diff options
author | arseny.kapoulkine <arseny.kapoulkine@99668b35-9821-0410-8761-19e4c4f06640> | 2010-06-06 19:06:15 +0000 |
---|---|---|
committer | arseny.kapoulkine <arseny.kapoulkine@99668b35-9821-0410-8761-19e4c4f06640> | 2010-06-06 19:06:15 +0000 |
commit | 45a0392656faaac09a8c7e322ead46a84bb1f1a8 (patch) | |
tree | c3ffd0f921fae0311291ad6fbe5d87b6b83f513f /src | |
parent | 9fa82b15f53f0f20363f50b5b1adf1a762ed96d6 (diff) |
Replaced charN_t types with uintN_t (C++0x compatibility)
git-svn-id: http://pugixml.googlecode.com/svn/trunk@504 99668b35-9821-0410-8761-19e4c4f06640
Diffstat (limited to 'src')
-rw-r--r-- | src/pugixml.cpp | 182 |
1 files changed, 89 insertions, 93 deletions
diff --git a/src/pugixml.cpp b/src/pugixml.cpp index 7911689..758d951 100644 --- a/src/pugixml.cpp +++ b/src/pugixml.cpp @@ -598,16 +598,12 @@ namespace pugi {
namespace impl
{
- typedef uint8_t char8_t;
- typedef uint16_t char16_t;
- typedef uint32_t char32_t;
-
- inline char16_t endian_swap(char16_t value)
+ inline uint16_t endian_swap(uint16_t value)
{
- return static_cast<char16_t>(((value & 0xff) << 8) | (value >> 8));
+ return static_cast<uint16_t>(((value & 0xff) << 8) | (value >> 8));
}
- inline char32_t endian_swap(char32_t value)
+ inline uint32_t endian_swap(uint32_t value)
{
return ((value & 0xff) << 24) | ((value & 0xff00) << 8) | ((value & 0xff0000) >> 8) | (value >> 24);
}
@@ -616,7 +612,7 @@ namespace pugi {
typedef size_t value_type;
- static value_type low(value_type result, char32_t ch)
+ static value_type low(value_type result, uint32_t ch)
{
// U+0000..U+007F
if (ch < 0x80) return result + 1;
@@ -626,7 +622,7 @@ namespace pugi else return result + 3;
}
- static value_type high(value_type result, char32_t)
+ static value_type high(value_type result, uint32_t)
{
// U+10000..U+10FFFF
return result + 4;
@@ -635,44 +631,44 @@ namespace pugi struct utf8_writer
{
- typedef char8_t* value_type;
+ typedef uint8_t* value_type;
- static value_type low(value_type result, char32_t ch)
+ static value_type low(value_type result, uint32_t ch)
{
// U+0000..U+007F
if (ch < 0x80)
{
- *result = static_cast<char8_t>(ch);
+ *result = static_cast<uint8_t>(ch);
return result + 1;
}
// U+0080..U+07FF
else if (ch < 0x800)
{
- result[0] = static_cast<char8_t>(0xC0 | (ch >> 6));
- result[1] = static_cast<char8_t>(0x80 | (ch & 0x3F));
+ result[0] = static_cast<uint8_t>(0xC0 | (ch >> 6));
+ result[1] = static_cast<uint8_t>(0x80 | (ch & 0x3F));
return result + 2;
}
// U+0800..U+FFFF
else
{
- result[0] = static_cast<char8_t>(0xE0 | (ch >> 12));
- result[1] = static_cast<char8_t>(0x80 | ((ch >> 6) & 0x3F));
- result[2] = static_cast<char8_t>(0x80 | (ch & 0x3F));
+ result[0] = static_cast<uint8_t>(0xE0 | (ch >> 12));
+ result[1] = static_cast<uint8_t>(0x80 | ((ch >> 6) & 0x3F));
+ result[2] = static_cast<uint8_t>(0x80 | (ch & 0x3F));
return result + 3;
}
}
- static value_type high(value_type result, char32_t ch)
+ static value_type high(value_type result, uint32_t ch)
{
// U+10000..U+10FFFF
- result[0] = static_cast<char8_t>(0xF0 | (ch >> 18));
- result[1] = static_cast<char8_t>(0x80 | ((ch >> 12) & 0x3F));
- result[2] = static_cast<char8_t>(0x80 | ((ch >> 6) & 0x3F));
- result[3] = static_cast<char8_t>(0x80 | (ch & 0x3F));
+ result[0] = static_cast<uint8_t>(0xF0 | (ch >> 18));
+ result[1] = static_cast<uint8_t>(0x80 | ((ch >> 12) & 0x3F));
+ result[2] = static_cast<uint8_t>(0x80 | ((ch >> 6) & 0x3F));
+ result[3] = static_cast<uint8_t>(0x80 | (ch & 0x3F));
return result + 4;
}
- static value_type any(value_type result, char32_t ch)
+ static value_type any(value_type result, uint32_t ch)
{
return (ch < 0x10000) ? low(result, ch) : high(result, ch);
}
@@ -682,12 +678,12 @@ namespace pugi {
typedef size_t value_type;
- static value_type low(value_type result, char32_t)
+ static value_type low(value_type result, uint32_t)
{
return result + 1;
}
- static value_type high(value_type result, char32_t)
+ static value_type high(value_type result, uint32_t)
{
return result + 2;
}
@@ -695,27 +691,27 @@ namespace pugi struct utf16_writer
{
- typedef char16_t* value_type;
+ typedef uint16_t* value_type;
- static value_type low(value_type result, char32_t ch)
+ static value_type low(value_type result, uint32_t ch)
{
- *result = static_cast<char16_t>(ch);
+ *result = static_cast<uint16_t>(ch);
return result + 1;
}
- static value_type high(value_type result, char32_t ch)
+ static value_type high(value_type result, uint32_t ch)
{
- char32_t msh = (char32_t)(ch - 0x10000) >> 10;
- char32_t lsh = (char32_t)(ch - 0x10000) & 0x3ff;
+ uint32_t msh = (uint32_t)(ch - 0x10000) >> 10;
+ uint32_t lsh = (uint32_t)(ch - 0x10000) & 0x3ff;
- result[0] = static_cast<char16_t>(0xD800 + msh);
- result[1] = static_cast<char16_t>(0xDC00 + lsh);
+ result[0] = static_cast<uint16_t>(0xD800 + msh);
+ result[1] = static_cast<uint16_t>(0xDC00 + lsh);
return result + 2;
}
- static value_type any(value_type result, char32_t ch)
+ static value_type any(value_type result, uint32_t ch)
{
return (ch < 0x10000) ? low(result, ch) : high(result, ch);
}
@@ -725,12 +721,12 @@ namespace pugi {
typedef size_t value_type;
- static value_type low(value_type result, char32_t)
+ static value_type low(value_type result, uint32_t)
{
return result + 1;
}
- static value_type high(value_type result, char32_t)
+ static value_type high(value_type result, uint32_t)
{
return result + 1;
}
@@ -738,23 +734,23 @@ namespace pugi struct utf32_writer
{
- typedef char32_t* value_type;
+ typedef uint32_t* value_type;
- static value_type low(value_type result, char32_t ch)
+ static value_type low(value_type result, uint32_t ch)
{
*result = ch;
return result + 1;
}
- static value_type high(value_type result, char32_t ch)
+ static value_type high(value_type result, uint32_t ch)
{
*result = ch;
return result + 1;
}
- static value_type any(value_type result, char32_t ch)
+ static value_type any(value_type result, uint32_t ch)
{
*result = ch;
@@ -766,14 +762,14 @@ namespace pugi template <> struct wchar_selector<2>
{
- typedef char16_t type;
+ typedef uint16_t type;
typedef utf16_counter counter;
typedef utf16_writer writer;
};
template <> struct wchar_selector<4>
{
- typedef char32_t type;
+ typedef uint32_t type;
typedef utf32_counter counter;
typedef utf32_writer writer;
};
@@ -781,13 +777,13 @@ namespace pugi typedef wchar_selector<sizeof(wchar_t)>::counter wchar_counter;
typedef wchar_selector<sizeof(wchar_t)>::writer wchar_writer;
- template <typename Traits> static inline typename Traits::value_type decode_utf8_block(const char8_t* data, size_t size, typename Traits::value_type result, Traits = Traits())
+ template <typename Traits> static inline typename Traits::value_type decode_utf8_block(const uint8_t* data, size_t size, typename Traits::value_type result, Traits = Traits())
{
- const char8_t utf8_byte_mask = 0x3f;
+ const uint8_t utf8_byte_mask = 0x3f;
while (size)
{
- char8_t lead = *data;
+ uint8_t lead = *data;
// 0xxxxxxx -> U+0000..U+007F
if (lead < 0x80)
@@ -799,7 +795,7 @@ namespace pugi // process aligned single-byte (ascii) blocks
if ((reinterpret_cast<uintptr_t>(data) & 3) == 0)
{
- while (size >= 4 && (*reinterpret_cast<const char32_t*>(data) & 0x80808080) == 0)
+ while (size >= 4 && (*reinterpret_cast<const uint32_t*>(data) & 0x80808080) == 0)
{
result = Traits::low(result, data[0]);
result = Traits::low(result, data[1]);
@@ -842,15 +838,15 @@ namespace pugi return result;
}
- template <typename Traits, typename opt1> static inline typename Traits::value_type decode_utf16_block(const char16_t* data, size_t size, typename Traits::value_type result, opt1, Traits = Traits())
+ template <typename Traits, typename opt1> static inline typename Traits::value_type decode_utf16_block(const uint16_t* data, size_t size, typename Traits::value_type result, opt1, Traits = Traits())
{
const bool swap = opt1::o1;
- const char16_t* end = data + size;
+ const uint16_t* end = data + size;
while (data < end)
{
- char16_t lead = swap ? endian_swap(*data) : *data;
+ uint16_t lead = swap ? endian_swap(*data) : *data;
// U+0000..U+D7FF
if (lead < 0xD800)
@@ -867,7 +863,7 @@ namespace pugi // surrogate pair lead
else if ((unsigned)(lead - 0xD800) < 0x400 && data + 1 < end)
{
- char16_t next = swap ? endian_swap(data[1]) : data[1];
+ uint16_t next = swap ? endian_swap(data[1]) : data[1];
if ((unsigned)(next - 0xDC00) < 0x400)
{
@@ -888,15 +884,15 @@ namespace pugi return result;
}
- template <typename Traits, typename opt1> static inline typename Traits::value_type decode_utf32_block(const char32_t* data, size_t size, typename Traits::value_type result, opt1, Traits = Traits())
+ template <typename Traits, typename opt1> static inline typename Traits::value_type decode_utf32_block(const uint32_t* data, size_t size, typename Traits::value_type result, opt1, Traits = Traits())
{
const bool swap = opt1::o1;
- const char32_t* end = data + size;
+ const uint32_t* end = data + size;
while (data < end)
{
- char32_t lead = swap ? endian_swap(*data) : *data;
+ uint32_t lead = swap ? endian_swap(*data) : *data;
// U+0000..U+FFFF
if (lead < 0x10000)
@@ -1052,7 +1048,7 @@ namespace if (encoding != encoding_auto) return encoding;
// try to guess encoding (based on XML specification, Appendix F.1)
- const impl::char8_t* data = static_cast<const impl::char8_t*>(contents);
+ const uint8_t* data = static_cast<const uint8_t*>(contents);
// look for BOM in first few bytes
if (size > 4 && data[0] == 0 && data[1] == 0 && data[2] == 0xfe && data[3] == 0xff) return encoding_utf32_be;
@@ -1127,7 +1123,7 @@ namespace bool convert_buffer_utf8(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size)
{
- const impl::char8_t* data = static_cast<const impl::char8_t*>(contents);
+ const uint8_t* data = static_cast<const uint8_t*>(contents);
// first pass: get length in wchar_t units
out_length = impl::decode_utf8_block<impl::wchar_counter>(data, size, 0);
@@ -1148,8 +1144,8 @@ namespace template <typename opt1> bool convert_buffer_utf16(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size, opt1)
{
- const impl::char16_t* data = static_cast<const impl::char16_t*>(contents);
- size_t length = size / sizeof(impl::char16_t);
+ const uint16_t* data = static_cast<const uint16_t*>(contents);
+ size_t length = size / sizeof(uint16_t);
// first pass: get length in wchar_t units
out_length = impl::decode_utf16_block<impl::wchar_counter>(data, length, 0, opt1());
@@ -1170,8 +1166,8 @@ namespace template <typename opt1> bool convert_buffer_utf32(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size, opt1)
{
- const impl::char32_t* data = static_cast<const impl::char32_t*>(contents);
- size_t length = size / sizeof(impl::char32_t);
+ const uint32_t* data = static_cast<const uint32_t*>(contents);
+ size_t length = size / sizeof(uint32_t);
// first pass: get length in wchar_t units
out_length = impl::decode_utf32_block<impl::wchar_counter>(data, length, 0, opt1());
@@ -1232,8 +1228,8 @@ namespace #else
template <typename opt1> bool convert_buffer_utf16(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size, opt1)
{
- const impl::char16_t* data = static_cast<const impl::char16_t*>(contents);
- size_t length = size / sizeof(impl::char16_t);
+ const uint16_t* data = static_cast<const uint16_t*>(contents);
+ size_t length = size / sizeof(uint16_t);
// first pass: get length in utf8 units
out_length = impl::decode_utf16_block<impl::utf8_counter>(data, length, 0, opt1());
@@ -1243,8 +1239,8 @@ namespace if (!out_buffer) return false;
// second pass: convert utf16 input to utf8
- impl::char8_t* out_begin = reinterpret_cast<impl::char8_t*>(out_buffer);
- impl::char8_t* out_end = impl::decode_utf16_block<impl::utf8_writer>(data, length, out_begin, opt1());
+ uint8_t* out_begin = reinterpret_cast<uint8_t*>(out_buffer);
+ uint8_t* out_end = impl::decode_utf16_block<impl::utf8_writer>(data, length, out_begin, opt1());
assert(out_end == out_begin + out_length);
(void)!out_end;
@@ -1254,8 +1250,8 @@ namespace template <typename opt1> bool convert_buffer_utf32(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size, opt1)
{
- const impl::char32_t* data = static_cast<const impl::char32_t*>(contents);
- size_t length = size / sizeof(impl::char32_t);
+ const uint32_t* data = static_cast<const uint32_t*>(contents);
+ size_t length = size / sizeof(uint32_t);
// first pass: get length in utf8 units
out_length = impl::decode_utf32_block<impl::utf8_counter>(data, length, 0, opt1());
@@ -1265,8 +1261,8 @@ namespace if (!out_buffer) return false;
// second pass: convert utf32 input to utf8
- impl::char8_t* out_begin = reinterpret_cast<impl::char8_t*>(out_buffer);
- impl::char8_t* out_end = impl::decode_utf32_block<impl::utf8_writer>(data, length, out_begin, opt1());
+ uint8_t* out_begin = reinterpret_cast<uint8_t*>(out_buffer);
+ uint8_t* out_end = impl::decode_utf32_block<impl::utf8_writer>(data, length, out_begin, opt1());
assert(out_end == out_begin + out_length);
(void)!out_end;
@@ -1421,7 +1417,7 @@ namespace #ifdef PUGIXML_WCHAR_MODE
s = reinterpret_cast<char_t*>(impl::wchar_writer::any(reinterpret_cast<impl::wchar_writer::value_type>(s), ucsc));
#else
- s = reinterpret_cast<char_t*>(impl::utf8_writer::any(reinterpret_cast<impl::char8_t*>(s), ucsc));
+ s = reinterpret_cast<char_t*>(impl::utf8_writer::any(reinterpret_cast<uint8_t*>(s), ucsc));
#endif
g.push(s, stre - s);
@@ -2457,7 +2453,7 @@ namespace assert(length > 0);
// discard last character if it's the lead of a surrogate pair
- return (sizeof(wchar_t) == 2 && (unsigned)(static_cast<impl::char16_t>(data[length - 1]) - 0xD800) < 0x400) ? length - 1 : length;
+ return (sizeof(wchar_t) == 2 && (unsigned)(static_cast<uint16_t>(data[length - 1]) - 0xD800) < 0x400) ? length - 1 : length;
}
size_t convert_buffer(char* result, const char_t* data, size_t length, encoding_t encoding)
@@ -2473,11 +2469,11 @@ namespace // convert to utf8
if (encoding == encoding_utf8)
{
- impl::char8_t* dest = reinterpret_cast<impl::char8_t*>(result);
+ uint8_t* dest = reinterpret_cast<uint8_t*>(result);
- impl::char8_t* end = sizeof(wchar_t) == 2 ?
- impl::decode_utf16_block<impl::utf8_writer>(reinterpret_cast<const impl::char16_t*>(data), length, dest, opt1_to_type<false>()) :
- impl::decode_utf32_block<impl::utf8_writer>(reinterpret_cast<const impl::char32_t*>(data), length, dest, opt1_to_type<false>());
+ uint8_t* end = sizeof(wchar_t) == 2 ?
+ impl::decode_utf16_block<impl::utf8_writer>(reinterpret_cast<const uint16_t*>(data), length, dest, opt1_to_type<false>()) :
+ impl::decode_utf32_block<impl::utf8_writer>(reinterpret_cast<const uint32_t*>(data), length, dest, opt1_to_type<false>());
return static_cast<size_t>(end - dest);
}
@@ -2485,33 +2481,33 @@ namespace // convert to utf16
if (encoding == encoding_utf16_be || encoding == encoding_utf16_le)
{
- impl::char16_t* dest = reinterpret_cast<impl::char16_t*>(result);
+ uint16_t* dest = reinterpret_cast<uint16_t*>(result);
// convert to native utf16
- impl::char16_t* end = impl::decode_utf32_block<impl::utf16_writer>(reinterpret_cast<const impl::char32_t*>(data), length, dest, opt1_to_type<false>());
+ uint16_t* end = impl::decode_utf32_block<impl::utf16_writer>(reinterpret_cast<const uint32_t*>(data), length, dest, opt1_to_type<false>());
// swap if necessary
encoding_t native_encoding = is_little_endian() ? encoding_utf16_le : encoding_utf16_be;
if (native_encoding != encoding) impl::convert_utf_endian_swap(dest, dest, static_cast<size_t>(end - dest));
- return static_cast<size_t>(end - dest) * sizeof(impl::char16_t);
+ return static_cast<size_t>(end - dest) * sizeof(uint16_t);
}
// convert to utf32
if (encoding == encoding_utf32_be || encoding == encoding_utf32_le)
{
- impl::char32_t* dest = reinterpret_cast<impl::char32_t*>(result);
+ uint32_t* dest = reinterpret_cast<uint32_t*>(result);
// convert to native utf32
- impl::char32_t* end = impl::decode_utf16_block<impl::utf32_writer>(reinterpret_cast<const impl::char16_t*>(data), length, dest, opt1_to_type<false>());
+ uint32_t* end = impl::decode_utf16_block<impl::utf32_writer>(reinterpret_cast<const uint16_t*>(data), length, dest, opt1_to_type<false>());
// swap if necessary
encoding_t native_encoding = is_little_endian() ? encoding_utf32_le : encoding_utf32_be;
if (native_encoding != encoding) impl::convert_utf_endian_swap(dest, dest, static_cast<size_t>(end - dest));
- return static_cast<size_t>(end - dest) * sizeof(impl::char32_t);
+ return static_cast<size_t>(end - dest) * sizeof(uint32_t);
}
// invalid encoding combination (this can't happen)
@@ -2526,7 +2522,7 @@ namespace for (size_t i = 1; i <= 4; ++i)
{
- impl::char8_t ch = static_cast<impl::char8_t>(data[length - i]);
+ uint8_t ch = static_cast<uint8_t>(data[length - i]);
// either a standalone character or a leading one
if ((ch & 0xc0) != 0x80) return length - i;
@@ -2540,32 +2536,32 @@ namespace {
if (encoding == encoding_utf16_be || encoding == encoding_utf16_le)
{
- impl::char16_t* dest = reinterpret_cast<impl::char16_t*>(result);
+ uint16_t* dest = reinterpret_cast<uint16_t*>(result);
// convert to native utf16
- impl::char16_t* end = impl::decode_utf8_block<impl::utf16_writer>(reinterpret_cast<const impl::char8_t*>(data), length, dest);
+ uint16_t* end = impl::decode_utf8_block<impl::utf16_writer>(reinterpret_cast<const uint8_t*>(data), length, dest);
// swap if necessary
encoding_t native_encoding = is_little_endian() ? encoding_utf16_le : encoding_utf16_be;
if (native_encoding != encoding) impl::convert_utf_endian_swap(dest, dest, static_cast<size_t>(end - dest));
- return static_cast<size_t>(end - dest) * sizeof(impl::char16_t);
+ return static_cast<size_t>(end - dest) * sizeof(uint16_t);
}
if (encoding == encoding_utf32_be || encoding == encoding_utf32_le)
{
- impl::char32_t* dest = reinterpret_cast<impl::char32_t*>(result);
+ uint32_t* dest = reinterpret_cast<uint32_t*>(result);
// convert to native utf32
- impl::char32_t* end = impl::decode_utf8_block<impl::utf32_writer>(reinterpret_cast<const impl::char8_t*>(data), length, dest);
+ uint32_t* end = impl::decode_utf8_block<impl::utf32_writer>(reinterpret_cast<const uint8_t*>(data), length, dest);
// swap if necessary
encoding_t native_encoding = is_little_endian() ? encoding_utf32_le : encoding_utf32_be;
if (native_encoding != encoding) impl::convert_utf_endian_swap(dest, dest, static_cast<size_t>(end - dest));
- return static_cast<size_t>(end - dest) * sizeof(impl::char32_t);
+ return static_cast<size_t>(end - dest) * sizeof(uint32_t);
}
// invalid encoding combination (this can't happen)
@@ -4472,8 +4468,8 @@ namespace pugi // first pass: get length in utf8 characters
size_t size = sizeof(wchar_t) == 2 ?
- impl::decode_utf16_block<impl::utf8_counter>(reinterpret_cast<const impl::char16_t*>(str), length, 0, opt1_to_type<false>()) :
- impl::decode_utf32_block<impl::utf8_counter>(reinterpret_cast<const impl::char32_t*>(str), length, 0, opt1_to_type<false>());
+ impl::decode_utf16_block<impl::utf8_counter>(reinterpret_cast<const uint16_t*>(str), length, 0, opt1_to_type<false>()) :
+ impl::decode_utf32_block<impl::utf8_counter>(reinterpret_cast<const uint32_t*>(str), length, 0, opt1_to_type<false>());
// allocate resulting string
std::string result;
@@ -4482,10 +4478,10 @@ namespace pugi // second pass: convert to utf8
if (size > 0)
{
- impl::char8_t* begin = reinterpret_cast<impl::char8_t*>(&result[0]);
- impl::char8_t* end = sizeof(wchar_t) == 2 ?
- impl::decode_utf16_block<impl::utf8_writer>(reinterpret_cast<const impl::char16_t*>(str), length, begin, opt1_to_type<false>()) :
- impl::decode_utf32_block<impl::utf8_writer>(reinterpret_cast<const impl::char32_t*>(str), length, begin, opt1_to_type<false>());
+ uint8_t* begin = reinterpret_cast<uint8_t*>(&result[0]);
+ uint8_t* end = sizeof(wchar_t) == 2 ?
+ impl::decode_utf16_block<impl::utf8_writer>(reinterpret_cast<const uint16_t*>(str), length, begin, opt1_to_type<false>()) :
+ impl::decode_utf32_block<impl::utf8_writer>(reinterpret_cast<const uint32_t*>(str), length, begin, opt1_to_type<false>());
// truncate invalid output
assert(begin <= end && static_cast<size_t>(end - begin) <= result.size());
@@ -4502,7 +4498,7 @@ namespace pugi std::wstring PUGIXML_FUNCTION as_wide(const char* str)
{
- const impl::char8_t* data = reinterpret_cast<const impl::char8_t*>(str);
+ const uint8_t* data = reinterpret_cast<const uint8_t*>(str);
size_t size = strlen(str);
// first pass: get length in wchar_t
|