1 // Copyright (C) 1997, 1998, 1999, 2000, 2001 Free Software Foundation, Inc.
3 // This file is part of the GNU ISO C++ Library. This library is free
4 // software; you can redistribute it and/or modify it under the
5 // terms of the GNU General Public License as published by the
6 // Free Software Foundation; either version 2, or (at your option)
9 // This library is distributed in the hope that it will be useful,
10 // but WITHOUT ANY WARRANTY; without even the implied warranty of
11 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 // GNU General Public License for more details.
14 // You should have received a copy of the GNU General Public License along
15 // with this library; see the file COPYING. If not, write to the Free
16 // Software Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307,
19 // As a special exception, you may use this file as part of a free software
20 // library without restriction. Specifically, if other files instantiate
21 // templates or use macros or inline functions from this file, or you compile
22 // this file and link it with other files to produce an executable, this
23 // file does not by itself cause the resulting executable to be covered by
24 // the GNU General Public License. This exception does not however
25 // invalidate any other reasons why the executable file might be covered by
26 // the GNU General Public License.
28 #include <bits/std_clocale.h>
29 #include <bits/std_cstring.h>
30 #include <bits/std_cassert.h>
31 #include <bits/std_cctype.h>
32 #include <bits/std_limits.h>
34 #include <bits/std_stdexcept.h>
35 #include <bits/std_locale.h>
36 #include <bits/std_istream.h>
37 #include <bits/std_ostream.h>
38 #include <bits/std_vector.h>
39 #include <bits/std_memory.h> // for auto_ptr
40 #ifdef _GLIBCPP_USE_WCHAR_T
41 # include <bits/std_cwctype.h> // for towupper, etc.
46 // Defined in globals.cc.
47 extern locale::_Impl locale_impl_c;
48 extern locale locale_c;
50 // Definitions for static const data members of locale.
51 const locale::category locale::none;
52 const locale::category locale::ctype;
53 const locale::category locale::numeric;
54 const locale::category locale::collate;
55 const locale::category locale::time;
56 const locale::category locale::monetary;
57 const locale::category locale::messages;
58 const locale::category locale::all;
60 locale::_Impl* locale::_S_classic;
61 locale::_Impl* locale::_S_global;
62 const size_t locale::_S_num_categories;
63 const size_t locale::_S_num_facets;
65 // Definitions for locale::id of standard facets.
66 locale::id ctype<char>::id;
67 locale::id codecvt<char, char, mbstate_t>::id;
69 #ifdef _GLIBCPP_USE_WCHAR_T
70 locale::id ctype<wchar_t>::id;
71 locale::id codecvt<wchar_t, char, mbstate_t>::id;
74 // Definitions for static const data members of locale::id
75 size_t locale::id::_S_highwater; // init'd to 0 by linker
77 // Definitions for static const data members of locale::_Impl
78 const locale::id* const
79 locale::_Impl::_S_id_ctype[] =
81 &std::ctype<char>::id,
82 &codecvt<char, char, mbstate_t>::id,
83 #ifdef _GLIBCPP_USE_WCHAR_T
84 &std::ctype<wchar_t>::id,
85 &codecvt<wchar_t, char, mbstate_t>::id,
90 const locale::id* const
91 locale::_Impl::_S_id_numeric[] =
96 #ifdef _GLIBCPP_USE_WCHAR_T
97 &num_get<wchar_t>::id,
98 &num_put<wchar_t>::id,
99 &numpunct<wchar_t>::id,
104 const locale::id* const
105 locale::_Impl::_S_id_collate[] =
107 &std::collate<char>::id,
108 #ifdef _GLIBCPP_USE_WCHAR_T
109 &std::collate<wchar_t>::id,
114 const locale::id* const
115 locale::_Impl::_S_id_time[] =
119 #ifdef _GLIBCPP_USE_WCHAR_T
120 &time_get<wchar_t>::id,
121 &time_put<wchar_t>::id,
126 const locale::id* const
127 locale::_Impl::_S_id_monetary[] =
129 &money_get<char>::id,
130 &money_put<char>::id,
131 &moneypunct<char, false>::id,
132 &moneypunct<char, true >::id,
133 #ifdef _GLIBCPP_USE_WCHAR_T
134 &money_get<wchar_t>::id,
135 &money_put<wchar_t>::id,
136 &moneypunct<wchar_t, false>::id,
137 &moneypunct<wchar_t, true >::id,
142 const locale::id* const
143 locale::_Impl::_S_id_messages[] =
145 &std::messages<char>::id,
146 #ifdef _GLIBCPP_USE_WCHAR_T
147 &std::messages<wchar_t>::id,
152 const locale::id* const* const
153 locale::_Impl::_S_facet_categories[] =
155 // Order must match the decl order in class locale.
156 locale::_Impl::_S_id_ctype,
157 locale::_Impl::_S_id_numeric,
158 locale::_Impl::_S_id_collate,
159 locale::_Impl::_S_id_time,
160 locale::_Impl::_S_id_monetary,
161 locale::_Impl::_S_id_messages,
165 locale::~locale() throw()
166 { _M_impl->_M_remove_reference(); }
169 locale::_M_coalesce(const locale& __base, const locale& __add,
172 __cat = _S_normalize_category(__cat);
173 _M_impl = new _Impl(*__base._M_impl, 1);
176 { _M_impl->_M_replace_categories(__add._M_impl, __cat); }
179 _M_impl->_M_remove_reference();
180 __throw_exception_again;
184 locale::locale() throw()
187 (_M_impl = _S_global)->_M_add_reference();
190 locale::locale(const locale& __other) throw()
191 { (_M_impl = __other._M_impl)->_M_add_reference(); }
193 // This is used to initialize global and classic locales, and
194 // assumes that the _Impl objects are constructed correctly.
195 locale::locale(_Impl* __ip) throw() : _M_impl(__ip)
198 locale::locale(const char* __s)
203 if (strcmp(__s, "C") == 0 || strcmp(__s, "POSIX") == 0)
204 (_M_impl = _S_classic)->_M_add_reference();
206 _M_impl = new _Impl(__s, 1);
209 __throw_runtime_error("attempt to create locale from NULL name");
212 locale::locale(const locale& __base, const char* __s, category __cat)
214 // NB: There are complicated, yet more efficient ways to do
215 // this. Building up locales on a per-category way is tedious, so
216 // let's do it this way until people complain.
218 _M_coalesce(__base, __add, __cat);
221 locale::locale(const locale& __base, const locale& __add, category __cat)
222 { _M_coalesce(__base, __add, __cat); }
225 locale::operator==(const locale& __rhs) const throw()
227 string __name = this->name();
228 return (_M_impl == __rhs._M_impl
229 || (__name != "*" && __name == __rhs.name()));
233 locale::operator=(const locale& __other) throw()
235 __other._M_impl->_M_add_reference();
236 _M_impl->_M_remove_reference();
237 _M_impl = __other._M_impl;
242 locale::global(const locale& __other)
246 _Impl* __old = _S_global;
247 __other._M_impl->_M_add_reference();
248 _S_global = __other._M_impl;
249 if (_S_global->_M_check_same_name() && _S_global->_M_names[0] != "*")
250 setlocale(LC_ALL, __other.name().c_str());
252 // Reference count sanity check: one reference removed for the
253 // subsition of __other locale, one added by return-by-value. Net
254 // difference: zero. When the returned locale object's destrutor
255 // is called, then the reference count is decremented and possibly
257 return locale(__old);
264 // Need some kind of separator character. This one was pretty much
265 // arbitrarily chosen as to not conflict with glibc locales: the
266 // exact formatting is not set in stone.
267 const char __separator = '|';
269 if (_M_impl->_M_check_same_name())
270 __ret = _M_impl->_M_names[0];
273 for (size_t i = 0; i < _S_num_categories; ++i)
274 __ret += __separator + _M_impl->_M_names[i];
287 // 26 Standard facets, 2 references.
288 // One reference for _M_classic, one for _M_global
289 _S_classic = new (&locale_impl_c) _Impl("C", 2);
290 _S_global = _S_classic;
291 new (&locale_c) locale(_S_classic);
295 // Just call destructor, so that locale_impl_c's memory is
296 // not deallocated via a call to delete.
298 _S_classic->~_Impl();
299 _S_classic = _S_global = 0;
300 __throw_exception_again;
307 locale::_S_normalize_category(category __cat)
310 if (__cat == none || (__cat & all) && !(__cat & ~all))
314 // NB: May be a C-style "LC_ALL" category; convert.
332 #ifdef _GLIBCPP_HAVE_LC_MESSAGES
341 __throw_runtime_error("bad locale category");
348 facet(size_t __refs) throw() : _M_references(__refs)
353 _M_add_reference() throw()
354 { ++_M_references; } // XXX MT
358 _M_remove_reference() throw()
360 if (_M_references-- == 0)
369 // Definitions for static const data members of ctype_base.
370 const ctype_base::mask ctype_base::space;
371 const ctype_base::mask ctype_base::print;
372 const ctype_base::mask ctype_base::cntrl;
373 const ctype_base::mask ctype_base::upper;
374 const ctype_base::mask ctype_base::lower;
375 const ctype_base::mask ctype_base::alpha;
376 const ctype_base::mask ctype_base::digit;
377 const ctype_base::mask ctype_base::punct;
378 const ctype_base::mask ctype_base::xdigit;
379 const ctype_base::mask ctype_base::alnum;
380 const ctype_base::mask ctype_base::graph;
382 // Platform-specific initialization code for ctype tables.
383 #include <bits/ctype_noninline.h>
385 const size_t ctype<char>::table_size;
387 ctype<char>::~ctype()
388 { if (_M_del) delete[] this->table(); }
390 // These are dummy placeholders as these virtual functions are never called.
392 ctype<char>::do_is(mask, char_type) const
396 ctype<char>::do_is(const char_type* __c, const char_type*, mask*) const
400 ctype<char>::do_scan_is(mask, const char_type* __c, const char_type*) const
404 ctype<char>::do_scan_not(mask, const char_type* __c, const char_type*) const
408 ctype<char>::do_widen(char __c) const
412 ctype<char>::do_widen(const char* __lo, const char* __hi, char* __dest) const
414 memcpy(__dest, __lo, __hi - __lo);
419 ctype<char>::do_narrow(char __c, char /*__dfault*/) const
423 ctype<char>::do_narrow(const char* __lo, const char* __hi,
424 char /*__dfault*/, char* __dest) const
426 memcpy(__dest, __lo, __hi - __lo);
431 ctype_byname<char>::ctype_byname(const char* /*__s*/, size_t __refs)
432 : ctype<char>(new mask[table_size], true, __refs)
435 // Definitions for static const data members of money_base
436 const money_base::pattern
437 money_base::_S_default_pattern = {{symbol, sign, none, value}};
440 _Format_cache<char>::_Format_cache()
442 _M_decimal_point('.'), _M_thousands_sep(','),
443 _M_truename("true"), _M_falsename("false"), _M_use_grouping(false)
446 #ifdef _GLIBCPP_USE_WCHAR_T
448 _Format_cache<wchar_t>::_Format_cache()
450 _M_decimal_point(L'.'), _M_thousands_sep(L','),
451 _M_truename(L"true"), _M_falsename(L"false"), _M_use_grouping(false)
457 use_facet<ctype<char> >(const locale& __loc)
459 size_t __i = ctype<char>::id._M_index;
460 const locale::_Impl* __tmp = __loc._M_impl;
461 return static_cast<const ctype<char>&>(* (*(__tmp->_M_facets))[__i]);
464 #ifdef _GLIBCPP_USE_WCHAR_T
466 const ctype<wchar_t>&
467 use_facet<ctype<wchar_t> >(const locale& __loc)
469 size_t __i = ctype<wchar_t>::id._M_index;
470 const locale::_Impl* __tmp = __loc._M_impl;
471 return static_cast<const ctype<wchar_t>&>(* (*(__tmp->_M_facets))[__i]);
477 num_get<char, istreambuf_iterator<char> >::
478 _M_extract(istreambuf_iterator<char> __beg,
479 istreambuf_iterator<char> __end, ios_base& __io,
480 ios_base::iostate& __err, char* __xtrc, int& __base,
483 typedef _Format_cache<char> __cache_type;
485 // Prepare for possible failure
488 // Stage 1: determine a conversion specifier.
489 ios_base::fmtflags __basefield = __io.flags() & ios_base::basefield;
490 if (__basefield == ios_base::dec)
492 else if (__basefield == ios_base::oct)
494 else if (__basefield == ios_base::hex)
498 // As far as I can tell, bases other than 10 are not available for
499 // floating point types
503 // Stage 2: extract characters.
504 __cache_type const* __fmt = __cache_type::_S_get(__io);
506 // Fail quickly if !__valid
509 __err |= (ios_base::eofbit | ios_base::failbit);
513 // Acceptable formats for numbers here are based on 22.2.3.1
517 const char* __lits = __fmt->_S_literals;
520 // Check first for sign
521 bool __testsign = false;
522 if ((__c == __lits[__cache_type::_S_minus])
523 || (__c == __lits[__cache_type::_S_plus]))
526 __xtrc[__pos++] = __c;
530 // Whitespace may follow a sign
531 while ((__beg != __end) && (isspace(__c)))
537 // There had better be more to come...
540 __xtrc[__pos] = '\0';
541 __err |= (ios_base::eofbit | ios_base::failbit);
546 // Now check if first character is a zero.
547 bool __testzero = false;
548 if (__c == __lits[__cache_type::_S_digits])
554 // We have to check for __beg == __end here. If so,
555 // a plain '0' (possibly with a sign) can be got rid of now
558 __xtrc[__pos++] = __lits[__cache_type::_S_digits];
559 __xtrc[__pos] = '\0';
560 __err |= ios_base::eofbit;
564 // Figure out base for integer types only
565 // Based on Table 55 of 22.2.2.1.2
566 if (!__fp && __base != 10 && __base != 8)
568 // Here, __base == 0 or 16
569 if ((__c == __lits[__cache_type::_S_x])
570 || (__c == __lits[__cache_type::_S_X]))
575 __testzero = false; // "0x" is not a leading zero
577 else if (__base == 0)
581 // Remove any more leading zeros
582 while (__beg != __end)
584 if (__c == __lits[__cache_type::_S_digits])
594 else if (__base == 0) // 1st character is not zero
597 // We now seek "units", i.e. digits and thousands separators.
598 // We may need to know if anything is found here. A leading zero
599 // (removed by now) would count.
600 bool __testunits = __testzero;
601 while (__beg != __end)
603 const char* __p = strchr(__lits, __c);
605 // NB: strchr returns true for __c == 0x0
607 &&((__p >= &__lits[__cache_type::_S_digits]
608 && __p < &__lits[__cache_type::_S_digits + __base])
609 || (__p >= &__lits[__cache_type::_S_udigits]
610 && __p < &__lits[__cache_type::_S_udigits + __base])))
612 // Try first for acceptable digit; record it if found.
613 __xtrc[__pos++] = __c;
619 else if (__c == __fmt->_M_thousands_sep && __fmt->_M_use_grouping)
621 // NB: Thousands separator at the beginning of a string
622 // is a no-no, as is two consecutive thousands
626 __grp += static_cast<char>(__sep_pos);
633 __err |= ios_base::failbit;
638 // Not a valid input item.
642 // Digit grouping is checked. If _M_groupings() doesn't
643 // match, then get very very upset, and set failbit.
644 if (__fmt->_M_use_grouping && !__grp.empty())
646 // Add the ending grouping
647 __grp += static_cast<char>(__sep_pos);
649 // __grp is parsed L to R
650 // 1,222,444 == __grp of "/1/3/3"
651 // __fmt->_M_grouping is parsed R to L
652 // 1,222,444 == __fmt->_M_grouping of "/3" == "/3/3/3"
655 const int __len = __fmt->_M_grouping.size();
656 int __n = __grp.size();
659 // Parsed number groupings have to match the
660 // numpunct::grouping string exactly, starting at the
661 // right-most point of the parsed sequence of elements ...
662 while (__test && __i < __n - 1)
663 for (__j = 0; __test && __j < __len && __i < __n - 1; ++__j,++__i)
664 __test &= __fmt->_M_grouping[__j] == __grp[__n - __i - 1];
665 // ... but the last parsed grouping can be <= numpunct
667 __j == __len ? __j = 0 : __j;
668 __test &= __fmt->_M_grouping[__j] >= __grp[__n - __i - 1];
672 __err |= ios_base::failbit;
673 __xtrc[__pos] = '\0';
675 __err |= ios_base::eofbit;
680 // If there was nothing but zeros, put one in the output string
681 if (__testzero && (__pos == 0 || (__pos == 1 && __testsign)))
682 __xtrc[__pos++] = __lits[__cache_type::_S_digits];
684 // That's it for integer types. Remaining code is for floating point
685 if (__fp && __beg != __end)
687 // Check first for decimal point. There MUST be one if
688 // __testunits is false.
689 bool __testdec = false; // Is there a decimal point
690 // with digits following it?
691 if (__c == __fmt->_M_decimal_point)
693 __xtrc[__pos++] = '.';
697 // Now we get any digits after the decimal point
698 // There MUST be some if __testunits is false.
699 while (__beg != __end)
701 const char* __p = strchr(__lits, __c);
702 if ((__p >= &__lits[__cache_type::_S_digits]
703 && __p < &__lits[__cache_type::_S_digits + __base])
704 || (__p >= &__lits[__cache_type::_S_udigits]
705 && __p < &__lits[__cache_type::_S_udigits + __base]))
707 __xtrc[__pos++] = __c;
716 if (!__testunits && !__testdec) // Ill formed
718 __err |= ios_base::failbit;
719 __xtrc[__pos] = '\0';
721 __err |= ios_base::eofbit;
725 // Now we may find an exponent
728 if ((__c == __lits[__cache_type::_S_ee])
729 || (__c == __lits[__cache_type::_S_Ee]))
731 __xtrc[__pos++] = __c;
735 // Now there may be a sign
738 if ((__c == __lits[__cache_type::_S_minus])
739 || (__c == __lits[__cache_type::_S_plus]))
741 __xtrc[__pos++] = __c;
744 // whitespace may follow a sign
745 while ((__beg != __end) && (isspace(__c)))
752 // And now there must be some digits
755 __xtrc[__pos] = '\0';
756 __err |= (ios_base::eofbit | ios_base::failbit);
759 while (__beg != __end)
761 const char* __p = strchr(__lits, __c);
762 if ((__p >= &__lits[__cache_type::_S_digits]
763 && __p < &__lits[__cache_type::_S_digits + __base])
764 || (__p >= &__lits[__cache_type::_S_udigits]
765 && __p < &__lits[__cache_type::_S_udigits + __base]))
767 __xtrc[__pos++] = __c;
776 // Finally, that's it for floating point
780 __xtrc[__pos] = '\0';
782 __err |= ios_base::eofbit;
785 // The following code uses sprintf() to convert floating point
786 // values for insertion into a stream. The current implementation
787 // replicates the code in _S_pad_numeric() (in _S_output_float()) in
788 // order to prevent having to create a "wide" buffer in addition to
789 // the "narrow" buffer passed to sprintf(). An optimization would be
790 // to replace sprintf() with code that works directly on a wide
791 // buffer and then use _S_pad_numeric() to do the padding. It would
792 // be good to replace sprintf() anyway to avoid accidental buffer
793 // overruns and to gain back the efficiency that C++ provides by
794 // knowing up front the type of the values to insert. This
795 // implementation follows the C++ standard fairly directly as
796 // outlined in 22.2.2.2 [lib.locale.num.put]
798 __build_float_format(ios_base& __io, char* __fptr, char __modifier,
801 bool __incl_prec = false;
802 ios_base::fmtflags __flags = __io.flags();
804 // [22.2.2.2.2] Table 60
805 if (__flags & ios_base::showpos)
807 if (__flags & ios_base::showpoint)
809 // As per [22.2.2.2.2.11]
810 if (__flags & ios_base::fixed || __prec > 0)
817 *__fptr++ = __modifier;
818 ios_base::fmtflags __fltfield = __flags & ios_base::floatfield;
819 // [22.2.2.2.2] Table 58
820 if (__fltfield == ios_base::fixed)
822 else if (__fltfield == ios_base::scientific)
823 *__fptr++ = (__flags & ios_base::uppercase) ? 'E' : 'e';
825 *__fptr++ = (__flags & ios_base::uppercase) ? 'G' : 'g';
831 moneypunct_byname<char, false>::moneypunct_byname(const char* /*__s*/,
833 : moneypunct<char, false>(__refs) { }
836 moneypunct_byname<char, true>::moneypunct_byname(const char* /*__s*/,
838 : moneypunct<char, true>(__refs) { }
840 #ifdef _GLIBCPP_USE_WCHAR_T
841 ctype<wchar_t>::__wmask_type
842 ctype<wchar_t>::_M_convert_to_wmask(const mask __m) const
848 __ret = wctype("space");
851 __ret = wctype("print");
854 __ret = wctype("cntrl");
857 __ret = wctype("upper");
860 __ret = wctype("lower");
863 __ret = wctype("alpha");
866 __ret = wctype("digit");
869 __ret = wctype("punct");
872 __ret = wctype("xdigit");
875 __ret = wctype("alnum");
878 __ret = wctype("graph");
886 ctype<wchar_t>::~ctype() { }
888 // NB: These ctype<wchar_t> methods are not configuration-specific,
889 // unlike the ctype<char> bits.
890 ctype<wchar_t>::ctype(size_t __refs) : __ctype_abstract_base<wchar_t>(__refs)
894 ctype<wchar_t>::do_toupper(wchar_t __c) const
895 { return towupper(__c); }
898 ctype<wchar_t>::do_toupper(wchar_t* __lo, const wchar_t* __hi) const
902 *__lo = towupper(*__lo);
909 ctype<wchar_t>::do_tolower(wchar_t __c) const
910 { return towlower(__c); }
913 ctype<wchar_t>::do_tolower(wchar_t* __lo, const wchar_t* __hi) const
917 *__lo = towlower(*__lo);
925 do_is(mask __m, char_type __c) const
926 { return static_cast<bool>(iswctype(__c, _M_convert_to_wmask(__m))); }
930 do_is(const wchar_t* __lo, const wchar_t* __hi, mask* __m) const
932 while (__lo < __hi && !this->is(*__m, *__lo))
939 do_scan_is(mask __m, const wchar_t* __lo, const wchar_t* __hi) const
941 while (__lo < __hi && !this->is(__m, *__lo))
948 do_scan_not(mask __m, const char_type* __lo, const char_type* __hi) const
950 while (__lo < __hi && this->is(__m, *__lo) != 0)
957 do_widen(char __c) const
958 { return btowc(__c); }
962 do_widen(const char* __lo, const char* __hi, wchar_t* __dest) const
965 memset(static_cast<void*>(&__state), 0, sizeof(mbstate_t));
966 mbsrtowcs(__dest, &__lo, __hi - __lo, &__state);
972 do_narrow(wchar_t __wc, char __dfault) const
974 int __c = wctob(__wc);
975 return (__c == EOF ? __dfault : static_cast<char>(__c));
980 do_narrow(const wchar_t* __lo, const wchar_t* __hi, char __dfault,
984 memset(static_cast<void*>(&__state), 0, sizeof(mbstate_t));
985 size_t __len = __hi - __lo;
986 size_t __conv = wcsrtombs(__dest, &__lo, __len, &__state);
993 ctype_byname<wchar_t>::
994 ctype_byname(const char* /*__s*/, size_t __refs)
995 : ctype<wchar_t>(__refs) { }
996 #endif // _GLIBCPP_USE_WCHAR_T