Fix moneypunct_byname algorithm to more accurately represent C locales in C++.

git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@152501 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Jeffrey Yasskin
2012-03-10 18:31:43 +00:00
parent 05b57d5cdf
commit 558ae17391
8 changed files with 412 additions and 222 deletions

View File

@@ -5275,116 +5275,200 @@ __time_put::__do_put(wchar_t* __wb, wchar_t*& __we, const tm* __tm,
// moneypunct_byname
template <class charT>
static
void
__init_pat(money_base::pattern& pat, char cs_precedes, char sep_by_space, char sign_posn)
__init_pat(money_base::pattern& pat, basic_string<charT>& __curr_symbol_,
bool intl, char cs_precedes, char sep_by_space, char sign_posn,
charT space_char)
{
const char sign = static_cast<char>(money_base::sign);
const char space = static_cast<char>(money_base::space);
const char none = static_cast<char>(money_base::none);
const char symbol = static_cast<char>(money_base::symbol);
const char value = static_cast<char>(money_base::value);
const bool symbol_contains_sep = intl && __curr_symbol_.size() == 4;
// Comments on case branches reflect 'C11 7.11.2.1 The localeconv
// function'. "Space between sign and symbol or value" means that
// if the sign is adjacent to the symbol, there's a space between
// them, and otherwise there's a space between the sign and value.
//
// C11's localeconv specifies that the fourth character of an
// international curr_symbol is used to separate the sign and
// value when sep_by_space says to do so. C++ can't represent
// that, so we just use a space. When sep_by_space says to
// separate the symbol and value-or-sign with a space, we rearrange the
// curr_symbol to put its spacing character on the correct side of
// the symbol.
//
// We also need to avoid adding an extra space between the sign
// and value when the currency symbol is suppressed (by not
// setting showbase). We match glibc's strfmon by interpreting
// sep_by_space==1 as "omit the space when the currency symbol is
// absent".
//
// Users who want to get this right should use ICU instead.
switch (cs_precedes)
{
case 0:
case 0: // value before curr_symbol
if (symbol_contains_sep) {
// Move the separator to before the symbol, to place it
// between the value and symbol.
rotate(__curr_symbol_.begin(), __curr_symbol_.begin() + 3,
__curr_symbol_.end());
}
switch (sign_posn)
{
case 0:
case 0: // Parentheses surround the quantity and currency symbol.
pat.field[0] = sign;
pat.field[1] = value;
pat.field[2] = none; // Any space appears in the symbol.
pat.field[3] = symbol;
switch (sep_by_space)
{
case 0:
pat.field[2] = none;
case 0: // No space separates the currency symbol and value.
// This case may have changed between C99 and C11;
// assume the currency symbol matches the intention.
case 2: // Space between sign and currency or value.
// The "sign" is two parentheses, so no space here either.
return;
case 1:
case 2:
pat.field[2] = space;
case 1: // Space between currency-and-sign or currency and value.
if (!symbol_contains_sep) {
// We insert the space into the symbol instead of
// setting pat.field[2]=space so that when
// showbase is not set, the space goes away too.
__curr_symbol_.insert(0, 1, space_char);
}
return;
default:
break;
}
break;
case 1:
case 1: // The sign string precedes the quantity and currency symbol.
pat.field[0] = sign;
pat.field[3] = symbol;
switch (sep_by_space)
{
case 0:
case 0: // No space separates the currency symbol and value.
pat.field[1] = value;
pat.field[2] = none;
return;
case 1:
case 1: // Space between currency-and-sign or currency and value.
pat.field[1] = value;
pat.field[2] = space;
pat.field[2] = none;
if (!symbol_contains_sep) {
// We insert the space into the symbol instead of
// setting pat.field[2]=space so that when
// showbase is not set, the space goes away too.
__curr_symbol_.insert(0, 1, space_char);
}
return;
case 2:
case 2: // Space between sign and currency or value.
pat.field[1] = space;
pat.field[2] = value;
if (symbol_contains_sep) {
// Remove the separator from the symbol, since it
// has already appeared after the sign.
__curr_symbol_.erase(__curr_symbol_.begin());
}
return;
default:
break;
}
break;
case 2:
case 2: // The sign string succeeds the quantity and currency symbol.
pat.field[0] = value;
pat.field[3] = sign;
switch (sep_by_space)
{
case 0:
case 0: // No space separates the currency symbol and value.
pat.field[1] = none;
pat.field[2] = symbol;
return;
case 1:
pat.field[1] = space;
case 1: // Space between currency-and-sign or currency and value.
if (!symbol_contains_sep) {
// We insert the space into the symbol instead of
// setting pat.field[1]=space so that when
// showbase is not set, the space goes away too.
__curr_symbol_.insert(0, 1, space_char);
}
pat.field[1] = none;
pat.field[2] = symbol;
return;
case 2:
case 2: // Space between sign and currency or value.
pat.field[1] = symbol;
pat.field[2] = space;
if (symbol_contains_sep) {
// Remove the separator from the symbol, since it
// should not be removed if showbase is absent.
__curr_symbol_.erase(__curr_symbol_.begin());
}
return;
default:
break;
}
break;
case 3:
case 3: // The sign string immediately precedes the currency symbol.
pat.field[0] = value;
pat.field[3] = symbol;
switch (sep_by_space)
{
case 0:
case 0: // No space separates the currency symbol and value.
pat.field[1] = none;
pat.field[2] = sign;
return;
case 1:
case 1: // Space between currency-and-sign or currency and value.
pat.field[1] = space;
pat.field[2] = sign;
if (symbol_contains_sep) {
// Remove the separator from the symbol, since it
// has already appeared before the sign.
__curr_symbol_.erase(__curr_symbol_.begin());
}
return;
case 2:
case 2: // Space between sign and currency or value.
pat.field[1] = sign;
pat.field[2] = space;
pat.field[2] = none;
if (!symbol_contains_sep) {
// We insert the space into the symbol instead of
// setting pat.field[2]=space so that when
// showbase is not set, the space goes away too.
__curr_symbol_.insert(0, 1, space_char);
}
return;
default:
break;
}
break;
case 4:
case 4: // The sign string immediately succeeds the currency symbol.
pat.field[0] = value;
pat.field[3] = sign;
switch (sep_by_space)
{
case 0:
case 0: // No space separates the currency symbol and value.
pat.field[1] = none;
pat.field[2] = symbol;
return;
case 1:
pat.field[1] = space;
case 1: // Space between currency-and-sign or currency and value.
pat.field[1] = none;
pat.field[2] = symbol;
if (!symbol_contains_sep) {
// We insert the space into the symbol instead of
// setting pat.field[1]=space so that when
// showbase is not set, the space goes away too.
__curr_symbol_.insert(0, 1, space_char);
}
return;
case 2:
case 2: // Space between sign and currency or value.
pat.field[1] = symbol;
pat.field[2] = space;
if (symbol_contains_sep) {
// Remove the separator from the symbol, since it
// should not disappear when showbase is absent.
__curr_symbol_.erase(__curr_symbol_.begin());
}
return;
default:
break;
@@ -5394,105 +5478,157 @@ __init_pat(money_base::pattern& pat, char cs_precedes, char sep_by_space, char s
break;
}
break;
case 1:
case 1: // curr_symbol before value
switch (sign_posn)
{
case 0:
case 0: // Parentheses surround the quantity and currency symbol.
pat.field[0] = sign;
pat.field[1] = symbol;
pat.field[2] = none; // Any space appears in the symbol.
pat.field[3] = value;
switch (sep_by_space)
{
case 0:
pat.field[2] = none;
case 0: // No space separates the currency symbol and value.
// This case may have changed between C99 and C11;
// assume the currency symbol matches the intention.
case 2: // Space between sign and currency or value.
// The "sign" is two parentheses, so no space here either.
return;
case 1:
case 2:
pat.field[2] = space;
case 1: // Space between currency-and-sign or currency and value.
if (!symbol_contains_sep) {
// We insert the space into the symbol instead of
// setting pat.field[2]=space so that when
// showbase is not set, the space goes away too.
__curr_symbol_.insert(0, 1, space_char);
}
return;
default:
break;
}
break;
case 1:
case 1: // The sign string precedes the quantity and currency symbol.
pat.field[0] = sign;
pat.field[3] = value;
switch (sep_by_space)
{
case 0:
case 0: // No space separates the currency symbol and value.
pat.field[1] = symbol;
pat.field[2] = none;
return;
case 1:
case 1: // Space between currency-and-sign or currency and value.
pat.field[1] = symbol;
pat.field[2] = space;
pat.field[2] = none;
if (!symbol_contains_sep) {
// We insert the space into the symbol instead of
// setting pat.field[2]=space so that when
// showbase is not set, the space goes away too.
__curr_symbol_.push_back(space_char);
}
return;
case 2:
case 2: // Space between sign and currency or value.
pat.field[1] = space;
pat.field[2] = symbol;
if (symbol_contains_sep) {
// Remove the separator from the symbol, since it
// has already appeared after the sign.
__curr_symbol_.pop_back();
}
return;
default:
break;
}
break;
case 2:
case 2: // The sign string succeeds the quantity and currency symbol.
pat.field[0] = symbol;
pat.field[3] = sign;
switch (sep_by_space)
{
case 0:
case 0: // No space separates the currency symbol and value.
pat.field[1] = none;
pat.field[2] = value;
return;
case 1:
pat.field[1] = space;
case 1: // Space between currency-and-sign or currency and value.
pat.field[1] = none;
pat.field[2] = value;
if (!symbol_contains_sep) {
// We insert the space into the symbol instead of
// setting pat.field[1]=space so that when
// showbase is not set, the space goes away too.
__curr_symbol_.push_back(space_char);
}
return;
case 2:
case 2: // Space between sign and currency or value.
pat.field[1] = value;
pat.field[2] = space;
if (symbol_contains_sep) {
// Remove the separator from the symbol, since it
// will appear before the sign.
__curr_symbol_.pop_back();
}
return;
default:
break;
}
break;
case 3:
case 3: // The sign string immediately precedes the currency symbol.
pat.field[0] = sign;
pat.field[3] = value;
switch (sep_by_space)
{
case 0:
case 0: // No space separates the currency symbol and value.
pat.field[1] = symbol;
pat.field[2] = none;
return;
case 1:
case 1: // Space between currency-and-sign or currency and value.
pat.field[1] = symbol;
pat.field[2] = space;
pat.field[2] = none;
if (!symbol_contains_sep) {
// We insert the space into the symbol instead of
// setting pat.field[2]=space so that when
// showbase is not set, the space goes away too.
__curr_symbol_.push_back(space_char);
}
return;
case 2:
case 2: // Space between sign and currency or value.
pat.field[1] = space;
pat.field[2] = symbol;
if (symbol_contains_sep) {
// Remove the separator from the symbol, since it
// has already appeared after the sign.
__curr_symbol_.pop_back();
}
return;
default:
break;
}
break;
case 4:
case 4: // The sign string immediately succeeds the currency symbol.
pat.field[0] = symbol;
pat.field[3] = value;
switch (sep_by_space)
{
case 0:
case 0: // No space separates the currency symbol and value.
pat.field[1] = sign;
pat.field[2] = none;
return;
case 1:
case 1: // Space between currency-and-sign or currency and value.
pat.field[1] = sign;
pat.field[2] = space;
if (symbol_contains_sep) {
// Remove the separator from the symbol, since it
// should not disappear when showbase is absent.
__curr_symbol_.pop_back();
}
return;
case 2:
pat.field[1] = space;
case 2: // Space between sign and currency or value.
pat.field[1] = none;
pat.field[2] = sign;
if (!symbol_contains_sep) {
// We insert the space into the symbol instead of
// setting pat.field[1]=space so that when
// showbase is not set, the space goes away too.
__curr_symbol_.push_back(space_char);
}
return;
default:
break;
@@ -5549,8 +5685,14 @@ moneypunct_byname<char, false>::init(const char* nm)
__negative_sign_ = "()";
else
__negative_sign_ = lc->negative_sign;
__init_pat(__pos_format_, lc->p_cs_precedes, lc->p_sep_by_space, lc->p_sign_posn);
__init_pat(__neg_format_, lc->n_cs_precedes, lc->n_sep_by_space, lc->n_sign_posn);
// Assume the positive and negative formats will want spaces in
// the same places in curr_symbol since there's no way to
// represent anything else.
string_type __dummy_curr_symbol = __curr_symbol_;
__init_pat(__pos_format_, __dummy_curr_symbol, false,
lc->p_cs_precedes, lc->p_sep_by_space, lc->p_sign_posn, ' ');
__init_pat(__neg_format_, __curr_symbol_, false,
lc->n_cs_precedes, lc->n_sep_by_space, lc->n_sign_posn, ' ');
}
template<>
@@ -5599,12 +5741,22 @@ moneypunct_byname<char, true>::init(const char* nm)
__negative_sign_ = "()";
else
__negative_sign_ = lc->negative_sign;
// Assume the positive and negative formats will want spaces in
// the same places in curr_symbol since there's no way to
// represent anything else.
string_type __dummy_curr_symbol = __curr_symbol_;
#if _WIN32
__init_pat(__pos_format_, lc->p_cs_precedes, lc->p_sep_by_space, lc->p_sign_posn);
__init_pat(__neg_format_, lc->n_cs_precedes, lc->n_sep_by_space, lc->n_sign_posn);
__init_pat(__pos_format_, __dummy_curr_symbol, true,
lc->p_cs_precedes, lc->p_sep_by_space, lc->p_sign_posn, ' ');
__init_pat(__neg_format_, __curr_symbol_, true,
lc->n_cs_precedes, lc->n_sep_by_space, lc->n_sign_posn, ' ');
#else
__init_pat(__pos_format_, lc->int_p_cs_precedes, lc->int_p_sep_by_space, lc->int_p_sign_posn);
__init_pat(__neg_format_, lc->int_n_cs_precedes, lc->int_n_sep_by_space, lc->int_n_sign_posn);
__init_pat(__pos_format_, __dummy_curr_symbol, true,
lc->int_p_cs_precedes, lc->int_p_sep_by_space,
lc->int_p_sign_posn, ' ');
__init_pat(__neg_format_, __curr_symbol_, true,
lc->int_n_cs_precedes, lc->int_n_sep_by_space,
lc->int_n_sign_posn, ' ');
#endif // _WIN32
}
@@ -5681,8 +5833,14 @@ moneypunct_byname<wchar_t, false>::init(const char* nm)
wbe = wbuf + j;
__negative_sign_.assign(wbuf, wbe);
}
__init_pat(__pos_format_, lc->p_cs_precedes, lc->p_sep_by_space, lc->p_sign_posn);
__init_pat(__neg_format_, lc->n_cs_precedes, lc->n_sep_by_space, lc->n_sign_posn);
// Assume the positive and negative formats will want spaces in
// the same places in curr_symbol since there's no way to
// represent anything else.
string_type __dummy_curr_symbol = __curr_symbol_;
__init_pat(__pos_format_, __dummy_curr_symbol, false,
lc->p_cs_precedes, lc->p_sep_by_space, lc->p_sign_posn, L' ');
__init_pat(__neg_format_, __curr_symbol_, false,
lc->n_cs_precedes, lc->n_sep_by_space, lc->n_sign_posn, L' ');
}
template<>
@@ -5766,12 +5924,22 @@ moneypunct_byname<wchar_t, true>::init(const char* nm)
wbe = wbuf + j;
__negative_sign_.assign(wbuf, wbe);
}
// Assume the positive and negative formats will want spaces in
// the same places in curr_symbol since there's no way to
// represent anything else.
string_type __dummy_curr_symbol = __curr_symbol_;
#if _WIN32
__init_pat(__pos_format_, lc->p_cs_precedes, lc->p_sep_by_space, lc->p_sign_posn);
__init_pat(__neg_format_, lc->n_cs_precedes, lc->n_sep_by_space, lc->n_sign_posn);
__init_pat(__pos_format_, __dummy_curr_symbol, true,
lc->p_cs_precedes, lc->p_sep_by_space, lc->p_sign_posn, L' ');
__init_pat(__neg_format_, __curr_symbol_, true,
lc->n_cs_precedes, lc->n_sep_by_space, lc->n_sign_posn, L' ');
#else // _WIN32
__init_pat(__pos_format_, lc->int_p_cs_precedes, lc->int_p_sep_by_space, lc->int_p_sign_posn);
__init_pat(__neg_format_, lc->int_n_cs_precedes, lc->int_n_sep_by_space, lc->int_n_sign_posn);
__init_pat(__pos_format_, __dummy_curr_symbol, true,
lc->int_p_cs_precedes, lc->int_p_sep_by_space,
lc->int_p_sign_posn, L' ');
__init_pat(__neg_format_, __curr_symbol_, true,
lc->int_n_cs_precedes, lc->int_n_sep_by_space,
lc->int_n_sign_posn, L' ');
#endif // _WIN32
}