c_;
sl@0:     string_type  quote_;      
sl@0:     bool last_;
sl@0: 
sl@0:     bool is_escape(Char e) {
sl@0:       char_eq f(e);
sl@0:       return std::find_if(escape_.begin(),escape_.end(),f)!=escape_.end();
sl@0:     }
sl@0:     bool is_c(Char e) {
sl@0:       char_eq f(e);
sl@0:       return std::find_if(c_.begin(),c_.end(),f)!=c_.end();
sl@0:     }
sl@0:     bool is_quote(Char e) {
sl@0:       char_eq f(e);
sl@0:       return std::find_if(quote_.begin(),quote_.end(),f)!=quote_.end();
sl@0:     }
sl@0:     template <typename iterator, typename Token>
sl@0:     void do_escape(iterator& next,iterator end,Token& tok) {
sl@0:       if (++next == end)
sl@0:         throw escaped_list_error(std::string("cannot end with escape"));
sl@0:       if (Traits::eq(*next,'n')) {
sl@0:         tok+='\n';
sl@0:         return;
sl@0:       }
sl@0:       else if (is_quote(*next)) {
sl@0:         tok+=*next;
sl@0:         return;
sl@0:       }
sl@0:       else if (is_c(*next)) {
sl@0:         tok+=*next;
sl@0:         return;
sl@0:       }
sl@0:       else if (is_escape(*next)) {
sl@0:         tok+=*next;
sl@0:         return;
sl@0:       }
sl@0:       else
sl@0:         throw escaped_list_error(std::string("unknown escape sequence"));
sl@0:     }
sl@0: 
sl@0:     public:
sl@0:     
sl@0:     explicit escaped_list_separator(Char  e = '\\',
sl@0:                                     Char c = ',',Char  q = '\"')
sl@0:       : escape_(1,e), c_(1,c), quote_(1,q), last_(false) { }
sl@0:     
sl@0:     escaped_list_separator(string_type e, string_type c, string_type q)
sl@0:       : escape_(e), c_(c), quote_(q), last_(false) { }
sl@0:     
sl@0:     void reset() {last_=false;}
sl@0: 
sl@0:     template <typename InputIterator, typename Token>
sl@0:     bool operator()(InputIterator& next,InputIterator end,Token& tok) {
sl@0:       bool bInQuote = false;
sl@0:       tok = Token();
sl@0:       
sl@0:       if (next == end) {
sl@0:         if (last_) {
sl@0:           last_ = false;
sl@0:           return true;
sl@0:         }
sl@0:         else
sl@0:           return false;
sl@0:       }
sl@0:       last_ = false;
sl@0:       for (;next != end;++next) {
sl@0:         if (is_escape(*next)) {
sl@0:           do_escape(next,end,tok);
sl@0:         }
sl@0:         else if (is_c(*next)) {
sl@0:           if (!bInQuote) {
sl@0:             // If we are not in quote, then we are done
sl@0:             ++next;
sl@0:             // The last character was a c, that means there is
sl@0:             // 1 more blank field
sl@0:             last_ = true; 
sl@0:             return true;
sl@0:           }
sl@0:           else tok+=*next;
sl@0:         }
sl@0:         else if (is_quote(*next)) {
sl@0:           bInQuote=!bInQuote;
sl@0:         }
sl@0:         else {
sl@0:           tok += *next;
sl@0:         }
sl@0:       }
sl@0:       return true;
sl@0:     }
sl@0:   };
sl@0: 
sl@0:   //===========================================================================
sl@0:   // The classes here are used by offset_separator and char_separator to implement
sl@0:   // faster assigning of tokens using assign instead of +=
sl@0:   
sl@0:   namespace tokenizer_detail {
sl@0: 
sl@0:   // The assign_or_plus_equal struct contains functions that implement
sl@0:   // assign, +=, and clearing based on the iterator type.  The
sl@0:   // generic case does nothing for plus_equal and clearing, while
sl@0:   // passing through the call for assign.
sl@0:   //
sl@0:   // When an input iterator is being used, the situation is reversed.
sl@0:   // The assign method does nothing, plus_equal invokes operator +=,
sl@0:   // and the clearing method sets the supplied token to the default
sl@0:   // token constructor's result.
sl@0:   //
sl@0: 
sl@0:   template<class IteratorTag>
sl@0:   struct assign_or_plus_equal {
sl@0:     template<class Iterator, class Token>
sl@0:     static void assign(Iterator b, Iterator e, Token &t) {
sl@0: 
sl@0: #if BOOST_WORKAROUND(BOOST_MSVC, < 1300) &&\
sl@0:     BOOST_WORKAROUND(__SGI_STL_PORT, < 0x500) &&\
sl@0:     defined(_STLP_DEBUG) &&\
sl@0:     (defined(_STLP_USE_DYNAMIC_LIB) || defined(_DLL))
sl@0:     // Problem with string::assign for msvc-stlport in debug mode: the
sl@0:     // linker tries to import the templatized version of this memfun,
sl@0:     // which is obviously not exported.
sl@0:     // See http://www.stlport.com/dcforum/DCForumID6/1763.html for details.
sl@0: 
sl@0:       t = Token();
sl@0:       while(b != e) t += *b++;
sl@0: #else
sl@0:       t.assign(b, e);
sl@0: #endif
sl@0: 
sl@0:     }
sl@0: 
sl@0:     template<class Token, class Value> 
sl@0:     static void plus_equal(Token &, const Value &) {
sl@0:     
sl@0:   }
sl@0: 
sl@0:     // If we are doing an assign, there is no need for the
sl@0:     // the clear. 
sl@0:     //
sl@0:     template<class Token>
sl@0:     static void clear(Token &) {
sl@0: 
sl@0:     }
sl@0:   };
sl@0: 
sl@0:   template <>
sl@0:   struct assign_or_plus_equal<std::input_iterator_tag> {
sl@0:     template<class Iterator, class Token>
sl@0:     static void assign(Iterator b, Iterator e, Token &t) {
sl@0: 
sl@0:     }
sl@0:     template<class Token, class Value> 
sl@0:     static void plus_equal(Token &t, const Value &v) {
sl@0:       t += v;
sl@0:     }
sl@0:     template<class Token>
sl@0:     static void clear(Token &t) {
sl@0:       t = Token();
sl@0:     }
sl@0:   };
sl@0: 
sl@0: 
sl@0:   template<class Iterator>
sl@0:   struct pointer_iterator_category{
sl@0:     typedef std::random_access_iterator_tag type;
sl@0:   };
sl@0: 
sl@0: 
sl@0:   template<class Iterator>
sl@0:   struct class_iterator_category{
sl@0:     typedef typename Iterator::iterator_category type;
sl@0:   };
sl@0: 
sl@0: 
sl@0: 
sl@0:   // This portably gets the iterator_tag without partial template specialization
sl@0:   template<class Iterator>
sl@0:     struct get_iterator_category{
sl@0:     typedef typename mpl::if_<is_pointer<Iterator>,
sl@0:       pointer_iterator_category<Iterator>,
sl@0:       class_iterator_category<Iterator>
sl@0:     >::type cat;
sl@0: 
sl@0:     typedef typename cat::type iterator_category;
sl@0:   };
sl@0: 
sl@0:   
sl@0: }
sl@0: 
sl@0:    
sl@0:   //===========================================================================
sl@0:   // The offset_separator class, which is a model of TokenizerFunction.
sl@0:   // Offset breaks a string into tokens based on a range of offsets
sl@0: 
sl@0:   class offset_separator {
sl@0:   private:
sl@0: 
sl@0:     std::vector<int> offsets_;
sl@0:     unsigned int current_offset_;
sl@0:     bool wrap_offsets_;
sl@0:     bool return_partial_last_;
sl@0:     
sl@0:   public:
sl@0:     template <typename Iter>
sl@0:     offset_separator(Iter begin, Iter end, bool wrap_offsets = true,
sl@0:                      bool return_partial_last = true)
sl@0:       : offsets_(begin,end), current_offset_(0),
sl@0:         wrap_offsets_(wrap_offsets),
sl@0:         return_partial_last_(return_partial_last) { }
sl@0:    
sl@0:     offset_separator()
sl@0:       : offsets_(1,1), current_offset_(),
sl@0:         wrap_offsets_(true), return_partial_last_(true) { }
sl@0: 
sl@0:     void reset() {
sl@0:       current_offset_ = 0;
sl@0:     }
sl@0: 
sl@0:     template <typename InputIterator, typename Token>
sl@0:     bool operator()(InputIterator& next, InputIterator end, Token& tok)
sl@0:     {
sl@0:       typedef tokenizer_detail::assign_or_plus_equal<
sl@0: #if     !defined(BOOST_MSVC) || BOOST_MSVC > 1300
sl@0:         typename
sl@0: #endif
sl@0:         tokenizer_detail::get_iterator_category<
sl@0:         InputIterator>::iterator_category> assigner;
sl@0: 
sl@0: 
sl@0:       BOOST_ASSERT(!offsets_.empty());
sl@0:     
sl@0:       assigner::clear(tok);
sl@0:       InputIterator start(next);
sl@0:       
sl@0:       if (next == end)
sl@0:         return false;
sl@0: 
sl@0:       if (current_offset_ == offsets_.size())
sl@0:         if (wrap_offsets_)
sl@0:           current_offset_=0;
sl@0:         else
sl@0:           return false;
sl@0:       
sl@0:       int c = offsets_[current_offset_];
sl@0:       int i = 0;
sl@0:       for (; i < c; ++i) {
sl@0:         if (next == end)break;
sl@0:         assigner::plus_equal(tok,*next++);
sl@0:       }
sl@0:       assigner::assign(start,next,tok);
sl@0:     
sl@0:       if (!return_partial_last_)
sl@0:         if (i < (c-1) )
sl@0:           return false;
sl@0:       
sl@0:       ++current_offset_;
sl@0:       return true;
sl@0:     }
sl@0:   };
sl@0: 
sl@0: 
sl@0:   //===========================================================================
sl@0:   // The char_separator class breaks a sequence of characters into
sl@0:   // tokens based on the character delimiters (very much like bad old
sl@0:   // strtok). A delimiter character can either be kept or dropped. A
sl@0:   // kept delimiter shows up as an output token, whereas a dropped
sl@0:   // delimiter does not.
sl@0: 
sl@0:   // This class replaces the char_delimiters_separator class. The
sl@0:   // constructor for the char_delimiters_separator class was too
sl@0:   // confusing and needed to be deprecated. However, because of the
sl@0:   // default arguments to the constructor, adding the new constructor
sl@0:   // would cause ambiguity, so instead I deprecated the whole class.
sl@0:   // The implementation of the class was also simplified considerably.
sl@0: 
sl@0:   enum empty_token_policy { drop_empty_tokens, keep_empty_tokens };
sl@0: 
sl@0:   // The out of the box GCC 2.95 on cygwin does not have a char_traits class.
sl@0: #if !defined(BOOST_MSVC) || BOOST_MSVC > 1300
sl@0:   template <typename Char, 
sl@0:     typename Traits = typename std::basic_string<Char>::traits_type >
sl@0: #else
sl@0:   template <typename Char, 
sl@0:     typename Traits = std::basic_string<Char>::traits_type >
sl@0: #endif
sl@0:   class char_separator
sl@0:   {
sl@0:     typedef std::basic_string<Char,Traits> string_type;
sl@0:   public:
sl@0:     explicit 
sl@0:     char_separator(const Char* dropped_delims,
sl@0:                    const Char* kept_delims = 0,
sl@0:                    empty_token_policy empty_tokens = drop_empty_tokens)
sl@0:       : m_dropped_delims(dropped_delims),
sl@0:         m_use_ispunct(false),
sl@0:         m_use_isspace(false),
sl@0:         m_empty_tokens(empty_tokens),
sl@0:         m_output_done(false)
sl@0:     {
sl@0:       // Borland workaround
sl@0:       if (kept_delims)
sl@0:         m_kept_delims = kept_delims;
sl@0:     }
sl@0: 
sl@0:                 // use ispunct() for kept delimiters and isspace for dropped.
sl@0:     explicit
sl@0:     char_separator()
sl@0:       : m_use_ispunct(true), 
sl@0:         m_use_isspace(true), 
sl@0:         m_empty_tokens(drop_empty_tokens) { }
sl@0: 
sl@0:     void reset() { }
sl@0: 
sl@0:     template <typename InputIterator, typename Token>
sl@0:     bool operator()(InputIterator& next, InputIterator end, Token& tok)
sl@0:     {
sl@0:       typedef tokenizer_detail::assign_or_plus_equal<
sl@0: #if     !defined(BOOST_MSVC) || BOOST_MSVC > 1300
sl@0:         typename
sl@0: #endif
sl@0:         tokenizer_detail::get_iterator_category<
sl@0:         InputIterator>::iterator_category> assigner;
sl@0: 
sl@0:       assigner::clear(tok);
sl@0: 
sl@0:       // skip past all dropped_delims
sl@0:       if (m_empty_tokens == drop_empty_tokens)
sl@0:         for (; next != end  && is_dropped(*next); ++next)
sl@0:           { }
sl@0:       
sl@0:       InputIterator start(next);
sl@0: 
sl@0:       if (m_empty_tokens == drop_empty_tokens) {
sl@0: 
sl@0:         if (next == end)
sl@0:           return false;
sl@0: 
sl@0: 
sl@0:         // if we are on a kept_delims move past it and stop
sl@0:         if (is_kept(*next)) {
sl@0:           assigner::plus_equal(tok,*next);
sl@0:           ++next;
sl@0:         } else
sl@0:           // append all the non delim characters
sl@0:           for (; next != end && !is_dropped(*next) && !is_kept(*next); ++next)
sl@0:             assigner::plus_equal(tok,*next);
sl@0:       } 
sl@0:       else { // m_empty_tokens == keep_empty_tokens
sl@0:         
sl@0:         // Handle empty token at the end
sl@0:         if (next == end)
sl@0:           if (m_output_done == false) {
sl@0:             m_output_done = true;
sl@0:             assigner::assign(start,next,tok);
sl@0:             return true;
sl@0:           } else
sl@0:             return false;
sl@0:         
sl@0:         if (is_kept(*next)) {
sl@0:           if (m_output_done == false)
sl@0:             m_output_done = true;
sl@0:           else {
sl@0:             assigner::plus_equal(tok,*next);
sl@0:             ++next;
sl@0:             m_output_done = false;
sl@0:           }
sl@0:         } 
sl@0:         else if (m_output_done == false && is_dropped(*next)) {
sl@0:           m_output_done = true;
sl@0:         } 
sl@0:         else {
sl@0:           if (is_dropped(*next))
sl@0:             start=++next;
sl@0:           for (; next != end && !is_dropped(*next) && !is_kept(*next); ++next)
sl@0:             assigner::plus_equal(tok,*next);
sl@0:           m_output_done = true;
sl@0:         }
sl@0:       }
sl@0:       assigner::assign(start,next,tok);
sl@0:       return true;
sl@0:     }
sl@0: 
sl@0:   private:
sl@0:     string_type m_kept_delims;
sl@0:     string_type m_dropped_delims;
sl@0:     bool m_use_ispunct;
sl@0:     bool m_use_isspace;
sl@0:     empty_token_policy m_empty_tokens;
sl@0:     bool m_output_done;
sl@0:     
sl@0:     bool is_kept(Char E) const
sl@0:     {  
sl@0:       if (m_kept_delims.length())
sl@0:         return m_kept_delims.find(E) != string_type::npos;
sl@0:       else if (m_use_ispunct) {
sl@0:         return std::ispunct(E) != 0;
sl@0:       } else
sl@0:         return false;
sl@0:     }
sl@0:     bool is_dropped(Char E) const
sl@0:     {
sl@0:       if (m_dropped_delims.length())
sl@0:         return m_dropped_delims.find(E) != string_type::npos;
sl@0:       else if (m_use_isspace) {
sl@0:         return std::isspace(E) != 0;
sl@0:       } else
sl@0:         return false;
sl@0:     }
sl@0:   };
sl@0: 
sl@0:   //===========================================================================
sl@0:   // The following class is DEPRECATED, use class char_separators instead.
sl@0:   //
sl@0:   // The char_delimiters_separator class, which is a model of
sl@0:   // TokenizerFunction.  char_delimiters_separator breaks a string
sl@0:   // into tokens based on character delimiters. There are 2 types of
sl@0:   // delimiters. returnable delimiters can be returned as
sl@0:   // tokens. These are often punctuation. nonreturnable delimiters
sl@0:   // cannot be returned as tokens. These are often whitespace
sl@0: 
sl@0:   // The out of the box GCC 2.95 on cygwin does not have a char_traits class.
sl@0: #if !defined(BOOST_MSVC) || BOOST_MSVC > 1300
sl@0:   template <class Char, 
sl@0:     class Traits = typename std::basic_string<Char>::traits_type >
sl@0: #else
sl@0:   template <class Char, 
sl@0:     class Traits = std::basic_string<Char>::traits_type >
sl@0: #endif
sl@0:   class char_delimiters_separator {
sl@0:   private:  
sl@0: 
sl@0:     typedef std::basic_string<Char,Traits> string_type;
sl@0:     string_type returnable_;
sl@0:     string_type nonreturnable_;
sl@0:     bool return_delims_;
sl@0:     bool no_ispunct_;
sl@0:     bool no_isspace_;
sl@0:     
sl@0:     bool is_ret(Char E)const
sl@0:     {  
sl@0:       if (returnable_.length())
sl@0:         return  returnable_.find(E) != string_type::npos;
sl@0:       else{
sl@0:         if (no_ispunct_) {return false;}
sl@0:         else{
sl@0:           int r = std::ispunct(E);
sl@0:           return r != 0;
sl@0:         }
sl@0:       }
sl@0:     }
sl@0:     bool is_nonret(Char E)const
sl@0:     {
sl@0:       if (nonreturnable_.length())
sl@0:         return  nonreturnable_.find(E) != string_type::npos;
sl@0:       else{
sl@0:         if (no_isspace_) {return false;}
sl@0:         else{
sl@0:           int r = std::isspace(E);
sl@0:           return r != 0;
sl@0:         }
sl@0:       }
sl@0:     }
sl@0:     
sl@0:   public:
sl@0:     explicit char_delimiters_separator(bool return_delims = false, 
sl@0:                                        const Char* returnable = 0,
sl@0:                                        const Char* nonreturnable = 0)
sl@0:       : returnable_(returnable ? returnable : string_type().c_str()),
sl@0:         nonreturnable_(nonreturnable ? nonreturnable:string_type().c_str()),
sl@0:         return_delims_(return_delims), no_ispunct_(returnable!=0),
sl@0:         no_isspace_(nonreturnable!=0) { }
sl@0:     
sl@0:     void reset() { }
sl@0: 
sl@0:   public:
sl@0: 
sl@0:      template <typename InputIterator, typename Token>
sl@0:      bool operator()(InputIterator& next, InputIterator end,Token& tok) {
sl@0:      tok = Token();
sl@0:      
sl@0:      // skip past all nonreturnable delims
sl@0:      // skip past the returnable only if we are not returning delims
sl@0:      for (;next!=end && ( is_nonret(*next) || (is_ret(*next) 
sl@0:        && !return_delims_ ) );++next) { }
sl@0:      
sl@0:      if (next == end) {
sl@0:        return false;
sl@0:      }
sl@0:      
sl@0:      // if we are to return delims and we are one a returnable one
sl@0:      // move past it and stop
sl@0:      if (is_ret(*next) && return_delims_) {
sl@0:        tok+=*next;
sl@0:        ++next;
sl@0:      }
sl@0:      else
sl@0:        // append all the non delim characters
sl@0:        for (;next!=end && !is_nonret(*next) && !is_ret(*next);++next)
sl@0:          tok+=*next;
sl@0:        
sl@0:        
sl@0:      return true;
sl@0:    }
sl@0:   };
sl@0: 
sl@0: 
sl@0: } //namespace boost
sl@0: 
sl@0: 
sl@0: #endif 
sl@0: 
sl@0: 
sl@0: 
sl@0: 
sl@0: