sl@0: // ----------------------------------------------------------------------------
sl@0: // parsing.hpp :  implementation of the parsing member functions
sl@0: //                      ( parse, parse_printf_directive)
sl@0: // ----------------------------------------------------------------------------
sl@0: 
sl@0: //  Copyright Samuel Krempp 2003. Use, modification, and distribution are
sl@0: //  subject to the Boost Software License, Version 1.0. (See accompanying
sl@0: //  file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
sl@0: 
sl@0: // see http://www.boost.org/libs/format for library home page
sl@0: 
sl@0: // ----------------------------------------------------------------------------
sl@0: 
sl@0: #ifndef BOOST_FORMAT_PARSING_HPP
sl@0: #define BOOST_FORMAT_PARSING_HPP
sl@0: 
sl@0: 
sl@0: #include <boost/format/format_class.hpp>
sl@0: #include <boost/throw_exception.hpp>
sl@0: #include <boost/assert.hpp>
sl@0: 
sl@0: 
sl@0: namespace boost {
sl@0: namespace io {
sl@0: namespace detail {
sl@0: 
sl@0: #if defined(BOOST_NO_STD_LOCALE)
sl@0:     // streams will be used for narrow / widen. but these methods are not const
sl@0:     template<class T>
sl@0:     T& const_or_not(const T& x) { 
sl@0:         return const_cast<T&> (x);
sl@0:     }
sl@0: #else
sl@0:     template<class T>
sl@0:     const T& const_or_not(const T& x) { 
sl@0:         return x;
sl@0:     }
sl@0: #endif
sl@0: 
sl@0:     template<class Ch, class Facet> inline
sl@0:     char wrap_narrow(const Facet& fac, Ch c, char deflt) {
sl@0:         return const_or_not(fac).narrow(c, deflt);
sl@0:     }
sl@0: 
sl@0:     template<class Ch, class Facet> inline
sl@0:     bool wrap_isdigit(const Facet& fac, Ch c) {
sl@0: #if ! defined( BOOST_NO_LOCALE_ISDIGIT )
sl@0:         return fac.is(std::ctype<Ch>::digit, c);
sl@0: # else
sl@0:         using namespace std;
sl@0:         return isdigit(c); 
sl@0: #endif 
sl@0:     }
sl@0:  
sl@0:     template<class Iter, class Facet> 
sl@0:     Iter wrap_scan_notdigit(const Facet & fac, Iter beg, Iter end) {
sl@0:         using namespace std;
sl@0:         for( ; beg!=end && wrap_isdigit(fac, *beg); ++beg) ;
sl@0:         return beg;
sl@0:     }
sl@0: 
sl@0: 
sl@0:     // Input : [start, last) iterators range and a
sl@0:     //          a Facet to use its widen/narrow member function
sl@0:     // Effects : read sequence and convert digits into integral n, of type Res
sl@0:     // Returns : n
sl@0:     template<class Res, class Iter, class Facet>
sl@0:     Iter str2int (const Iter & start, const Iter & last, Res & res, 
sl@0:                  const Facet& fac) 
sl@0:     {
sl@0:         using namespace std;
sl@0:         Iter it;
sl@0:         res=0;
sl@0:         for(it=start; it != last && wrap_isdigit(fac, *it); ++it ) {
sl@0:             char cur_ch = wrap_narrow(fac, *it, 0); // cant fail.
sl@0:             res *= 10;
sl@0:             res += cur_ch - '0'; // 22.2.1.1.2.13 of the C++ standard
sl@0:         }
sl@0:         return it;
sl@0:     }
sl@0: 
sl@0:     // skip printf's "asterisk-fields" directives in the format-string buf
sl@0:     // Input : char string, with starting index *pos_p
sl@0:     //         a Facet merely to use its widen/narrow member function
sl@0:     // Effects : advance *pos_p by skipping printf's asterisk fields.
sl@0:     // Returns : nothing
sl@0:     template<class Iter, class Facet>
sl@0:     Iter skip_asterisk(Iter start, Iter last, const Facet& fac) 
sl@0:     {
sl@0:         using namespace std;
sl@0:         ++ start;
sl@0:         start = wrap_scan_notdigit(fac, start, last);
sl@0:         if(start!=last && *start== const_or_not(fac).widen( '$') )
sl@0:             ++start;
sl@0:         return start;
sl@0:     }
sl@0: 
sl@0: 
sl@0:     // auxiliary func called by parse_printf_directive
sl@0:     // for centralising error handling
sl@0:     // it either throws if user sets the corresponding flag, or does nothing.
sl@0:     inline void maybe_throw_exception(unsigned char exceptions, 
sl@0:                                       std::size_t pos, std::size_t size)
sl@0:     {
sl@0:         if(exceptions & io::bad_format_string_bit)
sl@0:             boost::throw_exception(io::bad_format_string(pos, size) );
sl@0:     }
sl@0:     
sl@0: 
sl@0:     // Input: the position of a printf-directive in the format-string
sl@0:     //    a basic_ios& merely to use its widen/narrow member function
sl@0:     //    a bitset'exceptions' telling whether to throw exceptions on errors.
sl@0:     // Returns:
sl@0:     //  true if parse succeeded (ignore some errors if exceptions disabled)
sl@0:     //  false if it failed so bad that the directive should be printed verbatim
sl@0:     // Effects:
sl@0:     //  start is incremented so that *start is the first char after
sl@0:     //     this directive
sl@0:     //  *fpar is set with the parameters read in the directive
sl@0:     template<class Ch, class Tr, class Alloc, class Iter, class Facet>
sl@0:     bool parse_printf_directive(Iter & start, const Iter& last, 
sl@0:                                 detail::format_item<Ch, Tr, Alloc> * fpar,
sl@0:                                 const Facet& fac,
sl@0:                                 std::size_t offset, unsigned char exceptions)
sl@0:     {
sl@0:         typedef typename basic_format<Ch, Tr, Alloc>::format_item_t format_item_t;
sl@0: 
sl@0:         fpar->argN_ = format_item_t::argN_no_posit;  // if no positional-directive
sl@0:         bool precision_set = false;
sl@0:         bool in_brackets=false;
sl@0:         Iter start0 = start;
sl@0:         std::size_t fstring_size = last-start0+offset;
sl@0:         if(*start== const_or_not(fac).widen( '|')) {
sl@0:             in_brackets=true;
sl@0:             if( ++start >= last ) {
sl@0:                 maybe_throw_exception(exceptions, start-start0 + offset, fstring_size);
sl@0:                 return false;
sl@0:             }
sl@0:         }
sl@0: 
sl@0:         // the flag '0' would be picked as a digit for argument order, but here it's a flag :
sl@0:         if(*start== const_or_not(fac).widen( '0')) 
sl@0:             goto parse_flags;
sl@0: 
sl@0:         // handle argument order (%2$d)  or possibly width specification: %2d
sl@0:         if(wrap_isdigit(fac, *start)) {
sl@0:             int n;
sl@0:             start = str2int(start, last, n, fac);
sl@0:             if( start >= last ) {
sl@0:                 maybe_throw_exception(exceptions, start-start0+offset, fstring_size);
sl@0:                 return false;
sl@0:             }
sl@0:             
sl@0:             // %N% case : this is already the end of the directive
sl@0:             if( *start ==  const_or_not(fac).widen( '%') ) {
sl@0:                 fpar->argN_ = n-1;
sl@0:                 ++start;
sl@0:                 if( in_brackets) 
sl@0:                     maybe_throw_exception(exceptions, start-start0+offset, fstring_size); 
sl@0:                 // but don't return.  maybe "%" was used in lieu of '$', so we go on.
sl@0:                 else
sl@0:                     return true;
sl@0:             }
sl@0: 
sl@0:             if ( *start== const_or_not(fac).widen( '$') ) {
sl@0:                 fpar->argN_ = n-1;
sl@0:                 ++start;
sl@0:             } 
sl@0:             else {
sl@0:                 // non-positionnal directive
sl@0:                 fpar->fmtstate_.width_ = n;
sl@0:                 fpar->argN_  = format_item_t::argN_no_posit;
sl@0:                 goto parse_precision;
sl@0:             }
sl@0:         }
sl@0:     
sl@0:       parse_flags: 
sl@0:         // handle flags
sl@0:         while ( start != last) { // as long as char is one of + - = _ # 0 l h   or ' '
sl@0:             // misc switches
sl@0:             switch ( wrap_narrow(fac, *start, 0)) {
sl@0:             case '\'' : break; // no effect yet. (painful to implement)
sl@0:             case 'l':
sl@0:             case 'h':  // short/long modifier : for printf-comaptibility (no action needed)
sl@0:                 break;
sl@0:             case '-':
sl@0:                 fpar->fmtstate_.flags_ |= std::ios_base::left;
sl@0:                 break;
sl@0:             case '=':
sl@0:                 fpar->pad_scheme_ |= format_item_t::centered;
sl@0:                 break;
sl@0:             case '_':
sl@0:                 fpar->fmtstate_.flags_ |= std::ios_base::internal;
sl@0:                 break;
sl@0:             case ' ':
sl@0:                 fpar->pad_scheme_ |= format_item_t::spacepad;
sl@0:                 break;
sl@0:             case '+':
sl@0:                 fpar->fmtstate_.flags_ |= std::ios_base::showpos;
sl@0:                 break;
sl@0:             case '0':
sl@0:                 fpar->pad_scheme_ |= format_item_t::zeropad;
sl@0:                 // need to know alignment before really setting flags,
sl@0:                 // so just add 'zeropad' flag for now, it will be processed later.
sl@0:                 break;
sl@0:             case '#':
sl@0:                 fpar->fmtstate_.flags_ |= std::ios_base::showpoint | std::ios_base::showbase;
sl@0:                 break;
sl@0:             default:
sl@0:                 goto parse_width;
sl@0:             }
sl@0:             ++start;
sl@0:         } // loop on flag.
sl@0: 
sl@0:         if( start>=last) {
sl@0:             maybe_throw_exception(exceptions, start-start0+offset, fstring_size);
sl@0:             return true; 
sl@0:         }
sl@0:       parse_width:
sl@0:         // handle width spec
sl@0:         // first skip 'asterisk fields' :  *, or *N$
sl@0:         if(*start == const_or_not(fac).widen( '*') )
sl@0:             start = skip_asterisk(start, last, fac); 
sl@0:         if(start!=last && wrap_isdigit(fac, *start))
sl@0:             start = str2int(start, last, fpar->fmtstate_.width_, fac);
sl@0: 
sl@0:       parse_precision:
sl@0:         if( start>= last) { 
sl@0:             maybe_throw_exception(exceptions, start-start0+offset, fstring_size);
sl@0:             return true;
sl@0:         }
sl@0:         // handle precision spec
sl@0:         if (*start== const_or_not(fac).widen( '.')) {
sl@0:             ++start;
sl@0:             if(start != last && *start == const_or_not(fac).widen( '*') )
sl@0:                 start = skip_asterisk(start, last, fac); 
sl@0:             if(start != last && wrap_isdigit(fac, *start)) {
sl@0:                 start = str2int(start, last, fpar->fmtstate_.precision_, fac);
sl@0:                 precision_set = true;
sl@0:             }
sl@0:             else
sl@0:                 fpar->fmtstate_.precision_ =0;
sl@0:         }
sl@0:     
sl@0:         // handle  formatting-type flags :
sl@0:         while( start != last && ( *start== const_or_not(fac).widen( 'l') 
sl@0:                                   || *start== const_or_not(fac).widen( 'L') 
sl@0:                                   || *start== const_or_not(fac).widen( 'h')) )
sl@0:             ++start;
sl@0:         if( start>=last) {
sl@0:             maybe_throw_exception(exceptions, start-start0+offset, fstring_size);
sl@0:             return true;
sl@0:         }
sl@0: 
sl@0:         if( in_brackets && *start== const_or_not(fac).widen( '|') ) {
sl@0:             ++start;
sl@0:             return true;
sl@0:         }
sl@0:         switch ( wrap_narrow(fac, *start, 0) ) {
sl@0:         case 'X':
sl@0:             fpar->fmtstate_.flags_ |= std::ios_base::uppercase;
sl@0:         case 'p': // pointer => set hex.
sl@0:         case 'x':
sl@0:             fpar->fmtstate_.flags_ &= ~std::ios_base::basefield;
sl@0:             fpar->fmtstate_.flags_ |= std::ios_base::hex;
sl@0:             break;
sl@0: 
sl@0:         case 'o':
sl@0:             fpar->fmtstate_.flags_ &= ~std::ios_base::basefield;
sl@0:             fpar->fmtstate_.flags_ |=  std::ios_base::oct;
sl@0:             break;
sl@0: 
sl@0:         case 'E':
sl@0:             fpar->fmtstate_.flags_ |=  std::ios_base::uppercase;
sl@0:         case 'e':
sl@0:             fpar->fmtstate_.flags_ &= ~std::ios_base::floatfield;
sl@0:             fpar->fmtstate_.flags_ |=  std::ios_base::scientific;
sl@0: 
sl@0:             fpar->fmtstate_.flags_ &= ~std::ios_base::basefield;
sl@0:             fpar->fmtstate_.flags_ |=  std::ios_base::dec;
sl@0:             break;
sl@0:       
sl@0:         case 'f':
sl@0:             fpar->fmtstate_.flags_ &= ~std::ios_base::floatfield;
sl@0:             fpar->fmtstate_.flags_ |=  std::ios_base::fixed;
sl@0:         case 'u':
sl@0:         case 'd':
sl@0:         case 'i':
sl@0:             fpar->fmtstate_.flags_ &= ~std::ios_base::basefield;
sl@0:             fpar->fmtstate_.flags_ |=  std::ios_base::dec;
sl@0:             break;
sl@0: 
sl@0:         case 'T':
sl@0:             ++start;
sl@0:             if( start >= last)
sl@0:                 maybe_throw_exception(exceptions, start-start0+offset, fstring_size);
sl@0:             else
sl@0:                 fpar->fmtstate_.fill_ = *start;
sl@0:             fpar->pad_scheme_ |= format_item_t::tabulation;
sl@0:             fpar->argN_ = format_item_t::argN_tabulation; 
sl@0:             break;
sl@0:         case 't': 
sl@0:             fpar->fmtstate_.fill_ = const_or_not(fac).widen( ' ');
sl@0:             fpar->pad_scheme_ |= format_item_t::tabulation;
sl@0:             fpar->argN_ = format_item_t::argN_tabulation; 
sl@0:             break;
sl@0: 
sl@0:         case 'G':
sl@0:             fpar->fmtstate_.flags_ |= std::ios_base::uppercase;
sl@0:             break;
sl@0:         case 'g': // 'g' conversion is default for floats.
sl@0:             fpar->fmtstate_.flags_ &= ~std::ios_base::basefield;
sl@0:             fpar->fmtstate_.flags_ |=  std::ios_base::dec;
sl@0: 
sl@0:             // CLEAR all floatield flags, so stream will CHOOSE
sl@0:             fpar->fmtstate_.flags_ &= ~std::ios_base::floatfield; 
sl@0:             break;
sl@0: 
sl@0:         case 'C':
sl@0:         case 'c': 
sl@0:             fpar->truncate_ = 1;
sl@0:             break;
sl@0:         case 'S':
sl@0:         case 's': 
sl@0:             if(precision_set) // handle truncation manually, with own parameter.
sl@0:                 fpar->truncate_ = fpar->fmtstate_.precision_;
sl@0:             fpar->fmtstate_.precision_ = 6; // default stream precision.
sl@0:             break;
sl@0:         case 'n' :  
sl@0:             fpar->argN_ = format_item_t::argN_ignored;
sl@0:             break;
sl@0:         default: 
sl@0:             maybe_throw_exception(exceptions, start-start0+offset, fstring_size);
sl@0:         }
sl@0:         ++start;
sl@0: 
sl@0:         if( in_brackets ) {
sl@0:             if( start != last && *start== const_or_not(fac).widen( '|') ) {
sl@0:                 ++start;
sl@0:                 return true;
sl@0:             }
sl@0:             else  maybe_throw_exception(exceptions, start-start0+offset, fstring_size);
sl@0:         }
sl@0:         return true;
sl@0:     }
sl@0:     // -end parse_printf_directive()
sl@0: 
sl@0:     template<class String, class Facet>
sl@0:     int upper_bound_from_fstring(const String& buf, 
sl@0:                                  const typename String::value_type arg_mark,
sl@0:                                  const Facet& fac, 
sl@0:                                  unsigned char exceptions) 
sl@0:     {
sl@0:         // quick-parsing of the format-string to count arguments mark (arg_mark, '%')
sl@0:         // returns : upper bound on the number of format items in the format strings
sl@0:         using namespace boost::io;
sl@0:         typename String::size_type i1=0;
sl@0:         int num_items=0;
sl@0:         while( (i1=buf.find(arg_mark,i1)) != String::npos ) {
sl@0:             if( i1+1 >= buf.size() ) {
sl@0:                 if(exceptions & bad_format_string_bit)
sl@0:                     boost::throw_exception(bad_format_string(i1, buf.size() )); // must not end in ".. %"
sl@0:                 else break; // stop there, ignore last '%'
sl@0:             }
sl@0:             if(buf[i1+1] == buf[i1] ) {// escaped "%%"
sl@0:                 i1+=2; continue; 
sl@0:             }
sl@0: 
sl@0:             ++i1;
sl@0:             // in case of %N% directives, dont count it double (wastes allocations..) :
sl@0:             i1 = detail::wrap_scan_notdigit(fac, buf.begin()+i1, buf.end()) - buf.begin();
sl@0:             if( i1 < buf.size() && buf[i1] == arg_mark )
sl@0:                 ++i1;
sl@0:             ++num_items;
sl@0:         }
sl@0:         return num_items;
sl@0:     }
sl@0:     template<class String> inline
sl@0:     void append_string(String& dst, const String& src, 
sl@0:                        const typename String::size_type beg, 
sl@0:                        const typename String::size_type end) {
sl@0: #if !defined(BOOST_NO_STRING_APPEND)
sl@0:         dst.append(src.begin()+beg, src.begin()+end);
sl@0: #else
sl@0:         dst += src.substr(beg, end-beg);
sl@0: #endif
sl@0:     }
sl@0: 
sl@0: } // detail namespace
sl@0: } // io namespace
sl@0: 
sl@0: 
sl@0: 
sl@0: // -----------------------------------------------
sl@0: //  format :: parse(..)
sl@0: 
sl@0:     template<class Ch, class Tr, class Alloc>
sl@0:     basic_format<Ch, Tr, Alloc>& basic_format<Ch, Tr, Alloc>:: 
sl@0:     parse (const string_type& buf) {
sl@0:         // parse the format-string 
sl@0:         using namespace std;
sl@0: #if !defined(BOOST_NO_STD_LOCALE)
sl@0:         const std::ctype<Ch> & fac = BOOST_USE_FACET( std::ctype<Ch>, getloc());
sl@0: #else
sl@0:         io::basic_oaltstringstream<Ch, Tr, Alloc> fac; 
sl@0:         //has widen and narrow even on compilers without locale
sl@0: #endif
sl@0: 
sl@0:         const Ch arg_mark = io::detail::const_or_not(fac).widen( '%');
sl@0:         bool ordered_args=true; 
sl@0:         int max_argN=-1;
sl@0: 
sl@0:         // A: find upper_bound on num_items and allocates arrays
sl@0:         int num_items = io::detail::upper_bound_from_fstring(buf, arg_mark, fac, exceptions());
sl@0:         make_or_reuse_data(num_items);
sl@0: 
sl@0:         // B: Now the real parsing of the format string :
sl@0:         num_items=0;
sl@0:         typename string_type::size_type i0=0, i1=0;
sl@0:         typename string_type::const_iterator it;
sl@0:         bool special_things=false;
sl@0:         int cur_item=0;
sl@0:         while( (i1=buf.find(arg_mark,i1)) != string_type::npos ) {
sl@0:             string_type & piece = (cur_item==0) ? prefix_ : items_[cur_item-1].appendix_;
sl@0:             if( buf[i1+1] == buf[i1] ) { // escaped mark, '%%' 
sl@0:                 io::detail::append_string(piece, buf, i0, i1+1);
sl@0:                 i1+=2; i0=i1;
sl@0:                 continue; 
sl@0:             }
sl@0:             BOOST_ASSERT(  static_cast<unsigned int>(cur_item) < items_.size() || cur_item==0);
sl@0: 
sl@0:             if(i1!=i0)
sl@0:                 io::detail::append_string(piece, buf, i0, i1);
sl@0:             ++i1;
sl@0:             it = buf.begin()+i1;
sl@0:             bool parse_ok = io::detail::parse_printf_directive(
sl@0:                 it, buf.end(), &items_[cur_item], fac, i1, exceptions());
sl@0:             i1 = it - buf.begin();
sl@0:             if( ! parse_ok ) // the directive will be printed verbatim
sl@0:                 continue; 
sl@0:             i0=i1;
sl@0:             items_[cur_item].compute_states(); // process complex options, like zeropad, into params
sl@0: 
sl@0:             int argN=items_[cur_item].argN_;
sl@0:             if(argN == format_item_t::argN_ignored)
sl@0:                 continue;
sl@0:             if(argN ==format_item_t::argN_no_posit)
sl@0:                 ordered_args=false;
sl@0:             else if(argN == format_item_t::argN_tabulation) special_things=true;
sl@0:             else if(argN > max_argN) max_argN = argN;
sl@0:             ++num_items;
sl@0:             ++cur_item;
sl@0:         } // loop on %'s
sl@0:         BOOST_ASSERT(cur_item == num_items);
sl@0:     
sl@0:         // store the final piece of string
sl@0:         {
sl@0:             string_type & piece = (cur_item==0) ? prefix_ : items_[cur_item-1].appendix_;
sl@0:             io::detail::append_string(piece, buf, i0, buf.size());
sl@0:         }
sl@0:     
sl@0:         if( !ordered_args) {
sl@0:             if(max_argN >= 0 ) {  // dont mix positional with non-positionnal directives
sl@0:                 if(exceptions() & io::bad_format_string_bit)
sl@0:                     boost::throw_exception(io::bad_format_string(max_argN, 0));
sl@0:                 // else do nothing. => positionnal arguments are processed as non-positionnal
sl@0:             }
sl@0:             // set things like it would have been with positional directives :
sl@0:             int non_ordered_items = 0;
sl@0:             for(int i=0; i< num_items; ++i)
sl@0:                 if(items_[i].argN_ == format_item_t::argN_no_posit) {
sl@0:                     items_[i].argN_ = non_ordered_items;
sl@0:                     ++non_ordered_items;
sl@0:                 }
sl@0:             max_argN = non_ordered_items-1;
sl@0:         }
sl@0:     
sl@0:         // C: set some member data :
sl@0:         items_.resize(num_items, format_item_t(io::detail::const_or_not(fac).widen( ' ')) );
sl@0: 
sl@0:         if(special_things) style_ |= special_needs;
sl@0:         num_args_ = max_argN + 1;
sl@0:         if(ordered_args) style_ |=  ordered;
sl@0:         else style_ &= ~ordered;
sl@0:         return *this;
sl@0:     }
sl@0: 
sl@0: } // namespace boost
sl@0: 
sl@0: 
sl@0: #endif //  BOOST_FORMAT_PARSING_HPP