mirror of
https://github.com/autc04/Retro68.git
synced 2024-11-24 23:32:06 +00:00
937 lines
25 KiB
C++
937 lines
25 KiB
C++
// std::moneypunct implementation details, GNU version -*- C++ -*-
|
|
|
|
// Copyright (C) 2001-2014 Free Software Foundation, Inc.
|
|
//
|
|
// This file is part of the GNU ISO C++ Library. This library is free
|
|
// software; you can redistribute it and/or modify it under the
|
|
// terms of the GNU General Public License as published by the
|
|
// Free Software Foundation; either version 3, or (at your option)
|
|
// any later version.
|
|
|
|
// This library is distributed in the hope that it will be useful,
|
|
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
// GNU General Public License for more details.
|
|
|
|
// Under Section 7 of GPL version 3, you are granted additional
|
|
// permissions described in the GCC Runtime Library Exception, version
|
|
// 3.1, as published by the Free Software Foundation.
|
|
|
|
// You should have received a copy of the GNU General Public License and
|
|
// a copy of the GCC Runtime Library Exception along with this program;
|
|
// see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
|
|
// <http://www.gnu.org/licenses/>.
|
|
|
|
//
|
|
// ISO C++ 14882: 22.2.6.3.2 moneypunct virtual functions
|
|
//
|
|
|
|
// Written by Benjamin Kosnik <bkoz@redhat.com>
|
|
|
|
#include <locale>
|
|
#include <bits/c++locale_internal.h>
|
|
|
|
namespace std _GLIBCXX_VISIBILITY(default)
|
|
{
|
|
_GLIBCXX_BEGIN_NAMESPACE_VERSION
|
|
|
|
// Construct and return valid pattern consisting of some combination of:
|
|
// space none symbol sign value
|
|
money_base::pattern
|
|
money_base::_S_construct_pattern(char __precedes, char __space,
|
|
char __posn) throw()
|
|
{
|
|
pattern __ret;
|
|
|
|
// This insanely complicated routine attempts to construct a valid
|
|
// pattern for use with monyepunct. A couple of invariants:
|
|
|
|
// if (__precedes) symbol -> value
|
|
// else value -> symbol
|
|
|
|
// if (__space) space
|
|
// else none
|
|
|
|
// none == never first
|
|
// space never first or last
|
|
|
|
// Any elegant implementations of this are welcome.
|
|
switch (__posn)
|
|
{
|
|
case 0:
|
|
case 1:
|
|
// 1 The sign precedes the value and symbol.
|
|
__ret.field[0] = sign;
|
|
if (__space)
|
|
{
|
|
// Pattern starts with sign.
|
|
if (__precedes)
|
|
{
|
|
__ret.field[1] = symbol;
|
|
__ret.field[3] = value;
|
|
}
|
|
else
|
|
{
|
|
__ret.field[1] = value;
|
|
__ret.field[3] = symbol;
|
|
}
|
|
__ret.field[2] = space;
|
|
}
|
|
else
|
|
{
|
|
// Pattern starts with sign and ends with none.
|
|
if (__precedes)
|
|
{
|
|
__ret.field[1] = symbol;
|
|
__ret.field[2] = value;
|
|
}
|
|
else
|
|
{
|
|
__ret.field[1] = value;
|
|
__ret.field[2] = symbol;
|
|
}
|
|
__ret.field[3] = none;
|
|
}
|
|
break;
|
|
case 2:
|
|
// 2 The sign follows the value and symbol.
|
|
if (__space)
|
|
{
|
|
// Pattern either ends with sign.
|
|
if (__precedes)
|
|
{
|
|
__ret.field[0] = symbol;
|
|
__ret.field[2] = value;
|
|
}
|
|
else
|
|
{
|
|
__ret.field[0] = value;
|
|
__ret.field[2] = symbol;
|
|
}
|
|
__ret.field[1] = space;
|
|
__ret.field[3] = sign;
|
|
}
|
|
else
|
|
{
|
|
// Pattern ends with sign then none.
|
|
if (__precedes)
|
|
{
|
|
__ret.field[0] = symbol;
|
|
__ret.field[1] = value;
|
|
}
|
|
else
|
|
{
|
|
__ret.field[0] = value;
|
|
__ret.field[1] = symbol;
|
|
}
|
|
__ret.field[2] = sign;
|
|
__ret.field[3] = none;
|
|
}
|
|
break;
|
|
case 3:
|
|
// 3 The sign immediately precedes the symbol.
|
|
if (__precedes)
|
|
{
|
|
__ret.field[0] = sign;
|
|
__ret.field[1] = symbol;
|
|
if (__space)
|
|
{
|
|
__ret.field[2] = space;
|
|
__ret.field[3] = value;
|
|
}
|
|
else
|
|
{
|
|
__ret.field[2] = value;
|
|
__ret.field[3] = none;
|
|
}
|
|
}
|
|
else
|
|
{
|
|
__ret.field[0] = value;
|
|
if (__space)
|
|
{
|
|
__ret.field[1] = space;
|
|
__ret.field[2] = sign;
|
|
__ret.field[3] = symbol;
|
|
}
|
|
else
|
|
{
|
|
__ret.field[1] = sign;
|
|
__ret.field[2] = symbol;
|
|
__ret.field[3] = none;
|
|
}
|
|
}
|
|
break;
|
|
case 4:
|
|
// 4 The sign immediately follows the symbol.
|
|
if (__precedes)
|
|
{
|
|
__ret.field[0] = symbol;
|
|
__ret.field[1] = sign;
|
|
if (__space)
|
|
{
|
|
__ret.field[2] = space;
|
|
__ret.field[3] = value;
|
|
}
|
|
else
|
|
{
|
|
__ret.field[2] = value;
|
|
__ret.field[3] = none;
|
|
}
|
|
}
|
|
else
|
|
{
|
|
__ret.field[0] = value;
|
|
if (__space)
|
|
{
|
|
__ret.field[1] = space;
|
|
__ret.field[2] = symbol;
|
|
__ret.field[3] = sign;
|
|
}
|
|
else
|
|
{
|
|
__ret.field[1] = symbol;
|
|
__ret.field[2] = sign;
|
|
__ret.field[3] = none;
|
|
}
|
|
}
|
|
break;
|
|
default:
|
|
__ret = pattern();
|
|
}
|
|
return __ret;
|
|
}
|
|
|
|
template<>
|
|
void
|
|
moneypunct<char, true>::_M_initialize_moneypunct(__c_locale __cloc,
|
|
const char*)
|
|
{
|
|
if (!_M_data)
|
|
_M_data = new __moneypunct_cache<char, true>;
|
|
|
|
if (!__cloc)
|
|
{
|
|
// "C" locale
|
|
_M_data->_M_decimal_point = '.';
|
|
_M_data->_M_thousands_sep = ',';
|
|
_M_data->_M_grouping = "";
|
|
_M_data->_M_grouping_size = 0;
|
|
_M_data->_M_use_grouping = false;
|
|
_M_data->_M_curr_symbol = "";
|
|
_M_data->_M_curr_symbol_size = 0;
|
|
_M_data->_M_positive_sign = "";
|
|
_M_data->_M_positive_sign_size = 0;
|
|
_M_data->_M_negative_sign = "";
|
|
_M_data->_M_negative_sign_size = 0;
|
|
_M_data->_M_frac_digits = 0;
|
|
_M_data->_M_pos_format = money_base::_S_default_pattern;
|
|
_M_data->_M_neg_format = money_base::_S_default_pattern;
|
|
|
|
for (size_t __i = 0; __i < money_base::_S_end; ++__i)
|
|
_M_data->_M_atoms[__i] = money_base::_S_atoms[__i];
|
|
}
|
|
else
|
|
{
|
|
// Named locale.
|
|
_M_data->_M_decimal_point = *(__nl_langinfo_l(__MON_DECIMAL_POINT,
|
|
__cloc));
|
|
_M_data->_M_thousands_sep = *(__nl_langinfo_l(__MON_THOUSANDS_SEP,
|
|
__cloc));
|
|
|
|
// Check for NULL, which implies no fractional digits.
|
|
if (_M_data->_M_decimal_point == '\0')
|
|
{
|
|
// Like in "C" locale.
|
|
_M_data->_M_frac_digits = 0;
|
|
_M_data->_M_decimal_point = '.';
|
|
}
|
|
else
|
|
_M_data->_M_frac_digits = *(__nl_langinfo_l(__INT_FRAC_DIGITS,
|
|
__cloc));
|
|
|
|
const char* __cgroup = __nl_langinfo_l(__MON_GROUPING, __cloc);
|
|
const char* __cpossign = __nl_langinfo_l(__POSITIVE_SIGN, __cloc);
|
|
const char* __cnegsign = __nl_langinfo_l(__NEGATIVE_SIGN, __cloc);
|
|
// _Intl == true
|
|
const char* __ccurr = __nl_langinfo_l(__INT_CURR_SYMBOL, __cloc);
|
|
|
|
char* __group = 0;
|
|
char* __ps = 0;
|
|
char* __ns = 0;
|
|
const char __nposn = *(__nl_langinfo_l(__INT_N_SIGN_POSN, __cloc));
|
|
__try
|
|
{
|
|
size_t __len;
|
|
|
|
// Check for NULL, which implies no grouping.
|
|
if (_M_data->_M_thousands_sep == '\0')
|
|
{
|
|
// Like in "C" locale.
|
|
_M_data->_M_grouping = "";
|
|
_M_data->_M_grouping_size = 0;
|
|
_M_data->_M_use_grouping = false;
|
|
_M_data->_M_thousands_sep = ',';
|
|
}
|
|
else
|
|
{
|
|
__len = strlen(__cgroup);
|
|
if (__len)
|
|
{
|
|
__group = new char[__len + 1];
|
|
memcpy(__group, __cgroup, __len + 1);
|
|
_M_data->_M_grouping = __group;
|
|
}
|
|
else
|
|
{
|
|
_M_data->_M_grouping = "";
|
|
_M_data->_M_use_grouping = false;
|
|
}
|
|
_M_data->_M_grouping_size = __len;
|
|
}
|
|
|
|
__len = strlen(__cpossign);
|
|
if (__len)
|
|
{
|
|
__ps = new char[__len + 1];
|
|
memcpy(__ps, __cpossign, __len + 1);
|
|
_M_data->_M_positive_sign = __ps;
|
|
}
|
|
else
|
|
_M_data->_M_positive_sign = "";
|
|
_M_data->_M_positive_sign_size = __len;
|
|
|
|
if (!__nposn)
|
|
{
|
|
_M_data->_M_negative_sign = "()";
|
|
_M_data->_M_negative_sign_size = 2;
|
|
}
|
|
else
|
|
{
|
|
__len = strlen(__cnegsign);
|
|
if (__len)
|
|
{
|
|
__ns = new char[__len + 1];
|
|
memcpy(__ns, __cnegsign, __len + 1);
|
|
_M_data->_M_negative_sign = __ns;
|
|
}
|
|
else
|
|
_M_data->_M_negative_sign = "";
|
|
_M_data->_M_negative_sign_size = __len;
|
|
}
|
|
|
|
__len = strlen(__ccurr);
|
|
if (__len)
|
|
{
|
|
char* __curr = new char[__len + 1];
|
|
memcpy(__curr, __ccurr, __len + 1);
|
|
_M_data->_M_curr_symbol = __curr;
|
|
}
|
|
else
|
|
_M_data->_M_curr_symbol = "";
|
|
_M_data->_M_curr_symbol_size = __len;
|
|
}
|
|
__catch(...)
|
|
{
|
|
delete _M_data;
|
|
_M_data = 0;
|
|
delete [] __group;
|
|
delete [] __ps;
|
|
delete [] __ns;
|
|
__throw_exception_again;
|
|
}
|
|
|
|
char __pprecedes = *(__nl_langinfo_l(__INT_P_CS_PRECEDES, __cloc));
|
|
char __pspace = *(__nl_langinfo_l(__INT_P_SEP_BY_SPACE, __cloc));
|
|
char __pposn = *(__nl_langinfo_l(__INT_P_SIGN_POSN, __cloc));
|
|
_M_data->_M_pos_format = _S_construct_pattern(__pprecedes, __pspace,
|
|
__pposn);
|
|
char __nprecedes = *(__nl_langinfo_l(__INT_N_CS_PRECEDES, __cloc));
|
|
char __nspace = *(__nl_langinfo_l(__INT_N_SEP_BY_SPACE, __cloc));
|
|
_M_data->_M_neg_format = _S_construct_pattern(__nprecedes, __nspace,
|
|
__nposn);
|
|
}
|
|
}
|
|
|
|
template<>
|
|
void
|
|
moneypunct<char, false>::_M_initialize_moneypunct(__c_locale __cloc,
|
|
const char*)
|
|
{
|
|
if (!_M_data)
|
|
_M_data = new __moneypunct_cache<char, false>;
|
|
|
|
if (!__cloc)
|
|
{
|
|
// "C" locale
|
|
_M_data->_M_decimal_point = '.';
|
|
_M_data->_M_thousands_sep = ',';
|
|
_M_data->_M_grouping = "";
|
|
_M_data->_M_grouping_size = 0;
|
|
_M_data->_M_use_grouping = false;
|
|
_M_data->_M_curr_symbol = "";
|
|
_M_data->_M_curr_symbol_size = 0;
|
|
_M_data->_M_positive_sign = "";
|
|
_M_data->_M_positive_sign_size = 0;
|
|
_M_data->_M_negative_sign = "";
|
|
_M_data->_M_negative_sign_size = 0;
|
|
_M_data->_M_frac_digits = 0;
|
|
_M_data->_M_pos_format = money_base::_S_default_pattern;
|
|
_M_data->_M_neg_format = money_base::_S_default_pattern;
|
|
|
|
for (size_t __i = 0; __i < money_base::_S_end; ++__i)
|
|
_M_data->_M_atoms[__i] = money_base::_S_atoms[__i];
|
|
}
|
|
else
|
|
{
|
|
// Named locale.
|
|
_M_data->_M_decimal_point = *(__nl_langinfo_l(__MON_DECIMAL_POINT,
|
|
__cloc));
|
|
_M_data->_M_thousands_sep = *(__nl_langinfo_l(__MON_THOUSANDS_SEP,
|
|
__cloc));
|
|
|
|
// Check for NULL, which implies no fractional digits.
|
|
if (_M_data->_M_decimal_point == '\0')
|
|
{
|
|
// Like in "C" locale.
|
|
_M_data->_M_frac_digits = 0;
|
|
_M_data->_M_decimal_point = '.';
|
|
}
|
|
else
|
|
_M_data->_M_frac_digits = *(__nl_langinfo_l(__FRAC_DIGITS,
|
|
__cloc));
|
|
|
|
const char* __cgroup = __nl_langinfo_l(__MON_GROUPING, __cloc);
|
|
const char* __cpossign = __nl_langinfo_l(__POSITIVE_SIGN, __cloc);
|
|
const char* __cnegsign = __nl_langinfo_l(__NEGATIVE_SIGN, __cloc);
|
|
// _Intl == false
|
|
const char* __ccurr = __nl_langinfo_l(__CURRENCY_SYMBOL, __cloc);
|
|
|
|
char* __group = 0;
|
|
char* __ps = 0;
|
|
char* __ns = 0;
|
|
const char __nposn = *(__nl_langinfo_l(__N_SIGN_POSN, __cloc));
|
|
__try
|
|
{
|
|
size_t __len;
|
|
|
|
// Check for NULL, which implies no grouping.
|
|
if (_M_data->_M_thousands_sep == '\0')
|
|
{
|
|
// Like in "C" locale.
|
|
_M_data->_M_grouping = "";
|
|
_M_data->_M_grouping_size = 0;
|
|
_M_data->_M_use_grouping = false;
|
|
_M_data->_M_thousands_sep = ',';
|
|
}
|
|
else
|
|
{
|
|
__len = strlen(__cgroup);
|
|
if (__len)
|
|
{
|
|
__group = new char[__len + 1];
|
|
memcpy(__group, __cgroup, __len + 1);
|
|
_M_data->_M_grouping = __group;
|
|
}
|
|
else
|
|
{
|
|
_M_data->_M_grouping = "";
|
|
_M_data->_M_use_grouping = false;
|
|
}
|
|
_M_data->_M_grouping_size = __len;
|
|
}
|
|
|
|
__len = strlen(__cpossign);
|
|
if (__len)
|
|
{
|
|
__ps = new char[__len + 1];
|
|
memcpy(__ps, __cpossign, __len + 1);
|
|
_M_data->_M_positive_sign = __ps;
|
|
}
|
|
else
|
|
_M_data->_M_positive_sign = "";
|
|
_M_data->_M_positive_sign_size = __len;
|
|
|
|
if (!__nposn)
|
|
{
|
|
_M_data->_M_negative_sign = "()";
|
|
_M_data->_M_negative_sign_size = 2;
|
|
}
|
|
else
|
|
{
|
|
__len = strlen(__cnegsign);
|
|
if (__len)
|
|
{
|
|
__ns = new char[__len + 1];
|
|
memcpy(__ns, __cnegsign, __len + 1);
|
|
_M_data->_M_negative_sign = __ns;
|
|
}
|
|
else
|
|
_M_data->_M_negative_sign = "";
|
|
_M_data->_M_negative_sign_size = __len;
|
|
}
|
|
|
|
__len = strlen(__ccurr);
|
|
if (__len)
|
|
{
|
|
char* __curr = new char[__len + 1];
|
|
memcpy(__curr, __ccurr, __len + 1);
|
|
_M_data->_M_curr_symbol = __curr;
|
|
}
|
|
else
|
|
_M_data->_M_curr_symbol = "";
|
|
_M_data->_M_curr_symbol_size = __len;
|
|
}
|
|
__catch(...)
|
|
{
|
|
delete _M_data;
|
|
_M_data = 0;
|
|
delete [] __group;
|
|
delete [] __ps;
|
|
delete [] __ns;
|
|
__throw_exception_again;
|
|
}
|
|
|
|
char __pprecedes = *(__nl_langinfo_l(__P_CS_PRECEDES, __cloc));
|
|
char __pspace = *(__nl_langinfo_l(__P_SEP_BY_SPACE, __cloc));
|
|
char __pposn = *(__nl_langinfo_l(__P_SIGN_POSN, __cloc));
|
|
_M_data->_M_pos_format = _S_construct_pattern(__pprecedes, __pspace,
|
|
__pposn);
|
|
char __nprecedes = *(__nl_langinfo_l(__N_CS_PRECEDES, __cloc));
|
|
char __nspace = *(__nl_langinfo_l(__N_SEP_BY_SPACE, __cloc));
|
|
_M_data->_M_neg_format = _S_construct_pattern(__nprecedes, __nspace,
|
|
__nposn);
|
|
}
|
|
}
|
|
|
|
template<>
|
|
moneypunct<char, true>::~moneypunct()
|
|
{
|
|
if (_M_data->_M_grouping_size)
|
|
delete [] _M_data->_M_grouping;
|
|
if (_M_data->_M_positive_sign_size)
|
|
delete [] _M_data->_M_positive_sign;
|
|
if (_M_data->_M_negative_sign_size
|
|
&& strcmp(_M_data->_M_negative_sign, "()") != 0)
|
|
delete [] _M_data->_M_negative_sign;
|
|
if (_M_data->_M_curr_symbol_size)
|
|
delete [] _M_data->_M_curr_symbol;
|
|
delete _M_data;
|
|
}
|
|
|
|
template<>
|
|
moneypunct<char, false>::~moneypunct()
|
|
{
|
|
if (_M_data->_M_grouping_size)
|
|
delete [] _M_data->_M_grouping;
|
|
if (_M_data->_M_positive_sign_size)
|
|
delete [] _M_data->_M_positive_sign;
|
|
if (_M_data->_M_negative_sign_size
|
|
&& strcmp(_M_data->_M_negative_sign, "()") != 0)
|
|
delete [] _M_data->_M_negative_sign;
|
|
if (_M_data->_M_curr_symbol_size)
|
|
delete [] _M_data->_M_curr_symbol;
|
|
delete _M_data;
|
|
}
|
|
|
|
#ifdef _GLIBCXX_USE_WCHAR_T
|
|
template<>
|
|
void
|
|
moneypunct<wchar_t, true>::_M_initialize_moneypunct(__c_locale __cloc,
|
|
#if __GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ > 2)
|
|
const char*)
|
|
#else
|
|
const char* __name)
|
|
#endif
|
|
{
|
|
if (!_M_data)
|
|
_M_data = new __moneypunct_cache<wchar_t, true>;
|
|
|
|
if (!__cloc)
|
|
{
|
|
// "C" locale
|
|
_M_data->_M_decimal_point = L'.';
|
|
_M_data->_M_thousands_sep = L',';
|
|
_M_data->_M_grouping = "";
|
|
_M_data->_M_grouping_size = 0;
|
|
_M_data->_M_use_grouping = false;
|
|
_M_data->_M_curr_symbol = L"";
|
|
_M_data->_M_curr_symbol_size = 0;
|
|
_M_data->_M_positive_sign = L"";
|
|
_M_data->_M_positive_sign_size = 0;
|
|
_M_data->_M_negative_sign = L"";
|
|
_M_data->_M_negative_sign_size = 0;
|
|
_M_data->_M_frac_digits = 0;
|
|
_M_data->_M_pos_format = money_base::_S_default_pattern;
|
|
_M_data->_M_neg_format = money_base::_S_default_pattern;
|
|
|
|
// Use ctype::widen code without the facet...
|
|
for (size_t __i = 0; __i < money_base::_S_end; ++__i)
|
|
_M_data->_M_atoms[__i] =
|
|
static_cast<wchar_t>(money_base::_S_atoms[__i]);
|
|
}
|
|
else
|
|
{
|
|
// Named locale.
|
|
#if __GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ > 2)
|
|
__c_locale __old = __uselocale(__cloc);
|
|
#else
|
|
// Switch to named locale so that mbsrtowcs will work.
|
|
char* __old = setlocale(LC_ALL, 0);
|
|
const size_t __llen = strlen(__old) + 1;
|
|
char* __sav = new char[__llen];
|
|
memcpy(__sav, __old, __llen);
|
|
setlocale(LC_ALL, __name);
|
|
#endif
|
|
|
|
union { char *__s; wchar_t __w; } __u;
|
|
__u.__s = __nl_langinfo_l(_NL_MONETARY_DECIMAL_POINT_WC, __cloc);
|
|
_M_data->_M_decimal_point = __u.__w;
|
|
|
|
__u.__s = __nl_langinfo_l(_NL_MONETARY_THOUSANDS_SEP_WC, __cloc);
|
|
_M_data->_M_thousands_sep = __u.__w;
|
|
|
|
// Check for NULL, which implies no fractional digits.
|
|
if (_M_data->_M_decimal_point == L'\0')
|
|
{
|
|
// Like in "C" locale.
|
|
_M_data->_M_frac_digits = 0;
|
|
_M_data->_M_decimal_point = L'.';
|
|
}
|
|
else
|
|
_M_data->_M_frac_digits = *(__nl_langinfo_l(__INT_FRAC_DIGITS,
|
|
__cloc));
|
|
|
|
const char* __cgroup = __nl_langinfo_l(__MON_GROUPING, __cloc);
|
|
const char* __cpossign = __nl_langinfo_l(__POSITIVE_SIGN, __cloc);
|
|
const char* __cnegsign = __nl_langinfo_l(__NEGATIVE_SIGN, __cloc);
|
|
const char* __ccurr = __nl_langinfo_l(__INT_CURR_SYMBOL, __cloc);
|
|
|
|
char* __group = 0;
|
|
wchar_t* __wcs_ps = 0;
|
|
wchar_t* __wcs_ns = 0;
|
|
const char __nposn = *(__nl_langinfo_l(__INT_N_SIGN_POSN, __cloc));
|
|
__try
|
|
{
|
|
size_t __len;
|
|
|
|
// Check for NULL, which implies no grouping.
|
|
if (_M_data->_M_thousands_sep == L'\0')
|
|
{
|
|
// Like in "C" locale.
|
|
_M_data->_M_grouping = "";
|
|
_M_data->_M_grouping_size = 0;
|
|
_M_data->_M_use_grouping = false;
|
|
_M_data->_M_thousands_sep = L',';
|
|
}
|
|
else
|
|
{
|
|
__len = strlen(__cgroup);
|
|
if (__len)
|
|
{
|
|
__group = new char[__len + 1];
|
|
memcpy(__group, __cgroup, __len + 1);
|
|
_M_data->_M_grouping = __group;
|
|
}
|
|
else
|
|
{
|
|
_M_data->_M_grouping = "";
|
|
_M_data->_M_use_grouping = false;
|
|
}
|
|
_M_data->_M_grouping_size = __len;
|
|
}
|
|
|
|
mbstate_t __state;
|
|
__len = strlen(__cpossign);
|
|
if (__len)
|
|
{
|
|
memset(&__state, 0, sizeof(mbstate_t));
|
|
__wcs_ps = new wchar_t[__len + 1];
|
|
mbsrtowcs(__wcs_ps, &__cpossign, __len + 1, &__state);
|
|
_M_data->_M_positive_sign = __wcs_ps;
|
|
}
|
|
else
|
|
_M_data->_M_positive_sign = L"";
|
|
_M_data->_M_positive_sign_size =
|
|
wcslen(_M_data->_M_positive_sign);
|
|
|
|
__len = strlen(__cnegsign);
|
|
if (!__nposn)
|
|
_M_data->_M_negative_sign = L"()";
|
|
else if (__len)
|
|
{
|
|
memset(&__state, 0, sizeof(mbstate_t));
|
|
__wcs_ns = new wchar_t[__len + 1];
|
|
mbsrtowcs(__wcs_ns, &__cnegsign, __len + 1, &__state);
|
|
_M_data->_M_negative_sign = __wcs_ns;
|
|
}
|
|
else
|
|
_M_data->_M_negative_sign = L"";
|
|
_M_data->_M_negative_sign_size =
|
|
wcslen(_M_data->_M_negative_sign);
|
|
|
|
// _Intl == true.
|
|
__len = strlen(__ccurr);
|
|
if (__len)
|
|
{
|
|
memset(&__state, 0, sizeof(mbstate_t));
|
|
wchar_t* __wcs = new wchar_t[__len + 1];
|
|
mbsrtowcs(__wcs, &__ccurr, __len + 1, &__state);
|
|
_M_data->_M_curr_symbol = __wcs;
|
|
}
|
|
else
|
|
_M_data->_M_curr_symbol = L"";
|
|
_M_data->_M_curr_symbol_size = wcslen(_M_data->_M_curr_symbol);
|
|
}
|
|
__catch(...)
|
|
{
|
|
delete _M_data;
|
|
_M_data = 0;
|
|
delete [] __group;
|
|
delete [] __wcs_ps;
|
|
delete [] __wcs_ns;
|
|
#if __GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ > 2)
|
|
__uselocale(__old);
|
|
#else
|
|
setlocale(LC_ALL, __sav);
|
|
delete [] __sav;
|
|
#endif
|
|
__throw_exception_again;
|
|
}
|
|
|
|
char __pprecedes = *(__nl_langinfo_l(__INT_P_CS_PRECEDES, __cloc));
|
|
char __pspace = *(__nl_langinfo_l(__INT_P_SEP_BY_SPACE, __cloc));
|
|
char __pposn = *(__nl_langinfo_l(__INT_P_SIGN_POSN, __cloc));
|
|
_M_data->_M_pos_format = _S_construct_pattern(__pprecedes, __pspace,
|
|
__pposn);
|
|
char __nprecedes = *(__nl_langinfo_l(__INT_N_CS_PRECEDES, __cloc));
|
|
char __nspace = *(__nl_langinfo_l(__INT_N_SEP_BY_SPACE, __cloc));
|
|
_M_data->_M_neg_format = _S_construct_pattern(__nprecedes, __nspace,
|
|
__nposn);
|
|
|
|
#if __GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ > 2)
|
|
__uselocale(__old);
|
|
#else
|
|
setlocale(LC_ALL, __sav);
|
|
delete [] __sav;
|
|
#endif
|
|
}
|
|
}
|
|
|
|
template<>
|
|
void
|
|
moneypunct<wchar_t, false>::_M_initialize_moneypunct(__c_locale __cloc,
|
|
#if __GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ > 2)
|
|
const char*)
|
|
#else
|
|
const char* __name)
|
|
#endif
|
|
{
|
|
if (!_M_data)
|
|
_M_data = new __moneypunct_cache<wchar_t, false>;
|
|
|
|
if (!__cloc)
|
|
{
|
|
// "C" locale
|
|
_M_data->_M_decimal_point = L'.';
|
|
_M_data->_M_thousands_sep = L',';
|
|
_M_data->_M_grouping = "";
|
|
_M_data->_M_grouping_size = 0;
|
|
_M_data->_M_use_grouping = false;
|
|
_M_data->_M_curr_symbol = L"";
|
|
_M_data->_M_curr_symbol_size = 0;
|
|
_M_data->_M_positive_sign = L"";
|
|
_M_data->_M_positive_sign_size = 0;
|
|
_M_data->_M_negative_sign = L"";
|
|
_M_data->_M_negative_sign_size = 0;
|
|
_M_data->_M_frac_digits = 0;
|
|
_M_data->_M_pos_format = money_base::_S_default_pattern;
|
|
_M_data->_M_neg_format = money_base::_S_default_pattern;
|
|
|
|
// Use ctype::widen code without the facet...
|
|
for (size_t __i = 0; __i < money_base::_S_end; ++__i)
|
|
_M_data->_M_atoms[__i] =
|
|
static_cast<wchar_t>(money_base::_S_atoms[__i]);
|
|
}
|
|
else
|
|
{
|
|
// Named locale.
|
|
#if __GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ > 2)
|
|
__c_locale __old = __uselocale(__cloc);
|
|
#else
|
|
// Switch to named locale so that mbsrtowcs will work.
|
|
char* __old = setlocale(LC_ALL, 0);
|
|
const size_t __llen = strlen(__old) + 1;
|
|
char* __sav = new char[__llen];
|
|
memcpy(__sav, __old, __llen);
|
|
setlocale(LC_ALL, __name);
|
|
#endif
|
|
|
|
union { char *__s; wchar_t __w; } __u;
|
|
__u.__s = __nl_langinfo_l(_NL_MONETARY_DECIMAL_POINT_WC, __cloc);
|
|
_M_data->_M_decimal_point = __u.__w;
|
|
|
|
__u.__s = __nl_langinfo_l(_NL_MONETARY_THOUSANDS_SEP_WC, __cloc);
|
|
_M_data->_M_thousands_sep = __u.__w;
|
|
|
|
// Check for NULL, which implies no fractional digits.
|
|
if (_M_data->_M_decimal_point == L'\0')
|
|
{
|
|
// Like in "C" locale.
|
|
_M_data->_M_frac_digits = 0;
|
|
_M_data->_M_decimal_point = L'.';
|
|
}
|
|
else
|
|
_M_data->_M_frac_digits = *(__nl_langinfo_l(__FRAC_DIGITS,
|
|
__cloc));
|
|
|
|
const char* __cgroup = __nl_langinfo_l(__MON_GROUPING, __cloc);
|
|
const char* __cpossign = __nl_langinfo_l(__POSITIVE_SIGN, __cloc);
|
|
const char* __cnegsign = __nl_langinfo_l(__NEGATIVE_SIGN, __cloc);
|
|
const char* __ccurr = __nl_langinfo_l(__CURRENCY_SYMBOL, __cloc);
|
|
|
|
char* __group = 0;
|
|
wchar_t* __wcs_ps = 0;
|
|
wchar_t* __wcs_ns = 0;
|
|
const char __nposn = *(__nl_langinfo_l(__N_SIGN_POSN, __cloc));
|
|
__try
|
|
{
|
|
size_t __len;
|
|
|
|
// Check for NULL, which implies no grouping.
|
|
if (_M_data->_M_thousands_sep == L'\0')
|
|
{
|
|
// Like in "C" locale.
|
|
_M_data->_M_grouping = "";
|
|
_M_data->_M_grouping_size = 0;
|
|
_M_data->_M_use_grouping = false;
|
|
_M_data->_M_thousands_sep = L',';
|
|
}
|
|
else
|
|
{
|
|
__len = strlen(__cgroup);
|
|
if (__len)
|
|
{
|
|
__group = new char[__len + 1];
|
|
memcpy(__group, __cgroup, __len + 1);
|
|
_M_data->_M_grouping = __group;
|
|
}
|
|
else
|
|
{
|
|
_M_data->_M_grouping = "";
|
|
_M_data->_M_use_grouping = false;
|
|
}
|
|
_M_data->_M_grouping_size = __len;
|
|
}
|
|
|
|
mbstate_t __state;
|
|
__len = strlen(__cpossign);
|
|
if (__len)
|
|
{
|
|
memset(&__state, 0, sizeof(mbstate_t));
|
|
__wcs_ps = new wchar_t[__len + 1];
|
|
mbsrtowcs(__wcs_ps, &__cpossign, __len + 1, &__state);
|
|
_M_data->_M_positive_sign = __wcs_ps;
|
|
}
|
|
else
|
|
_M_data->_M_positive_sign = L"";
|
|
_M_data->_M_positive_sign_size =
|
|
wcslen(_M_data->_M_positive_sign);
|
|
|
|
__len = strlen(__cnegsign);
|
|
if (!__nposn)
|
|
_M_data->_M_negative_sign = L"()";
|
|
else if (__len)
|
|
{
|
|
memset(&__state, 0, sizeof(mbstate_t));
|
|
__wcs_ns = new wchar_t[__len + 1];
|
|
mbsrtowcs(__wcs_ns, &__cnegsign, __len + 1, &__state);
|
|
_M_data->_M_negative_sign = __wcs_ns;
|
|
}
|
|
else
|
|
_M_data->_M_negative_sign = L"";
|
|
_M_data->_M_negative_sign_size =
|
|
wcslen(_M_data->_M_negative_sign);
|
|
|
|
// _Intl == true.
|
|
__len = strlen(__ccurr);
|
|
if (__len)
|
|
{
|
|
memset(&__state, 0, sizeof(mbstate_t));
|
|
wchar_t* __wcs = new wchar_t[__len + 1];
|
|
mbsrtowcs(__wcs, &__ccurr, __len + 1, &__state);
|
|
_M_data->_M_curr_symbol = __wcs;
|
|
}
|
|
else
|
|
_M_data->_M_curr_symbol = L"";
|
|
_M_data->_M_curr_symbol_size = wcslen(_M_data->_M_curr_symbol);
|
|
}
|
|
__catch(...)
|
|
{
|
|
delete _M_data;
|
|
_M_data = 0;
|
|
delete [] __group;
|
|
delete [] __wcs_ps;
|
|
delete [] __wcs_ns;
|
|
#if __GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ > 2)
|
|
__uselocale(__old);
|
|
#else
|
|
setlocale(LC_ALL, __sav);
|
|
delete [] __sav;
|
|
#endif
|
|
__throw_exception_again;
|
|
}
|
|
|
|
char __pprecedes = *(__nl_langinfo_l(__P_CS_PRECEDES, __cloc));
|
|
char __pspace = *(__nl_langinfo_l(__P_SEP_BY_SPACE, __cloc));
|
|
char __pposn = *(__nl_langinfo_l(__P_SIGN_POSN, __cloc));
|
|
_M_data->_M_pos_format = _S_construct_pattern(__pprecedes, __pspace,
|
|
__pposn);
|
|
char __nprecedes = *(__nl_langinfo_l(__N_CS_PRECEDES, __cloc));
|
|
char __nspace = *(__nl_langinfo_l(__N_SEP_BY_SPACE, __cloc));
|
|
_M_data->_M_neg_format = _S_construct_pattern(__nprecedes, __nspace,
|
|
__nposn);
|
|
|
|
#if __GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ > 2)
|
|
__uselocale(__old);
|
|
#else
|
|
setlocale(LC_ALL, __sav);
|
|
delete [] __sav;
|
|
#endif
|
|
}
|
|
}
|
|
|
|
template<>
|
|
moneypunct<wchar_t, true>::~moneypunct()
|
|
{
|
|
if (_M_data->_M_grouping_size)
|
|
delete [] _M_data->_M_grouping;
|
|
if (_M_data->_M_positive_sign_size)
|
|
delete [] _M_data->_M_positive_sign;
|
|
if (_M_data->_M_negative_sign_size
|
|
&& wcscmp(_M_data->_M_negative_sign, L"()") != 0)
|
|
delete [] _M_data->_M_negative_sign;
|
|
if (_M_data->_M_curr_symbol_size)
|
|
delete [] _M_data->_M_curr_symbol;
|
|
delete _M_data;
|
|
}
|
|
|
|
template<>
|
|
moneypunct<wchar_t, false>::~moneypunct()
|
|
{
|
|
if (_M_data->_M_grouping_size)
|
|
delete [] _M_data->_M_grouping;
|
|
if (_M_data->_M_positive_sign_size)
|
|
delete [] _M_data->_M_positive_sign;
|
|
if (_M_data->_M_negative_sign_size
|
|
&& wcscmp(_M_data->_M_negative_sign, L"()") != 0)
|
|
delete [] _M_data->_M_negative_sign;
|
|
if (_M_data->_M_curr_symbol_size)
|
|
delete [] _M_data->_M_curr_symbol;
|
|
delete _M_data;
|
|
}
|
|
#endif
|
|
|
|
_GLIBCXX_END_NAMESPACE_VERSION
|
|
} // namespace
|