diff --git a/libpugixml/pugiconfig.hpp b/libpugixml/pugiconfig.hpp index 90b81e1..f90a480 100755 --- a/libpugixml/pugiconfig.hpp +++ b/libpugixml/pugiconfig.hpp @@ -1,7 +1,7 @@ /** - * pugixml parser - version 1.0 + * pugixml parser - version 1.9 * -------------------------------------------------------- - * Copyright (C) 2006-2010, by Arseny Kapoulkine (arseny.kapoulkine@gmail.com) + * Copyright (C) 2006-2018, by Arseny Kapoulkine (arseny.kapoulkine@gmail.com) * Report bugs and download new versions at http://pugixml.org/ * * This library is distributed under the MIT License. See notice at the end @@ -17,15 +17,16 @@ // Uncomment this to enable wchar_t mode // #define PUGIXML_WCHAR_MODE +// Uncomment this to enable compact mode +// #define PUGIXML_COMPACT + // Uncomment this to disable XPath // #define PUGIXML_NO_XPATH // Uncomment this to disable STL -// Note: you can't use XPath with PUGIXML_NO_STL // #define PUGIXML_NO_STL // Uncomment this to disable exceptions -// Note: you can't use XPath with PUGIXML_NO_EXCEPTIONS // #define PUGIXML_NO_EXCEPTIONS // Set this to control attributes for public classes/functions, i.e.: @@ -34,10 +35,21 @@ // #define PUGIXML_FUNCTION __fastcall // to set calling conventions to all public functions to fastcall // In absence of PUGIXML_CLASS/PUGIXML_FUNCTION definitions PUGIXML_API is used instead +// Tune these constants to adjust memory-related behavior +// #define PUGIXML_MEMORY_PAGE_SIZE 32768 +// #define PUGIXML_MEMORY_OUTPUT_STACK 10240 +// #define PUGIXML_MEMORY_XPATH_PAGE_SIZE 4096 + +// Uncomment this to switch to header-only version +// #define PUGIXML_HEADER_ONLY + +// Uncomment this to enable long long support +// #define PUGIXML_HAS_LONG_LONG + #endif /** - * Copyright (c) 2006-2010 Arseny Kapoulkine + * Copyright (c) 2006-2018 Arseny Kapoulkine * * Permission is hereby granted, free of charge, to any person * obtaining a copy of this software and associated documentation @@ -50,7 +62,7 @@ * * The above copyright notice and this permission notice shall be * included in all copies or substantial portions of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND @@ -60,3 +72,4 @@ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR * OTHER DEALINGS IN THE SOFTWARE. */ + diff --git a/libpugixml/pugixml.cpp b/libpugixml/pugixml.cpp index 901731a..01ed2c0 100755 --- a/libpugixml/pugixml.cpp +++ b/libpugixml/pugixml.cpp @@ -1,7 +1,7 @@ /** - * pugixml parser - version 1.0 + * pugixml parser - version 1.9 * -------------------------------------------------------- - * Copyright (C) 2006-2010, by Arseny Kapoulkine (arseny.kapoulkine@gmail.com) + * Copyright (C) 2006-2018, by Arseny Kapoulkine (arseny.kapoulkine@gmail.com) * Report bugs and download new versions at http://pugixml.org/ * * This library is distributed under the MIT License. See notice at the end @@ -11,14 +11,20 @@ * Copyright (C) 2003, by Kristen Wegner (kristen@tima.net) */ +#ifndef SOURCE_PUGIXML_CPP +#define SOURCE_PUGIXML_CPP + #include "pugixml.hpp" #include #include #include #include -#include -#include +#include + +#ifdef PUGIXML_WCHAR_MODE +# include +#endif #ifndef PUGIXML_NO_XPATH # include @@ -35,85 +41,175 @@ #include #ifdef _MSC_VER +# pragma warning(push) # pragma warning(disable: 4127) // conditional expression is constant # pragma warning(disable: 4324) // structure was padded due to __declspec(align()) -# pragma warning(disable: 4611) // interaction between '_setjmp' and C++ object destruction is non-portable # pragma warning(disable: 4702) // unreachable code # pragma warning(disable: 4996) // this function or variable may be unsafe #endif +#if defined(_MSC_VER) && defined(__c2__) +# pragma clang diagnostic push +# pragma clang diagnostic ignored "-Wdeprecated" // this function or variable may be unsafe +#endif + #ifdef __INTEL_COMPILER -# pragma warning(disable: 177) // function was declared but never referenced +# pragma warning(disable: 177) // function was declared but never referenced +# pragma warning(disable: 279) // controlling expression is constant # pragma warning(disable: 1478 1786) // function was declared "deprecated" +# pragma warning(disable: 1684) // conversion from pointer to same-sized integral type +#endif + +#if defined(__BORLANDC__) && defined(PUGIXML_HEADER_ONLY) +# pragma warn -8080 // symbol is declared but never used; disabling this inside push/pop bracket does not make the warning go away #endif #ifdef __BORLANDC__ +# pragma option push # pragma warn -8008 // condition is always false # pragma warn -8066 // unreachable code #endif #ifdef __SNC__ +// Using diag_push/diag_pop does not disable the warnings inside templates due to a compiler bug # pragma diag_suppress=178 // function was declared but never referenced # pragma diag_suppress=237 // controlling expression is constant #endif -// uintptr_t -#if !defined(_MSC_VER) || _MSC_VER >= 1600 -# include -#else -# if _MSC_VER < 1300 -// No native uintptr_t in MSVC6 -typedef size_t uintptr_t; -# endif -typedef unsigned __int8 uint8_t; -typedef unsigned __int16 uint16_t; -typedef unsigned __int32 uint32_t; -typedef __int32 int32_t; +#ifdef __TI_COMPILER_VERSION__ +# pragma diag_suppress 179 // function was declared but never referenced #endif // Inlining controls #if defined(_MSC_VER) && _MSC_VER >= 1300 -# define PUGIXML_NO_INLINE __declspec(noinline) +# define PUGI__NO_INLINE __declspec(noinline) #elif defined(__GNUC__) -# define PUGIXML_NO_INLINE __attribute__((noinline)) +# define PUGI__NO_INLINE __attribute__((noinline)) #else -# define PUGIXML_NO_INLINE +# define PUGI__NO_INLINE +#endif + +// Branch weight controls +#if defined(__GNUC__) && !defined(__c2__) +# define PUGI__UNLIKELY(cond) __builtin_expect(cond, 0) +#else +# define PUGI__UNLIKELY(cond) (cond) #endif // Simple static assertion -#define STATIC_ASSERT(cond) { static const char condition_failed[(cond) ? 1 : -1] = {0}; (void)condition_failed[0]; } +#define PUGI__STATIC_ASSERT(cond) { static const char condition_failed[(cond) ? 1 : -1] = {0}; (void)condition_failed[0]; } // Digital Mars C++ bug workaround for passing char loaded from memory via stack #ifdef __DMC__ -# define DMC_VOLATILE volatile +# define PUGI__DMC_VOLATILE volatile +#else +# define PUGI__DMC_VOLATILE +#endif + +// Integer sanitizer workaround; we only apply this for clang since gcc8 has no_sanitize but not unsigned-integer-overflow and produces "attribute directive ignored" warnings +#if defined(__clang__) && defined(__has_attribute) +# if __has_attribute(no_sanitize) +# define PUGI__UNSIGNED_OVERFLOW __attribute__((no_sanitize("unsigned-integer-overflow"))) +# else +# define PUGI__UNSIGNED_OVERFLOW +# endif #else -# define DMC_VOLATILE +# define PUGI__UNSIGNED_OVERFLOW +#endif + +// Borland C++ bug workaround for not defining ::memcpy depending on header include order (can't always use std::memcpy because some compilers don't have it at all) +#if defined(__BORLANDC__) && !defined(__MEM_H_USING_LIST) +using std::memcpy; +using std::memmove; +using std::memset; #endif -using namespace pugi; +// Some MinGW/GCC versions have headers that erroneously omit LLONG_MIN/LLONG_MAX/ULLONG_MAX definitions from limits.h in some configurations +#if defined(PUGIXML_HAS_LONG_LONG) && defined(__GNUC__) && !defined(LLONG_MAX) && !defined(LLONG_MIN) && !defined(ULLONG_MAX) +# define LLONG_MIN (-LLONG_MAX - 1LL) +# define LLONG_MAX __LONG_LONG_MAX__ +# define ULLONG_MAX (LLONG_MAX * 2ULL + 1ULL) +#endif -// Memory allocation -namespace +// In some environments MSVC is a compiler but the CRT lacks certain MSVC-specific features +#if defined(_MSC_VER) && !defined(__S3E__) +# define PUGI__MSVC_CRT_VERSION _MSC_VER +#endif + +// Not all platforms have snprintf; we define a wrapper that uses snprintf if possible. This only works with buffers with a known size. +#if __cplusplus >= 201103 +# define PUGI__SNPRINTF(buf, ...) snprintf(buf, sizeof(buf), __VA_ARGS__) +#elif defined(PUGI__MSVC_CRT_VERSION) && PUGI__MSVC_CRT_VERSION >= 1400 +# define PUGI__SNPRINTF(buf, ...) _snprintf_s(buf, _countof(buf), _TRUNCATE, __VA_ARGS__) +#else +# define PUGI__SNPRINTF sprintf +#endif + +// We put implementation details into an anonymous namespace in source mode, but have to keep it in non-anonymous namespace in header-only mode to prevent binary bloat. +#ifdef PUGIXML_HEADER_ONLY +# define PUGI__NS_BEGIN namespace pugi { namespace impl { +# define PUGI__NS_END } } +# define PUGI__FN inline +# define PUGI__FN_NO_INLINE inline +#else +# if defined(_MSC_VER) && _MSC_VER < 1300 // MSVC6 seems to have an amusing bug with anonymous namespaces inside namespaces +# define PUGI__NS_BEGIN namespace pugi { namespace impl { +# define PUGI__NS_END } } +# else +# define PUGI__NS_BEGIN namespace pugi { namespace impl { namespace { +# define PUGI__NS_END } } } +# endif +# define PUGI__FN +# define PUGI__FN_NO_INLINE PUGI__NO_INLINE +#endif + +// uintptr_t +#if (defined(_MSC_VER) && _MSC_VER < 1600) || (defined(__BORLANDC__) && __BORLANDC__ < 0x561) +namespace pugi { - void* default_allocate(size_t size) +# ifndef _UINTPTR_T_DEFINED + typedef size_t uintptr_t; +# endif + + typedef unsigned __int8 uint8_t; + typedef unsigned __int16 uint16_t; + typedef unsigned __int32 uint32_t; +} +#else +# include +#endif + +// Memory allocation +PUGI__NS_BEGIN + PUGI__FN void* default_allocate(size_t size) { return malloc(size); } - void default_deallocate(void* ptr) + PUGI__FN void default_deallocate(void* ptr) { free(ptr); } - allocation_function global_allocate = default_allocate; - deallocation_function global_deallocate = default_deallocate; -} + template + struct xml_memory_management_function_storage + { + static allocation_function allocate; + static deallocation_function deallocate; + }; + + // Global allocation functions are stored in class statics so that in header mode linker deduplicates them + // Without a template<> we'll get multiple definitions of the same static + template allocation_function xml_memory_management_function_storage::allocate = default_allocate; + template deallocation_function xml_memory_management_function_storage::deallocate = default_deallocate; + + typedef xml_memory_management_function_storage xml_memory; +PUGI__NS_END // String utilities -namespace -{ +PUGI__NS_BEGIN // Get string length - size_t strlength(const char_t* s) + PUGI__FN size_t strlength(const char_t* s) { assert(s); @@ -125,7 +221,7 @@ namespace } // Compare two strings - bool strequal(const char_t* src, const char_t* dst) + PUGI__FN bool strequal(const char_t* src, const char_t* dst) { assert(src && dst); @@ -137,62 +233,225 @@ namespace } // Compare lhs with [rhs_begin, rhs_end) - bool strequalrange(const char_t* lhs, const char_t* rhs, size_t count) + PUGI__FN bool strequalrange(const char_t* lhs, const char_t* rhs, size_t count) { for (size_t i = 0; i < count; ++i) if (lhs[i] != rhs[i]) return false; - + return lhs[count] == 0; } - -#ifdef PUGIXML_WCHAR_MODE - // Convert string to wide string, assuming all symbols are ASCII - void widen_ascii(wchar_t* dest, const char* source) + + // Get length of wide string, even if CRT lacks wide character support + PUGI__FN size_t strlength_wide(const wchar_t* s) { - for (const char* i = source; *i; ++i) *dest++ = *i; - *dest = 0; + assert(s); + + #ifdef PUGIXML_WCHAR_MODE + return wcslen(s); + #else + const wchar_t* end = s; + while (*end) end++; + return static_cast(end - s); + #endif } -#endif -} +PUGI__NS_END -#if !defined(PUGIXML_NO_STL) || !defined(PUGIXML_NO_XPATH) -// auto_ptr-like buffer holder for exception recovery -namespace -{ - struct buffer_holder +// auto_ptr-like object for exception recovery +PUGI__NS_BEGIN + template struct auto_deleter { - void* data; - void (*deleter)(void*); + typedef void (*D)(T*); - buffer_holder(void* data, void (*deleter)(void*)): data(data), deleter(deleter) + T* data; + D deleter; + + auto_deleter(T* data_, D deleter_): data(data_), deleter(deleter_) { } - ~buffer_holder() + ~auto_deleter() { if (data) deleter(data); } - void* release() + T* release() { - void* result = data; + T* result = data; data = 0; return result; } }; -} +PUGI__NS_END + +#ifdef PUGIXML_COMPACT +PUGI__NS_BEGIN + class compact_hash_table + { + public: + compact_hash_table(): _items(0), _capacity(0), _count(0) + { + } + + void clear() + { + if (_items) + { + xml_memory::deallocate(_items); + _items = 0; + _capacity = 0; + _count = 0; + } + } + + void* find(const void* key) + { + if (_capacity == 0) return 0; + + item_t* item = get_item(key); + assert(item); + assert(item->key == key || (item->key == 0 && item->value == 0)); + + return item->value; + } + + void insert(const void* key, void* value) + { + assert(_capacity != 0 && _count < _capacity - _capacity / 4); + + item_t* item = get_item(key); + assert(item); + + if (item->key == 0) + { + _count++; + item->key = key; + } + + item->value = value; + } + + bool reserve(size_t extra = 16) + { + if (_count + extra >= _capacity - _capacity / 4) + return rehash(_count + extra); + + return true; + } + + private: + struct item_t + { + const void* key; + void* value; + }; + + item_t* _items; + size_t _capacity; + + size_t _count; + + bool rehash(size_t count); + + item_t* get_item(const void* key) + { + assert(key); + assert(_capacity > 0); + + size_t hashmod = _capacity - 1; + size_t bucket = hash(key) & hashmod; + + for (size_t probe = 0; probe <= hashmod; ++probe) + { + item_t& probe_item = _items[bucket]; + + if (probe_item.key == key || probe_item.key == 0) + return &probe_item; + + // hash collision, quadratic probing + bucket = (bucket + probe + 1) & hashmod; + } + + assert(false && "Hash table is full"); // unreachable + return 0; + } + + static PUGI__UNSIGNED_OVERFLOW unsigned int hash(const void* key) + { + unsigned int h = static_cast(reinterpret_cast(key)); + + // MurmurHash3 32-bit finalizer + h ^= h >> 16; + h *= 0x85ebca6bu; + h ^= h >> 13; + h *= 0xc2b2ae35u; + h ^= h >> 16; + + return h; + } + }; + + PUGI__FN_NO_INLINE bool compact_hash_table::rehash(size_t count) + { + size_t capacity = 32; + while (count >= capacity - capacity / 4) + capacity *= 2; + + compact_hash_table rt; + rt._capacity = capacity; + rt._items = static_cast(xml_memory::allocate(sizeof(item_t) * capacity)); + + if (!rt._items) + return false; + + memset(rt._items, 0, sizeof(item_t) * capacity); + + for (size_t i = 0; i < _capacity; ++i) + if (_items[i].key) + rt.insert(_items[i].key, _items[i].value); + + if (_items) + xml_memory::deallocate(_items); + + _capacity = capacity; + _items = rt._items; + + assert(_count == rt._count); + + return true; + } + +PUGI__NS_END #endif -namespace -{ - static const size_t xml_memory_page_size = 32768; +PUGI__NS_BEGIN +#ifdef PUGIXML_COMPACT + static const uintptr_t xml_memory_block_alignment = 4; +#else + static const uintptr_t xml_memory_block_alignment = sizeof(void*); +#endif + + // extra metadata bits + static const uintptr_t xml_memory_page_contents_shared_mask = 64; + static const uintptr_t xml_memory_page_name_allocated_mask = 32; + static const uintptr_t xml_memory_page_value_allocated_mask = 16; + static const uintptr_t xml_memory_page_type_mask = 15; + + // combined masks for string uniqueness + static const uintptr_t xml_memory_page_name_allocated_or_shared_mask = xml_memory_page_name_allocated_mask | xml_memory_page_contents_shared_mask; + static const uintptr_t xml_memory_page_value_allocated_or_shared_mask = xml_memory_page_value_allocated_mask | xml_memory_page_contents_shared_mask; - static const uintptr_t xml_memory_page_alignment = 32; - static const uintptr_t xml_memory_page_pointer_mask = ~(xml_memory_page_alignment - 1); - static const uintptr_t xml_memory_page_name_allocated_mask = 16; - static const uintptr_t xml_memory_page_value_allocated_mask = 8; - static const uintptr_t xml_memory_page_type_mask = 7; +#ifdef PUGIXML_COMPACT + #define PUGI__GETHEADER_IMPL(object, page, flags) // unused + #define PUGI__GETPAGE_IMPL(header) (header).get_page() +#else + #define PUGI__GETHEADER_IMPL(object, page, flags) (((reinterpret_cast(object) - reinterpret_cast(page)) << 8) | (flags)) + // this macro casts pointers through void* to avoid 'cast increases required alignment of target type' warnings + #define PUGI__GETPAGE_IMPL(header) static_cast(const_cast(static_cast(reinterpret_cast(&header) - (header >> 8)))) +#endif + + #define PUGI__GETPAGE(n) PUGI__GETPAGE_IMPL((n)->header) + #define PUGI__NODETYPE(n) static_cast((n)->header & impl::xml_memory_page_type_mask) struct xml_allocator; @@ -200,33 +459,46 @@ namespace { static xml_memory_page* construct(void* memory) { - if (!memory) return 0; //$ redundant, left for performance - xml_memory_page* result = static_cast(memory); result->allocator = 0; - result->memory = 0; result->prev = 0; result->next = 0; result->busy_size = 0; result->freed_size = 0; + #ifdef PUGIXML_COMPACT + result->compact_string_base = 0; + result->compact_shared_parent = 0; + result->compact_page_marker = 0; + #endif + return result; } xml_allocator* allocator; - void* memory; - xml_memory_page* prev; xml_memory_page* next; size_t busy_size; size_t freed_size; - char data[1]; + #ifdef PUGIXML_COMPACT + char_t* compact_string_base; + void* compact_shared_parent; + uint32_t* compact_page_marker; + #endif }; + static const size_t xml_memory_page_size = + #ifdef PUGIXML_MEMORY_PAGE_SIZE + (PUGIXML_MEMORY_PAGE_SIZE) + #else + 32768 + #endif + - sizeof(xml_memory_page); + struct xml_memory_string_header { uint16_t page_offset; // offset from page->data @@ -237,23 +509,23 @@ namespace { xml_allocator(xml_memory_page* root): _root(root), _busy_size(root->busy_size) { + #ifdef PUGIXML_COMPACT + _hash = 0; + #endif } xml_memory_page* allocate_page(size_t data_size) { - size_t size = offsetof(xml_memory_page, data) + data_size; + size_t size = sizeof(xml_memory_page) + data_size; // allocate block with some alignment, leaving memory for worst-case padding - void* memory = global_allocate(size + xml_memory_page_alignment); + void* memory = xml_memory::allocate(size); if (!memory) return 0; - // align upwards to page boundary - void* page_memory = reinterpret_cast((reinterpret_cast(memory) + (xml_memory_page_alignment - 1)) & ~(xml_memory_page_alignment - 1)); - // prepare page structure - xml_memory_page* page = xml_memory_page::construct(page_memory); + xml_memory_page* page = xml_memory_page::construct(memory); + assert(page); - page->memory = memory; page->allocator = _root->allocator; return page; @@ -261,16 +533,17 @@ namespace static void deallocate_page(xml_memory_page* page) { - global_deallocate(page->memory); + xml_memory::deallocate(page); } void* allocate_memory_oob(size_t size, xml_memory_page*& out_page); void* allocate_memory(size_t size, xml_memory_page*& out_page) { - if (_busy_size + size > xml_memory_page_size) return allocate_memory_oob(size, out_page); + if (PUGI__UNLIKELY(_busy_size + size > xml_memory_page_size)) + return allocate_memory_oob(size, out_page); - void* buf = _root->data + _busy_size; + void* buf = reinterpret_cast(_root) + sizeof(xml_memory_page) + _busy_size; _busy_size += size; @@ -279,11 +552,49 @@ namespace return buf; } + #ifdef PUGIXML_COMPACT + void* allocate_object(size_t size, xml_memory_page*& out_page) + { + void* result = allocate_memory(size + sizeof(uint32_t), out_page); + if (!result) return 0; + + // adjust for marker + ptrdiff_t offset = static_cast(result) - reinterpret_cast(out_page->compact_page_marker); + + if (PUGI__UNLIKELY(static_cast(offset) >= 256 * xml_memory_block_alignment)) + { + // insert new marker + uint32_t* marker = static_cast(result); + + *marker = static_cast(reinterpret_cast(marker) - reinterpret_cast(out_page)); + out_page->compact_page_marker = marker; + + // since we don't reuse the page space until we reallocate it, we can just pretend that we freed the marker block + // this will make sure deallocate_memory correctly tracks the size + out_page->freed_size += sizeof(uint32_t); + + return marker + 1; + } + else + { + // roll back uint32_t part + _busy_size -= sizeof(uint32_t); + + return result; + } + } + #else + void* allocate_object(size_t size, xml_memory_page*& out_page) + { + return allocate_memory(size, out_page); + } + #endif + void deallocate_memory(void* ptr, size_t size, xml_memory_page* page) { if (page == _root) page->busy_size = _busy_size; - assert(ptr >= page->data && ptr < page->data + page->busy_size); + assert(ptr >= reinterpret_cast(page) + sizeof(xml_memory_page) && ptr < reinterpret_cast(page) + sizeof(xml_memory_page) + page->busy_size); (void)!ptr; page->freed_size += size; @@ -296,7 +607,16 @@ namespace assert(_root == page); // top page freed, just reset sizes - page->busy_size = page->freed_size = 0; + page->busy_size = 0; + page->freed_size = 0; + + #ifdef PUGIXML_COMPACT + // reset compact state to maximize efficiency + page->compact_string_base = 0; + page->compact_shared_parent = 0; + page->compact_page_marker = 0; + #endif + _busy_size = 0; } else @@ -316,11 +636,15 @@ namespace char_t* allocate_string(size_t length) { + static const size_t max_encoded_offset = (1 << 16) * xml_memory_block_alignment; + + PUGI__STATIC_ASSERT(xml_memory_page_size <= max_encoded_offset); + // allocate memory for string and header block size_t size = sizeof(xml_memory_string_header) + length * sizeof(char_t); - - // round size up to pointer alignment boundary - size_t full_size = (size + (sizeof(void*) - 1)) & ~(sizeof(void*) - 1); + + // round size up to block alignment boundary + size_t full_size = (size + (xml_memory_block_alignment - 1)) & ~(xml_memory_block_alignment - 1); xml_memory_page* page; xml_memory_string_header* header = static_cast(allocate_memory(full_size, page)); @@ -328,42 +652,65 @@ namespace if (!header) return 0; // setup header - ptrdiff_t page_offset = reinterpret_cast(header) - page->data; + ptrdiff_t page_offset = reinterpret_cast(header) - reinterpret_cast(page) - sizeof(xml_memory_page); - assert(page_offset >= 0 && page_offset < (1 << 16)); - header->page_offset = static_cast(page_offset); + assert(page_offset % xml_memory_block_alignment == 0); + assert(page_offset >= 0 && static_cast(page_offset) < max_encoded_offset); + header->page_offset = static_cast(static_cast(page_offset) / xml_memory_block_alignment); // full_size == 0 for large strings that occupy the whole page - assert(full_size < (1 << 16) || (page->busy_size == full_size && page_offset == 0)); - header->full_size = static_cast(full_size < (1 << 16) ? full_size : 0); + assert(full_size % xml_memory_block_alignment == 0); + assert(full_size < max_encoded_offset || (page->busy_size == full_size && page_offset == 0)); + header->full_size = static_cast(full_size < max_encoded_offset ? full_size / xml_memory_block_alignment : 0); - return reinterpret_cast(header + 1); + // round-trip through void* to avoid 'cast increases required alignment of target type' warning + // header is guaranteed a pointer-sized alignment, which should be enough for char_t + return static_cast(static_cast(header + 1)); } void deallocate_string(char_t* string) { + // this function casts pointers through void* to avoid 'cast increases required alignment of target type' warnings + // we're guaranteed the proper (pointer-sized) alignment on the input string if it was allocated via allocate_string + // get header - xml_memory_string_header* header = reinterpret_cast(string) - 1; + xml_memory_string_header* header = static_cast(static_cast(string)) - 1; + assert(header); // deallocate - size_t page_offset = offsetof(xml_memory_page, data) + header->page_offset; - xml_memory_page* page = reinterpret_cast(reinterpret_cast(header) - page_offset); + size_t page_offset = sizeof(xml_memory_page) + header->page_offset * xml_memory_block_alignment; + xml_memory_page* page = reinterpret_cast(static_cast(reinterpret_cast(header) - page_offset)); // if full_size == 0 then this string occupies the whole page - size_t full_size = header->full_size == 0 ? page->busy_size : header->full_size; + size_t full_size = header->full_size == 0 ? page->busy_size : header->full_size * xml_memory_block_alignment; deallocate_memory(header, full_size, page); } + bool reserve() + { + #ifdef PUGIXML_COMPACT + return _hash->reserve(); + #else + return true; + #endif + } + xml_memory_page* _root; size_t _busy_size; + + #ifdef PUGIXML_COMPACT + compact_hash_table* _hash; + #endif }; - PUGIXML_NO_INLINE void* xml_allocator::allocate_memory_oob(size_t size, xml_memory_page*& out_page) + PUGI__FN_NO_INLINE void* xml_allocator::allocate_memory_oob(size_t size, xml_memory_page*& out_page) { const size_t large_allocation_threshold = xml_memory_page_size / 4; xml_memory_page* page = allocate_page(size <= large_allocation_threshold ? xml_memory_page_size : size); + out_page = page; + if (!page) return 0; if (size <= large_allocation_threshold) @@ -388,4081 +735,6075 @@ namespace _root->prev->next = page; _root->prev = page; - } - // allocate inside page - page->busy_size = size; + page->busy_size = size; + } - out_page = page; - return page->data; + return reinterpret_cast(page) + sizeof(xml_memory_page); } -} +PUGI__NS_END -namespace pugi -{ - /// A 'name=value' XML attribute structure. - struct xml_attribute_struct +#ifdef PUGIXML_COMPACT +PUGI__NS_BEGIN + static const uintptr_t compact_alignment_log2 = 2; + static const uintptr_t compact_alignment = 1 << compact_alignment_log2; + + class compact_header { - /// Default ctor - xml_attribute_struct(xml_memory_page* page): header(reinterpret_cast(page)), name(0), value(0), prev_attribute_c(0), next_attribute(0) + public: + compact_header(xml_memory_page* page, unsigned int flags) { - } - - uintptr_t header; + PUGI__STATIC_ASSERT(xml_memory_block_alignment == compact_alignment); - char_t* name; ///< Pointer to attribute name. - char_t* value; ///< Pointer to attribute value. + ptrdiff_t offset = (reinterpret_cast(this) - reinterpret_cast(page->compact_page_marker)); + assert(offset % compact_alignment == 0 && static_cast(offset) < 256 * compact_alignment); - xml_attribute_struct* prev_attribute_c; ///< Previous attribute (cyclic list) - xml_attribute_struct* next_attribute; ///< Next attribute - }; + _page = static_cast(offset >> compact_alignment_log2); + _flags = static_cast(flags); + } - /// An XML document tree node. - struct xml_node_struct - { - /// Default ctor - /// \param type - node type - xml_node_struct(xml_memory_page* page, xml_node_type type): header(reinterpret_cast(page) | (type - 1)), parent(0), name(0), value(0), first_child(0), prev_sibling_c(0), next_sibling(0), first_attribute(0) + void operator&=(uintptr_t mod) { + _flags &= static_cast(mod); } - uintptr_t header; - - xml_node_struct* parent; ///< Pointer to parent - - char_t* name; ///< Pointer to element name. - char_t* value; ///< Pointer to any associated string data. - - xml_node_struct* first_child; ///< First child - - xml_node_struct* prev_sibling_c; ///< Left brother (cyclic list) - xml_node_struct* next_sibling; ///< Right brother - - xml_attribute_struct* first_attribute; ///< First attribute - }; -} + void operator|=(uintptr_t mod) + { + _flags |= static_cast(mod); + } -namespace -{ - struct xml_document_struct: public xml_node_struct, public xml_allocator - { - xml_document_struct(xml_memory_page* page): xml_node_struct(page, node_document), xml_allocator(page), buffer(0) + uintptr_t operator&(uintptr_t mod) const { + return _flags & mod; } - const char_t* buffer; - }; + xml_memory_page* get_page() const + { + // round-trip through void* to silence 'cast increases required alignment of target type' warnings + const char* page_marker = reinterpret_cast(this) - (_page << compact_alignment_log2); + const char* page = page_marker - *reinterpret_cast(static_cast(page_marker)); - static inline xml_allocator& get_allocator(const xml_node_struct* node) - { - assert(node); + return const_cast(reinterpret_cast(static_cast(page))); + } - return *reinterpret_cast(node->header & xml_memory_page_pointer_mask)->allocator; - } -} + private: + unsigned char _page; + unsigned char _flags; + }; -// Low-level DOM operations -namespace -{ - inline xml_attribute_struct* allocate_attribute(xml_allocator& alloc) + PUGI__FN xml_memory_page* compact_get_page(const void* object, int header_offset) { - xml_memory_page* page; - void* memory = alloc.allocate_memory(sizeof(xml_attribute_struct), page); + const compact_header* header = reinterpret_cast(static_cast(object) - header_offset); - return new (memory) xml_attribute_struct(page); + return header->get_page(); } - inline xml_node_struct* allocate_node(xml_allocator& alloc, xml_node_type type) + template PUGI__FN_NO_INLINE T* compact_get_value(const void* object) { - xml_memory_page* page; - void* memory = alloc.allocate_memory(sizeof(xml_node_struct), page); - - return new (memory) xml_node_struct(page, type); + return static_cast(compact_get_page(object, header_offset)->allocator->_hash->find(object)); } - inline void destroy_attribute(xml_attribute_struct* a, xml_allocator& alloc) + template PUGI__FN_NO_INLINE void compact_set_value(const void* object, T* value) { - uintptr_t header = a->header; - - if (header & xml_memory_page_name_allocated_mask) alloc.deallocate_string(a->name); - if (header & xml_memory_page_value_allocated_mask) alloc.deallocate_string(a->value); - - alloc.deallocate_memory(a, sizeof(xml_attribute_struct), reinterpret_cast(header & xml_memory_page_pointer_mask)); + compact_get_page(object, header_offset)->allocator->_hash->insert(object, value); } - inline void destroy_node(xml_node_struct* n, xml_allocator& alloc) + template class compact_pointer { - uintptr_t header = n->header; + public: + compact_pointer(): _data(0) + { + } - if (header & xml_memory_page_name_allocated_mask) alloc.deallocate_string(n->name); - if (header & xml_memory_page_value_allocated_mask) alloc.deallocate_string(n->value); + void operator=(const compact_pointer& rhs) + { + *this = rhs + 0; + } - for (xml_attribute_struct* attr = n->first_attribute; attr; ) + void operator=(T* value) { - xml_attribute_struct* next = attr->next_attribute; + if (value) + { + // value is guaranteed to be compact-aligned; 'this' is not + // our decoding is based on 'this' aligned to compact alignment downwards (see operator T*) + // so for negative offsets (e.g. -3) we need to adjust the diff by compact_alignment - 1 to + // compensate for arithmetic shift rounding for negative values + ptrdiff_t diff = reinterpret_cast(value) - reinterpret_cast(this); + ptrdiff_t offset = ((diff + int(compact_alignment - 1)) >> compact_alignment_log2) - start; - destroy_attribute(attr, alloc); + if (static_cast(offset) <= 253) + _data = static_cast(offset + 1); + else + { + compact_set_value(this, value); - attr = next; + _data = 255; + } + } + else + _data = 0; } - for (xml_node_struct* child = n->first_child; child; ) + operator T*() const { - xml_node_struct* next = child->next_sibling; - - destroy_node(child, alloc); + if (_data) + { + if (_data < 255) + { + uintptr_t base = reinterpret_cast(this) & ~(compact_alignment - 1); - child = next; + return reinterpret_cast(base + (_data - 1 + start) * compact_alignment); + } + else + return compact_get_value(this); + } + else + return 0; } - alloc.deallocate_memory(n, sizeof(xml_node_struct), reinterpret_cast(header & xml_memory_page_pointer_mask)); - } - - PUGIXML_NO_INLINE xml_node_struct* append_node(xml_node_struct* node, xml_allocator& alloc, xml_node_type type = node_element) - { - xml_node_struct* child = allocate_node(alloc, type); - if (!child) return 0; + T* operator->() const + { + return *this; + } - child->parent = node; + private: + unsigned char _data; + }; - xml_node_struct* first_child = node->first_child; - - if (first_child) + template class compact_pointer_parent + { + public: + compact_pointer_parent(): _data(0) { - xml_node_struct* last_child = first_child->prev_sibling_c; - - last_child->next_sibling = child; - child->prev_sibling_c = last_child; - first_child->prev_sibling_c = child; } - else + + void operator=(const compact_pointer_parent& rhs) { - node->first_child = child; - child->prev_sibling_c = child; + *this = rhs + 0; } - - return child; - } - PUGIXML_NO_INLINE xml_attribute_struct* append_attribute_ll(xml_node_struct* node, xml_allocator& alloc) - { - xml_attribute_struct* a = allocate_attribute(alloc); - if (!a) return 0; + void operator=(T* value) + { + if (value) + { + // value is guaranteed to be compact-aligned; 'this' is not + // our decoding is based on 'this' aligned to compact alignment downwards (see operator T*) + // so for negative offsets (e.g. -3) we need to adjust the diff by compact_alignment - 1 to + // compensate for arithmetic shift behavior for negative values + ptrdiff_t diff = reinterpret_cast(value) - reinterpret_cast(this); + ptrdiff_t offset = ((diff + int(compact_alignment - 1)) >> compact_alignment_log2) + 65533; + + if (static_cast(offset) <= 65533) + { + _data = static_cast(offset + 1); + } + else + { + xml_memory_page* page = compact_get_page(this, header_offset); - xml_attribute_struct* first_attribute = node->first_attribute; + if (PUGI__UNLIKELY(page->compact_shared_parent == 0)) + page->compact_shared_parent = value; - if (first_attribute) - { - xml_attribute_struct* last_attribute = first_attribute->prev_attribute_c; + if (page->compact_shared_parent == value) + { + _data = 65534; + } + else + { + compact_set_value(this, value); - last_attribute->next_attribute = a; - a->prev_attribute_c = last_attribute; - first_attribute->prev_attribute_c = a; + _data = 65535; + } + } + } + else + { + _data = 0; + } } - else + + operator T*() const { - node->first_attribute = a; - a->prev_attribute_c = a; + if (_data) + { + if (_data < 65534) + { + uintptr_t base = reinterpret_cast(this) & ~(compact_alignment - 1); + + return reinterpret_cast(base + (_data - 1 - 65533) * compact_alignment); + } + else if (_data == 65534) + return static_cast(compact_get_page(this, header_offset)->compact_shared_parent); + else + return compact_get_value(this); + } + else + return 0; } - - return a; - } -} -// Helper classes for code generation -namespace -{ - struct opt_false - { - enum { value = 0 }; - }; + T* operator->() const + { + return *this; + } - struct opt_true - { - enum { value = 1 }; + private: + uint16_t _data; }; -} -// Unicode utilities -namespace -{ - inline uint16_t endian_swap(uint16_t value) - { - return static_cast(((value & 0xff) << 8) | (value >> 8)); - } - - inline uint32_t endian_swap(uint32_t value) + template class compact_string { - return ((value & 0xff) << 24) | ((value & 0xff00) << 8) | ((value & 0xff0000) >> 8) | (value >> 24); - } - - struct utf8_counter - { - typedef size_t value_type; - - static value_type low(value_type result, uint32_t ch) + public: + compact_string(): _data(0) { - // U+0000..U+007F - if (ch < 0x80) return result + 1; - // U+0080..U+07FF - else if (ch < 0x800) return result + 2; - // U+0800..U+FFFF - else return result + 3; } - static value_type high(value_type result, uint32_t) + void operator=(const compact_string& rhs) { - // U+10000..U+10FFFF - return result + 4; + *this = rhs + 0; } - }; - - struct utf8_writer - { - typedef uint8_t* value_type; - static value_type low(value_type result, uint32_t ch) + void operator=(char_t* value) { - // U+0000..U+007F - if (ch < 0x80) - { - *result = static_cast(ch); - return result + 1; - } - // U+0080..U+07FF - else if (ch < 0x800) + if (value) { - result[0] = static_cast(0xC0 | (ch >> 6)); - result[1] = static_cast(0x80 | (ch & 0x3F)); - return result + 2; + xml_memory_page* page = compact_get_page(this, header_offset); + + if (PUGI__UNLIKELY(page->compact_string_base == 0)) + page->compact_string_base = value; + + ptrdiff_t offset = value - page->compact_string_base; + + if (static_cast(offset) < (65535 << 7)) + { + // round-trip through void* to silence 'cast increases required alignment of target type' warnings + uint16_t* base = reinterpret_cast(static_cast(reinterpret_cast(this) - base_offset)); + + if (*base == 0) + { + *base = static_cast((offset >> 7) + 1); + _data = static_cast((offset & 127) + 1); + } + else + { + ptrdiff_t remainder = offset - ((*base - 1) << 7); + + if (static_cast(remainder) <= 253) + { + _data = static_cast(remainder + 1); + } + else + { + compact_set_value(this, value); + + _data = 255; + } + } + } + else + { + compact_set_value(this, value); + + _data = 255; + } } - // U+0800..U+FFFF else { - result[0] = static_cast(0xE0 | (ch >> 12)); - result[1] = static_cast(0x80 | ((ch >> 6) & 0x3F)); - result[2] = static_cast(0x80 | (ch & 0x3F)); - return result + 3; + _data = 0; } } - static value_type high(value_type result, uint32_t ch) + operator char_t*() const { - // U+10000..U+10FFFF - result[0] = static_cast(0xF0 | (ch >> 18)); - result[1] = static_cast(0x80 | ((ch >> 12) & 0x3F)); - result[2] = static_cast(0x80 | ((ch >> 6) & 0x3F)); - result[3] = static_cast(0x80 | (ch & 0x3F)); - return result + 4; - } + if (_data) + { + if (_data < 255) + { + xml_memory_page* page = compact_get_page(this, header_offset); - static value_type any(value_type result, uint32_t ch) - { - return (ch < 0x10000) ? low(result, ch) : high(result, ch); - } - }; + // round-trip through void* to silence 'cast increases required alignment of target type' warnings + const uint16_t* base = reinterpret_cast(static_cast(reinterpret_cast(this) - base_offset)); + assert(*base); - struct utf16_counter - { - typedef size_t value_type; + ptrdiff_t offset = ((*base - 1) << 7) + (_data - 1); - static value_type low(value_type result, uint32_t) - { - return result + 1; + return page->compact_string_base + offset; + } + else + { + return compact_get_value(this); + } + } + else + return 0; } - static value_type high(value_type result, uint32_t) - { - return result + 2; - } + private: + unsigned char _data; }; +PUGI__NS_END +#endif - struct utf16_writer +#ifdef PUGIXML_COMPACT +namespace pugi +{ + struct xml_attribute_struct { - typedef uint16_t* value_type; - - static value_type low(value_type result, uint32_t ch) + xml_attribute_struct(impl::xml_memory_page* page): header(page, 0), namevalue_base(0) { - *result = static_cast(ch); - - return result + 1; + PUGI__STATIC_ASSERT(sizeof(xml_attribute_struct) == 8); } - static value_type high(value_type result, uint32_t ch) - { - uint32_t msh = (uint32_t)(ch - 0x10000) >> 10; - uint32_t lsh = (uint32_t)(ch - 0x10000) & 0x3ff; + impl::compact_header header; - result[0] = static_cast(0xD800 + msh); - result[1] = static_cast(0xDC00 + lsh); + uint16_t namevalue_base; - return result + 2; - } + impl::compact_string<4, 2> name; + impl::compact_string<5, 3> value; - static value_type any(value_type result, uint32_t ch) - { - return (ch < 0x10000) ? low(result, ch) : high(result, ch); - } + impl::compact_pointer prev_attribute_c; + impl::compact_pointer next_attribute; }; - struct utf32_counter + struct xml_node_struct { - typedef size_t value_type; - - static value_type low(value_type result, uint32_t) + xml_node_struct(impl::xml_memory_page* page, xml_node_type type): header(page, type), namevalue_base(0) { - return result + 1; + PUGI__STATIC_ASSERT(sizeof(xml_node_struct) == 12); } - static value_type high(value_type result, uint32_t) + impl::compact_header header; + + uint16_t namevalue_base; + + impl::compact_string<4, 2> name; + impl::compact_string<5, 3> value; + + impl::compact_pointer_parent parent; + + impl::compact_pointer first_child; + + impl::compact_pointer prev_sibling_c; + impl::compact_pointer next_sibling; + + impl::compact_pointer first_attribute; + }; +} +#else +namespace pugi +{ + struct xml_attribute_struct + { + xml_attribute_struct(impl::xml_memory_page* page): name(0), value(0), prev_attribute_c(0), next_attribute(0) { - return result + 1; + header = PUGI__GETHEADER_IMPL(this, page, 0); } + + uintptr_t header; + + char_t* name; + char_t* value; + + xml_attribute_struct* prev_attribute_c; + xml_attribute_struct* next_attribute; }; - struct utf32_writer + struct xml_node_struct { - typedef uint32_t* value_type; - - static value_type low(value_type result, uint32_t ch) + xml_node_struct(impl::xml_memory_page* page, xml_node_type type): name(0), value(0), parent(0), first_child(0), prev_sibling_c(0), next_sibling(0), first_attribute(0) { - *result = ch; - - return result + 1; + header = PUGI__GETHEADER_IMPL(this, page, type); } - static value_type high(value_type result, uint32_t ch) - { - *result = ch; + uintptr_t header; - return result + 1; - } + char_t* name; + char_t* value; - static value_type any(value_type result, uint32_t ch) - { - *result = ch; + xml_node_struct* parent; - return result + 1; - } - }; + xml_node_struct* first_child; - template struct wchar_selector; + xml_node_struct* prev_sibling_c; + xml_node_struct* next_sibling; - template <> struct wchar_selector<2> - { - typedef uint16_t type; - typedef utf16_counter counter; - typedef utf16_writer writer; + xml_attribute_struct* first_attribute; }; +} +#endif - template <> struct wchar_selector<4> +PUGI__NS_BEGIN + struct xml_extra_buffer { - typedef uint32_t type; - typedef utf32_counter counter; - typedef utf32_writer writer; + char_t* buffer; + xml_extra_buffer* next; }; - typedef wchar_selector::counter wchar_counter; - typedef wchar_selector::writer wchar_writer; - - template struct utf_decoder + struct xml_document_struct: public xml_node_struct, public xml_allocator { - static inline typename Traits::value_type decode_utf8_block(const uint8_t* data, size_t size, typename Traits::value_type result) + xml_document_struct(xml_memory_page* page): xml_node_struct(page, node_document), xml_allocator(page), buffer(0), extra_buffers(0) { - const uint8_t utf8_byte_mask = 0x3f; + } - while (size) - { - uint8_t lead = *data; + const char_t* buffer; - // 0xxxxxxx -> U+0000..U+007F - if (lead < 0x80) - { - result = Traits::low(result, lead); - data += 1; - size -= 1; + xml_extra_buffer* extra_buffers; - // process aligned single-byte (ascii) blocks - if ((reinterpret_cast(data) & 3) == 0) - { - while (size >= 4 && (*reinterpret_cast(data) & 0x80808080) == 0) - { - result = Traits::low(result, data[0]); - result = Traits::low(result, data[1]); - result = Traits::low(result, data[2]); - result = Traits::low(result, data[3]); - data += 4; - size -= 4; - } - } - } - // 110xxxxx -> U+0080..U+07FF - else if ((unsigned)(lead - 0xC0) < 0x20 && size >= 2 && (data[1] & 0xc0) == 0x80) - { - result = Traits::low(result, ((lead & ~0xC0) << 6) | (data[1] & utf8_byte_mask)); - data += 2; - size -= 2; - } - // 1110xxxx -> U+0800-U+FFFF - else if ((unsigned)(lead - 0xE0) < 0x10 && size >= 3 && (data[1] & 0xc0) == 0x80 && (data[2] & 0xc0) == 0x80) - { - result = Traits::low(result, ((lead & ~0xE0) << 12) | ((data[1] & utf8_byte_mask) << 6) | (data[2] & utf8_byte_mask)); - data += 3; - size -= 3; - } - // 11110xxx -> U+10000..U+10FFFF - else if ((unsigned)(lead - 0xF0) < 0x08 && size >= 4 && (data[1] & 0xc0) == 0x80 && (data[2] & 0xc0) == 0x80 && (data[3] & 0xc0) == 0x80) - { - result = Traits::high(result, ((lead & ~0xF0) << 18) | ((data[1] & utf8_byte_mask) << 12) | ((data[2] & utf8_byte_mask) << 6) | (data[3] & utf8_byte_mask)); - data += 4; - size -= 4; - } - // 10xxxxxx or 11111xxx -> invalid - else - { - data += 1; - size -= 1; - } - } + #ifdef PUGIXML_COMPACT + compact_hash_table hash; + #endif + }; - return result; - } + template inline xml_allocator& get_allocator(const Object* object) + { + assert(object); - static inline typename Traits::value_type decode_utf16_block(const uint16_t* data, size_t size, typename Traits::value_type result) - { - const uint16_t* end = data + size; + return *PUGI__GETPAGE(object)->allocator; + } - while (data < end) - { - uint16_t lead = opt_swap::value ? endian_swap(*data) : *data; + template inline xml_document_struct& get_document(const Object* object) + { + assert(object); - // U+0000..U+D7FF - if (lead < 0xD800) - { - result = Traits::low(result, lead); - data += 1; - } - // U+E000..U+FFFF - else if ((unsigned)(lead - 0xE000) < 0x2000) - { - result = Traits::low(result, lead); - data += 1; - } - // surrogate pair lead - else if ((unsigned)(lead - 0xD800) < 0x400 && data + 1 < end) - { - uint16_t next = opt_swap::value ? endian_swap(data[1]) : data[1]; + return *static_cast(PUGI__GETPAGE(object)->allocator); + } +PUGI__NS_END - if ((unsigned)(next - 0xDC00) < 0x400) - { - result = Traits::high(result, 0x10000 + ((lead & 0x3ff) << 10) + (next & 0x3ff)); - data += 2; - } - else - { - data += 1; - } - } - else - { - data += 1; - } - } +// Low-level DOM operations +PUGI__NS_BEGIN + inline xml_attribute_struct* allocate_attribute(xml_allocator& alloc) + { + xml_memory_page* page; + void* memory = alloc.allocate_object(sizeof(xml_attribute_struct), page); + if (!memory) return 0; - return result; - } + return new (memory) xml_attribute_struct(page); + } - static inline typename Traits::value_type decode_utf32_block(const uint32_t* data, size_t size, typename Traits::value_type result) - { - const uint32_t* end = data + size; + inline xml_node_struct* allocate_node(xml_allocator& alloc, xml_node_type type) + { + xml_memory_page* page; + void* memory = alloc.allocate_object(sizeof(xml_node_struct), page); + if (!memory) return 0; - while (data < end) - { - uint32_t lead = opt_swap::value ? endian_swap(*data) : *data; + return new (memory) xml_node_struct(page, type); + } - // U+0000..U+FFFF - if (lead < 0x10000) - { - result = Traits::low(result, lead); - data += 1; - } - // U+10000..U+10FFFF - else - { - result = Traits::high(result, lead); - data += 1; - } - } + inline void destroy_attribute(xml_attribute_struct* a, xml_allocator& alloc) + { + if (a->header & impl::xml_memory_page_name_allocated_mask) + alloc.deallocate_string(a->name); - return result; - } - }; + if (a->header & impl::xml_memory_page_value_allocated_mask) + alloc.deallocate_string(a->value); - template inline void convert_utf_endian_swap(T* result, const T* data, size_t length) - { - for (size_t i = 0; i < length; ++i) result[i] = endian_swap(data[i]); + alloc.deallocate_memory(a, sizeof(xml_attribute_struct), PUGI__GETPAGE(a)); } - inline void convert_wchar_endian_swap(wchar_t* result, const wchar_t* data, size_t length) + inline void destroy_node(xml_node_struct* n, xml_allocator& alloc) { - for (size_t i = 0; i < length; ++i) result[i] = static_cast(endian_swap(static_cast::type>(data[i]))); - } -} + if (n->header & impl::xml_memory_page_name_allocated_mask) + alloc.deallocate_string(n->name); -namespace -{ - enum chartype_t - { - ct_parse_pcdata = 1, // \0, &, \r, < - ct_parse_attr = 2, // \0, &, \r, ', " - ct_parse_attr_ws = 4, // \0, &, \r, ', ", \n, tab - ct_space = 8, // \r, \n, space, tab - ct_parse_cdata = 16, // \0, ], >, \r - ct_parse_comment = 32, // \0, -, >, \r - ct_symbol = 64, // Any symbol > 127, a-z, A-Z, 0-9, _, :, -, . - ct_start_symbol = 128 // Any symbol > 127, a-z, A-Z, _, : - }; + if (n->header & impl::xml_memory_page_value_allocated_mask) + alloc.deallocate_string(n->value); - const unsigned char chartype_table[256] = - { - 55, 0, 0, 0, 0, 0, 0, 0, 0, 12, 12, 0, 0, 63, 0, 0, // 0-15 - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 16-31 - 8, 0, 6, 0, 0, 0, 7, 6, 0, 0, 0, 0, 0, 96, 64, 0, // 32-47 - 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 192, 0, 1, 0, 48, 0, // 48-63 - 0, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, // 64-79 - 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 0, 0, 16, 0, 192, // 80-95 - 0, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, // 96-111 - 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 0, 0, 0, 0, 0, // 112-127 + for (xml_attribute_struct* attr = n->first_attribute; attr; ) + { + xml_attribute_struct* next = attr->next_attribute; - 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, // 128+ - 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, - 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, - 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, - 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, - 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, - 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, - 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192 - }; + destroy_attribute(attr, alloc); - enum chartypex_t - { - ctx_special_pcdata = 1, // Any symbol >= 0 and < 32 (except \t, \r, \n), &, <, > - ctx_special_attr = 2, // Any symbol >= 0 and < 32 (except \t), &, <, >, " - ctx_start_symbol = 4, // Any symbol > 127, a-z, A-Z, _ - ctx_digit = 8, // 0-9 - ctx_symbol = 16 // Any symbol > 127, a-z, A-Z, 0-9, _, -, . - }; - - const unsigned char chartypex_table[256] = - { - 3, 3, 3, 3, 3, 3, 3, 3, 3, 0, 2, 3, 3, 2, 3, 3, // 0-15 - 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, // 16-31 - 0, 0, 2, 0, 0, 0, 3, 0, 0, 0, 0, 0, 0, 16, 16, 0, // 32-47 - 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 0, 0, 3, 0, 3, 0, // 48-63 + attr = next; + } - 0, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, // 64-79 - 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 0, 0, 0, 0, 20, // 80-95 - 0, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, // 96-111 - 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 0, 0, 0, 0, 0, // 112-127 + for (xml_node_struct* child = n->first_child; child; ) + { + xml_node_struct* next = child->next_sibling; - 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, // 128+ - 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, - 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, - 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, - 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, - 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, - 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, - 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20 - }; - -#ifdef PUGIXML_WCHAR_MODE - #define IS_CHARTYPE_IMPL(c, ct, table) ((static_cast(c) < 128 ? table[static_cast(c)] : table[128]) & (ct)) -#else - #define IS_CHARTYPE_IMPL(c, ct, table) (table[static_cast(c)] & (ct)) -#endif + destroy_node(child, alloc); + + child = next; + } - #define IS_CHARTYPE(c, ct) IS_CHARTYPE_IMPL(c, ct, chartype_table) - #define IS_CHARTYPEX(c, ct) IS_CHARTYPE_IMPL(c, ct, chartypex_table) + alloc.deallocate_memory(n, sizeof(xml_node_struct), PUGI__GETPAGE(n)); + } - bool is_little_endian() + inline void append_node(xml_node_struct* child, xml_node_struct* node) { - unsigned int ui = 1; + child->parent = node; - return *reinterpret_cast(&ui) == 1; + xml_node_struct* head = node->first_child; + + if (head) + { + xml_node_struct* tail = head->prev_sibling_c; + + tail->next_sibling = child; + child->prev_sibling_c = tail; + head->prev_sibling_c = child; + } + else + { + node->first_child = child; + child->prev_sibling_c = child; + } } - xml_encoding get_wchar_encoding() + inline void prepend_node(xml_node_struct* child, xml_node_struct* node) { - STATIC_ASSERT(sizeof(wchar_t) == 2 || sizeof(wchar_t) == 4); + child->parent = node; - if (sizeof(wchar_t) == 2) - return is_little_endian() ? encoding_utf16_le : encoding_utf16_be; - else - return is_little_endian() ? encoding_utf32_le : encoding_utf32_be; + xml_node_struct* head = node->first_child; + + if (head) + { + child->prev_sibling_c = head->prev_sibling_c; + head->prev_sibling_c = child; + } + else + child->prev_sibling_c = child; + + child->next_sibling = head; + node->first_child = child; } - xml_encoding guess_buffer_encoding(uint8_t d0, uint8_t d1, uint8_t d2, uint8_t d3) + inline void insert_node_after(xml_node_struct* child, xml_node_struct* node) { - // look for BOM in first few bytes - if (d0 == 0 && d1 == 0 && d2 == 0xfe && d3 == 0xff) return encoding_utf32_be; - if (d0 == 0xff && d1 == 0xfe && d2 == 0 && d3 == 0) return encoding_utf32_le; - if (d0 == 0xfe && d1 == 0xff) return encoding_utf16_be; - if (d0 == 0xff && d1 == 0xfe) return encoding_utf16_le; - if (d0 == 0xef && d1 == 0xbb && d2 == 0xbf) return encoding_utf8; + xml_node_struct* parent = node->parent; - // look for <, parent = parent; - // look for utf16 < followed by node name (this may fail, but is better than utf8 since it's zero terminated so early) - if (d0 == 0 && d1 == 0x3c) return encoding_utf16_be; - if (d0 == 0x3c && d1 == 0) return encoding_utf16_le; + if (node->next_sibling) + node->next_sibling->prev_sibling_c = child; + else + parent->first_child->prev_sibling_c = child; - // no known BOM detected, assume utf8 - return encoding_utf8; + child->next_sibling = node->next_sibling; + child->prev_sibling_c = node; + + node->next_sibling = child; } - xml_encoding get_buffer_encoding(xml_encoding encoding, const void* contents, size_t size) + inline void insert_node_before(xml_node_struct* child, xml_node_struct* node) { - // replace wchar encoding with utf implementation - if (encoding == encoding_wchar) return get_wchar_encoding(); + xml_node_struct* parent = node->parent; - // replace utf16 encoding with utf16 with specific endianness - if (encoding == encoding_utf16) return is_little_endian() ? encoding_utf16_le : encoding_utf16_be; + child->parent = parent; - // replace utf32 encoding with utf32 with specific endianness - if (encoding == encoding_utf32) return is_little_endian() ? encoding_utf32_le : encoding_utf32_be; + if (node->prev_sibling_c->next_sibling) + node->prev_sibling_c->next_sibling = child; + else + parent->first_child = child; - // only do autodetection if no explicit encoding is requested - if (encoding != encoding_auto) return encoding; + child->prev_sibling_c = node->prev_sibling_c; + child->next_sibling = node; - // skip encoding autodetection if input buffer is too small - if (size < 4) return encoding_utf8; + node->prev_sibling_c = child; + } - // try to guess encoding (based on XML specification, Appendix F.1) - const uint8_t* data = static_cast(contents); + inline void remove_node(xml_node_struct* node) + { + xml_node_struct* parent = node->parent; + + if (node->next_sibling) + node->next_sibling->prev_sibling_c = node->prev_sibling_c; + else + parent->first_child->prev_sibling_c = node->prev_sibling_c; - DMC_VOLATILE uint8_t d0 = data[0], d1 = data[1], d2 = data[2], d3 = data[3]; + if (node->prev_sibling_c->next_sibling) + node->prev_sibling_c->next_sibling = node->next_sibling; + else + parent->first_child = node->next_sibling; - return guess_buffer_encoding(d0, d1, d2, d3); + node->parent = 0; + node->prev_sibling_c = 0; + node->next_sibling = 0; } - bool get_mutable_buffer(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size, bool is_mutable) + inline void append_attribute(xml_attribute_struct* attr, xml_node_struct* node) { - if (is_mutable) - { - out_buffer = static_cast(const_cast(contents)); + xml_attribute_struct* head = node->first_attribute; + + if (head) + { + xml_attribute_struct* tail = head->prev_attribute_c; + + tail->next_attribute = attr; + attr->prev_attribute_c = tail; + head->prev_attribute_c = attr; } else { - void* buffer = global_allocate(size > 0 ? size : 1); - if (!buffer) return false; - - memcpy(buffer, contents, size); - - out_buffer = static_cast(buffer); + node->first_attribute = attr; + attr->prev_attribute_c = attr; } - - out_length = size / sizeof(char_t); - - return true; } -#ifdef PUGIXML_WCHAR_MODE - inline bool need_endian_swap_utf(xml_encoding le, xml_encoding re) + inline void prepend_attribute(xml_attribute_struct* attr, xml_node_struct* node) { - return (le == encoding_utf16_be && re == encoding_utf16_le) || (le == encoding_utf16_le && re == encoding_utf16_be) || - (le == encoding_utf32_be && re == encoding_utf32_le) || (le == encoding_utf32_le && re == encoding_utf32_be); - } + xml_attribute_struct* head = node->first_attribute; - bool convert_buffer_endian_swap(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size, bool is_mutable) - { - const char_t* data = static_cast(contents); - - if (is_mutable) + if (head) { - out_buffer = const_cast(data); + attr->prev_attribute_c = head->prev_attribute_c; + head->prev_attribute_c = attr; } else - { - out_buffer = static_cast(global_allocate(size > 0 ? size : 1)); - if (!out_buffer) return false; - } + attr->prev_attribute_c = attr; - out_length = size / sizeof(char_t); + attr->next_attribute = head; + node->first_attribute = attr; + } - convert_wchar_endian_swap(out_buffer, data, out_length); + inline void insert_attribute_after(xml_attribute_struct* attr, xml_attribute_struct* place, xml_node_struct* node) + { + if (place->next_attribute) + place->next_attribute->prev_attribute_c = attr; + else + node->first_attribute->prev_attribute_c = attr; - return true; + attr->next_attribute = place->next_attribute; + attr->prev_attribute_c = place; + place->next_attribute = attr; } - bool convert_buffer_utf8(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size) + inline void insert_attribute_before(xml_attribute_struct* attr, xml_attribute_struct* place, xml_node_struct* node) { - const uint8_t* data = static_cast(contents); - - // first pass: get length in wchar_t units - out_length = utf_decoder::decode_utf8_block(data, size, 0); + if (place->prev_attribute_c->next_attribute) + place->prev_attribute_c->next_attribute = attr; + else + node->first_attribute = attr; - // allocate buffer of suitable length - out_buffer = static_cast(global_allocate((out_length > 0 ? out_length : 1) * sizeof(char_t))); - if (!out_buffer) return false; + attr->prev_attribute_c = place->prev_attribute_c; + attr->next_attribute = place; + place->prev_attribute_c = attr; + } - // second pass: convert utf8 input to wchar_t - wchar_writer::value_type out_begin = reinterpret_cast(out_buffer); - wchar_writer::value_type out_end = utf_decoder::decode_utf8_block(data, size, out_begin); + inline void remove_attribute(xml_attribute_struct* attr, xml_node_struct* node) + { + if (attr->next_attribute) + attr->next_attribute->prev_attribute_c = attr->prev_attribute_c; + else + node->first_attribute->prev_attribute_c = attr->prev_attribute_c; - assert(out_end == out_begin + out_length); - (void)!out_end; + if (attr->prev_attribute_c->next_attribute) + attr->prev_attribute_c->next_attribute = attr->next_attribute; + else + node->first_attribute = attr->next_attribute; - return true; + attr->prev_attribute_c = 0; + attr->next_attribute = 0; } - template bool convert_buffer_utf16(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size, opt_swap) + PUGI__FN_NO_INLINE xml_node_struct* append_new_node(xml_node_struct* node, xml_allocator& alloc, xml_node_type type = node_element) { - const uint16_t* data = static_cast(contents); - size_t length = size / sizeof(uint16_t); - - // first pass: get length in wchar_t units - out_length = utf_decoder::decode_utf16_block(data, length, 0); + if (!alloc.reserve()) return 0; - // allocate buffer of suitable length - out_buffer = static_cast(global_allocate((out_length > 0 ? out_length : 1) * sizeof(char_t))); - if (!out_buffer) return false; - - // second pass: convert utf16 input to wchar_t - wchar_writer::value_type out_begin = reinterpret_cast(out_buffer); - wchar_writer::value_type out_end = utf_decoder::decode_utf16_block(data, length, out_begin); + xml_node_struct* child = allocate_node(alloc, type); + if (!child) return 0; - assert(out_end == out_begin + out_length); - (void)!out_end; + append_node(child, node); - return true; + return child; } - template bool convert_buffer_utf32(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size, opt_swap) + PUGI__FN_NO_INLINE xml_attribute_struct* append_new_attribute(xml_node_struct* node, xml_allocator& alloc) { - const uint32_t* data = static_cast(contents); - size_t length = size / sizeof(uint32_t); - - // first pass: get length in wchar_t units - out_length = utf_decoder::decode_utf32_block(data, length, 0); - - // allocate buffer of suitable length - out_buffer = static_cast(global_allocate((out_length > 0 ? out_length : 1) * sizeof(char_t))); - if (!out_buffer) return false; + if (!alloc.reserve()) return 0; - // second pass: convert utf32 input to wchar_t - wchar_writer::value_type out_begin = reinterpret_cast(out_buffer); - wchar_writer::value_type out_end = utf_decoder::decode_utf32_block(data, length, out_begin); + xml_attribute_struct* attr = allocate_attribute(alloc); + if (!attr) return 0; - assert(out_end == out_begin + out_length); - (void)!out_end; + append_attribute(attr, node); - return true; + return attr; } +PUGI__NS_END - bool convert_buffer(char_t*& out_buffer, size_t& out_length, xml_encoding encoding, const void* contents, size_t size, bool is_mutable) +// Helper classes for code generation +PUGI__NS_BEGIN + struct opt_false { - // get native encoding - xml_encoding wchar_encoding = get_wchar_encoding(); + enum { value = 0 }; + }; - // fast path: no conversion required - if (encoding == wchar_encoding) return get_mutable_buffer(out_buffer, out_length, contents, size, is_mutable); + struct opt_true + { + enum { value = 1 }; + }; +PUGI__NS_END - // only endian-swapping is required - if (need_endian_swap_utf(encoding, wchar_encoding)) return convert_buffer_endian_swap(out_buffer, out_length, contents, size, is_mutable); +// Unicode utilities +PUGI__NS_BEGIN + inline uint16_t endian_swap(uint16_t value) + { + return static_cast(((value & 0xff) << 8) | (value >> 8)); + } - // source encoding is utf8 - if (encoding == encoding_utf8) return convert_buffer_utf8(out_buffer, out_length, contents, size); + inline uint32_t endian_swap(uint32_t value) + { + return ((value & 0xff) << 24) | ((value & 0xff00) << 8) | ((value & 0xff0000) >> 8) | (value >> 24); + } - // source encoding is utf16 - if (encoding == encoding_utf16_be || encoding == encoding_utf16_le) - { - xml_encoding native_encoding = is_little_endian() ? encoding_utf16_le : encoding_utf16_be; + struct utf8_counter + { + typedef size_t value_type; - return (native_encoding == encoding) ? - convert_buffer_utf16(out_buffer, out_length, contents, size, opt_false()) : - convert_buffer_utf16(out_buffer, out_length, contents, size, opt_true()); + static value_type low(value_type result, uint32_t ch) + { + // U+0000..U+007F + if (ch < 0x80) return result + 1; + // U+0080..U+07FF + else if (ch < 0x800) return result + 2; + // U+0800..U+FFFF + else return result + 3; } - // source encoding is utf32 - if (encoding == encoding_utf32_be || encoding == encoding_utf32_le) + static value_type high(value_type result, uint32_t) { - xml_encoding native_encoding = is_little_endian() ? encoding_utf32_le : encoding_utf32_be; - - return (native_encoding == encoding) ? - convert_buffer_utf32(out_buffer, out_length, contents, size, opt_false()) : - convert_buffer_utf32(out_buffer, out_length, contents, size, opt_true()); + // U+10000..U+10FFFF + return result + 4; } + }; - assert(!"Invalid encoding"); - return false; - } -#else - template bool convert_buffer_utf16(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size, opt_swap) + struct utf8_writer { - const uint16_t* data = static_cast(contents); - size_t length = size / sizeof(uint16_t); + typedef uint8_t* value_type; - // first pass: get length in utf8 units - out_length = utf_decoder::decode_utf16_block(data, length, 0); + static value_type low(value_type result, uint32_t ch) + { + // U+0000..U+007F + if (ch < 0x80) + { + *result = static_cast(ch); + return result + 1; + } + // U+0080..U+07FF + else if (ch < 0x800) + { + result[0] = static_cast(0xC0 | (ch >> 6)); + result[1] = static_cast(0x80 | (ch & 0x3F)); + return result + 2; + } + // U+0800..U+FFFF + else + { + result[0] = static_cast(0xE0 | (ch >> 12)); + result[1] = static_cast(0x80 | ((ch >> 6) & 0x3F)); + result[2] = static_cast(0x80 | (ch & 0x3F)); + return result + 3; + } + } - // allocate buffer of suitable length - out_buffer = static_cast(global_allocate((out_length > 0 ? out_length : 1) * sizeof(char_t))); - if (!out_buffer) return false; + static value_type high(value_type result, uint32_t ch) + { + // U+10000..U+10FFFF + result[0] = static_cast(0xF0 | (ch >> 18)); + result[1] = static_cast(0x80 | ((ch >> 12) & 0x3F)); + result[2] = static_cast(0x80 | ((ch >> 6) & 0x3F)); + result[3] = static_cast(0x80 | (ch & 0x3F)); + return result + 4; + } - // second pass: convert utf16 input to utf8 - uint8_t* out_begin = reinterpret_cast(out_buffer); - uint8_t* out_end = utf_decoder::decode_utf16_block(data, length, out_begin); + static value_type any(value_type result, uint32_t ch) + { + return (ch < 0x10000) ? low(result, ch) : high(result, ch); + } + }; + + struct utf16_counter + { + typedef size_t value_type; - assert(out_end == out_begin + out_length); - (void)!out_end; + static value_type low(value_type result, uint32_t) + { + return result + 1; + } - return true; - } + static value_type high(value_type result, uint32_t) + { + return result + 2; + } + }; - template bool convert_buffer_utf32(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size, opt_swap) + struct utf16_writer { - const uint32_t* data = static_cast(contents); - size_t length = size / sizeof(uint32_t); + typedef uint16_t* value_type; - // first pass: get length in utf8 units - out_length = utf_decoder::decode_utf32_block(data, length, 0); + static value_type low(value_type result, uint32_t ch) + { + *result = static_cast(ch); - // allocate buffer of suitable length - out_buffer = static_cast(global_allocate((out_length > 0 ? out_length : 1) * sizeof(char_t))); - if (!out_buffer) return false; + return result + 1; + } + + static value_type high(value_type result, uint32_t ch) + { + uint32_t msh = static_cast(ch - 0x10000) >> 10; + uint32_t lsh = static_cast(ch - 0x10000) & 0x3ff; - // second pass: convert utf32 input to utf8 - uint8_t* out_begin = reinterpret_cast(out_buffer); - uint8_t* out_end = utf_decoder::decode_utf32_block(data, length, out_begin); + result[0] = static_cast(0xD800 + msh); + result[1] = static_cast(0xDC00 + lsh); - assert(out_end == out_begin + out_length); - (void)!out_end; + return result + 2; + } - return true; - } + static value_type any(value_type result, uint32_t ch) + { + return (ch < 0x10000) ? low(result, ch) : high(result, ch); + } + }; - bool convert_buffer(char_t*& out_buffer, size_t& out_length, xml_encoding encoding, const void* contents, size_t size, bool is_mutable) + struct utf32_counter { - // fast path: no conversion required - if (encoding == encoding_utf8) return get_mutable_buffer(out_buffer, out_length, contents, size, is_mutable); + typedef size_t value_type; - // source encoding is utf16 - if (encoding == encoding_utf16_be || encoding == encoding_utf16_le) + static value_type low(value_type result, uint32_t) { - xml_encoding native_encoding = is_little_endian() ? encoding_utf16_le : encoding_utf16_be; - - return (native_encoding == encoding) ? - convert_buffer_utf16(out_buffer, out_length, contents, size, opt_false()) : - convert_buffer_utf16(out_buffer, out_length, contents, size, opt_true()); + return result + 1; } - // source encoding is utf32 - if (encoding == encoding_utf32_be || encoding == encoding_utf32_le) + static value_type high(value_type result, uint32_t) { - xml_encoding native_encoding = is_little_endian() ? encoding_utf32_le : encoding_utf32_be; - - return (native_encoding == encoding) ? - convert_buffer_utf32(out_buffer, out_length, contents, size, opt_false()) : - convert_buffer_utf32(out_buffer, out_length, contents, size, opt_true()); + return result + 1; } + }; - assert(!"Invalid encoding"); - return false; - } -#endif - - size_t as_utf8_begin(const wchar_t* str, size_t length) + struct utf32_writer { - STATIC_ASSERT(sizeof(wchar_t) == 2 || sizeof(wchar_t) == 4); + typedef uint32_t* value_type; - // get length in utf8 characters - return sizeof(wchar_t) == 2 ? - utf_decoder::decode_utf16_block(reinterpret_cast(str), length, 0) : - utf_decoder::decode_utf32_block(reinterpret_cast(str), length, 0); - } - - void as_utf8_end(char* buffer, size_t size, const wchar_t* str, size_t length) - { - STATIC_ASSERT(sizeof(wchar_t) == 2 || sizeof(wchar_t) == 4); - - // convert to utf8 - uint8_t* begin = reinterpret_cast(buffer); - uint8_t* end = sizeof(wchar_t) == 2 ? - utf_decoder::decode_utf16_block(reinterpret_cast(str), length, begin) : - utf_decoder::decode_utf32_block(reinterpret_cast(str), length, begin); - - assert(begin + size == end); - (void)!end; + static value_type low(value_type result, uint32_t ch) + { + *result = ch; - // zero-terminate - buffer[size] = 0; - } - -#ifndef PUGIXML_NO_STL - std::string as_utf8_impl(const wchar_t* str, size_t length) - { - // first pass: get length in utf8 characters - size_t size = as_utf8_begin(str, length); + return result + 1; + } - // allocate resulting string - std::string result; - result.resize(size); + static value_type high(value_type result, uint32_t ch) + { + *result = ch; - // second pass: convert to utf8 - if (size > 0) as_utf8_end(&result[0], size, str, length); + return result + 1; + } + + static value_type any(value_type result, uint32_t ch) + { + *result = ch; - return result; - } + return result + 1; + } + }; - std::wstring as_wide_impl(const char* str, size_t size) + struct latin1_writer { - const uint8_t* data = reinterpret_cast(str); + typedef uint8_t* value_type; - // first pass: get length in wchar_t units - size_t length = utf_decoder::decode_utf8_block(data, size, 0); + static value_type low(value_type result, uint32_t ch) + { + *result = static_cast(ch > 255 ? '?' : ch); - // allocate resulting string - std::wstring result; - result.resize(length); + return result + 1; + } - // second pass: convert to wchar_t - if (length > 0) + static value_type high(value_type result, uint32_t ch) { - wchar_writer::value_type begin = reinterpret_cast(&result[0]); - wchar_writer::value_type end = utf_decoder::decode_utf8_block(data, size, begin); + (void)ch; - assert(begin + length == end); - (void)!end; - } + *result = '?'; - return result; - } -#endif + return result + 1; + } + }; - inline bool strcpy_insitu_allow(size_t length, uintptr_t allocated, char_t* target) + struct utf8_decoder { - assert(target); - size_t target_length = strlength(target); + typedef uint8_t type; - // always reuse document buffer memory if possible - if (!allocated) return target_length >= length; + template static inline typename Traits::value_type process(const uint8_t* data, size_t size, typename Traits::value_type result, Traits) + { + const uint8_t utf8_byte_mask = 0x3f; - // reuse heap memory if waste is not too great - const size_t reuse_threshold = 32; + while (size) + { + uint8_t lead = *data; - return target_length >= length && (target_length < reuse_threshold || target_length - length < target_length / 2); + // 0xxxxxxx -> U+0000..U+007F + if (lead < 0x80) + { + result = Traits::low(result, lead); + data += 1; + size -= 1; + + // process aligned single-byte (ascii) blocks + if ((reinterpret_cast(data) & 3) == 0) + { + // round-trip through void* to silence 'cast increases required alignment of target type' warnings + while (size >= 4 && (*static_cast(static_cast(data)) & 0x80808080) == 0) + { + result = Traits::low(result, data[0]); + result = Traits::low(result, data[1]); + result = Traits::low(result, data[2]); + result = Traits::low(result, data[3]); + data += 4; + size -= 4; + } + } + } + // 110xxxxx -> U+0080..U+07FF + else if (static_cast(lead - 0xC0) < 0x20 && size >= 2 && (data[1] & 0xc0) == 0x80) + { + result = Traits::low(result, ((lead & ~0xC0) << 6) | (data[1] & utf8_byte_mask)); + data += 2; + size -= 2; + } + // 1110xxxx -> U+0800-U+FFFF + else if (static_cast(lead - 0xE0) < 0x10 && size >= 3 && (data[1] & 0xc0) == 0x80 && (data[2] & 0xc0) == 0x80) + { + result = Traits::low(result, ((lead & ~0xE0) << 12) | ((data[1] & utf8_byte_mask) << 6) | (data[2] & utf8_byte_mask)); + data += 3; + size -= 3; + } + // 11110xxx -> U+10000..U+10FFFF + else if (static_cast(lead - 0xF0) < 0x08 && size >= 4 && (data[1] & 0xc0) == 0x80 && (data[2] & 0xc0) == 0x80 && (data[3] & 0xc0) == 0x80) + { + result = Traits::high(result, ((lead & ~0xF0) << 18) | ((data[1] & utf8_byte_mask) << 12) | ((data[2] & utf8_byte_mask) << 6) | (data[3] & utf8_byte_mask)); + data += 4; + size -= 4; + } + // 10xxxxxx or 11111xxx -> invalid + else + { + data += 1; + size -= 1; + } + } + + return result; + } + }; + + template struct utf16_decoder + { + typedef uint16_t type; + + template static inline typename Traits::value_type process(const uint16_t* data, size_t size, typename Traits::value_type result, Traits) + { + while (size) + { + uint16_t lead = opt_swap::value ? endian_swap(*data) : *data; + + // U+0000..U+D7FF + if (lead < 0xD800) + { + result = Traits::low(result, lead); + data += 1; + size -= 1; + } + // U+E000..U+FFFF + else if (static_cast(lead - 0xE000) < 0x2000) + { + result = Traits::low(result, lead); + data += 1; + size -= 1; + } + // surrogate pair lead + else if (static_cast(lead - 0xD800) < 0x400 && size >= 2) + { + uint16_t next = opt_swap::value ? endian_swap(data[1]) : data[1]; + + if (static_cast(next - 0xDC00) < 0x400) + { + result = Traits::high(result, 0x10000 + ((lead & 0x3ff) << 10) + (next & 0x3ff)); + data += 2; + size -= 2; + } + else + { + data += 1; + size -= 1; + } + } + else + { + data += 1; + size -= 1; + } + } + + return result; + } + }; + + template struct utf32_decoder + { + typedef uint32_t type; + + template static inline typename Traits::value_type process(const uint32_t* data, size_t size, typename Traits::value_type result, Traits) + { + while (size) + { + uint32_t lead = opt_swap::value ? endian_swap(*data) : *data; + + // U+0000..U+FFFF + if (lead < 0x10000) + { + result = Traits::low(result, lead); + data += 1; + size -= 1; + } + // U+10000..U+10FFFF + else + { + result = Traits::high(result, lead); + data += 1; + size -= 1; + } + } + + return result; + } + }; + + struct latin1_decoder + { + typedef uint8_t type; + + template static inline typename Traits::value_type process(const uint8_t* data, size_t size, typename Traits::value_type result, Traits) + { + while (size) + { + result = Traits::low(result, *data); + data += 1; + size -= 1; + } + + return result; + } + }; + + template struct wchar_selector; + + template <> struct wchar_selector<2> + { + typedef uint16_t type; + typedef utf16_counter counter; + typedef utf16_writer writer; + typedef utf16_decoder decoder; + }; + + template <> struct wchar_selector<4> + { + typedef uint32_t type; + typedef utf32_counter counter; + typedef utf32_writer writer; + typedef utf32_decoder decoder; + }; + + typedef wchar_selector::counter wchar_counter; + typedef wchar_selector::writer wchar_writer; + + struct wchar_decoder + { + typedef wchar_t type; + + template static inline typename Traits::value_type process(const wchar_t* data, size_t size, typename Traits::value_type result, Traits traits) + { + typedef wchar_selector::decoder decoder; + + return decoder::process(reinterpret_cast(data), size, result, traits); + } + }; + +#ifdef PUGIXML_WCHAR_MODE + PUGI__FN void convert_wchar_endian_swap(wchar_t* result, const wchar_t* data, size_t length) + { + for (size_t i = 0; i < length; ++i) + result[i] = static_cast(endian_swap(static_cast::type>(data[i]))); + } +#endif +PUGI__NS_END + +PUGI__NS_BEGIN + enum chartype_t + { + ct_parse_pcdata = 1, // \0, &, \r, < + ct_parse_attr = 2, // \0, &, \r, ', " + ct_parse_attr_ws = 4, // \0, &, \r, ', ", \n, tab + ct_space = 8, // \r, \n, space, tab + ct_parse_cdata = 16, // \0, ], >, \r + ct_parse_comment = 32, // \0, -, >, \r + ct_symbol = 64, // Any symbol > 127, a-z, A-Z, 0-9, _, :, -, . + ct_start_symbol = 128 // Any symbol > 127, a-z, A-Z, _, : + }; + + static const unsigned char chartype_table[256] = + { + 55, 0, 0, 0, 0, 0, 0, 0, 0, 12, 12, 0, 0, 63, 0, 0, // 0-15 + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 16-31 + 8, 0, 6, 0, 0, 0, 7, 6, 0, 0, 0, 0, 0, 96, 64, 0, // 32-47 + 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 192, 0, 1, 0, 48, 0, // 48-63 + 0, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, // 64-79 + 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 0, 0, 16, 0, 192, // 80-95 + 0, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, // 96-111 + 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 0, 0, 0, 0, 0, // 112-127 + + 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, // 128+ + 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, + 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, + 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, + 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, + 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, + 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, + 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192 + }; + + enum chartypex_t + { + ctx_special_pcdata = 1, // Any symbol >= 0 and < 32 (except \t, \r, \n), &, <, > + ctx_special_attr = 2, // Any symbol >= 0 and < 32 (except \t), &, <, >, " + ctx_start_symbol = 4, // Any symbol > 127, a-z, A-Z, _ + ctx_digit = 8, // 0-9 + ctx_symbol = 16 // Any symbol > 127, a-z, A-Z, 0-9, _, -, . + }; + + static const unsigned char chartypex_table[256] = + { + 3, 3, 3, 3, 3, 3, 3, 3, 3, 0, 2, 3, 3, 2, 3, 3, // 0-15 + 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, // 16-31 + 0, 0, 2, 0, 0, 0, 3, 0, 0, 0, 0, 0, 0, 16, 16, 0, // 32-47 + 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 0, 0, 3, 0, 3, 0, // 48-63 + + 0, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, // 64-79 + 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 0, 0, 0, 0, 20, // 80-95 + 0, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, // 96-111 + 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 0, 0, 0, 0, 0, // 112-127 + + 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, // 128+ + 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, + 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, + 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, + 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, + 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, + 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, + 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20 + }; + +#ifdef PUGIXML_WCHAR_MODE + #define PUGI__IS_CHARTYPE_IMPL(c, ct, table) ((static_cast(c) < 128 ? table[static_cast(c)] : table[128]) & (ct)) +#else + #define PUGI__IS_CHARTYPE_IMPL(c, ct, table) (table[static_cast(c)] & (ct)) +#endif + + #define PUGI__IS_CHARTYPE(c, ct) PUGI__IS_CHARTYPE_IMPL(c, ct, chartype_table) + #define PUGI__IS_CHARTYPEX(c, ct) PUGI__IS_CHARTYPE_IMPL(c, ct, chartypex_table) + + PUGI__FN bool is_little_endian() + { + unsigned int ui = 1; + + return *reinterpret_cast(&ui) == 1; + } + + PUGI__FN xml_encoding get_wchar_encoding() + { + PUGI__STATIC_ASSERT(sizeof(wchar_t) == 2 || sizeof(wchar_t) == 4); + + if (sizeof(wchar_t) == 2) + return is_little_endian() ? encoding_utf16_le : encoding_utf16_be; + else + return is_little_endian() ? encoding_utf32_le : encoding_utf32_be; } - bool strcpy_insitu(char_t*& dest, uintptr_t& header, uintptr_t header_mask, const char_t* source) - { - size_t source_length = strlength(source); + PUGI__FN bool parse_declaration_encoding(const uint8_t* data, size_t size, const uint8_t*& out_encoding, size_t& out_length) + { + #define PUGI__SCANCHAR(ch) { if (offset >= size || data[offset] != ch) return false; offset++; } + #define PUGI__SCANCHARTYPE(ct) { while (offset < size && PUGI__IS_CHARTYPE(data[offset], ct)) offset++; } + + // check if we have a non-empty XML declaration + if (size < 6 || !((data[0] == '<') & (data[1] == '?') & (data[2] == 'x') & (data[3] == 'm') & (data[4] == 'l') && PUGI__IS_CHARTYPE(data[5], ct_space))) + return false; + + // scan XML declaration until the encoding field + for (size_t i = 6; i + 1 < size; ++i) + { + // declaration can not contain ? in quoted values + if (data[i] == '?') + return false; + + if (data[i] == 'e' && data[i + 1] == 'n') + { + size_t offset = i; + + // encoding follows the version field which can't contain 'en' so this has to be the encoding if XML is well formed + PUGI__SCANCHAR('e'); PUGI__SCANCHAR('n'); PUGI__SCANCHAR('c'); PUGI__SCANCHAR('o'); + PUGI__SCANCHAR('d'); PUGI__SCANCHAR('i'); PUGI__SCANCHAR('n'); PUGI__SCANCHAR('g'); + + // S? = S? + PUGI__SCANCHARTYPE(ct_space); + PUGI__SCANCHAR('='); + PUGI__SCANCHARTYPE(ct_space); + + // the only two valid delimiters are ' and " + uint8_t delimiter = (offset < size && data[offset] == '"') ? '"' : '\''; + + PUGI__SCANCHAR(delimiter); + + size_t start = offset; + + out_encoding = data + offset; + + PUGI__SCANCHARTYPE(ct_symbol); + + out_length = offset - start; + + PUGI__SCANCHAR(delimiter); + + return true; + } + } + + return false; + + #undef PUGI__SCANCHAR + #undef PUGI__SCANCHARTYPE + } + + PUGI__FN xml_encoding guess_buffer_encoding(const uint8_t* data, size_t size) + { + // skip encoding autodetection if input buffer is too small + if (size < 4) return encoding_utf8; + + uint8_t d0 = data[0], d1 = data[1], d2 = data[2], d3 = data[3]; + + // look for BOM in first few bytes + if (d0 == 0 && d1 == 0 && d2 == 0xfe && d3 == 0xff) return encoding_utf32_be; + if (d0 == 0xff && d1 == 0xfe && d2 == 0 && d3 == 0) return encoding_utf32_le; + if (d0 == 0xfe && d1 == 0xff) return encoding_utf16_be; + if (d0 == 0xff && d1 == 0xfe) return encoding_utf16_le; + if (d0 == 0xef && d1 == 0xbb && d2 == 0xbf) return encoding_utf8; + + // look for <, (contents); + + return guess_buffer_encoding(data, size); + } + + PUGI__FN bool get_mutable_buffer(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size, bool is_mutable) + { + size_t length = size / sizeof(char_t); + + if (is_mutable) + { + out_buffer = static_cast(const_cast(contents)); + out_length = length; + } + else + { + char_t* buffer = static_cast(xml_memory::allocate((length + 1) * sizeof(char_t))); + if (!buffer) return false; + + if (contents) + memcpy(buffer, contents, length * sizeof(char_t)); + else + assert(length == 0); + + buffer[length] = 0; + + out_buffer = buffer; + out_length = length + 1; + } + + return true; + } + +#ifdef PUGIXML_WCHAR_MODE + PUGI__FN bool need_endian_swap_utf(xml_encoding le, xml_encoding re) + { + return (le == encoding_utf16_be && re == encoding_utf16_le) || (le == encoding_utf16_le && re == encoding_utf16_be) || + (le == encoding_utf32_be && re == encoding_utf32_le) || (le == encoding_utf32_le && re == encoding_utf32_be); + } + + PUGI__FN bool convert_buffer_endian_swap(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size, bool is_mutable) + { + const char_t* data = static_cast(contents); + size_t length = size / sizeof(char_t); + + if (is_mutable) + { + char_t* buffer = const_cast(data); + + convert_wchar_endian_swap(buffer, data, length); + + out_buffer = buffer; + out_length = length; + } + else + { + char_t* buffer = static_cast(xml_memory::allocate((length + 1) * sizeof(char_t))); + if (!buffer) return false; + + convert_wchar_endian_swap(buffer, data, length); + buffer[length] = 0; + + out_buffer = buffer; + out_length = length + 1; + } + + return true; + } + + template PUGI__FN bool convert_buffer_generic(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size, D) + { + const typename D::type* data = static_cast(contents); + size_t data_length = size / sizeof(typename D::type); + + // first pass: get length in wchar_t units + size_t length = D::process(data, data_length, 0, wchar_counter()); + + // allocate buffer of suitable length + char_t* buffer = static_cast(xml_memory::allocate((length + 1) * sizeof(char_t))); + if (!buffer) return false; + + // second pass: convert utf16 input to wchar_t + wchar_writer::value_type obegin = reinterpret_cast(buffer); + wchar_writer::value_type oend = D::process(data, data_length, obegin, wchar_writer()); + + assert(oend == obegin + length); + *oend = 0; + + out_buffer = buffer; + out_length = length + 1; + + return true; + } + + PUGI__FN bool convert_buffer(char_t*& out_buffer, size_t& out_length, xml_encoding encoding, const void* contents, size_t size, bool is_mutable) + { + // get native encoding + xml_encoding wchar_encoding = get_wchar_encoding(); + + // fast path: no conversion required + if (encoding == wchar_encoding) + return get_mutable_buffer(out_buffer, out_length, contents, size, is_mutable); + + // only endian-swapping is required + if (need_endian_swap_utf(encoding, wchar_encoding)) + return convert_buffer_endian_swap(out_buffer, out_length, contents, size, is_mutable); + + // source encoding is utf8 + if (encoding == encoding_utf8) + return convert_buffer_generic(out_buffer, out_length, contents, size, utf8_decoder()); + + // source encoding is utf16 + if (encoding == encoding_utf16_be || encoding == encoding_utf16_le) + { + xml_encoding native_encoding = is_little_endian() ? encoding_utf16_le : encoding_utf16_be; + + return (native_encoding == encoding) ? + convert_buffer_generic(out_buffer, out_length, contents, size, utf16_decoder()) : + convert_buffer_generic(out_buffer, out_length, contents, size, utf16_decoder()); + } + + // source encoding is utf32 + if (encoding == encoding_utf32_be || encoding == encoding_utf32_le) + { + xml_encoding native_encoding = is_little_endian() ? encoding_utf32_le : encoding_utf32_be; + + return (native_encoding == encoding) ? + convert_buffer_generic(out_buffer, out_length, contents, size, utf32_decoder()) : + convert_buffer_generic(out_buffer, out_length, contents, size, utf32_decoder()); + } + + // source encoding is latin1 + if (encoding == encoding_latin1) + return convert_buffer_generic(out_buffer, out_length, contents, size, latin1_decoder()); + + assert(false && "Invalid encoding"); // unreachable + return false; + } +#else + template PUGI__FN bool convert_buffer_generic(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size, D) + { + const typename D::type* data = static_cast(contents); + size_t data_length = size / sizeof(typename D::type); + + // first pass: get length in utf8 units + size_t length = D::process(data, data_length, 0, utf8_counter()); + + // allocate buffer of suitable length + char_t* buffer = static_cast(xml_memory::allocate((length + 1) * sizeof(char_t))); + if (!buffer) return false; + + // second pass: convert utf16 input to utf8 + uint8_t* obegin = reinterpret_cast(buffer); + uint8_t* oend = D::process(data, data_length, obegin, utf8_writer()); + + assert(oend == obegin + length); + *oend = 0; + + out_buffer = buffer; + out_length = length + 1; + + return true; + } + + PUGI__FN size_t get_latin1_7bit_prefix_length(const uint8_t* data, size_t size) + { + for (size_t i = 0; i < size; ++i) + if (data[i] > 127) + return i; + + return size; + } + + PUGI__FN bool convert_buffer_latin1(char_t*& out_buffer, size_t& out_length, const void* contents, size_t size, bool is_mutable) + { + const uint8_t* data = static_cast(contents); + size_t data_length = size; + + // get size of prefix that does not need utf8 conversion + size_t prefix_length = get_latin1_7bit_prefix_length(data, data_length); + assert(prefix_length <= data_length); + + const uint8_t* postfix = data + prefix_length; + size_t postfix_length = data_length - prefix_length; + + // if no conversion is needed, just return the original buffer + if (postfix_length == 0) return get_mutable_buffer(out_buffer, out_length, contents, size, is_mutable); + + // first pass: get length in utf8 units + size_t length = prefix_length + latin1_decoder::process(postfix, postfix_length, 0, utf8_counter()); + + // allocate buffer of suitable length + char_t* buffer = static_cast(xml_memory::allocate((length + 1) * sizeof(char_t))); + if (!buffer) return false; + + // second pass: convert latin1 input to utf8 + memcpy(buffer, data, prefix_length); + + uint8_t* obegin = reinterpret_cast(buffer); + uint8_t* oend = latin1_decoder::process(postfix, postfix_length, obegin + prefix_length, utf8_writer()); + + assert(oend == obegin + length); + *oend = 0; + + out_buffer = buffer; + out_length = length + 1; + + return true; + } + + PUGI__FN bool convert_buffer(char_t*& out_buffer, size_t& out_length, xml_encoding encoding, const void* contents, size_t size, bool is_mutable) + { + // fast path: no conversion required + if (encoding == encoding_utf8) + return get_mutable_buffer(out_buffer, out_length, contents, size, is_mutable); + + // source encoding is utf16 + if (encoding == encoding_utf16_be || encoding == encoding_utf16_le) + { + xml_encoding native_encoding = is_little_endian() ? encoding_utf16_le : encoding_utf16_be; + + return (native_encoding == encoding) ? + convert_buffer_generic(out_buffer, out_length, contents, size, utf16_decoder()) : + convert_buffer_generic(out_buffer, out_length, contents, size, utf16_decoder()); + } + + // source encoding is utf32 + if (encoding == encoding_utf32_be || encoding == encoding_utf32_le) + { + xml_encoding native_encoding = is_little_endian() ? encoding_utf32_le : encoding_utf32_be; + + return (native_encoding == encoding) ? + convert_buffer_generic(out_buffer, out_length, contents, size, utf32_decoder()) : + convert_buffer_generic(out_buffer, out_length, contents, size, utf32_decoder()); + } + + // source encoding is latin1 + if (encoding == encoding_latin1) + return convert_buffer_latin1(out_buffer, out_length, contents, size, is_mutable); + + assert(false && "Invalid encoding"); // unreachable + return false; + } +#endif + + PUGI__FN size_t as_utf8_begin(const wchar_t* str, size_t length) + { + // get length in utf8 characters + return wchar_decoder::process(str, length, 0, utf8_counter()); + } + + PUGI__FN void as_utf8_end(char* buffer, size_t size, const wchar_t* str, size_t length) + { + // convert to utf8 + uint8_t* begin = reinterpret_cast(buffer); + uint8_t* end = wchar_decoder::process(str, length, begin, utf8_writer()); + + assert(begin + size == end); + (void)!end; + (void)!size; + } + +#ifndef PUGIXML_NO_STL + PUGI__FN std::string as_utf8_impl(const wchar_t* str, size_t length) + { + // first pass: get length in utf8 characters + size_t size = as_utf8_begin(str, length); + + // allocate resulting string + std::string result; + result.resize(size); + + // second pass: convert to utf8 + if (size > 0) as_utf8_end(&result[0], size, str, length); + + return result; + } + + PUGI__FN std::basic_string as_wide_impl(const char* str, size_t size) + { + const uint8_t* data = reinterpret_cast(str); + + // first pass: get length in wchar_t units + size_t length = utf8_decoder::process(data, size, 0, wchar_counter()); + + // allocate resulting string + std::basic_string result; + result.resize(length); + + // second pass: convert to wchar_t + if (length > 0) + { + wchar_writer::value_type begin = reinterpret_cast(&result[0]); + wchar_writer::value_type end = utf8_decoder::process(data, size, begin, wchar_writer()); + + assert(begin + length == end); + (void)!end; + } + + return result; + } +#endif + + template + inline bool strcpy_insitu_allow(size_t length, const Header& header, uintptr_t header_mask, char_t* target) + { + // never reuse shared memory + if (header & xml_memory_page_contents_shared_mask) return false; + + size_t target_length = strlength(target); + + // always reuse document buffer memory if possible + if ((header & header_mask) == 0) return target_length >= length; + + // reuse heap memory if waste is not too great + const size_t reuse_threshold = 32; + + return target_length >= length && (target_length < reuse_threshold || target_length - length < target_length / 2); + } + + template + PUGI__FN bool strcpy_insitu(String& dest, Header& header, uintptr_t header_mask, const char_t* source, size_t source_length) + { + if (source_length == 0) + { + // empty string and null pointer are equivalent, so just deallocate old memory + xml_allocator* alloc = PUGI__GETPAGE_IMPL(header)->allocator; + + if (header & header_mask) alloc->deallocate_string(dest); + + // mark the string as not allocated + dest = 0; + header &= ~header_mask; + + return true; + } + else if (dest && strcpy_insitu_allow(source_length, header, header_mask, dest)) + { + // we can reuse old buffer, so just copy the new data (including zero terminator) + memcpy(dest, source, source_length * sizeof(char_t)); + dest[source_length] = 0; + + return true; + } + else + { + xml_allocator* alloc = PUGI__GETPAGE_IMPL(header)->allocator; + + if (!alloc->reserve()) return false; + + // allocate new buffer + char_t* buf = alloc->allocate_string(source_length + 1); + if (!buf) return false; + + // copy the string (including zero terminator) + memcpy(buf, source, source_length * sizeof(char_t)); + buf[source_length] = 0; + + // deallocate old buffer (*after* the above to protect against overlapping memory and/or allocation failures) + if (header & header_mask) alloc->deallocate_string(dest); + + // the string is now allocated, so set the flag + dest = buf; + header |= header_mask; + + return true; + } + } + + struct gap + { + char_t* end; + size_t size; + + gap(): end(0), size(0) + { + } + + // Push new gap, move s count bytes further (skipping the gap). + // Collapse previous gap. + void push(char_t*& s, size_t count) + { + if (end) // there was a gap already; collapse it + { + // Move [old_gap_end, new_gap_start) to [old_gap_start, ...) + assert(s >= end); + memmove(end - size, end, reinterpret_cast(s) - reinterpret_cast(end)); + } + + s += count; // end of current gap + + // "merge" two gaps + end = s; + size += count; + } + + // Collapse all gaps, return past-the-end pointer + char_t* flush(char_t* s) + { + if (end) + { + // Move [old_gap_end, current_pos) to [old_gap_start, ...) + assert(s >= end); + memmove(end - size, end, reinterpret_cast(s) - reinterpret_cast(end)); + + return s - size; + } + else return s; + } + }; + + PUGI__FN char_t* strconv_escape(char_t* s, gap& g) + { + char_t* stre = s + 1; + + switch (*stre) + { + case '#': // &#... + { + unsigned int ucsc = 0; + + if (stre[1] == 'x') // &#x... (hex code) + { + stre += 2; + + char_t ch = *stre; + + if (ch == ';') return stre; + + for (;;) + { + if (static_cast(ch - '0') <= 9) + ucsc = 16 * ucsc + (ch - '0'); + else if (static_cast((ch | ' ') - 'a') <= 5) + ucsc = 16 * ucsc + ((ch | ' ') - 'a' + 10); + else if (ch == ';') + break; + else // cancel + return stre; + + ch = *++stre; + } + + ++stre; + } + else // &#... (dec code) + { + char_t ch = *++stre; + + if (ch == ';') return stre; + + for (;;) + { + if (static_cast(ch - '0') <= 9) + ucsc = 10 * ucsc + (ch - '0'); + else if (ch == ';') + break; + else // cancel + return stre; + + ch = *++stre; + } + + ++stre; + } + + #ifdef PUGIXML_WCHAR_MODE + s = reinterpret_cast(wchar_writer::any(reinterpret_cast(s), ucsc)); + #else + s = reinterpret_cast(utf8_writer::any(reinterpret_cast(s), ucsc)); + #endif + + g.push(s, stre - s); + return stre; + } + + case 'a': // &a + { + ++stre; + + if (*stre == 'm') // &am + { + if (*++stre == 'p' && *++stre == ';') // & + { + *s++ = '&'; + ++stre; + + g.push(s, stre - s); + return stre; + } + } + else if (*stre == 'p') // &ap + { + if (*++stre == 'o' && *++stre == 's' && *++stre == ';') // ' + { + *s++ = '\''; + ++stre; + + g.push(s, stre - s); + return stre; + } + } + break; + } + + case 'g': // &g + { + if (*++stre == 't' && *++stre == ';') // > + { + *s++ = '>'; + ++stre; + + g.push(s, stre - s); + return stre; + } + break; + } + + case 'l': // &l + { + if (*++stre == 't' && *++stre == ';') // < + { + *s++ = '<'; + ++stre; + + g.push(s, stre - s); + return stre; + } + break; + } + + case 'q': // &q + { + if (*++stre == 'u' && *++stre == 'o' && *++stre == 't' && *++stre == ';') // " + { + *s++ = '"'; + ++stre; + + g.push(s, stre - s); + return stre; + } + break; + } + + default: + break; + } + + return stre; + } + + // Parser utilities + #define PUGI__ENDSWITH(c, e) ((c) == (e) || ((c) == 0 && endch == (e))) + #define PUGI__SKIPWS() { while (PUGI__IS_CHARTYPE(*s, ct_space)) ++s; } + #define PUGI__OPTSET(OPT) ( optmsk & (OPT) ) + #define PUGI__PUSHNODE(TYPE) { cursor = append_new_node(cursor, *alloc, TYPE); if (!cursor) PUGI__THROW_ERROR(status_out_of_memory, s); } + #define PUGI__POPNODE() { cursor = cursor->parent; } + #define PUGI__SCANFOR(X) { while (*s != 0 && !(X)) ++s; } + #define PUGI__SCANWHILE(X) { while (X) ++s; } + #define PUGI__SCANWHILE_UNROLL(X) { for (;;) { char_t ss = s[0]; if (PUGI__UNLIKELY(!(X))) { break; } ss = s[1]; if (PUGI__UNLIKELY(!(X))) { s += 1; break; } ss = s[2]; if (PUGI__UNLIKELY(!(X))) { s += 2; break; } ss = s[3]; if (PUGI__UNLIKELY(!(X))) { s += 3; break; } s += 4; } } + #define PUGI__ENDSEG() { ch = *s; *s = 0; ++s; } + #define PUGI__THROW_ERROR(err, m) return error_offset = m, error_status = err, static_cast(0) + #define PUGI__CHECK_ERROR(err, m) { if (*s == 0) PUGI__THROW_ERROR(err, m); } + + PUGI__FN char_t* strconv_comment(char_t* s, char_t endch) + { + gap g; + + while (true) + { + PUGI__SCANWHILE_UNROLL(!PUGI__IS_CHARTYPE(ss, ct_parse_comment)); + + if (*s == '\r') // Either a single 0x0d or 0x0d 0x0a pair + { + *s++ = '\n'; // replace first one with 0x0a + + if (*s == '\n') g.push(s, 1); + } + else if (s[0] == '-' && s[1] == '-' && PUGI__ENDSWITH(s[2], '>')) // comment ends here + { + *g.flush(s) = 0; + + return s + (s[2] == '>' ? 3 : 2); + } + else if (*s == 0) + { + return 0; + } + else ++s; + } + } + + PUGI__FN char_t* strconv_cdata(char_t* s, char_t endch) + { + gap g; + + while (true) + { + PUGI__SCANWHILE_UNROLL(!PUGI__IS_CHARTYPE(ss, ct_parse_cdata)); + + if (*s == '\r') // Either a single 0x0d or 0x0d 0x0a pair + { + *s++ = '\n'; // replace first one with 0x0a + + if (*s == '\n') g.push(s, 1); + } + else if (s[0] == ']' && s[1] == ']' && PUGI__ENDSWITH(s[2], '>')) // CDATA ends here + { + *g.flush(s) = 0; + + return s + 1; + } + else if (*s == 0) + { + return 0; + } + else ++s; + } + } + + typedef char_t* (*strconv_pcdata_t)(char_t*); + + template struct strconv_pcdata_impl + { + static char_t* parse(char_t* s) + { + gap g; + + char_t* begin = s; + + while (true) + { + PUGI__SCANWHILE_UNROLL(!PUGI__IS_CHARTYPE(ss, ct_parse_pcdata)); + + if (*s == '<') // PCDATA ends here + { + char_t* end = g.flush(s); + + if (opt_trim::value) + while (end > begin && PUGI__IS_CHARTYPE(end[-1], ct_space)) + --end; + + *end = 0; + + return s + 1; + } + else if (opt_eol::value && *s == '\r') // Either a single 0x0d or 0x0d 0x0a pair + { + *s++ = '\n'; // replace first one with 0x0a + + if (*s == '\n') g.push(s, 1); + } + else if (opt_escape::value && *s == '&') + { + s = strconv_escape(s, g); + } + else if (*s == 0) + { + char_t* end = g.flush(s); + + if (opt_trim::value) + while (end > begin && PUGI__IS_CHARTYPE(end[-1], ct_space)) + --end; + + *end = 0; + + return s; + } + else ++s; + } + } + }; + + PUGI__FN strconv_pcdata_t get_strconv_pcdata(unsigned int optmask) + { + PUGI__STATIC_ASSERT(parse_escapes == 0x10 && parse_eol == 0x20 && parse_trim_pcdata == 0x0800); + + switch (((optmask >> 4) & 3) | ((optmask >> 9) & 4)) // get bitmask for flags (eol escapes trim) + { + case 0: return strconv_pcdata_impl::parse; + case 1: return strconv_pcdata_impl::parse; + case 2: return strconv_pcdata_impl::parse; + case 3: return strconv_pcdata_impl::parse; + case 4: return strconv_pcdata_impl::parse; + case 5: return strconv_pcdata_impl::parse; + case 6: return strconv_pcdata_impl::parse; + case 7: return strconv_pcdata_impl::parse; + default: assert(false); return 0; // unreachable + } + } + + typedef char_t* (*strconv_attribute_t)(char_t*, char_t); + + template struct strconv_attribute_impl + { + static char_t* parse_wnorm(char_t* s, char_t end_quote) + { + gap g; + + // trim leading whitespaces + if (PUGI__IS_CHARTYPE(*s, ct_space)) + { + char_t* str = s; + + do ++str; + while (PUGI__IS_CHARTYPE(*str, ct_space)); + + g.push(s, str - s); + } + + while (true) + { + PUGI__SCANWHILE_UNROLL(!PUGI__IS_CHARTYPE(ss, ct_parse_attr_ws | ct_space)); + + if (*s == end_quote) + { + char_t* str = g.flush(s); + + do *str-- = 0; + while (PUGI__IS_CHARTYPE(*str, ct_space)); + + return s + 1; + } + else if (PUGI__IS_CHARTYPE(*s, ct_space)) + { + *s++ = ' '; + + if (PUGI__IS_CHARTYPE(*s, ct_space)) + { + char_t* str = s + 1; + while (PUGI__IS_CHARTYPE(*str, ct_space)) ++str; + + g.push(s, str - s); + } + } + else if (opt_escape::value && *s == '&') + { + s = strconv_escape(s, g); + } + else if (!*s) + { + return 0; + } + else ++s; + } + } + + static char_t* parse_wconv(char_t* s, char_t end_quote) + { + gap g; + + while (true) + { + PUGI__SCANWHILE_UNROLL(!PUGI__IS_CHARTYPE(ss, ct_parse_attr_ws)); + + if (*s == end_quote) + { + *g.flush(s) = 0; + + return s + 1; + } + else if (PUGI__IS_CHARTYPE(*s, ct_space)) + { + if (*s == '\r') + { + *s++ = ' '; + + if (*s == '\n') g.push(s, 1); + } + else *s++ = ' '; + } + else if (opt_escape::value && *s == '&') + { + s = strconv_escape(s, g); + } + else if (!*s) + { + return 0; + } + else ++s; + } + } + + static char_t* parse_eol(char_t* s, char_t end_quote) + { + gap g; + + while (true) + { + PUGI__SCANWHILE_UNROLL(!PUGI__IS_CHARTYPE(ss, ct_parse_attr)); + + if (*s == end_quote) + { + *g.flush(s) = 0; + + return s + 1; + } + else if (*s == '\r') + { + *s++ = '\n'; + + if (*s == '\n') g.push(s, 1); + } + else if (opt_escape::value && *s == '&') + { + s = strconv_escape(s, g); + } + else if (!*s) + { + return 0; + } + else ++s; + } + } + + static char_t* parse_simple(char_t* s, char_t end_quote) + { + gap g; + + while (true) + { + PUGI__SCANWHILE_UNROLL(!PUGI__IS_CHARTYPE(ss, ct_parse_attr)); + + if (*s == end_quote) + { + *g.flush(s) = 0; + + return s + 1; + } + else if (opt_escape::value && *s == '&') + { + s = strconv_escape(s, g); + } + else if (!*s) + { + return 0; + } + else ++s; + } + } + }; + + PUGI__FN strconv_attribute_t get_strconv_attribute(unsigned int optmask) + { + PUGI__STATIC_ASSERT(parse_escapes == 0x10 && parse_eol == 0x20 && parse_wconv_attribute == 0x40 && parse_wnorm_attribute == 0x80); + + switch ((optmask >> 4) & 15) // get bitmask for flags (wconv wnorm eol escapes) + { + case 0: return strconv_attribute_impl::parse_simple; + case 1: return strconv_attribute_impl::parse_simple; + case 2: return strconv_attribute_impl::parse_eol; + case 3: return strconv_attribute_impl::parse_eol; + case 4: return strconv_attribute_impl::parse_wconv; + case 5: return strconv_attribute_impl::parse_wconv; + case 6: return strconv_attribute_impl::parse_wconv; + case 7: return strconv_attribute_impl::parse_wconv; + case 8: return strconv_attribute_impl::parse_wnorm; + case 9: return strconv_attribute_impl::parse_wnorm; + case 10: return strconv_attribute_impl::parse_wnorm; + case 11: return strconv_attribute_impl::parse_wnorm; + case 12: return strconv_attribute_impl::parse_wnorm; + case 13: return strconv_attribute_impl::parse_wnorm; + case 14: return strconv_attribute_impl::parse_wnorm; + case 15: return strconv_attribute_impl::parse_wnorm; + default: assert(false); return 0; // unreachable + } + } + + inline xml_parse_result make_parse_result(xml_parse_status status, ptrdiff_t offset = 0) + { + xml_parse_result result; + result.status = status; + result.offset = offset; + + return result; + } + + struct xml_parser + { + xml_allocator* alloc; + char_t* error_offset; + xml_parse_status error_status; + + xml_parser(xml_allocator* alloc_): alloc(alloc_), error_offset(0), error_status(status_ok) + { + } + + // DOCTYPE consists of nested sections of the following possible types: + // , , "...", '...' + // + // + // First group can not contain nested groups + // Second group can contain nested groups of the same type + // Third group can contain all other groups + char_t* parse_doctype_primitive(char_t* s) + { + if (*s == '"' || *s == '\'') + { + // quoted string + char_t ch = *s++; + PUGI__SCANFOR(*s == ch); + if (!*s) PUGI__THROW_ERROR(status_bad_doctype, s); + + s++; + } + else if (s[0] == '<' && s[1] == '?') + { + // + s += 2; + PUGI__SCANFOR(s[0] == '?' && s[1] == '>'); // no need for ENDSWITH because ?> can't terminate proper doctype + if (!*s) PUGI__THROW_ERROR(status_bad_doctype, s); + + s += 2; + } + else if (s[0] == '<' && s[1] == '!' && s[2] == '-' && s[3] == '-') + { + s += 4; + PUGI__SCANFOR(s[0] == '-' && s[1] == '-' && s[2] == '>'); // no need for ENDSWITH because --> can't terminate proper doctype + if (!*s) PUGI__THROW_ERROR(status_bad_doctype, s); + + s += 3; + } + else PUGI__THROW_ERROR(status_bad_doctype, s); + + return s; + } + + char_t* parse_doctype_ignore(char_t* s) + { + size_t depth = 0; + + assert(s[0] == '<' && s[1] == '!' && s[2] == '['); + s += 3; + + while (*s) + { + if (s[0] == '<' && s[1] == '!' && s[2] == '[') + { + // nested ignore section + s += 3; + depth++; + } + else if (s[0] == ']' && s[1] == ']' && s[2] == '>') + { + // ignore section end + s += 3; + + if (depth == 0) + return s; + + depth--; + } + else s++; + } + + PUGI__THROW_ERROR(status_bad_doctype, s); + } + + char_t* parse_doctype_group(char_t* s, char_t endch) + { + size_t depth = 0; + + assert((s[0] == '<' || s[0] == 0) && s[1] == '!'); + s += 2; + + while (*s) + { + if (s[0] == '<' && s[1] == '!' && s[2] != '-') + { + if (s[2] == '[') + { + // ignore + s = parse_doctype_ignore(s); + if (!s) return s; + } + else + { + // some control group + s += 2; + depth++; + } + } + else if (s[0] == '<' || s[0] == '"' || s[0] == '\'') + { + // unknown tag (forbidden), or some primitive group + s = parse_doctype_primitive(s); + if (!s) return s; + } + else if (*s == '>') + { + if (depth == 0) + return s; + + depth--; + s++; + } + else s++; + } + + if (depth != 0 || endch != '>') PUGI__THROW_ERROR(status_bad_doctype, s); + + return s; + } + + char_t* parse_exclamation(char_t* s, xml_node_struct* cursor, unsigned int optmsk, char_t endch) + { + // parse node contents, starting with exclamation mark + ++s; + + if (*s == '-') // 'value = s; // Save the offset. + } + + if (PUGI__OPTSET(parse_eol) && PUGI__OPTSET(parse_comments)) + { + s = strconv_comment(s, endch); + + if (!s) PUGI__THROW_ERROR(status_bad_comment, cursor->value); + } + else + { + // Scan for terminating '-->'. + PUGI__SCANFOR(s[0] == '-' && s[1] == '-' && PUGI__ENDSWITH(s[2], '>')); + PUGI__CHECK_ERROR(status_bad_comment, s); + + if (PUGI__OPTSET(parse_comments)) + *s = 0; // Zero-terminate this segment at the first terminating '-'. + + s += (s[2] == '>' ? 3 : 2); // Step over the '\0->'. + } + } + else PUGI__THROW_ERROR(status_bad_comment, s); + } + else if (*s == '[') + { + // 'value = s; // Save the offset. + + if (PUGI__OPTSET(parse_eol)) + { + s = strconv_cdata(s, endch); + + if (!s) PUGI__THROW_ERROR(status_bad_cdata, cursor->value); + } + else + { + // Scan for terminating ']]>'. + PUGI__SCANFOR(s[0] == ']' && s[1] == ']' && PUGI__ENDSWITH(s[2], '>')); + PUGI__CHECK_ERROR(status_bad_cdata, s); + + *s++ = 0; // Zero-terminate this segment. + } + } + else // Flagged for discard, but we still have to scan for the terminator. + { + // Scan for terminating ']]>'. + PUGI__SCANFOR(s[0] == ']' && s[1] == ']' && PUGI__ENDSWITH(s[2], '>')); + PUGI__CHECK_ERROR(status_bad_cdata, s); + + ++s; + } + + s += (s[1] == '>' ? 2 : 1); // Step over the last ']>'. + } + else PUGI__THROW_ERROR(status_bad_cdata, s); + } + else if (s[0] == 'D' && s[1] == 'O' && s[2] == 'C' && s[3] == 'T' && s[4] == 'Y' && s[5] == 'P' && PUGI__ENDSWITH(s[6], 'E')) + { + s -= 2; + + if (cursor->parent) PUGI__THROW_ERROR(status_bad_doctype, s); + + char_t* mark = s + 9; + + s = parse_doctype_group(s, endch); + if (!s) return s; + + assert((*s == 0 && endch == '>') || *s == '>'); + if (*s) *s++ = 0; + + if (PUGI__OPTSET(parse_doctype)) + { + while (PUGI__IS_CHARTYPE(*mark, ct_space)) ++mark; + + PUGI__PUSHNODE(node_doctype); + + cursor->value = mark; + } + } + else if (*s == 0 && endch == '-') PUGI__THROW_ERROR(status_bad_comment, s); + else if (*s == 0 && endch == '[') PUGI__THROW_ERROR(status_bad_cdata, s); + else PUGI__THROW_ERROR(status_unrecognized_tag, s); + + return s; + } + + char_t* parse_question(char_t* s, xml_node_struct*& ref_cursor, unsigned int optmsk, char_t endch) + { + // load into registers + xml_node_struct* cursor = ref_cursor; + char_t ch = 0; + + // parse node contents, starting with question mark + ++s; + + // read PI target + char_t* target = s; + + if (!PUGI__IS_CHARTYPE(*s, ct_start_symbol)) PUGI__THROW_ERROR(status_bad_pi, s); + + PUGI__SCANWHILE(PUGI__IS_CHARTYPE(*s, ct_symbol)); + PUGI__CHECK_ERROR(status_bad_pi, s); + + // determine node type; stricmp / strcasecmp is not portable + bool declaration = (target[0] | ' ') == 'x' && (target[1] | ' ') == 'm' && (target[2] | ' ') == 'l' && target + 3 == s; + + if (declaration ? PUGI__OPTSET(parse_declaration) : PUGI__OPTSET(parse_pi)) + { + if (declaration) + { + // disallow non top-level declarations + if (cursor->parent) PUGI__THROW_ERROR(status_bad_pi, s); + + PUGI__PUSHNODE(node_declaration); + } + else + { + PUGI__PUSHNODE(node_pi); + } + + cursor->name = target; + + PUGI__ENDSEG(); + + // parse value/attributes + if (ch == '?') + { + // empty node + if (!PUGI__ENDSWITH(*s, '>')) PUGI__THROW_ERROR(status_bad_pi, s); + s += (*s == '>'); + + PUGI__POPNODE(); + } + else if (PUGI__IS_CHARTYPE(ch, ct_space)) + { + PUGI__SKIPWS(); + + // scan for tag end + char_t* value = s; + + PUGI__SCANFOR(s[0] == '?' && PUGI__ENDSWITH(s[1], '>')); + PUGI__CHECK_ERROR(status_bad_pi, s); + + if (declaration) + { + // replace ending ? with / so that 'element' terminates properly + *s = '/'; + + // we exit from this function with cursor at node_declaration, which is a signal to parse() to go to LOC_ATTRIBUTES + s = value; + } + else + { + // store value and step over > + cursor->value = value; + + PUGI__POPNODE(); + + PUGI__ENDSEG(); + + s += (*s == '>'); + } + } + else PUGI__THROW_ERROR(status_bad_pi, s); + } + else + { + // scan for tag end + PUGI__SCANFOR(s[0] == '?' && PUGI__ENDSWITH(s[1], '>')); + PUGI__CHECK_ERROR(status_bad_pi, s); + + s += (s[1] == '>' ? 2 : 1); + } + + // store from registers + ref_cursor = cursor; + + return s; + } + + char_t* parse_tree(char_t* s, xml_node_struct* root, unsigned int optmsk, char_t endch) + { + strconv_attribute_t strconv_attribute = get_strconv_attribute(optmsk); + strconv_pcdata_t strconv_pcdata = get_strconv_pcdata(optmsk); + + char_t ch = 0; + xml_node_struct* cursor = root; + char_t* mark = s; + + while (*s != 0) + { + if (*s == '<') + { + ++s; + + LOC_TAG: + if (PUGI__IS_CHARTYPE(*s, ct_start_symbol)) // '<#...' + { + PUGI__PUSHNODE(node_element); // Append a new node to the tree. + + cursor->name = s; + + PUGI__SCANWHILE_UNROLL(PUGI__IS_CHARTYPE(ss, ct_symbol)); // Scan for a terminator. + PUGI__ENDSEG(); // Save char in 'ch', terminate & step over. + + if (ch == '>') + { + // end of tag + } + else if (PUGI__IS_CHARTYPE(ch, ct_space)) + { + LOC_ATTRIBUTES: + while (true) + { + PUGI__SKIPWS(); // Eat any whitespace. + + if (PUGI__IS_CHARTYPE(*s, ct_start_symbol)) // <... #... + { + xml_attribute_struct* a = append_new_attribute(cursor, *alloc); // Make space for this attribute. + if (!a) PUGI__THROW_ERROR(status_out_of_memory, s); + + a->name = s; // Save the offset. + + PUGI__SCANWHILE_UNROLL(PUGI__IS_CHARTYPE(ss, ct_symbol)); // Scan for a terminator. + PUGI__ENDSEG(); // Save char in 'ch', terminate & step over. + + if (PUGI__IS_CHARTYPE(ch, ct_space)) + { + PUGI__SKIPWS(); // Eat any whitespace. + + ch = *s; + ++s; + } + + if (ch == '=') // '<... #=...' + { + PUGI__SKIPWS(); // Eat any whitespace. + + if (*s == '"' || *s == '\'') // '<... #="...' + { + ch = *s; // Save quote char to avoid breaking on "''" -or- '""'. + ++s; // Step over the quote. + a->value = s; // Save the offset. + + s = strconv_attribute(s, ch); + + if (!s) PUGI__THROW_ERROR(status_bad_attribute, a->value); + + // After this line the loop continues from the start; + // Whitespaces, / and > are ok, symbols and EOF are wrong, + // everything else will be detected + if (PUGI__IS_CHARTYPE(*s, ct_start_symbol)) PUGI__THROW_ERROR(status_bad_attribute, s); + } + else PUGI__THROW_ERROR(status_bad_attribute, s); + } + else PUGI__THROW_ERROR(status_bad_attribute, s); + } + else if (*s == '/') + { + ++s; + + if (*s == '>') + { + PUGI__POPNODE(); + s++; + break; + } + else if (*s == 0 && endch == '>') + { + PUGI__POPNODE(); + break; + } + else PUGI__THROW_ERROR(status_bad_start_element, s); + } + else if (*s == '>') + { + ++s; + + break; + } + else if (*s == 0 && endch == '>') + { + break; + } + else PUGI__THROW_ERROR(status_bad_start_element, s); + } + + // !!! + } + else if (ch == '/') // '<#.../' + { + if (!PUGI__ENDSWITH(*s, '>')) PUGI__THROW_ERROR(status_bad_start_element, s); + + PUGI__POPNODE(); // Pop. + + s += (*s == '>'); + } + else if (ch == 0) + { + // we stepped over null terminator, backtrack & handle closing tag + --s; + + if (endch != '>') PUGI__THROW_ERROR(status_bad_start_element, s); + } + else PUGI__THROW_ERROR(status_bad_start_element, s); + } + else if (*s == '/') + { + ++s; + + mark = s; + + char_t* name = cursor->name; + if (!name) PUGI__THROW_ERROR(status_end_element_mismatch, mark); + + while (PUGI__IS_CHARTYPE(*s, ct_symbol)) + { + if (*s++ != *name++) PUGI__THROW_ERROR(status_end_element_mismatch, mark); + } + + if (*name) + { + if (*s == 0 && name[0] == endch && name[1] == 0) PUGI__THROW_ERROR(status_bad_end_element, s); + else PUGI__THROW_ERROR(status_end_element_mismatch, mark); + } + + PUGI__POPNODE(); // Pop. + + PUGI__SKIPWS(); + + if (*s == 0) + { + if (endch != '>') PUGI__THROW_ERROR(status_bad_end_element, s); + } + else + { + if (*s != '>') PUGI__THROW_ERROR(status_bad_end_element, s); + ++s; + } + } + else if (*s == '?') // 'first_child) continue; + } + } + + if (!PUGI__OPTSET(parse_trim_pcdata)) + s = mark; + + if (cursor->parent || PUGI__OPTSET(parse_fragment)) + { + if (PUGI__OPTSET(parse_embed_pcdata) && cursor->parent && !cursor->first_child && !cursor->value) + { + cursor->value = s; // Save the offset. + } + else + { + PUGI__PUSHNODE(node_pcdata); // Append a new node on the tree. + + cursor->value = s; // Save the offset. - if (source_length == 0) - { - // empty string and null pointer are equivalent, so just deallocate old memory - xml_allocator* alloc = reinterpret_cast(header & xml_memory_page_pointer_mask)->allocator; + PUGI__POPNODE(); // Pop since this is a standalone. + } - if (header & header_mask) alloc->deallocate_string(dest); - - // mark the string as not allocated - dest = 0; - header &= ~header_mask; + s = strconv_pcdata(s); - return true; - } - else if (dest && strcpy_insitu_allow(source_length, header & header_mask, dest)) - { - // we can reuse old buffer, so just copy the new data (including zero terminator) - memcpy(dest, source, (source_length + 1) * sizeof(char_t)); - - return true; - } - else - { - xml_allocator* alloc = reinterpret_cast(header & xml_memory_page_pointer_mask)->allocator; + if (!*s) break; + } + else + { + PUGI__SCANFOR(*s == '<'); // '...<' + if (!*s) break; - // allocate new buffer - char_t* buf = alloc->allocate_string(source_length + 1); - if (!buf) return false; + ++s; + } - // copy the string (including zero terminator) - memcpy(buf, source, (source_length + 1) * sizeof(char_t)); + // We're after '<' + goto LOC_TAG; + } + } - // deallocate old buffer (*after* the above to protect against overlapping memory and/or allocation failures) - if (header & header_mask) alloc->deallocate_string(dest); - - // the string is now allocated, so set the flag - dest = buf; - header |= header_mask; + // check that last tag is closed + if (cursor != root) PUGI__THROW_ERROR(status_end_element_mismatch, s); - return true; + return s; } - } - struct gap - { - char_t* end; - size_t size; - - gap(): end(0), size(0) + #ifdef PUGIXML_WCHAR_MODE + static char_t* parse_skip_bom(char_t* s) { + unsigned int bom = 0xfeff; + return (s[0] == static_cast(bom)) ? s + 1 : s; } - - // Push new gap, move s count bytes further (skipping the gap). - // Collapse previous gap. - void push(char_t*& s, size_t count) + #else + static char_t* parse_skip_bom(char_t* s) { - if (end) // there was a gap already; collapse it + return (s[0] == '\xef' && s[1] == '\xbb' && s[2] == '\xbf') ? s + 3 : s; + } + #endif + + static bool has_element_node_siblings(xml_node_struct* node) + { + while (node) { - // Move [old_gap_end, new_gap_start) to [old_gap_start, ...) - assert(s >= end); - memmove(end - size, end, reinterpret_cast(s) - reinterpret_cast(end)); + if (PUGI__NODETYPE(node) == node_element) return true; + + node = node->next_sibling; } - - s += count; // end of current gap - - // "merge" two gaps - end = s; - size += count; + + return false; } - - // Collapse all gaps, return past-the-end pointer - char_t* flush(char_t* s) + + static xml_parse_result parse(char_t* buffer, size_t length, xml_document_struct* xmldoc, xml_node_struct* root, unsigned int optmsk) { - if (end) + // early-out for empty documents + if (length == 0) + return make_parse_result(PUGI__OPTSET(parse_fragment) ? status_ok : status_no_document_element); + + // get last child of the root before parsing + xml_node_struct* last_root_child = root->first_child ? root->first_child->prev_sibling_c + 0 : 0; + + // create parser on stack + xml_parser parser(static_cast(xmldoc)); + + // save last character and make buffer zero-terminated (speeds up parsing) + char_t endch = buffer[length - 1]; + buffer[length - 1] = 0; + + // skip BOM to make sure it does not end up as part of parse output + char_t* buffer_data = parse_skip_bom(buffer); + + // perform actual parsing + parser.parse_tree(buffer_data, root, optmsk, endch); + + xml_parse_result result = make_parse_result(parser.error_status, parser.error_offset ? parser.error_offset - buffer : 0); + assert(result.offset >= 0 && static_cast(result.offset) <= length); + + if (result) { - // Move [old_gap_end, current_pos) to [old_gap_start, ...) - assert(s >= end); - memmove(end - size, end, reinterpret_cast(s) - reinterpret_cast(end)); + // since we removed last character, we have to handle the only possible false positive (stray <) + if (endch == '<') + return make_parse_result(status_unrecognized_tag, length - 1); - return s - size; + // check if there are any element nodes parsed + xml_node_struct* first_root_child_parsed = last_root_child ? last_root_child->next_sibling + 0 : root->first_child+ 0; + + if (!PUGI__OPTSET(parse_fragment) && !has_element_node_siblings(first_root_child_parsed)) + return make_parse_result(status_no_document_element, length - 1); } - else return s; + else + { + // roll back offset if it occurs on a null terminator in the source buffer + if (result.offset > 0 && static_cast(result.offset) == length - 1 && endch == 0) + result.offset--; + } + + return result; } }; - - char_t* strconv_escape(char_t* s, gap& g) + + // Output facilities + PUGI__FN xml_encoding get_write_native_encoding() { - char_t* stre = s + 1; + #ifdef PUGIXML_WCHAR_MODE + return get_wchar_encoding(); + #else + return encoding_utf8; + #endif + } - switch (*stre) - { - case '#': // &#... - { - unsigned int ucsc = 0; + PUGI__FN xml_encoding get_write_encoding(xml_encoding encoding) + { + // replace wchar encoding with utf implementation + if (encoding == encoding_wchar) return get_wchar_encoding(); - if (stre[1] == 'x') // &#x... (hex code) - { - stre += 2; + // replace utf16 encoding with utf16 with specific endianness + if (encoding == encoding_utf16) return is_little_endian() ? encoding_utf16_le : encoding_utf16_be; - char_t ch = *stre; + // replace utf32 encoding with utf32 with specific endianness + if (encoding == encoding_utf32) return is_little_endian() ? encoding_utf32_le : encoding_utf32_be; - if (ch == ';') return stre; + // only do autodetection if no explicit encoding is requested + if (encoding != encoding_auto) return encoding; - for (;;) - { - if (static_cast(ch - '0') <= 9) - ucsc = 16 * ucsc + (ch - '0'); - else if (static_cast((ch | ' ') - 'a') <= 5) - ucsc = 16 * ucsc + ((ch | ' ') - 'a' + 10); - else if (ch == ';') - break; - else // cancel - return stre; + // assume utf8 encoding + return encoding_utf8; + } - ch = *++stre; - } - - ++stre; - } - else // &#... (dec code) - { - char_t ch = *++stre; + template PUGI__FN size_t convert_buffer_output_generic(typename T::value_type dest, const char_t* data, size_t length, D, T) + { + PUGI__STATIC_ASSERT(sizeof(char_t) == sizeof(typename D::type)); - if (ch == ';') return stre; + typename T::value_type end = D::process(reinterpret_cast(data), length, dest, T()); - for (;;) - { - if (static_cast(ch - '0') <= 9) - ucsc = 10 * ucsc + (ch - '0'); - else if (ch == ';') - break; - else // cancel - return stre; + return static_cast(end - dest) * sizeof(*dest); + } - ch = *++stre; - } - - ++stre; - } + template PUGI__FN size_t convert_buffer_output_generic(typename T::value_type dest, const char_t* data, size_t length, D, T, bool opt_swap) + { + PUGI__STATIC_ASSERT(sizeof(char_t) == sizeof(typename D::type)); - #ifdef PUGIXML_WCHAR_MODE - s = reinterpret_cast(wchar_writer::any(reinterpret_cast(s), ucsc)); - #else - s = reinterpret_cast(utf8_writer::any(reinterpret_cast(s), ucsc)); - #endif - - g.push(s, stre - s); - return stre; - } - case 'a': // &a - { - ++stre; + typename T::value_type end = D::process(reinterpret_cast(data), length, dest, T()); - if (*stre == 'm') // &am - { - if (*++stre == 'p' && *++stre == ';') // & - { - *s++ = '&'; - ++stre; - - g.push(s, stre - s); - return stre; - } - } - else if (*stre == 'p') // &ap - { - if (*++stre == 'o' && *++stre == 's' && *++stre == ';') // ' - { - *s++ = '\''; - ++stre; + if (opt_swap) + { + for (typename T::value_type i = dest; i != end; ++i) + *i = endian_swap(*i); + } - g.push(s, stre - s); - return stre; - } - } - break; - } - case 'g': // &g - { - if (*++stre == 't' && *++stre == ';') // > - { - *s++ = '>'; - ++stre; - - g.push(s, stre - s); - return stre; - } - break; - } - case 'l': // &l - { - if (*++stre == 't' && *++stre == ';') // < - { - *s++ = '<'; - ++stre; - - g.push(s, stre - s); - return stre; - } - break; - } - case 'q': // &q - { - if (*++stre == 'u' && *++stre == 'o' && *++stre == 't' && *++stre == ';') // " - { - *s++ = '"'; - ++stre; - - g.push(s, stre - s); - return stre; - } - break; - } + return static_cast(end - dest) * sizeof(*dest); + } + +#ifdef PUGIXML_WCHAR_MODE + PUGI__FN size_t get_valid_length(const char_t* data, size_t length) + { + if (length < 1) return 0; + + // discard last character if it's the lead of a surrogate pair + return (sizeof(wchar_t) == 2 && static_cast(static_cast(data[length - 1]) - 0xD800) < 0x400) ? length - 1 : length; + } + + PUGI__FN size_t convert_buffer_output(char_t* r_char, uint8_t* r_u8, uint16_t* r_u16, uint32_t* r_u32, const char_t* data, size_t length, xml_encoding encoding) + { + // only endian-swapping is required + if (need_endian_swap_utf(encoding, get_wchar_encoding())) + { + convert_wchar_endian_swap(r_char, data, length); + + return length * sizeof(char_t); + } + + // convert to utf8 + if (encoding == encoding_utf8) + return convert_buffer_output_generic(r_u8, data, length, wchar_decoder(), utf8_writer()); + + // convert to utf16 + if (encoding == encoding_utf16_be || encoding == encoding_utf16_le) + { + xml_encoding native_encoding = is_little_endian() ? encoding_utf16_le : encoding_utf16_be; + + return convert_buffer_output_generic(r_u16, data, length, wchar_decoder(), utf16_writer(), native_encoding != encoding); + } + + // convert to utf32 + if (encoding == encoding_utf32_be || encoding == encoding_utf32_le) + { + xml_encoding native_encoding = is_little_endian() ? encoding_utf32_le : encoding_utf32_be; + + return convert_buffer_output_generic(r_u32, data, length, wchar_decoder(), utf32_writer(), native_encoding != encoding); } - - return stre; - } - // Utility macro for last character handling - #define ENDSWITH(c, e) ((c) == (e) || ((c) == 0 && endch == (e))) + // convert to latin1 + if (encoding == encoding_latin1) + return convert_buffer_output_generic(r_u8, data, length, wchar_decoder(), latin1_writer()); - char_t* strconv_comment(char_t* s, char_t endch) + assert(false && "Invalid encoding"); // unreachable + return 0; + } +#else + PUGI__FN size_t get_valid_length(const char_t* data, size_t length) { - gap g; - - while (true) + if (length < 5) return 0; + + for (size_t i = 1; i <= 4; ++i) { - while (!IS_CHARTYPE(*s, ct_parse_comment)) ++s; - - if (*s == '\r') // Either a single 0x0d or 0x0d 0x0a pair - { - *s++ = '\n'; // replace first one with 0x0a - - if (*s == '\n') g.push(s, 1); - } - else if (s[0] == '-' && s[1] == '-' && ENDSWITH(s[2], '>')) // comment ends here - { - *g.flush(s) = 0; - - return s + (s[2] == '>' ? 3 : 2); - } - else if (*s == 0) - { - return 0; - } - else ++s; + uint8_t ch = static_cast(data[length - i]); + + // either a standalone character or a leading one + if ((ch & 0xc0) != 0x80) return length - i; } + + // there are four non-leading characters at the end, sequence tail is broken so might as well process the whole chunk + return length; } - char_t* strconv_cdata(char_t* s, char_t endch) + PUGI__FN size_t convert_buffer_output(char_t* /* r_char */, uint8_t* r_u8, uint16_t* r_u16, uint32_t* r_u32, const char_t* data, size_t length, xml_encoding encoding) { - gap g; - - while (true) + if (encoding == encoding_utf16_be || encoding == encoding_utf16_le) { - while (!IS_CHARTYPE(*s, ct_parse_cdata)) ++s; - - if (*s == '\r') // Either a single 0x0d or 0x0d 0x0a pair - { - *s++ = '\n'; // replace first one with 0x0a - - if (*s == '\n') g.push(s, 1); - } - else if (s[0] == ']' && s[1] == ']' && ENDSWITH(s[2], '>')) // CDATA ends here - { - *g.flush(s) = 0; - - return s + 1; - } - else if (*s == 0) - { - return 0; - } - else ++s; + xml_encoding native_encoding = is_little_endian() ? encoding_utf16_le : encoding_utf16_be; + + return convert_buffer_output_generic(r_u16, data, length, utf8_decoder(), utf16_writer(), native_encoding != encoding); + } + + if (encoding == encoding_utf32_be || encoding == encoding_utf32_le) + { + xml_encoding native_encoding = is_little_endian() ? encoding_utf32_le : encoding_utf32_be; + + return convert_buffer_output_generic(r_u32, data, length, utf8_decoder(), utf32_writer(), native_encoding != encoding); } + + if (encoding == encoding_latin1) + return convert_buffer_output_generic(r_u8, data, length, utf8_decoder(), latin1_writer()); + + assert(false && "Invalid encoding"); // unreachable + return 0; } - - typedef char_t* (*strconv_pcdata_t)(char_t*); - - template struct strconv_pcdata_impl +#endif + + class xml_buffered_writer { - static char_t* parse(char_t* s) + xml_buffered_writer(const xml_buffered_writer&); + xml_buffered_writer& operator=(const xml_buffered_writer&); + + public: + xml_buffered_writer(xml_writer& writer_, xml_encoding user_encoding): writer(writer_), bufsize(0), encoding(get_write_encoding(user_encoding)) { - gap g; - - while (true) - { - while (!IS_CHARTYPE(*s, ct_parse_pcdata)) ++s; - - if (*s == '<') // PCDATA ends here - { - *g.flush(s) = 0; - - return s + 1; - } - else if (opt_eol::value && *s == '\r') // Either a single 0x0d or 0x0d 0x0a pair - { - *s++ = '\n'; // replace first one with 0x0a - - if (*s == '\n') g.push(s, 1); - } - else if (opt_escape::value && *s == '&') - { - s = strconv_escape(s, g); - } - else if (*s == 0) - { - return s; - } - else ++s; - } + PUGI__STATIC_ASSERT(bufcapacity >= 8); } - }; - - strconv_pcdata_t get_strconv_pcdata(unsigned int optmask) - { - STATIC_ASSERT(parse_escapes == 0x10 && parse_eol == 0x20); - switch ((optmask >> 4) & 3) // get bitmask for flags (eol escapes) + size_t flush() { - case 0: return strconv_pcdata_impl::parse; - case 1: return strconv_pcdata_impl::parse; - case 2: return strconv_pcdata_impl::parse; - case 3: return strconv_pcdata_impl::parse; - default: return 0; // should not get here + flush(buffer, bufsize); + bufsize = 0; + return 0; } - } - typedef char_t* (*strconv_attribute_t)(char_t*, char_t); - - template struct strconv_attribute_impl - { - static char_t* parse_wnorm(char_t* s, char_t end_quote) + void flush(const char_t* data, size_t size) { - gap g; + if (size == 0) return; - // trim leading whitespaces - if (IS_CHARTYPE(*s, ct_space)) + // fast path, just write data + if (encoding == get_write_native_encoding()) + writer.write(data, size * sizeof(char_t)); + else { - char_t* str = s; - - do ++str; - while (IS_CHARTYPE(*str, ct_space)); - - g.push(s, str - s); + // convert chunk + size_t result = convert_buffer_output(scratch.data_char, scratch.data_u8, scratch.data_u16, scratch.data_u32, data, size, encoding); + assert(result <= sizeof(scratch)); + + // write data + writer.write(scratch.data_u8, result); } + } - while (true) + void write_direct(const char_t* data, size_t length) + { + // flush the remaining buffer contents + flush(); + + // handle large chunks + if (length > bufcapacity) { - while (!IS_CHARTYPE(*s, ct_parse_attr_ws | ct_space)) ++s; - - if (*s == end_quote) - { - char_t* str = g.flush(s); - - do *str-- = 0; - while (IS_CHARTYPE(*str, ct_space)); - - return s + 1; - } - else if (IS_CHARTYPE(*s, ct_space)) - { - *s++ = ' '; - - if (IS_CHARTYPE(*s, ct_space)) - { - char_t* str = s + 1; - while (IS_CHARTYPE(*str, ct_space)) ++str; - - g.push(s, str - s); - } - } - else if (opt_escape::value && *s == '&') + if (encoding == get_write_native_encoding()) { - s = strconv_escape(s, g); + // fast path, can just write data chunk + writer.write(data, length * sizeof(char_t)); + return; } - else if (!*s) + + // need to convert in suitable chunks + while (length > bufcapacity) { - return 0; + // get chunk size by selecting such number of characters that are guaranteed to fit into scratch buffer + // and form a complete codepoint sequence (i.e. discard start of last codepoint if necessary) + size_t chunk_size = get_valid_length(data, bufcapacity); + assert(chunk_size); + + // convert chunk and write + flush(data, chunk_size); + + // iterate + data += chunk_size; + length -= chunk_size; } - else ++s; + + // small tail is copied below + bufsize = 0; } + + memcpy(buffer + bufsize, data, length * sizeof(char_t)); + bufsize += length; } - static char_t* parse_wconv(char_t* s, char_t end_quote) + void write_buffer(const char_t* data, size_t length) { - gap g; + size_t offset = bufsize; - while (true) + if (offset + length <= bufcapacity) { - while (!IS_CHARTYPE(*s, ct_parse_attr_ws)) ++s; - - if (*s == end_quote) - { - *g.flush(s) = 0; - - return s + 1; - } - else if (IS_CHARTYPE(*s, ct_space)) - { - if (*s == '\r') - { - *s++ = ' '; - - if (*s == '\n') g.push(s, 1); - } - else *s++ = ' '; - } - else if (opt_escape::value && *s == '&') - { - s = strconv_escape(s, g); - } - else if (!*s) - { - return 0; - } - else ++s; + memcpy(buffer + offset, data, length * sizeof(char_t)); + bufsize = offset + length; + } + else + { + write_direct(data, length); } } - static char_t* parse_eol(char_t* s, char_t end_quote) + void write_string(const char_t* data) { - gap g; + // write the part of the string that fits in the buffer + size_t offset = bufsize; - while (true) + while (*data && offset < bufcapacity) + buffer[offset++] = *data++; + + // write the rest + if (offset < bufcapacity) { - while (!IS_CHARTYPE(*s, ct_parse_attr)) ++s; - - if (*s == end_quote) - { - *g.flush(s) = 0; - - return s + 1; - } - else if (*s == '\r') - { - *s++ = '\n'; - - if (*s == '\n') g.push(s, 1); - } - else if (opt_escape::value && *s == '&') - { - s = strconv_escape(s, g); - } - else if (!*s) - { - return 0; - } - else ++s; + bufsize = offset; + } + else + { + // backtrack a bit if we have split the codepoint + size_t length = offset - bufsize; + size_t extra = length - get_valid_length(data - length, length); + + bufsize = offset - extra; + + write_direct(data - extra, strlength(data) + extra); } } - static char_t* parse_simple(char_t* s, char_t end_quote) + void write(char_t d0) + { + size_t offset = bufsize; + if (offset > bufcapacity - 1) offset = flush(); + + buffer[offset + 0] = d0; + bufsize = offset + 1; + } + + void write(char_t d0, char_t d1) + { + size_t offset = bufsize; + if (offset > bufcapacity - 2) offset = flush(); + + buffer[offset + 0] = d0; + buffer[offset + 1] = d1; + bufsize = offset + 2; + } + + void write(char_t d0, char_t d1, char_t d2) + { + size_t offset = bufsize; + if (offset > bufcapacity - 3) offset = flush(); + + buffer[offset + 0] = d0; + buffer[offset + 1] = d1; + buffer[offset + 2] = d2; + bufsize = offset + 3; + } + + void write(char_t d0, char_t d1, char_t d2, char_t d3) + { + size_t offset = bufsize; + if (offset > bufcapacity - 4) offset = flush(); + + buffer[offset + 0] = d0; + buffer[offset + 1] = d1; + buffer[offset + 2] = d2; + buffer[offset + 3] = d3; + bufsize = offset + 4; + } + + void write(char_t d0, char_t d1, char_t d2, char_t d3, char_t d4) + { + size_t offset = bufsize; + if (offset > bufcapacity - 5) offset = flush(); + + buffer[offset + 0] = d0; + buffer[offset + 1] = d1; + buffer[offset + 2] = d2; + buffer[offset + 3] = d3; + buffer[offset + 4] = d4; + bufsize = offset + 5; + } + + void write(char_t d0, char_t d1, char_t d2, char_t d3, char_t d4, char_t d5) + { + size_t offset = bufsize; + if (offset > bufcapacity - 6) offset = flush(); + + buffer[offset + 0] = d0; + buffer[offset + 1] = d1; + buffer[offset + 2] = d2; + buffer[offset + 3] = d3; + buffer[offset + 4] = d4; + buffer[offset + 5] = d5; + bufsize = offset + 6; + } + + // utf8 maximum expansion: x4 (-> utf32) + // utf16 maximum expansion: x2 (-> utf32) + // utf32 maximum expansion: x1 + enum + { + bufcapacitybytes = + #ifdef PUGIXML_MEMORY_OUTPUT_STACK + PUGIXML_MEMORY_OUTPUT_STACK + #else + 10240 + #endif + , + bufcapacity = bufcapacitybytes / (sizeof(char_t) + 4) + }; + + char_t buffer[bufcapacity]; + + union + { + uint8_t data_u8[4 * bufcapacity]; + uint16_t data_u16[2 * bufcapacity]; + uint32_t data_u32[bufcapacity]; + char_t data_char[bufcapacity]; + } scratch; + + xml_writer& writer; + size_t bufsize; + xml_encoding encoding; + }; + + PUGI__FN void text_output_escaped(xml_buffered_writer& writer, const char_t* s, chartypex_t type) + { + while (*s) { - gap g; + const char_t* prev = s; - while (true) + // While *s is a usual symbol + PUGI__SCANWHILE_UNROLL(!PUGI__IS_CHARTYPEX(ss, type)); + + writer.write_buffer(prev, static_cast(s - prev)); + + switch (*s) { - while (!IS_CHARTYPE(*s, ct_parse_attr)) ++s; - - if (*s == end_quote) - { - *g.flush(s) = 0; - - return s + 1; - } - else if (opt_escape::value && *s == '&') - { - s = strconv_escape(s, g); - } - else if (!*s) + case 0: break; + case '&': + writer.write('&', 'a', 'm', 'p', ';'); + ++s; + break; + case '<': + writer.write('&', 'l', 't', ';'); + ++s; + break; + case '>': + writer.write('&', 'g', 't', ';'); + ++s; + break; + case '"': + writer.write('&', 'q', 'u', 'o', 't', ';'); + ++s; + break; + default: // s is not a usual symbol { - return 0; + unsigned int ch = static_cast(*s++); + assert(ch < 32); + + writer.write('&', '#', static_cast((ch / 10) + '0'), static_cast((ch % 10) + '0'), ';'); } - else ++s; } } - }; + } - strconv_attribute_t get_strconv_attribute(unsigned int optmask) + PUGI__FN void text_output(xml_buffered_writer& writer, const char_t* s, chartypex_t type, unsigned int flags) { - STATIC_ASSERT(parse_escapes == 0x10 && parse_eol == 0x20 && parse_wconv_attribute == 0x40 && parse_wnorm_attribute == 0x80); - - switch ((optmask >> 4) & 15) // get bitmask for flags (wconv wnorm eol escapes) - { - case 0: return strconv_attribute_impl::parse_simple; - case 1: return strconv_attribute_impl::parse_simple; - case 2: return strconv_attribute_impl::parse_eol; - case 3: return strconv_attribute_impl::parse_eol; - case 4: return strconv_attribute_impl::parse_wconv; - case 5: return strconv_attribute_impl::parse_wconv; - case 6: return strconv_attribute_impl::parse_wconv; - case 7: return strconv_attribute_impl::parse_wconv; - case 8: return strconv_attribute_impl::parse_wnorm; - case 9: return strconv_attribute_impl::parse_wnorm; - case 10: return strconv_attribute_impl::parse_wnorm; - case 11: return strconv_attribute_impl::parse_wnorm; - case 12: return strconv_attribute_impl::parse_wnorm; - case 13: return strconv_attribute_impl::parse_wnorm; - case 14: return strconv_attribute_impl::parse_wnorm; - case 15: return strconv_attribute_impl::parse_wnorm; - default: return 0; // should not get here - } + if (flags & format_no_escapes) + writer.write_string(s); + else + text_output_escaped(writer, s, type); } - inline xml_parse_result make_parse_result(xml_parse_status status, ptrdiff_t offset = 0) + PUGI__FN void text_output_cdata(xml_buffered_writer& writer, const char_t* s) { - xml_parse_result result; - result.status = status; - result.offset = offset; + do + { + writer.write('<', '!', '[', 'C', 'D'); + writer.write('A', 'T', 'A', '['); - return result; + const char_t* prev = s; + + // look for ]]> sequence - we can't output it as is since it terminates CDATA + while (*s && !(s[0] == ']' && s[1] == ']' && s[2] == '>')) ++s; + + // skip ]] if we stopped at ]]>, > will go to the next CDATA section + if (*s) s += 2; + + writer.write_buffer(prev, static_cast(s - prev)); + + writer.write(']', ']', '>'); + } + while (*s); } - struct xml_parser + PUGI__FN void text_output_indent(xml_buffered_writer& writer, const char_t* indent, size_t indent_length, unsigned int depth) { - xml_allocator alloc; - char_t* error_offset; - jmp_buf error_handler; - - // Parser utilities. - #define SKIPWS() { while (IS_CHARTYPE(*s, ct_space)) ++s; } - #define OPTSET(OPT) ( optmsk & OPT ) - #define PUSHNODE(TYPE) { cursor = append_node(cursor, alloc, TYPE); if (!cursor) THROW_ERROR(status_out_of_memory, s); } - #define POPNODE() { cursor = cursor->parent; } - #define SCANFOR(X) { while (*s != 0 && !(X)) ++s; } - #define SCANWHILE(X) { while ((X)) ++s; } - #define ENDSEG() { ch = *s; *s = 0; ++s; } - #define THROW_ERROR(err, m) error_offset = m, longjmp(error_handler, err) - #define CHECK_ERROR(err, m) { if (*s == 0) THROW_ERROR(err, m); } - - xml_parser(const xml_allocator& alloc): alloc(alloc), error_offset(0) + switch (indent_length) { + case 1: + { + for (unsigned int i = 0; i < depth; ++i) + writer.write(indent[0]); + break; } - // DOCTYPE consists of nested sections of the following possible types: - // , , "...", '...' - // - // - // First group can not contain nested groups - // Second group can contain nested groups of the same type - // Third group can contain all other groups - char_t* parse_doctype_primitive(char_t* s) + case 2: { - if (*s == '"' || *s == '\'') - { - // quoted string - char_t ch = *s++; - SCANFOR(*s == ch); - if (!*s) THROW_ERROR(status_bad_doctype, s); - - s++; - } - else if (s[0] == '<' && s[1] == '?') - { - // - s += 2; - SCANFOR(s[0] == '?' && s[1] == '>'); // no need for ENDSWITH because ?> can't terminate proper doctype - if (!*s) THROW_ERROR(status_bad_doctype, s); + for (unsigned int i = 0; i < depth; ++i) + writer.write(indent[0], indent[1]); + break; + } - s += 2; - } - else if (s[0] == '<' && s[1] == '!' && s[2] == '-' && s[3] == '-') - { - s += 4; - SCANFOR(s[0] == '-' && s[1] == '-' && s[2] == '>'); // no need for ENDSWITH because --> can't terminate proper doctype - if (!*s) THROW_ERROR(status_bad_doctype, s); + case 3: + { + for (unsigned int i = 0; i < depth; ++i) + writer.write(indent[0], indent[1], indent[2]); + break; + } - s += 4; - } - else THROW_ERROR(status_bad_doctype, s); + case 4: + { + for (unsigned int i = 0; i < depth; ++i) + writer.write(indent[0], indent[1], indent[2], indent[3]); + break; + } - return s; + default: + { + for (unsigned int i = 0; i < depth; ++i) + writer.write_buffer(indent, indent_length); + } } + } - char_t* parse_doctype_ignore(char_t* s) + PUGI__FN void node_output_comment(xml_buffered_writer& writer, const char_t* s) + { + writer.write('<', '!', '-', '-'); + + while (*s) { - assert(s[0] == '<' && s[1] == '!' && s[2] == '['); - s++; + const char_t* prev = s; - while (*s) + // look for -\0 or -- sequence - we can't output it since -- is illegal in comment body + while (*s && !(s[0] == '-' && (s[1] == '-' || s[1] == 0))) ++s; + + writer.write_buffer(prev, static_cast(s - prev)); + + if (*s) { - if (s[0] == '<' && s[1] == '!' && s[2] == '[') - { - // nested ignore section - s = parse_doctype_ignore(s); - } - else if (s[0] == ']' && s[1] == ']' && s[2] == '>') - { - // ignore section end - s += 3; + assert(*s == '-'); - return s; - } - else s++; + writer.write('-', ' '); + ++s; } + } - THROW_ERROR(status_bad_doctype, s); + writer.write('-', '-', '>'); + } - return s; + PUGI__FN void node_output_pi_value(xml_buffered_writer& writer, const char_t* s) + { + while (*s) + { + const char_t* prev = s; + + // look for ?> sequence - we can't output it since ?> terminates PI + while (*s && !(s[0] == '?' && s[1] == '>')) ++s; + + writer.write_buffer(prev, static_cast(s - prev)); + + if (*s) + { + assert(s[0] == '?' && s[1] == '>'); + + writer.write('?', ' ', '>'); + s += 2; + } } + } - char_t* parse_doctype_group(char_t* s, char_t endch, bool toplevel) - { - assert(s[0] == '<' && s[1] == '!'); - s++; + PUGI__FN void node_output_attributes(xml_buffered_writer& writer, xml_node_struct* node, const char_t* indent, size_t indent_length, unsigned int flags, unsigned int depth) + { + const char_t* default_name = PUGIXML_TEXT(":anonymous"); - while (*s) + for (xml_attribute_struct* a = node->first_attribute; a; a = a->next_attribute) + { + if ((flags & (format_indent_attributes | format_raw)) == format_indent_attributes) { - if (s[0] == '<' && s[1] == '!' && s[2] != '-') - { - if (s[2] == '[') - { - // ignore - s = parse_doctype_ignore(s); - } - else - { - // some control group - s = parse_doctype_group(s, endch, false); - } - } - else if (s[0] == '<' || s[0] == '"' || s[0] == '\'') - { - // unknown tag (forbidden), or some primitive group - s = parse_doctype_primitive(s); - } - else if (*s == '>') - { - s++; + writer.write('\n'); - return s; - } - else s++; + text_output_indent(writer, indent, indent_length, depth + 1); + } + else + { + writer.write(' '); } - if (!toplevel || endch != '>') THROW_ERROR(status_bad_doctype, s); + writer.write_string(a->name ? a->name + 0 : default_name); + writer.write('=', '"'); + + if (a->value) + text_output(writer, a->value, ctx_special_attr, flags); - return s; + writer.write('"'); } + } - char_t* parse_exclamation(char_t* s, xml_node_struct* cursor, unsigned int optmsk, char_t endch) - { - // parse node contents, starting with exclamation mark - ++s; + PUGI__FN bool node_output_start(xml_buffered_writer& writer, xml_node_struct* node, const char_t* indent, size_t indent_length, unsigned int flags, unsigned int depth) + { + const char_t* default_name = PUGIXML_TEXT(":anonymous"); + const char_t* name = node->name ? node->name + 0 : default_name; - if (*s == '-') // 'first_attribute) + node_output_attributes(writer, node, indent, indent_length, flags, depth); + + // element nodes can have value if parse_embed_pcdata was used + if (!node->value) + { + if (!node->first_child) { - ++s; + if (flags & format_no_empty_element_tags) + { + writer.write('>', '<', '/'); + writer.write_string(name); + writer.write('>'); - if (*s == '-') // ''. - SCANFOR(s[0] == '-' && s[1] == '-' && ENDSWITH(s[2], '>')); - CHECK_ERROR(status_bad_comment, s); + return true; + } + } + else + { + writer.write('>'); - if (OPTSET(parse_comments)) - *s = 0; // Zero-terminate this segment at the first terminating '-'. + text_output(writer, node->value, ctx_special_pcdata, flags); - s += (s[2] == '>' ? 3 : 2); // Step over the '\0->'. - } - } - else THROW_ERROR(status_bad_comment, s); - } - else if (*s == '[') + if (!node->first_child) { - // ''); + + return false; + } + else + { + return true; + } + } + } - if (OPTSET(parse_cdata)) - { - PUSHNODE(node_cdata); // Append a new node on the tree. - cursor->value = s; // Save the offset. + PUGI__FN void node_output_end(xml_buffered_writer& writer, xml_node_struct* node) + { + const char_t* default_name = PUGIXML_TEXT(":anonymous"); + const char_t* name = node->name ? node->name + 0 : default_name; - if (OPTSET(parse_eol)) - { - s = strconv_cdata(s, endch); + writer.write('<', '/'); + writer.write_string(name); + writer.write('>'); + } - if (!s) THROW_ERROR(status_bad_cdata, cursor->value); - } - else - { - // Scan for terminating ']]>'. - SCANFOR(s[0] == ']' && s[1] == ']' && ENDSWITH(s[2], '>')); - CHECK_ERROR(status_bad_cdata, s); + PUGI__FN void node_output_simple(xml_buffered_writer& writer, xml_node_struct* node, unsigned int flags) + { + const char_t* default_name = PUGIXML_TEXT(":anonymous"); - *s++ = 0; // Zero-terminate this segment. - } - } - else // Flagged for discard, but we still have to scan for the terminator. - { - // Scan for terminating ']]>'. - SCANFOR(s[0] == ']' && s[1] == ']' && ENDSWITH(s[2], '>')); - CHECK_ERROR(status_bad_cdata, s); + switch (PUGI__NODETYPE(node)) + { + case node_pcdata: + text_output(writer, node->value ? node->value + 0 : PUGIXML_TEXT(""), ctx_special_pcdata, flags); + break; - ++s; - } + case node_cdata: + text_output_cdata(writer, node->value ? node->value + 0 : PUGIXML_TEXT("")); + break; - s += (s[1] == '>' ? 2 : 1); // Step over the last ']>'. + case node_comment: + node_output_comment(writer, node->value ? node->value + 0 : PUGIXML_TEXT("")); + break; + + case node_pi: + writer.write('<', '?'); + writer.write_string(node->name ? node->name + 0 : default_name); + + if (node->value) + { + writer.write(' '); + node_output_pi_value(writer, node->value); } - else THROW_ERROR(status_bad_cdata, s); - } - else if (s[0] == 'D' && s[1] == 'O' && s[2] == 'C' && s[3] == 'T' && s[4] == 'Y' && s[5] == 'P' && ENDSWITH(s[6], 'E')) - { - s -= 2; - if (cursor->parent) THROW_ERROR(status_bad_doctype, s); + writer.write('?', '>'); + break; - char_t* mark = s + 9; + case node_declaration: + writer.write('<', '?'); + writer.write_string(node->name ? node->name + 0 : default_name); + node_output_attributes(writer, node, PUGIXML_TEXT(""), 0, flags | format_raw, 0); + writer.write('?', '>'); + break; - s = parse_doctype_group(s, endch, true); + case node_doctype: + writer.write('<', '!', 'D', 'O', 'C'); + writer.write('T', 'Y', 'P', 'E'); - if (OPTSET(parse_doctype)) - { - while (IS_CHARTYPE(*mark, ct_space)) ++mark; + if (node->value) + { + writer.write(' '); + writer.write_string(node->value); + } - PUSHNODE(node_doctype); + writer.write('>'); + break; - cursor->value = mark; + default: + assert(false && "Invalid node type"); // unreachable + } + } - assert((s[0] == 0 && endch == '>') || s[-1] == '>'); - s[*s == 0 ? 0 : -1] = 0; + enum indent_flags_t + { + indent_newline = 1, + indent_indent = 2 + }; - POPNODE(); - } - } - else if (*s == 0 && endch == '-') THROW_ERROR(status_bad_comment, s); - else if (*s == 0 && endch == '[') THROW_ERROR(status_bad_cdata, s); - else THROW_ERROR(status_unrecognized_tag, s); + PUGI__FN void node_output(xml_buffered_writer& writer, xml_node_struct* root, const char_t* indent, unsigned int flags, unsigned int depth) + { + size_t indent_length = ((flags & (format_indent | format_indent_attributes)) && (flags & format_raw) == 0) ? strlength(indent) : 0; + unsigned int indent_flags = indent_indent; - return s; - } + xml_node_struct* node = root; - char_t* parse_question(char_t* s, xml_node_struct*& ref_cursor, unsigned int optmsk, char_t endch) + do { - // load into registers - xml_node_struct* cursor = ref_cursor; - char_t ch = 0; + assert(node); - // parse node contents, starting with question mark - ++s; + // begin writing current node + if (PUGI__NODETYPE(node) == node_pcdata || PUGI__NODETYPE(node) == node_cdata) + { + node_output_simple(writer, node, flags); - // read PI target - char_t* target = s; + indent_flags = 0; + } + else + { + if ((indent_flags & indent_newline) && (flags & format_raw) == 0) + writer.write('\n'); - if (!IS_CHARTYPE(*s, ct_start_symbol)) THROW_ERROR(status_bad_pi, s); + if ((indent_flags & indent_indent) && indent_length) + text_output_indent(writer, indent, indent_length, depth); - SCANWHILE(IS_CHARTYPE(*s, ct_symbol)); - CHECK_ERROR(status_bad_pi, s); + if (PUGI__NODETYPE(node) == node_element) + { + indent_flags = indent_newline | indent_indent; - // determine node type; stricmp / strcasecmp is not portable - bool declaration = (target[0] | ' ') == 'x' && (target[1] | ' ') == 'm' && (target[2] | ' ') == 'l' && target + 3 == s; + if (node_output_start(writer, node, indent, indent_length, flags, depth)) + { + // element nodes can have value if parse_embed_pcdata was used + if (node->value) + indent_flags = 0; - if (declaration ? OPTSET(parse_declaration) : OPTSET(parse_pi)) - { - if (declaration) + node = node->first_child; + depth++; + continue; + } + } + else if (PUGI__NODETYPE(node) == node_document) { - // disallow non top-level declarations - if (cursor->parent) THROW_ERROR(status_bad_pi, s); + indent_flags = indent_indent; - PUSHNODE(node_declaration); + if (node->first_child) + { + node = node->first_child; + continue; + } } else { - PUSHNODE(node_pi); - } - - cursor->name = target; + node_output_simple(writer, node, flags); - ENDSEG(); + indent_flags = indent_newline | indent_indent; + } + } - // parse value/attributes - if (ch == '?') + // continue to the next node + while (node != root) + { + if (node->next_sibling) { - // empty node - if (!ENDSWITH(*s, '>')) THROW_ERROR(status_bad_pi, s); - s += (*s == '>'); - - POPNODE(); + node = node->next_sibling; + break; } - else if (IS_CHARTYPE(ch, ct_space)) - { - SKIPWS(); - // scan for tag end - char_t* value = s; + node = node->parent; - SCANFOR(s[0] == '?' && ENDSWITH(s[1], '>')); - CHECK_ERROR(status_bad_pi, s); + // write closing node + if (PUGI__NODETYPE(node) == node_element) + { + depth--; - if (declaration) - { - // replace ending ? with / so that 'element' terminates properly - *s = '/'; + if ((indent_flags & indent_newline) && (flags & format_raw) == 0) + writer.write('\n'); - // we exit from this function with cursor at node_declaration, which is a signal to parse() to go to LOC_ATTRIBUTES - s = value; - } - else - { - // store value and step over > - cursor->value = value; - POPNODE(); + if ((indent_flags & indent_indent) && indent_length) + text_output_indent(writer, indent, indent_length, depth); - ENDSEG(); + node_output_end(writer, node); - s += (*s == '>'); - } + indent_flags = indent_newline | indent_indent; } - else THROW_ERROR(status_bad_pi, s); - } - else - { - // scan for tag end - SCANFOR(s[0] == '?' && ENDSWITH(s[1], '>')); - CHECK_ERROR(status_bad_pi, s); - - s += (s[1] == '>' ? 2 : 1); } - - // store from registers - ref_cursor = cursor; - - return s; } + while (node != root); - void parse(char_t* s, xml_node_struct* xmldoc, unsigned int optmsk, char_t endch) - { - strconv_attribute_t strconv_attribute = get_strconv_attribute(optmsk); - strconv_pcdata_t strconv_pcdata = get_strconv_pcdata(optmsk); - - char_t ch = 0; - xml_node_struct* cursor = xmldoc; - char_t* mark = s; - - while (*s != 0) - { - if (*s == '<') - { - ++s; + if ((indent_flags & indent_newline) && (flags & format_raw) == 0) + writer.write('\n'); + } - LOC_TAG: - if (IS_CHARTYPE(*s, ct_start_symbol)) // '<#...' - { - PUSHNODE(node_element); // Append a new node to the tree. + PUGI__FN bool has_declaration(xml_node_struct* node) + { + for (xml_node_struct* child = node->first_child; child; child = child->next_sibling) + { + xml_node_type type = PUGI__NODETYPE(child); - cursor->name = s; + if (type == node_declaration) return true; + if (type == node_element) return false; + } - SCANWHILE(IS_CHARTYPE(*s, ct_symbol)); // Scan for a terminator. - ENDSEG(); // Save char in 'ch', terminate & step over. + return false; + } - if (ch == '>') - { - // end of tag - } - else if (IS_CHARTYPE(ch, ct_space)) - { - LOC_ATTRIBUTES: - while (true) - { - SKIPWS(); // Eat any whitespace. - - if (IS_CHARTYPE(*s, ct_start_symbol)) // <... #... - { - xml_attribute_struct* a = append_attribute_ll(cursor, alloc); // Make space for this attribute. - if (!a) THROW_ERROR(status_out_of_memory, s); + PUGI__FN bool is_attribute_of(xml_attribute_struct* attr, xml_node_struct* node) + { + for (xml_attribute_struct* a = node->first_attribute; a; a = a->next_attribute) + if (a == attr) + return true; - a->name = s; // Save the offset. + return false; + } - SCANWHILE(IS_CHARTYPE(*s, ct_symbol)); // Scan for a terminator. - CHECK_ERROR(status_bad_attribute, s); //$ redundant, left for performance + PUGI__FN bool allow_insert_attribute(xml_node_type parent) + { + return parent == node_element || parent == node_declaration; + } - ENDSEG(); // Save char in 'ch', terminate & step over. - CHECK_ERROR(status_bad_attribute, s); //$ redundant, left for performance + PUGI__FN bool allow_insert_child(xml_node_type parent, xml_node_type child) + { + if (parent != node_document && parent != node_element) return false; + if (child == node_document || child == node_null) return false; + if (parent != node_document && (child == node_declaration || child == node_doctype)) return false; - if (IS_CHARTYPE(ch, ct_space)) - { - SKIPWS(); // Eat any whitespace. - CHECK_ERROR(status_bad_attribute, s); //$ redundant, left for performance + return true; + } - ch = *s; - ++s; - } - - if (ch == '=') // '<... #=...' - { - SKIPWS(); // Eat any whitespace. + PUGI__FN bool allow_move(xml_node parent, xml_node child) + { + // check that child can be a child of parent + if (!allow_insert_child(parent.type(), child.type())) + return false; - if (*s == '"' || *s == '\'') // '<... #="...' - { - ch = *s; // Save quote char to avoid breaking on "''" -or- '""'. - ++s; // Step over the quote. - a->value = s; // Save the offset. + // check that node is not moved between documents + if (parent.root() != child.root()) + return false; - s = strconv_attribute(s, ch); - - if (!s) THROW_ERROR(status_bad_attribute, a->value); + // check that new parent is not in the child subtree + xml_node cur = parent; - // After this line the loop continues from the start; - // Whitespaces, / and > are ok, symbols and EOF are wrong, - // everything else will be detected - if (IS_CHARTYPE(*s, ct_start_symbol)) THROW_ERROR(status_bad_attribute, s); - } - else THROW_ERROR(status_bad_attribute, s); - } - else THROW_ERROR(status_bad_attribute, s); - } - else if (*s == '/') - { - ++s; - - if (*s == '>') - { - POPNODE(); - s++; - break; - } - else if (*s == 0 && endch == '>') - { - POPNODE(); - break; - } - else THROW_ERROR(status_bad_start_element, s); - } - else if (*s == '>') - { - ++s; + while (cur) + { + if (cur == child) + return false; - break; - } - else if (*s == 0 && endch == '>') - { - break; - } - else THROW_ERROR(status_bad_start_element, s); - } + cur = cur.parent(); + } - // !!! - } - else if (ch == '/') // '<#.../' - { - if (!ENDSWITH(*s, '>')) THROW_ERROR(status_bad_start_element, s); + return true; + } - POPNODE(); // Pop. + template + PUGI__FN void node_copy_string(String& dest, Header& header, uintptr_t header_mask, char_t* source, Header& source_header, xml_allocator* alloc) + { + assert(!dest && (header & header_mask) == 0); - s += (*s == '>'); - } - else if (ch == 0) - { - // we stepped over null terminator, backtrack & handle closing tag - --s; - - if (endch != '>') THROW_ERROR(status_bad_start_element, s); - } - else THROW_ERROR(status_bad_start_element, s); - } - else if (*s == '/') - { - ++s; + if (source) + { + if (alloc && (source_header & header_mask) == 0) + { + dest = source; - char_t* name = cursor->name; - if (!name) THROW_ERROR(status_end_element_mismatch, s); - - while (IS_CHARTYPE(*s, ct_symbol)) - { - if (*s++ != *name++) THROW_ERROR(status_end_element_mismatch, s); - } + // since strcpy_insitu can reuse document buffer memory we need to mark both source and dest as shared + header |= xml_memory_page_contents_shared_mask; + source_header |= xml_memory_page_contents_shared_mask; + } + else + strcpy_insitu(dest, header, header_mask, source, strlength(source)); + } + } - if (*name) - { - if (*s == 0 && name[0] == endch && name[1] == 0) THROW_ERROR(status_bad_end_element, s); - else THROW_ERROR(status_end_element_mismatch, s); - } - - POPNODE(); // Pop. + PUGI__FN void node_copy_contents(xml_node_struct* dn, xml_node_struct* sn, xml_allocator* shared_alloc) + { + node_copy_string(dn->name, dn->header, xml_memory_page_name_allocated_mask, sn->name, sn->header, shared_alloc); + node_copy_string(dn->value, dn->header, xml_memory_page_value_allocated_mask, sn->value, sn->header, shared_alloc); - SKIPWS(); + for (xml_attribute_struct* sa = sn->first_attribute; sa; sa = sa->next_attribute) + { + xml_attribute_struct* da = append_new_attribute(dn, get_allocator(dn)); - if (*s == 0) - { - if (endch != '>') THROW_ERROR(status_bad_end_element, s); - } - else - { - if (*s != '>') THROW_ERROR(status_bad_end_element, s); - ++s; - } - } - else if (*s == '?') // 'name, da->header, xml_memory_page_name_allocated_mask, sa->name, sa->header, shared_alloc); + node_copy_string(da->value, da->header, xml_memory_page_value_allocated_mask, sa->value, sa->header, shared_alloc); + } + } + } - assert(cursor); - if ((cursor->header & xml_memory_page_type_mask) + 1 == node_declaration) goto LOC_ATTRIBUTES; - } - else if (*s == '!') // 'first_child; - s = mark; - - if (cursor->parent) - { - PUSHNODE(node_pcdata); // Append a new node on the tree. - cursor->value = s; // Save the offset. + while (sit && sit != sn) + { + // when a tree is copied into one of the descendants, we need to skip that subtree to avoid an infinite loop + if (sit != dn) + { + xml_node_struct* copy = append_new_node(dit, alloc, PUGI__NODETYPE(sit)); - s = strconv_pcdata(s); - - POPNODE(); // Pop since this is a standalone. - - if (!*s) break; - } - else + if (copy) + { + node_copy_contents(copy, sit, shared_alloc); + + if (sit->first_child) { - SCANFOR(*s == '<'); // '...<' - if (!*s) break; - - ++s; + dit = copy; + sit = sit->first_child; + continue; } - - // We're after '<' - goto LOC_TAG; } } - // check that last tag is closed - if (cursor != xmldoc) THROW_ERROR(status_end_element_mismatch, s); + // continue to the next node + do + { + if (sit->next_sibling) + { + sit = sit->next_sibling; + break; + } + + sit = sit->parent; + dit = dit->parent; + } + while (sit != sn); } + } - static xml_parse_result parse(char_t* buffer, size_t length, xml_node_struct* root, unsigned int optmsk) - { - xml_document_struct* xmldoc = static_cast(root); + PUGI__FN void node_copy_attribute(xml_attribute_struct* da, xml_attribute_struct* sa) + { + xml_allocator& alloc = get_allocator(da); + xml_allocator* shared_alloc = (&alloc == &get_allocator(sa)) ? &alloc : 0; - // store buffer for offset_debug - xmldoc->buffer = buffer; + node_copy_string(da->name, da->header, xml_memory_page_name_allocated_mask, sa->name, sa->header, shared_alloc); + node_copy_string(da->value, da->header, xml_memory_page_value_allocated_mask, sa->value, sa->header, shared_alloc); + } - // early-out for empty documents - if (length == 0) return make_parse_result(status_ok); + inline bool is_text_node(xml_node_struct* node) + { + xml_node_type type = PUGI__NODETYPE(node); - // create parser on stack - xml_parser parser(*xmldoc); + return type == node_pcdata || type == node_cdata; + } - // save last character and make buffer zero-terminated (speeds up parsing) - char_t endch = buffer[length - 1]; - buffer[length - 1] = 0; - - // perform actual parsing - int error = setjmp(parser.error_handler); + // get value with conversion functions + template PUGI__FN PUGI__UNSIGNED_OVERFLOW U string_to_integer(const char_t* value, U minv, U maxv) + { + U result = 0; + const char_t* s = value; + + while (PUGI__IS_CHARTYPE(*s, ct_space)) + s++; + + bool negative = (*s == '-'); + + s += (*s == '+' || *s == '-'); - if (error == 0) + bool overflow = false; + + if (s[0] == '0' && (s[1] | ' ') == 'x') + { + s += 2; + + // since overflow detection relies on length of the sequence skip leading zeros + while (*s == '0') + s++; + + const char_t* start = s; + + for (;;) { - parser.parse(buffer, xmldoc, optmsk, endch); + if (static_cast(*s - '0') < 10) + result = result * 16 + (*s - '0'); + else if (static_cast((*s | ' ') - 'a') < 6) + result = result * 16 + ((*s | ' ') - 'a' + 10); + else + break; + + s++; } - xml_parse_result result = make_parse_result(static_cast(error), parser.error_offset ? parser.error_offset - buffer : 0); - assert(result.offset >= 0 && static_cast(result.offset) <= length); + size_t digits = static_cast(s - start); + + overflow = digits > sizeof(U) * 2; + } + else + { + // since overflow detection relies on length of the sequence skip leading zeros + while (*s == '0') + s++; - // update allocator state - *static_cast(xmldoc) = parser.alloc; + const char_t* start = s; - // since we removed last character, we have to handle the only possible false positive - if (result && endch == '<') + for (;;) { - // there's no possible well-formed document with < at the end - return make_parse_result(status_unrecognized_tag, length); + if (static_cast(*s - '0') < 10) + result = result * 10 + (*s - '0'); + else + break; + + s++; } - return result; + size_t digits = static_cast(s - start); + + PUGI__STATIC_ASSERT(sizeof(U) == 8 || sizeof(U) == 4 || sizeof(U) == 2); + + const size_t max_digits10 = sizeof(U) == 8 ? 20 : sizeof(U) == 4 ? 10 : 5; + const char_t max_lead = sizeof(U) == 8 ? '1' : sizeof(U) == 4 ? '4' : '6'; + const size_t high_bit = sizeof(U) * 8 - 1; + + overflow = digits >= max_digits10 && !(digits == max_digits10 && (*start < max_lead || (*start == max_lead && result >> high_bit))); } - }; - // Output facilities - xml_encoding get_write_native_encoding() + if (negative) + { + // Workaround for crayc++ CC-3059: Expected no overflow in routine. + #ifdef _CRAYC + return (overflow || result > ~minv + 1) ? minv : ~result + 1; + #else + return (overflow || result > 0 - minv) ? minv : 0 - result; + #endif + } + else + return (overflow || result > maxv) ? maxv : result; + } + + PUGI__FN int get_value_int(const char_t* value) + { + return string_to_integer(value, static_cast(INT_MIN), INT_MAX); + } + + PUGI__FN unsigned int get_value_uint(const char_t* value) + { + return string_to_integer(value, 0, UINT_MAX); + } + + PUGI__FN double get_value_double(const char_t* value) { #ifdef PUGIXML_WCHAR_MODE - return get_wchar_encoding(); + return wcstod(value, 0); #else - return encoding_utf8; + return strtod(value, 0); #endif } - xml_encoding get_write_encoding(xml_encoding encoding) + PUGI__FN float get_value_float(const char_t* value) { - // replace wchar encoding with utf implementation - if (encoding == encoding_wchar) return get_wchar_encoding(); - - // replace utf16 encoding with utf16 with specific endianness - if (encoding == encoding_utf16) return is_little_endian() ? encoding_utf16_le : encoding_utf16_be; - - // replace utf32 encoding with utf32 with specific endianness - if (encoding == encoding_utf32) return is_little_endian() ? encoding_utf32_le : encoding_utf32_be; + #ifdef PUGIXML_WCHAR_MODE + return static_cast(wcstod(value, 0)); + #else + return static_cast(strtod(value, 0)); + #endif + } - // only do autodetection if no explicit encoding is requested - if (encoding != encoding_auto) return encoding; + PUGI__FN bool get_value_bool(const char_t* value) + { + // only look at first char + char_t first = *value; - // assume utf8 encoding - return encoding_utf8; + // 1*, t* (true), T* (True), y* (yes), Y* (YES) + return (first == '1' || first == 't' || first == 'T' || first == 'y' || first == 'Y'); } -#ifdef PUGIXML_WCHAR_MODE - size_t get_valid_length(const char_t* data, size_t length) +#ifdef PUGIXML_HAS_LONG_LONG + PUGI__FN long long get_value_llong(const char_t* value) { - assert(length > 0); + return string_to_integer(value, static_cast(LLONG_MIN), LLONG_MAX); + } - // discard last character if it's the lead of a surrogate pair - return (sizeof(wchar_t) == 2 && (unsigned)(static_cast(data[length - 1]) - 0xD800) < 0x400) ? length - 1 : length; + PUGI__FN unsigned long long get_value_ullong(const char_t* value) + { + return string_to_integer(value, 0, ULLONG_MAX); } +#endif - size_t convert_buffer(char* result, const char_t* data, size_t length, xml_encoding encoding) + template PUGI__FN PUGI__UNSIGNED_OVERFLOW char_t* integer_to_string(char_t* begin, char_t* end, U value, bool negative) { - // only endian-swapping is required - if (need_endian_swap_utf(encoding, get_wchar_encoding())) - { - convert_wchar_endian_swap(reinterpret_cast(result), data, length); + char_t* result = end - 1; + U rest = negative ? 0 - value : value; - return length * sizeof(char_t); - } - - // convert to utf8 - if (encoding == encoding_utf8) + do { - uint8_t* dest = reinterpret_cast(result); + *result-- = static_cast('0' + (rest % 10)); + rest /= 10; + } + while (rest); - uint8_t* end = sizeof(wchar_t) == 2 ? - utf_decoder::decode_utf16_block(reinterpret_cast(data), length, dest) : - utf_decoder::decode_utf32_block(reinterpret_cast(data), length, dest); + assert(result >= begin); + (void)begin; - return static_cast(end - dest); - } + *result = '-'; - // convert to utf16 - if (encoding == encoding_utf16_be || encoding == encoding_utf16_le) - { - uint16_t* dest = reinterpret_cast(result); + return result + !negative; + } - // convert to native utf16 - uint16_t* end = utf_decoder::decode_utf32_block(reinterpret_cast(data), length, dest); + // set value with conversion functions + template + PUGI__FN bool set_value_ascii(String& dest, Header& header, uintptr_t header_mask, char* buf) + { + #ifdef PUGIXML_WCHAR_MODE + char_t wbuf[128]; + assert(strlen(buf) < sizeof(wbuf) / sizeof(wbuf[0])); - // swap if necessary - xml_encoding native_encoding = is_little_endian() ? encoding_utf16_le : encoding_utf16_be; + size_t offset = 0; + for (; buf[offset]; ++offset) wbuf[offset] = buf[offset]; - if (native_encoding != encoding) convert_utf_endian_swap(dest, dest, static_cast(end - dest)); + return strcpy_insitu(dest, header, header_mask, wbuf, offset); + #else + return strcpy_insitu(dest, header, header_mask, buf, strlen(buf)); + #endif + } - return static_cast(end - dest) * sizeof(uint16_t); - } + template + PUGI__FN bool set_value_integer(String& dest, Header& header, uintptr_t header_mask, U value, bool negative) + { + char_t buf[64]; + char_t* end = buf + sizeof(buf) / sizeof(buf[0]); + char_t* begin = integer_to_string(buf, end, value, negative); - // convert to utf32 - if (encoding == encoding_utf32_be || encoding == encoding_utf32_le) - { - uint32_t* dest = reinterpret_cast(result); + return strcpy_insitu(dest, header, header_mask, begin, end - begin); + } - // convert to native utf32 - uint32_t* end = utf_decoder::decode_utf16_block(reinterpret_cast(data), length, dest); + template + PUGI__FN bool set_value_convert(String& dest, Header& header, uintptr_t header_mask, float value) + { + char buf[128]; + PUGI__SNPRINTF(buf, "%.9g", value); - // swap if necessary - xml_encoding native_encoding = is_little_endian() ? encoding_utf32_le : encoding_utf32_be; + return set_value_ascii(dest, header, header_mask, buf); + } - if (native_encoding != encoding) convert_utf_endian_swap(dest, dest, static_cast(end - dest)); + template + PUGI__FN bool set_value_convert(String& dest, Header& header, uintptr_t header_mask, double value) + { + char buf[128]; + PUGI__SNPRINTF(buf, "%.17g", value); - return static_cast(end - dest) * sizeof(uint32_t); - } + return set_value_ascii(dest, header, header_mask, buf); + } - assert(!"Invalid encoding"); - return 0; + template + PUGI__FN bool set_value_bool(String& dest, Header& header, uintptr_t header_mask, bool value) + { + return strcpy_insitu(dest, header, header_mask, value ? PUGIXML_TEXT("true") : PUGIXML_TEXT("false"), value ? 4 : 5); } -#else - size_t get_valid_length(const char_t* data, size_t length) + + PUGI__FN xml_parse_result load_buffer_impl(xml_document_struct* doc, xml_node_struct* root, void* contents, size_t size, unsigned int options, xml_encoding encoding, bool is_mutable, bool own, char_t** out_buffer) { - assert(length > 4); + // check input buffer + if (!contents && size) return make_parse_result(status_io_error); - for (size_t i = 1; i <= 4; ++i) - { - uint8_t ch = static_cast(data[length - i]); + // get actual encoding + xml_encoding buffer_encoding = impl::get_buffer_encoding(encoding, contents, size); - // either a standalone character or a leading one - if ((ch & 0xc0) != 0x80) return length - i; - } + // get private buffer + char_t* buffer = 0; + size_t length = 0; - // there are four non-leading characters at the end, sequence tail is broken so might as well process the whole chunk - return length; + if (!impl::convert_buffer(buffer, length, buffer_encoding, contents, size, is_mutable)) return impl::make_parse_result(status_out_of_memory); + + // delete original buffer if we performed a conversion + if (own && buffer != contents && contents) impl::xml_memory::deallocate(contents); + + // grab onto buffer if it's our buffer, user is responsible for deallocating contents himself + if (own || buffer != contents) *out_buffer = buffer; + + // store buffer for offset_debug + doc->buffer = buffer; + + // parse + xml_parse_result res = impl::xml_parser::parse(buffer, length, doc, root, options); + + // remember encoding + res.encoding = buffer_encoding; + + return res; } - size_t convert_buffer(char* result, const char_t* data, size_t length, xml_encoding encoding) + // we need to get length of entire file to load it in memory; the only (relatively) sane way to do it is via seek/tell trick + PUGI__FN xml_parse_status get_file_size(FILE* file, size_t& out_result) { - if (encoding == encoding_utf16_be || encoding == encoding_utf16_le) - { - uint16_t* dest = reinterpret_cast(result); + #if defined(PUGI__MSVC_CRT_VERSION) && PUGI__MSVC_CRT_VERSION >= 1400 && !defined(_WIN32_WCE) + // there are 64-bit versions of fseek/ftell, let's use them + typedef __int64 length_type; + + _fseeki64(file, 0, SEEK_END); + length_type length = _ftelli64(file); + _fseeki64(file, 0, SEEK_SET); + #elif defined(__MINGW32__) && !defined(__NO_MINGW_LFS) && (!defined(__STRICT_ANSI__) || defined(__MINGW64_VERSION_MAJOR)) + // there are 64-bit versions of fseek/ftell, let's use them + typedef off64_t length_type; + + fseeko64(file, 0, SEEK_END); + length_type length = ftello64(file); + fseeko64(file, 0, SEEK_SET); + #else + // if this is a 32-bit OS, long is enough; if this is a unix system, long is 64-bit, which is enough; otherwise we can't do anything anyway. + typedef long length_type; - // convert to native utf16 - uint16_t* end = utf_decoder::decode_utf8_block(reinterpret_cast(data), length, dest); + fseek(file, 0, SEEK_END); + length_type length = ftell(file); + fseek(file, 0, SEEK_SET); + #endif - // swap if necessary - xml_encoding native_encoding = is_little_endian() ? encoding_utf16_le : encoding_utf16_be; + // check for I/O errors + if (length < 0) return status_io_error; - if (native_encoding != encoding) convert_utf_endian_swap(dest, dest, static_cast(end - dest)); + // check for overflow + size_t result = static_cast(length); - return static_cast(end - dest) * sizeof(uint16_t); - } + if (static_cast(result) != length) return status_out_of_memory; - if (encoding == encoding_utf32_be || encoding == encoding_utf32_le) - { - uint32_t* dest = reinterpret_cast(result); + // finalize + out_result = result; - // convert to native utf32 - uint32_t* end = utf_decoder::decode_utf8_block(reinterpret_cast(data), length, dest); + return status_ok; + } - // swap if necessary - xml_encoding native_encoding = is_little_endian() ? encoding_utf32_le : encoding_utf32_be; + // This function assumes that buffer has extra sizeof(char_t) writable bytes after size + PUGI__FN size_t zero_terminate_buffer(void* buffer, size_t size, xml_encoding encoding) + { + // We only need to zero-terminate if encoding conversion does not do it for us + #ifdef PUGIXML_WCHAR_MODE + xml_encoding wchar_encoding = get_wchar_encoding(); - if (native_encoding != encoding) convert_utf_endian_swap(dest, dest, static_cast(end - dest)); + if (encoding == wchar_encoding || need_endian_swap_utf(encoding, wchar_encoding)) + { + size_t length = size / sizeof(char_t); - return static_cast(end - dest) * sizeof(uint32_t); + static_cast(buffer)[length] = 0; + return (length + 1) * sizeof(char_t); } + #else + if (encoding == encoding_utf8) + { + static_cast(buffer)[size] = 0; + return size + 1; + } + #endif - assert(!"Invalid encoding"); - return 0; + return size; } -#endif - class xml_buffered_writer + PUGI__FN xml_parse_result load_file_impl(xml_document_struct* doc, FILE* file, unsigned int options, xml_encoding encoding, char_t** out_buffer) { - xml_buffered_writer(const xml_buffered_writer&); - xml_buffered_writer& operator=(const xml_buffered_writer&); + if (!file) return make_parse_result(status_file_not_found); - public: - xml_buffered_writer(xml_writer& writer, xml_encoding user_encoding): writer(writer), bufsize(0), encoding(get_write_encoding(user_encoding)) - { - } + // get file size (can result in I/O errors) + size_t size = 0; + xml_parse_status size_status = get_file_size(file, size); + if (size_status != status_ok) return make_parse_result(size_status); + + size_t max_suffix_size = sizeof(char_t); + + // allocate buffer for the whole file + char* contents = static_cast(xml_memory::allocate(size + max_suffix_size)); + if (!contents) return make_parse_result(status_out_of_memory); - ~xml_buffered_writer() + // read file in memory + size_t read_size = fread(contents, 1, size, file); + + if (read_size != size) { - flush(); + xml_memory::deallocate(contents); + return make_parse_result(status_io_error); } - void flush() + xml_encoding real_encoding = get_buffer_encoding(encoding, contents, size); + + return load_buffer_impl(doc, doc, contents, zero_terminate_buffer(contents, size, real_encoding), options, real_encoding, true, true, out_buffer); + } + + PUGI__FN void close_file(FILE* file) + { + fclose(file); + } + +#ifndef PUGIXML_NO_STL + template struct xml_stream_chunk + { + static xml_stream_chunk* create() { - flush(buffer, bufsize); - bufsize = 0; + void* memory = xml_memory::allocate(sizeof(xml_stream_chunk)); + if (!memory) return 0; + + return new (memory) xml_stream_chunk(); } - void flush(const char_t* data, size_t size) + static void destroy(xml_stream_chunk* chunk) { - if (size == 0) return; - - // fast path, just write data - if (encoding == get_write_native_encoding()) - writer.write(data, size * sizeof(char_t)); - else + // free chunk chain + while (chunk) { - // convert chunk - size_t result = convert_buffer(scratch, data, size, encoding); - assert(result <= sizeof(scratch)); + xml_stream_chunk* next_ = chunk->next; - // write data - writer.write(scratch, result); + xml_memory::deallocate(chunk); + + chunk = next_; } } - void write(const char_t* data, size_t length) + xml_stream_chunk(): next(0), size(0) { - if (bufsize + length > bufcapacity) - { - // flush the remaining buffer contents - flush(); - - // handle large chunks - if (length > bufcapacity) - { - if (encoding == get_write_native_encoding()) - { - // fast path, can just write data chunk - writer.write(data, length * sizeof(char_t)); - return; - } - - // need to convert in suitable chunks - while (length > bufcapacity) - { - // get chunk size by selecting such number of characters that are guaranteed to fit into scratch buffer - // and form a complete codepoint sequence (i.e. discard start of last codepoint if necessary) - size_t chunk_size = get_valid_length(data, bufcapacity); + } - // convert chunk and write - flush(data, chunk_size); + xml_stream_chunk* next; + size_t size; - // iterate - data += chunk_size; - length -= chunk_size; - } + T data[xml_memory_page_size / sizeof(T)]; + }; - // small tail is copied below - bufsize = 0; - } - } + template PUGI__FN xml_parse_status load_stream_data_noseek(std::basic_istream& stream, void** out_buffer, size_t* out_size) + { + auto_deleter > chunks(0, xml_stream_chunk::destroy); - memcpy(buffer + bufsize, data, length * sizeof(char_t)); - bufsize += length; - } + // read file to a chunk list + size_t total = 0; + xml_stream_chunk* last = 0; - void write(const char_t* data) + while (!stream.eof()) { - write(data, strlength(data)); - } + // allocate new chunk + xml_stream_chunk* chunk = xml_stream_chunk::create(); + if (!chunk) return status_out_of_memory; - void write(char_t d0) - { - if (bufsize + 1 > bufcapacity) flush(); + // append chunk to list + if (last) last = last->next = chunk; + else chunks.data = last = chunk; - buffer[bufsize + 0] = d0; - bufsize += 1; - } + // read data to chunk + stream.read(chunk->data, static_cast(sizeof(chunk->data) / sizeof(T))); + chunk->size = static_cast(stream.gcount()) * sizeof(T); - void write(char_t d0, char_t d1) - { - if (bufsize + 2 > bufcapacity) flush(); + // read may set failbit | eofbit in case gcount() is less than read length, so check for other I/O errors + if (stream.bad() || (!stream.eof() && stream.fail())) return status_io_error; - buffer[bufsize + 0] = d0; - buffer[bufsize + 1] = d1; - bufsize += 2; + // guard against huge files (chunk size is small enough to make this overflow check work) + if (total + chunk->size < total) return status_out_of_memory; + total += chunk->size; } - void write(char_t d0, char_t d1, char_t d2) - { - if (bufsize + 3 > bufcapacity) flush(); + size_t max_suffix_size = sizeof(char_t); - buffer[bufsize + 0] = d0; - buffer[bufsize + 1] = d1; - buffer[bufsize + 2] = d2; - bufsize += 3; - } + // copy chunk list to a contiguous buffer + char* buffer = static_cast(xml_memory::allocate(total + max_suffix_size)); + if (!buffer) return status_out_of_memory; - void write(char_t d0, char_t d1, char_t d2, char_t d3) - { - if (bufsize + 4 > bufcapacity) flush(); + char* write = buffer; - buffer[bufsize + 0] = d0; - buffer[bufsize + 1] = d1; - buffer[bufsize + 2] = d2; - buffer[bufsize + 3] = d3; - bufsize += 4; + for (xml_stream_chunk* chunk = chunks.data; chunk; chunk = chunk->next) + { + assert(write + chunk->size <= buffer + total); + memcpy(write, chunk->data, chunk->size); + write += chunk->size; } - void write(char_t d0, char_t d1, char_t d2, char_t d3, char_t d4) - { - if (bufsize + 5 > bufcapacity) flush(); + assert(write == buffer + total); - buffer[bufsize + 0] = d0; - buffer[bufsize + 1] = d1; - buffer[bufsize + 2] = d2; - buffer[bufsize + 3] = d3; - buffer[bufsize + 4] = d4; - bufsize += 5; - } + // return buffer + *out_buffer = buffer; + *out_size = total; - void write(char_t d0, char_t d1, char_t d2, char_t d3, char_t d4, char_t d5) - { - if (bufsize + 6 > bufcapacity) flush(); + return status_ok; + } - buffer[bufsize + 0] = d0; - buffer[bufsize + 1] = d1; - buffer[bufsize + 2] = d2; - buffer[bufsize + 3] = d3; - buffer[bufsize + 4] = d4; - buffer[bufsize + 5] = d5; - bufsize += 6; - } + template PUGI__FN xml_parse_status load_stream_data_seek(std::basic_istream& stream, void** out_buffer, size_t* out_size) + { + // get length of remaining data in stream + typename std::basic_istream::pos_type pos = stream.tellg(); + stream.seekg(0, std::ios::end); + std::streamoff length = stream.tellg() - pos; + stream.seekg(pos); - // utf8 maximum expansion: x4 (-> utf32) - // utf16 maximum expansion: x2 (-> utf32) - // utf32 maximum expansion: x1 - enum { bufcapacity = 2048 }; + if (stream.fail() || pos < 0) return status_io_error; - char_t buffer[bufcapacity]; - char scratch[4 * bufcapacity]; + // guard against huge files + size_t read_length = static_cast(length); - xml_writer& writer; - size_t bufsize; - xml_encoding encoding; - }; + if (static_cast(read_length) != length || length < 0) return status_out_of_memory; - void write_bom(xml_writer& writer, xml_encoding encoding) - { - switch (encoding) - { - case encoding_utf8: - writer.write("\xef\xbb\xbf", 3); - break; + size_t max_suffix_size = sizeof(char_t); - case encoding_utf16_be: - writer.write("\xfe\xff", 2); - break; + // read stream data into memory (guard against stream exceptions with buffer holder) + auto_deleter buffer(xml_memory::allocate(read_length * sizeof(T) + max_suffix_size), xml_memory::deallocate); + if (!buffer.data) return status_out_of_memory; - case encoding_utf16_le: - writer.write("\xff\xfe", 2); - break; + stream.read(static_cast(buffer.data), static_cast(read_length)); - case encoding_utf32_be: - writer.write("\x00\x00\xfe\xff", 4); - break; + // read may set failbit | eofbit in case gcount() is less than read_length (i.e. line ending conversion), so check for other I/O errors + if (stream.bad() || (!stream.eof() && stream.fail())) return status_io_error; - case encoding_utf32_le: - writer.write("\xff\xfe\x00\x00", 4); - break; + // return buffer + size_t actual_length = static_cast(stream.gcount()); + assert(actual_length <= read_length); - default: - assert(!"Invalid encoding"); - } + *out_buffer = buffer.release(); + *out_size = actual_length * sizeof(T); + + return status_ok; } - void text_output_escaped(xml_buffered_writer& writer, const char_t* s, chartypex_t type) + template PUGI__FN xml_parse_result load_stream_impl(xml_document_struct* doc, std::basic_istream& stream, unsigned int options, xml_encoding encoding, char_t** out_buffer) { - while (*s) - { - const char_t* prev = s; - - // While *s is a usual symbol - while (!IS_CHARTYPEX(*s, type)) ++s; - - writer.write(prev, static_cast(s - prev)); + void* buffer = 0; + size_t size = 0; + xml_parse_status status = status_ok; - switch (*s) - { - case 0: break; - case '&': - writer.write('&', 'a', 'm', 'p', ';'); - ++s; - break; - case '<': - writer.write('&', 'l', 't', ';'); - ++s; - break; - case '>': - writer.write('&', 'g', 't', ';'); - ++s; - break; - case '"': - writer.write('&', 'q', 'u', 'o', 't', ';'); - ++s; - break; - default: // s is not a usual symbol - { - unsigned int ch = static_cast(*s++); - assert(ch < 32); + // if stream has an error bit set, bail out (otherwise tellg() can fail and we'll clear error bits) + if (stream.fail()) return make_parse_result(status_io_error); - writer.write('&', '#', static_cast((ch / 10) + '0'), static_cast((ch % 10) + '0'), ';'); - } - } + // load stream to memory (using seek-based implementation if possible, since it's faster and takes less memory) + if (stream.tellg() < 0) + { + stream.clear(); // clear error flags that could be set by a failing tellg + status = load_stream_data_noseek(stream, &buffer, &size); } + else + status = load_stream_data_seek(stream, &buffer, &size); + + if (status != status_ok) return make_parse_result(status); + + xml_encoding real_encoding = get_buffer_encoding(encoding, buffer, size); + + return load_buffer_impl(doc, doc, buffer, zero_terminate_buffer(buffer, size, real_encoding), options, real_encoding, true, true, out_buffer); } +#endif - void text_output_cdata(xml_buffered_writer& writer, const char_t* s) +#if defined(PUGI__MSVC_CRT_VERSION) || defined(__BORLANDC__) || (defined(__MINGW32__) && (!defined(__STRICT_ANSI__) || defined(__MINGW64_VERSION_MAJOR))) + PUGI__FN FILE* open_file_wide(const wchar_t* path, const wchar_t* mode) { - do - { - writer.write('<', '!', '[', 'C', 'D'); - writer.write('A', 'T', 'A', '['); + return _wfopen(path, mode); + } +#else + PUGI__FN char* convert_path_heap(const wchar_t* str) + { + assert(str); - const char_t* prev = s; + // first pass: get length in utf8 characters + size_t length = strlength_wide(str); + size_t size = as_utf8_begin(str, length); - // look for ]]> sequence - we can't output it as is since it terminates CDATA - while (*s && !(s[0] == ']' && s[1] == ']' && s[2] == '>')) ++s; + // allocate resulting string + char* result = static_cast(xml_memory::allocate(size + 1)); + if (!result) return 0; - // skip ]] if we stopped at ]]>, > will go to the next CDATA section - if (*s) s += 2; + // second pass: convert to utf8 + as_utf8_end(result, size, str, length); - writer.write(prev, static_cast(s - prev)); + // zero-terminate + result[size] = 0; - writer.write(']', ']', '>'); - } - while (*s); + return result; } - void node_output_attributes(xml_buffered_writer& writer, const xml_node& node) + PUGI__FN FILE* open_file_wide(const wchar_t* path, const wchar_t* mode) { - const char_t* default_name = PUGIXML_TEXT(":anonymous"); + // there is no standard function to open wide paths, so our best bet is to try utf8 path + char* path_utf8 = convert_path_heap(path); + if (!path_utf8) return 0; - for (xml_attribute a = node.first_attribute(); a; a = a.next_attribute()) - { - writer.write(' '); - writer.write(a.name()[0] ? a.name() : default_name); - writer.write('=', '"'); + // convert mode to ASCII (we mirror _wfopen interface) + char mode_ascii[4] = {0}; + for (size_t i = 0; mode[i]; ++i) mode_ascii[i] = static_cast(mode[i]); - text_output_escaped(writer, a.value(), ctx_special_attr); + // try to open the utf8 path + FILE* result = fopen(path_utf8, mode_ascii); - writer.write('"'); - } + // free dummy buffer + xml_memory::deallocate(path_utf8); + + return result; } +#endif - void node_output(xml_buffered_writer& writer, const xml_node& node, const char_t* indent, unsigned int flags, unsigned int depth) + PUGI__FN bool save_file_impl(const xml_document& doc, FILE* file, const char_t* indent, unsigned int flags, xml_encoding encoding) { - const char_t* default_name = PUGIXML_TEXT(":anonymous"); + if (!file) return false; + + xml_writer_file writer(file); + doc.save(writer, indent, flags, encoding); - if ((flags & format_indent) != 0 && (flags & format_raw) == 0) - for (unsigned int i = 0; i < depth; ++i) writer.write(indent); + return ferror(file) == 0; + } - switch (node.type()) - { - case node_document: + struct name_null_sentry + { + xml_node_struct* node; + char_t* name; + + name_null_sentry(xml_node_struct* node_): node(node_), name(node_->name) { - for (xml_node n = node.first_child(); n; n = n.next_sibling()) - node_output(writer, n, indent, flags, depth); - break; + node->name = 0; } - - case node_element: + + ~name_null_sentry() { - const char_t* name = node.name()[0] ? node.name() : default_name; + node->name = name; + } + }; +PUGI__NS_END - writer.write('<'); - writer.write(name); +namespace pugi +{ + PUGI__FN xml_writer_file::xml_writer_file(void* file_): file(file_) + { + } - node_output_attributes(writer, node); + PUGI__FN void xml_writer_file::write(const void* data, size_t size) + { + size_t result = fwrite(data, 1, size, static_cast(file)); + (void)!result; // unfortunately we can't do proper error handling here + } - if (flags & format_raw) - { - if (!node.first_child()) - writer.write(' ', '/', '>'); - else - { - writer.write('>'); +#ifndef PUGIXML_NO_STL + PUGI__FN xml_writer_stream::xml_writer_stream(std::basic_ostream >& stream): narrow_stream(&stream), wide_stream(0) + { + } - for (xml_node n = node.first_child(); n; n = n.next_sibling()) - node_output(writer, n, indent, flags, depth + 1); + PUGI__FN xml_writer_stream::xml_writer_stream(std::basic_ostream >& stream): narrow_stream(0), wide_stream(&stream) + { + } - writer.write('<', '/'); - writer.write(name); - writer.write('>'); - } - } - else if (!node.first_child()) - writer.write(' ', '/', '>', '\n'); - else if (node.first_child() == node.last_child() && (node.first_child().type() == node_pcdata || node.first_child().type() == node_cdata)) - { - writer.write('>'); + PUGI__FN void xml_writer_stream::write(const void* data, size_t size) + { + if (narrow_stream) + { + assert(!wide_stream); + narrow_stream->write(reinterpret_cast(data), static_cast(size)); + } + else + { + assert(wide_stream); + assert(size % sizeof(wchar_t) == 0); - if (node.first_child().type() == node_pcdata) - text_output_escaped(writer, node.first_child().value(), ctx_special_pcdata); - else - text_output_cdata(writer, node.first_child().value()); + wide_stream->write(reinterpret_cast(data), static_cast(size / sizeof(wchar_t))); + } + } +#endif - writer.write('<', '/'); - writer.write(name); - writer.write('>', '\n'); - } - else - { - writer.write('>', '\n'); - - for (xml_node n = node.first_child(); n; n = n.next_sibling()) - node_output(writer, n, indent, flags, depth + 1); + PUGI__FN xml_tree_walker::xml_tree_walker(): _depth(0) + { + } - if ((flags & format_indent) != 0 && (flags & format_raw) == 0) - for (unsigned int i = 0; i < depth; ++i) writer.write(indent); - - writer.write('<', '/'); - writer.write(name); - writer.write('>', '\n'); - } + PUGI__FN xml_tree_walker::~xml_tree_walker() + { + } - break; - } - - case node_pcdata: - text_output_escaped(writer, node.value(), ctx_special_pcdata); - if ((flags & format_raw) == 0) writer.write('\n'); - break; + PUGI__FN int xml_tree_walker::depth() const + { + return _depth; + } - case node_cdata: - text_output_cdata(writer, node.value()); - if ((flags & format_raw) == 0) writer.write('\n'); - break; + PUGI__FN bool xml_tree_walker::begin(xml_node&) + { + return true; + } - case node_comment: - writer.write('<', '!', '-', '-'); - writer.write(node.value()); - writer.write('-', '-', '>'); - if ((flags & format_raw) == 0) writer.write('\n'); - break; + PUGI__FN bool xml_tree_walker::end(xml_node&) + { + return true; + } - case node_pi: - case node_declaration: - writer.write('<', '?'); - writer.write(node.name()[0] ? node.name() : default_name); + PUGI__FN xml_attribute::xml_attribute(): _attr(0) + { + } - if (node.type() == node_declaration) - { - node_output_attributes(writer, node); - } - else if (node.value()[0]) - { - writer.write(' '); - writer.write(node.value()); - } + PUGI__FN xml_attribute::xml_attribute(xml_attribute_struct* attr): _attr(attr) + { + } - writer.write('?', '>'); - if ((flags & format_raw) == 0) writer.write('\n'); - break; + PUGI__FN static void unspecified_bool_xml_attribute(xml_attribute***) + { + } + + PUGI__FN xml_attribute::operator xml_attribute::unspecified_bool_type() const + { + return _attr ? unspecified_bool_xml_attribute : 0; + } - case node_doctype: - writer.write('<', '!', 'D', 'O', 'C'); - writer.write('T', 'Y', 'P', 'E'); + PUGI__FN bool xml_attribute::operator!() const + { + return !_attr; + } - if (node.value()[0]) - { - writer.write(' '); - writer.write(node.value()); - } + PUGI__FN bool xml_attribute::operator==(const xml_attribute& r) const + { + return (_attr == r._attr); + } - writer.write('>'); - if ((flags & format_raw) == 0) writer.write('\n'); - break; + PUGI__FN bool xml_attribute::operator!=(const xml_attribute& r) const + { + return (_attr != r._attr); + } - default: - assert(!"Invalid node type"); - } + PUGI__FN bool xml_attribute::operator<(const xml_attribute& r) const + { + return (_attr < r._attr); } - inline bool has_declaration(const xml_node& node) + PUGI__FN bool xml_attribute::operator>(const xml_attribute& r) const { - for (xml_node child = node.first_child(); child; child = child.next_sibling()) - { - xml_node_type type = child.type(); + return (_attr > r._attr); + } - if (type == node_declaration) return true; - if (type == node_element) return false; - } + PUGI__FN bool xml_attribute::operator<=(const xml_attribute& r) const + { + return (_attr <= r._attr); + } - return false; + PUGI__FN bool xml_attribute::operator>=(const xml_attribute& r) const + { + return (_attr >= r._attr); } - inline bool allow_insert_child(xml_node_type parent, xml_node_type child) + PUGI__FN xml_attribute xml_attribute::next_attribute() const { - if (parent != node_document && parent != node_element) return false; - if (child == node_document || child == node_null) return false; - if (parent != node_document && (child == node_declaration || child == node_doctype)) return false; + return _attr ? xml_attribute(_attr->next_attribute) : xml_attribute(); + } - return true; + PUGI__FN xml_attribute xml_attribute::previous_attribute() const + { + return _attr && _attr->prev_attribute_c->next_attribute ? xml_attribute(_attr->prev_attribute_c) : xml_attribute(); } - void recursive_copy_skip(xml_node& dest, const xml_node& source, const xml_node& skip) + PUGI__FN const char_t* xml_attribute::as_string(const char_t* def) const { - assert(dest.type() == source.type()); + return (_attr && _attr->value) ? _attr->value + 0 : def; + } - switch (source.type()) - { - case node_element: - { - dest.set_name(source.name()); + PUGI__FN int xml_attribute::as_int(int def) const + { + return (_attr && _attr->value) ? impl::get_value_int(_attr->value) : def; + } - for (xml_attribute a = source.first_attribute(); a; a = a.next_attribute()) - dest.append_attribute(a.name()).set_value(a.value()); + PUGI__FN unsigned int xml_attribute::as_uint(unsigned int def) const + { + return (_attr && _attr->value) ? impl::get_value_uint(_attr->value) : def; + } - for (xml_node c = source.first_child(); c; c = c.next_sibling()) - { - if (c == skip) continue; + PUGI__FN double xml_attribute::as_double(double def) const + { + return (_attr && _attr->value) ? impl::get_value_double(_attr->value) : def; + } - xml_node cc = dest.append_child(c.type()); - assert(cc); + PUGI__FN float xml_attribute::as_float(float def) const + { + return (_attr && _attr->value) ? impl::get_value_float(_attr->value) : def; + } - recursive_copy_skip(cc, c, skip); - } + PUGI__FN bool xml_attribute::as_bool(bool def) const + { + return (_attr && _attr->value) ? impl::get_value_bool(_attr->value) : def; + } - break; - } +#ifdef PUGIXML_HAS_LONG_LONG + PUGI__FN long long xml_attribute::as_llong(long long def) const + { + return (_attr && _attr->value) ? impl::get_value_llong(_attr->value) : def; + } - case node_pcdata: - case node_cdata: - case node_comment: - case node_doctype: - dest.set_value(source.value()); - break; + PUGI__FN unsigned long long xml_attribute::as_ullong(unsigned long long def) const + { + return (_attr && _attr->value) ? impl::get_value_ullong(_attr->value) : def; + } +#endif - case node_pi: - dest.set_name(source.name()); - dest.set_value(source.value()); - break; + PUGI__FN bool xml_attribute::empty() const + { + return !_attr; + } - case node_declaration: - { - dest.set_name(source.name()); + PUGI__FN const char_t* xml_attribute::name() const + { + return (_attr && _attr->name) ? _attr->name + 0 : PUGIXML_TEXT(""); + } - for (xml_attribute a = source.first_attribute(); a; a = a.next_attribute()) - dest.append_attribute(a.name()).set_value(a.value()); + PUGI__FN const char_t* xml_attribute::value() const + { + return (_attr && _attr->value) ? _attr->value + 0 : PUGIXML_TEXT(""); + } - break; - } + PUGI__FN size_t xml_attribute::hash_value() const + { + return static_cast(reinterpret_cast(_attr) / sizeof(xml_attribute_struct)); + } - default: - assert(!"Invalid node type"); - } + PUGI__FN xml_attribute_struct* xml_attribute::internal_object() const + { + return _attr; } - // we need to get length of entire file to load it in memory; the only (relatively) sane way to do it is via seek/tell trick - xml_parse_status get_file_size(FILE* file, size_t& out_result) + PUGI__FN xml_attribute& xml_attribute::operator=(const char_t* rhs) { - #if defined(_MSC_VER) && _MSC_VER >= 1400 - // there are 64-bit versions of fseek/ftell, let's use them - typedef __int64 length_type; + set_value(rhs); + return *this; + } - _fseeki64(file, 0, SEEK_END); - length_type length = _ftelli64(file); - _fseeki64(file, 0, SEEK_SET); - #elif defined(__MINGW32__) && !defined(__NO_MINGW_LFS) && !defined(__STRICT_ANSI__) - // there are 64-bit versions of fseek/ftell, let's use them - typedef off64_t length_type; + PUGI__FN xml_attribute& xml_attribute::operator=(int rhs) + { + set_value(rhs); + return *this; + } - fseeko64(file, 0, SEEK_END); - length_type length = ftello64(file); - fseeko64(file, 0, SEEK_SET); - #else - // if this is a 32-bit OS, long is enough; if this is a unix system, long is 64-bit, which is enough; otherwise we can't do anything anyway. - typedef long length_type; + PUGI__FN xml_attribute& xml_attribute::operator=(unsigned int rhs) + { + set_value(rhs); + return *this; + } - fseek(file, 0, SEEK_END); - length_type length = ftell(file); - fseek(file, 0, SEEK_SET); - #endif + PUGI__FN xml_attribute& xml_attribute::operator=(long rhs) + { + set_value(rhs); + return *this; + } - // check for I/O errors - if (length < 0) return status_io_error; - - // check for overflow - size_t result = static_cast(length); + PUGI__FN xml_attribute& xml_attribute::operator=(unsigned long rhs) + { + set_value(rhs); + return *this; + } - if (static_cast(result) != length) return status_out_of_memory; + PUGI__FN xml_attribute& xml_attribute::operator=(double rhs) + { + set_value(rhs); + return *this; + } - // finalize - out_result = result; + PUGI__FN xml_attribute& xml_attribute::operator=(float rhs) + { + set_value(rhs); + return *this; + } - return status_ok; + PUGI__FN xml_attribute& xml_attribute::operator=(bool rhs) + { + set_value(rhs); + return *this; } - xml_parse_result load_file_impl(xml_document& doc, FILE* file, unsigned int options, xml_encoding encoding) +#ifdef PUGIXML_HAS_LONG_LONG + PUGI__FN xml_attribute& xml_attribute::operator=(long long rhs) { - if (!file) return make_parse_result(status_file_not_found); + set_value(rhs); + return *this; + } - // get file size (can result in I/O errors) - size_t size = 0; - xml_parse_status size_status = get_file_size(file, size); + PUGI__FN xml_attribute& xml_attribute::operator=(unsigned long long rhs) + { + set_value(rhs); + return *this; + } +#endif - if (size_status != status_ok) - { - fclose(file); - return make_parse_result(size_status); - } - - // allocate buffer for the whole file - char* contents = static_cast(global_allocate(size > 0 ? size : 1)); + PUGI__FN bool xml_attribute::set_name(const char_t* rhs) + { + if (!_attr) return false; - if (!contents) - { - fclose(file); - return make_parse_result(status_out_of_memory); - } + return impl::strcpy_insitu(_attr->name, _attr->header, impl::xml_memory_page_name_allocated_mask, rhs, impl::strlength(rhs)); + } - // read file in memory - size_t read_size = fread(contents, 1, size, file); - fclose(file); + PUGI__FN bool xml_attribute::set_value(const char_t* rhs) + { + if (!_attr) return false; - if (read_size != size) - { - global_deallocate(contents); - return make_parse_result(status_io_error); - } - - return doc.load_buffer_inplace_own(contents, size, options, encoding); + return impl::strcpy_insitu(_attr->value, _attr->header, impl::xml_memory_page_value_allocated_mask, rhs, impl::strlength(rhs)); } -#ifndef PUGIXML_NO_STL - template xml_parse_result load_stream_impl(xml_document& doc, std::basic_istream& stream, unsigned int options, xml_encoding encoding) + PUGI__FN bool xml_attribute::set_value(int rhs) { - // get length of remaining data in stream - typename std::basic_istream::pos_type pos = stream.tellg(); - stream.seekg(0, std::ios::end); - std::streamoff length = stream.tellg() - pos; - stream.seekg(pos); + if (!_attr) return false; - if (stream.fail() || pos < 0) return make_parse_result(status_io_error); + return impl::set_value_integer(_attr->value, _attr->header, impl::xml_memory_page_value_allocated_mask, rhs, rhs < 0); + } - // guard against huge files - size_t read_length = static_cast(length); + PUGI__FN bool xml_attribute::set_value(unsigned int rhs) + { + if (!_attr) return false; - if (static_cast(read_length) != length || length < 0) return make_parse_result(status_out_of_memory); + return impl::set_value_integer(_attr->value, _attr->header, impl::xml_memory_page_value_allocated_mask, rhs, false); + } - // read stream data into memory (guard against stream exceptions with buffer holder) - buffer_holder buffer(global_allocate((read_length > 0 ? read_length : 1) * sizeof(T)), global_deallocate); - if (!buffer.data) return make_parse_result(status_out_of_memory); + PUGI__FN bool xml_attribute::set_value(long rhs) + { + if (!_attr) return false; - stream.read(static_cast(buffer.data), static_cast(read_length)); + return impl::set_value_integer(_attr->value, _attr->header, impl::xml_memory_page_value_allocated_mask, rhs, rhs < 0); + } - // read may set failbit | eofbit in case gcount() is less than read_length (i.e. line ending conversion), so check for other I/O errors - if (stream.bad()) return make_parse_result(status_io_error); + PUGI__FN bool xml_attribute::set_value(unsigned long rhs) + { + if (!_attr) return false; - // load data from buffer - size_t actual_length = static_cast(stream.gcount()); - assert(actual_length <= read_length); + return impl::set_value_integer(_attr->value, _attr->header, impl::xml_memory_page_value_allocated_mask, rhs, false); + } - return doc.load_buffer_inplace_own(buffer.release(), actual_length * sizeof(T), options, encoding); + PUGI__FN bool xml_attribute::set_value(double rhs) + { + if (!_attr) return false; + + return impl::set_value_convert(_attr->value, _attr->header, impl::xml_memory_page_value_allocated_mask, rhs); } -#endif -#if defined(_MSC_VER) || defined(__BORLANDC__) || defined(__MINGW32__) - FILE* open_file_wide(const wchar_t* path, const wchar_t* mode) + PUGI__FN bool xml_attribute::set_value(float rhs) { - return _wfopen(path, mode); + if (!_attr) return false; + + return impl::set_value_convert(_attr->value, _attr->header, impl::xml_memory_page_value_allocated_mask, rhs); } -#else - char* convert_path_heap(const wchar_t* str) + + PUGI__FN bool xml_attribute::set_value(bool rhs) { - assert(str); + if (!_attr) return false; - // first pass: get length in utf8 characters - size_t length = wcslen(str); - size_t size = as_utf8_begin(str, length); + return impl::set_value_bool(_attr->value, _attr->header, impl::xml_memory_page_value_allocated_mask, rhs); + } - // allocate resulting string - char* result = static_cast(global_allocate(size + 1)); - if (!result) return 0; +#ifdef PUGIXML_HAS_LONG_LONG + PUGI__FN bool xml_attribute::set_value(long long rhs) + { + if (!_attr) return false; - // second pass: convert to utf8 - as_utf8_end(result, size, str, length); + return impl::set_value_integer(_attr->value, _attr->header, impl::xml_memory_page_value_allocated_mask, rhs, rhs < 0); + } + + PUGI__FN bool xml_attribute::set_value(unsigned long long rhs) + { + if (!_attr) return false; - return result; + return impl::set_value_integer(_attr->value, _attr->header, impl::xml_memory_page_value_allocated_mask, rhs, false); } +#endif - FILE* open_file_wide(const wchar_t* path, const wchar_t* mode) +#ifdef __BORLANDC__ + PUGI__FN bool operator&&(const xml_attribute& lhs, bool rhs) { - // there is no standard function to open wide paths, so our best bet is to try utf8 path - char* path_utf8 = convert_path_heap(path); - if (!path_utf8) return 0; + return (bool)lhs && rhs; + } - // convert mode to ASCII (we mirror _wfopen interface) - char mode_ascii[4] = {0}; - for (size_t i = 0; mode[i]; ++i) mode_ascii[i] = static_cast(mode[i]); + PUGI__FN bool operator||(const xml_attribute& lhs, bool rhs) + { + return (bool)lhs || rhs; + } +#endif - // try to open the utf8 path - FILE* result = fopen(path_utf8, mode_ascii); + PUGI__FN xml_node::xml_node(): _root(0) + { + } - // free dummy buffer - global_deallocate(path_utf8); + PUGI__FN xml_node::xml_node(xml_node_struct* p): _root(p) + { + } - return result; + PUGI__FN static void unspecified_bool_xml_node(xml_node***) + { } -#endif -} -namespace pugi -{ - xml_writer_file::xml_writer_file(void* file): file(file) + PUGI__FN xml_node::operator xml_node::unspecified_bool_type() const { + return _root ? unspecified_bool_xml_node : 0; } - void xml_writer_file::write(const void* data, size_t size) + PUGI__FN bool xml_node::operator!() const { - fwrite(data, size, 1, static_cast(file)); + return !_root; } -#ifndef PUGIXML_NO_STL - xml_writer_stream::xml_writer_stream(std::basic_ostream >& stream): narrow_stream(&stream), wide_stream(0) + PUGI__FN xml_node::iterator xml_node::begin() const { + return iterator(_root ? _root->first_child + 0 : 0, _root); } - xml_writer_stream::xml_writer_stream(std::basic_ostream >& stream): narrow_stream(0), wide_stream(&stream) + PUGI__FN xml_node::iterator xml_node::end() const { + return iterator(0, _root); } - void xml_writer_stream::write(const void* data, size_t size) + PUGI__FN xml_node::attribute_iterator xml_node::attributes_begin() const { - if (narrow_stream) - { - assert(!wide_stream); - narrow_stream->write(reinterpret_cast(data), static_cast(size)); - } - else - { - assert(wide_stream); - assert(size % sizeof(wchar_t) == 0); + return attribute_iterator(_root ? _root->first_attribute + 0 : 0, _root); + } - wide_stream->write(reinterpret_cast(data), static_cast(size / sizeof(wchar_t))); - } + PUGI__FN xml_node::attribute_iterator xml_node::attributes_end() const + { + return attribute_iterator(0, _root); } -#endif - xml_tree_walker::xml_tree_walker(): _depth(0) + PUGI__FN xml_object_range xml_node::children() const { + return xml_object_range(begin(), end()); } - - xml_tree_walker::~xml_tree_walker() + + PUGI__FN xml_object_range xml_node::children(const char_t* name_) const { + return xml_object_range(xml_named_node_iterator(child(name_)._root, _root, name_), xml_named_node_iterator(0, _root, name_)); } - int xml_tree_walker::depth() const + PUGI__FN xml_object_range xml_node::attributes() const { - return _depth; + return xml_object_range(attributes_begin(), attributes_end()); } - bool xml_tree_walker::begin(xml_node&) + PUGI__FN bool xml_node::operator==(const xml_node& r) const { - return true; + return (_root == r._root); } - bool xml_tree_walker::end(xml_node&) + PUGI__FN bool xml_node::operator!=(const xml_node& r) const { - return true; + return (_root != r._root); } - xml_attribute::xml_attribute(): _attr(0) + PUGI__FN bool xml_node::operator<(const xml_node& r) const { + return (_root < r._root); } - xml_attribute::xml_attribute(xml_attribute_struct* attr): _attr(attr) + PUGI__FN bool xml_node::operator>(const xml_node& r) const { + return (_root > r._root); } - xml_attribute::operator xml_attribute::unspecified_bool_type() const + PUGI__FN bool xml_node::operator<=(const xml_node& r) const { - return _attr ? &xml_attribute::_attr : 0; - } - - bool xml_attribute::operator!() const - { - return !_attr; - } + return (_root <= r._root); + } - bool xml_attribute::operator==(const xml_attribute& r) const + PUGI__FN bool xml_node::operator>=(const xml_node& r) const { - return (_attr == r._attr); + return (_root >= r._root); } - - bool xml_attribute::operator!=(const xml_attribute& r) const + + PUGI__FN bool xml_node::empty() const { - return (_attr != r._attr); + return !_root; } - bool xml_attribute::operator<(const xml_attribute& r) const + PUGI__FN const char_t* xml_node::name() const { - return (_attr < r._attr); + return (_root && _root->name) ? _root->name + 0 : PUGIXML_TEXT(""); } - - bool xml_attribute::operator>(const xml_attribute& r) const + + PUGI__FN xml_node_type xml_node::type() const { - return (_attr > r._attr); + return _root ? PUGI__NODETYPE(_root) : node_null; } - - bool xml_attribute::operator<=(const xml_attribute& r) const + + PUGI__FN const char_t* xml_node::value() const { - return (_attr <= r._attr); + return (_root && _root->value) ? _root->value + 0 : PUGIXML_TEXT(""); } - - bool xml_attribute::operator>=(const xml_attribute& r) const + + PUGI__FN xml_node xml_node::child(const char_t* name_) const { - return (_attr >= r._attr); - } + if (!_root) return xml_node(); - xml_attribute xml_attribute::next_attribute() const - { - return _attr ? xml_attribute(_attr->next_attribute) : xml_attribute(); - } + for (xml_node_struct* i = _root->first_child; i; i = i->next_sibling) + if (i->name && impl::strequal(name_, i->name)) return xml_node(i); - xml_attribute xml_attribute::previous_attribute() const - { - return _attr && _attr->prev_attribute_c->next_attribute ? xml_attribute(_attr->prev_attribute_c) : xml_attribute(); - } + return xml_node(); + } - int xml_attribute::as_int() const + PUGI__FN xml_attribute xml_node::attribute(const char_t* name_) const { - if (!_attr || !_attr->value) return 0; + if (!_root) return xml_attribute(); - #ifdef PUGIXML_WCHAR_MODE - return (int)wcstol(_attr->value, 0, 10); - #else - return (int)strtol(_attr->value, 0, 10); - #endif + for (xml_attribute_struct* i = _root->first_attribute; i; i = i->next_attribute) + if (i->name && impl::strequal(name_, i->name)) + return xml_attribute(i); + + return xml_attribute(); } - unsigned int xml_attribute::as_uint() const + PUGI__FN xml_node xml_node::next_sibling(const char_t* name_) const { - if (!_attr || !_attr->value) return 0; + if (!_root) return xml_node(); - #ifdef PUGIXML_WCHAR_MODE - return (unsigned int)wcstoul(_attr->value, 0, 10); - #else - return (unsigned int)strtoul(_attr->value, 0, 10); - #endif + for (xml_node_struct* i = _root->next_sibling; i; i = i->next_sibling) + if (i->name && impl::strequal(name_, i->name)) return xml_node(i); + + return xml_node(); } - double xml_attribute::as_double() const + PUGI__FN xml_node xml_node::next_sibling() const { - if (!_attr || !_attr->value) return 0; - - #ifdef PUGIXML_WCHAR_MODE - return wcstod(_attr->value, 0); - #else - return strtod(_attr->value, 0); - #endif + return _root ? xml_node(_root->next_sibling) : xml_node(); } - float xml_attribute::as_float() const + PUGI__FN xml_node xml_node::previous_sibling(const char_t* name_) const { - if (!_attr || !_attr->value) return 0; + if (!_root) return xml_node(); - #ifdef PUGIXML_WCHAR_MODE - return (float)wcstod(_attr->value, 0); - #else - return (float)strtod(_attr->value, 0); - #endif + for (xml_node_struct* i = _root->prev_sibling_c; i->next_sibling; i = i->prev_sibling_c) + if (i->name && impl::strequal(name_, i->name)) return xml_node(i); + + return xml_node(); } - bool xml_attribute::as_bool() const + PUGI__FN xml_attribute xml_node::attribute(const char_t* name_, xml_attribute& hint_) const { - if (!_attr || !_attr->value) return false; + xml_attribute_struct* hint = hint_._attr; - // only look at first char - char_t first = *_attr->value; + // if hint is not an attribute of node, behavior is not defined + assert(!hint || (_root && impl::is_attribute_of(hint, _root))); - // 1*, t* (true), T* (True), y* (yes), Y* (YES) - return (first == '1' || first == 't' || first == 'T' || first == 'y' || first == 'Y'); + if (!_root) return xml_attribute(); + + // optimistically search from hint up until the end + for (xml_attribute_struct* i = hint; i; i = i->next_attribute) + if (i->name && impl::strequal(name_, i->name)) + { + // update hint to maximize efficiency of searching for consecutive attributes + hint_._attr = i->next_attribute; + + return xml_attribute(i); + } + + // wrap around and search from the first attribute until the hint + // 'j' null pointer check is technically redundant, but it prevents a crash in case the assertion above fails + for (xml_attribute_struct* j = _root->first_attribute; j && j != hint; j = j->next_attribute) + if (j->name && impl::strequal(name_, j->name)) + { + // update hint to maximize efficiency of searching for consecutive attributes + hint_._attr = j->next_attribute; + + return xml_attribute(j); + } + + return xml_attribute(); } - bool xml_attribute::empty() const + PUGI__FN xml_node xml_node::previous_sibling() const { - return !_attr; + if (!_root) return xml_node(); + + if (_root->prev_sibling_c->next_sibling) return xml_node(_root->prev_sibling_c); + else return xml_node(); } - const char_t* xml_attribute::name() const + PUGI__FN xml_node xml_node::parent() const { - return (_attr && _attr->name) ? _attr->name : PUGIXML_TEXT(""); + return _root ? xml_node(_root->parent) : xml_node(); } - const char_t* xml_attribute::value() const + PUGI__FN xml_node xml_node::root() const { - return (_attr && _attr->value) ? _attr->value : PUGIXML_TEXT(""); + return _root ? xml_node(&impl::get_document(_root)) : xml_node(); } - size_t xml_attribute::hash_value() const - { - return static_cast(reinterpret_cast(_attr) / sizeof(xml_attribute_struct)); - } - - xml_attribute_struct* xml_attribute::internal_object() const + PUGI__FN xml_text xml_node::text() const { - return _attr; + return xml_text(_root); } - xml_attribute& xml_attribute::operator=(const char_t* rhs) + PUGI__FN const char_t* xml_node::child_value() const { - set_value(rhs); - return *this; + if (!_root) return PUGIXML_TEXT(""); + + // element nodes can have value if parse_embed_pcdata was used + if (PUGI__NODETYPE(_root) == node_element && _root->value) + return _root->value; + + for (xml_node_struct* i = _root->first_child; i; i = i->next_sibling) + if (impl::is_text_node(i) && i->value) + return i->value; + + return PUGIXML_TEXT(""); } - - xml_attribute& xml_attribute::operator=(int rhs) + + PUGI__FN const char_t* xml_node::child_value(const char_t* name_) const { - set_value(rhs); - return *this; + return child(name_).child_value(); } - xml_attribute& xml_attribute::operator=(unsigned int rhs) + PUGI__FN xml_attribute xml_node::first_attribute() const { - set_value(rhs); - return *this; + return _root ? xml_attribute(_root->first_attribute) : xml_attribute(); } - xml_attribute& xml_attribute::operator=(double rhs) + PUGI__FN xml_attribute xml_node::last_attribute() const { - set_value(rhs); - return *this; + return _root && _root->first_attribute ? xml_attribute(_root->first_attribute->prev_attribute_c) : xml_attribute(); } - - xml_attribute& xml_attribute::operator=(bool rhs) + + PUGI__FN xml_node xml_node::first_child() const { - set_value(rhs); - return *this; + return _root ? xml_node(_root->first_child) : xml_node(); } - bool xml_attribute::set_name(const char_t* rhs) + PUGI__FN xml_node xml_node::last_child() const { - if (!_attr) return false; - - return strcpy_insitu(_attr->name, _attr->header, xml_memory_page_name_allocated_mask, rhs); + return _root && _root->first_child ? xml_node(_root->first_child->prev_sibling_c) : xml_node(); } - - bool xml_attribute::set_value(const char_t* rhs) + + PUGI__FN bool xml_node::set_name(const char_t* rhs) { - if (!_attr) return false; + xml_node_type type_ = _root ? PUGI__NODETYPE(_root) : node_null; + + if (type_ != node_element && type_ != node_pi && type_ != node_declaration) + return false; - return strcpy_insitu(_attr->value, _attr->header, xml_memory_page_value_allocated_mask, rhs); + return impl::strcpy_insitu(_root->name, _root->header, impl::xml_memory_page_name_allocated_mask, rhs, impl::strlength(rhs)); } - bool xml_attribute::set_value(int rhs) + PUGI__FN bool xml_node::set_value(const char_t* rhs) { - char buf[128]; - sprintf(buf, "%d", rhs); - - #ifdef PUGIXML_WCHAR_MODE - char_t wbuf[128]; - widen_ascii(wbuf, buf); + xml_node_type type_ = _root ? PUGI__NODETYPE(_root) : node_null; - return set_value(wbuf); - #else - return set_value(buf); - #endif + if (type_ != node_pcdata && type_ != node_cdata && type_ != node_comment && type_ != node_pi && type_ != node_doctype) + return false; + + return impl::strcpy_insitu(_root->value, _root->header, impl::xml_memory_page_value_allocated_mask, rhs, impl::strlength(rhs)); } - bool xml_attribute::set_value(unsigned int rhs) + PUGI__FN xml_attribute xml_node::append_attribute(const char_t* name_) { - char buf[128]; - sprintf(buf, "%u", rhs); + if (!impl::allow_insert_attribute(type())) return xml_attribute(); - #ifdef PUGIXML_WCHAR_MODE - char_t wbuf[128]; - widen_ascii(wbuf, buf); + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_attribute(); - return set_value(wbuf); - #else - return set_value(buf); - #endif - } + xml_attribute a(impl::allocate_attribute(alloc)); + if (!a) return xml_attribute(); - bool xml_attribute::set_value(double rhs) - { - char buf[128]; - sprintf(buf, "%g", rhs); + impl::append_attribute(a._attr, _root); - #ifdef PUGIXML_WCHAR_MODE - char_t wbuf[128]; - widen_ascii(wbuf, buf); + a.set_name(name_); - return set_value(wbuf); - #else - return set_value(buf); - #endif - } - - bool xml_attribute::set_value(bool rhs) - { - return set_value(rhs ? PUGIXML_TEXT("true") : PUGIXML_TEXT("false")); + return a; } -#ifdef __BORLANDC__ - bool operator&&(const xml_attribute& lhs, bool rhs) + PUGI__FN xml_attribute xml_node::prepend_attribute(const char_t* name_) { - return (bool)lhs && rhs; - } + if (!impl::allow_insert_attribute(type())) return xml_attribute(); - bool operator||(const xml_attribute& lhs, bool rhs) - { - return (bool)lhs || rhs; - } -#endif + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_attribute(); - xml_node::xml_node(): _root(0) - { - } + xml_attribute a(impl::allocate_attribute(alloc)); + if (!a) return xml_attribute(); - xml_node::xml_node(xml_node_struct* p): _root(p) - { - } - - xml_node::operator xml_node::unspecified_bool_type() const - { - return _root ? &xml_node::_root : 0; - } + impl::prepend_attribute(a._attr, _root); - bool xml_node::operator!() const - { - return !_root; - } + a.set_name(name_); - xml_node::iterator xml_node::begin() const - { - return iterator(_root ? _root->first_child : 0, _root); + return a; } - xml_node::iterator xml_node::end() const - { - return iterator(0, _root); - } - - xml_node::attribute_iterator xml_node::attributes_begin() const + PUGI__FN xml_attribute xml_node::insert_attribute_after(const char_t* name_, const xml_attribute& attr) { - return attribute_iterator(_root ? _root->first_attribute : 0, _root); - } + if (!impl::allow_insert_attribute(type())) return xml_attribute(); + if (!attr || !impl::is_attribute_of(attr._attr, _root)) return xml_attribute(); - xml_node::attribute_iterator xml_node::attributes_end() const - { - return attribute_iterator(0, _root); - } + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_attribute(); - bool xml_node::operator==(const xml_node& r) const - { - return (_root == r._root); - } + xml_attribute a(impl::allocate_attribute(alloc)); + if (!a) return xml_attribute(); - bool xml_node::operator!=(const xml_node& r) const - { - return (_root != r._root); - } + impl::insert_attribute_after(a._attr, attr._attr, _root); - bool xml_node::operator<(const xml_node& r) const - { - return (_root < r._root); - } - - bool xml_node::operator>(const xml_node& r) const - { - return (_root > r._root); - } - - bool xml_node::operator<=(const xml_node& r) const - { - return (_root <= r._root); + a.set_name(name_); + + return a; } - - bool xml_node::operator>=(const xml_node& r) const + + PUGI__FN xml_attribute xml_node::insert_attribute_before(const char_t* name_, const xml_attribute& attr) { - return (_root >= r._root); + if (!impl::allow_insert_attribute(type())) return xml_attribute(); + if (!attr || !impl::is_attribute_of(attr._attr, _root)) return xml_attribute(); + + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_attribute(); + + xml_attribute a(impl::allocate_attribute(alloc)); + if (!a) return xml_attribute(); + + impl::insert_attribute_before(a._attr, attr._attr, _root); + + a.set_name(name_); + + return a; } - bool xml_node::empty() const + PUGI__FN xml_attribute xml_node::append_copy(const xml_attribute& proto) { - return !_root; + if (!proto) return xml_attribute(); + if (!impl::allow_insert_attribute(type())) return xml_attribute(); + + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_attribute(); + + xml_attribute a(impl::allocate_attribute(alloc)); + if (!a) return xml_attribute(); + + impl::append_attribute(a._attr, _root); + impl::node_copy_attribute(a._attr, proto._attr); + + return a; } - - const char_t* xml_node::name() const + + PUGI__FN xml_attribute xml_node::prepend_copy(const xml_attribute& proto) { - return (_root && _root->name) ? _root->name : PUGIXML_TEXT(""); + if (!proto) return xml_attribute(); + if (!impl::allow_insert_attribute(type())) return xml_attribute(); + + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_attribute(); + + xml_attribute a(impl::allocate_attribute(alloc)); + if (!a) return xml_attribute(); + + impl::prepend_attribute(a._attr, _root); + impl::node_copy_attribute(a._attr, proto._attr); + + return a; } - xml_node_type xml_node::type() const + PUGI__FN xml_attribute xml_node::insert_copy_after(const xml_attribute& proto, const xml_attribute& attr) { - return _root ? static_cast((_root->header & xml_memory_page_type_mask) + 1) : node_null; + if (!proto) return xml_attribute(); + if (!impl::allow_insert_attribute(type())) return xml_attribute(); + if (!attr || !impl::is_attribute_of(attr._attr, _root)) return xml_attribute(); + + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_attribute(); + + xml_attribute a(impl::allocate_attribute(alloc)); + if (!a) return xml_attribute(); + + impl::insert_attribute_after(a._attr, attr._attr, _root); + impl::node_copy_attribute(a._attr, proto._attr); + + return a; } - - const char_t* xml_node::value() const + + PUGI__FN xml_attribute xml_node::insert_copy_before(const xml_attribute& proto, const xml_attribute& attr) { - return (_root && _root->value) ? _root->value : PUGIXML_TEXT(""); + if (!proto) return xml_attribute(); + if (!impl::allow_insert_attribute(type())) return xml_attribute(); + if (!attr || !impl::is_attribute_of(attr._attr, _root)) return xml_attribute(); + + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_attribute(); + + xml_attribute a(impl::allocate_attribute(alloc)); + if (!a) return xml_attribute(); + + impl::insert_attribute_before(a._attr, attr._attr, _root); + impl::node_copy_attribute(a._attr, proto._attr); + + return a; } - - xml_node xml_node::child(const char_t* name) const + + PUGI__FN xml_node xml_node::append_child(xml_node_type type_) { - if (!_root) return xml_node(); + if (!impl::allow_insert_child(type(), type_)) return xml_node(); - for (xml_node_struct* i = _root->first_child; i; i = i->next_sibling) - if (i->name && strequal(name, i->name)) return xml_node(i); + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_node(); - return xml_node(); + xml_node n(impl::allocate_node(alloc, type_)); + if (!n) return xml_node(); + + impl::append_node(n._root, _root); + + if (type_ == node_declaration) n.set_name(PUGIXML_TEXT("xml")); + + return n; } - xml_attribute xml_node::attribute(const char_t* name) const + PUGI__FN xml_node xml_node::prepend_child(xml_node_type type_) { - if (!_root) return xml_attribute(); + if (!impl::allow_insert_child(type(), type_)) return xml_node(); - for (xml_attribute_struct* i = _root->first_attribute; i; i = i->next_attribute) - if (i->name && strequal(name, i->name)) - return xml_attribute(i); - - return xml_attribute(); + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_node(); + + xml_node n(impl::allocate_node(alloc, type_)); + if (!n) return xml_node(); + + impl::prepend_node(n._root, _root); + + if (type_ == node_declaration) n.set_name(PUGIXML_TEXT("xml")); + + return n; } - - xml_node xml_node::next_sibling(const char_t* name) const + + PUGI__FN xml_node xml_node::insert_child_before(xml_node_type type_, const xml_node& node) { - if (!_root) return xml_node(); - - for (xml_node_struct* i = _root->next_sibling; i; i = i->next_sibling) - if (i->name && strequal(name, i->name)) return xml_node(i); + if (!impl::allow_insert_child(type(), type_)) return xml_node(); + if (!node._root || node._root->parent != _root) return xml_node(); - return xml_node(); + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_node(); + + xml_node n(impl::allocate_node(alloc, type_)); + if (!n) return xml_node(); + + impl::insert_node_before(n._root, node._root); + + if (type_ == node_declaration) n.set_name(PUGIXML_TEXT("xml")); + + return n; } - xml_node xml_node::next_sibling() const + PUGI__FN xml_node xml_node::insert_child_after(xml_node_type type_, const xml_node& node) { - if (!_root) return xml_node(); - - if (_root->next_sibling) return xml_node(_root->next_sibling); - else return xml_node(); + if (!impl::allow_insert_child(type(), type_)) return xml_node(); + if (!node._root || node._root->parent != _root) return xml_node(); + + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_node(); + + xml_node n(impl::allocate_node(alloc, type_)); + if (!n) return xml_node(); + + impl::insert_node_after(n._root, node._root); + + if (type_ == node_declaration) n.set_name(PUGIXML_TEXT("xml")); + + return n; } - xml_node xml_node::previous_sibling(const char_t* name) const + PUGI__FN xml_node xml_node::append_child(const char_t* name_) { - if (!_root) return xml_node(); - - for (xml_node_struct* i = _root->prev_sibling_c; i->next_sibling; i = i->prev_sibling_c) - if (i->name && strequal(name, i->name)) return xml_node(i); + xml_node result = append_child(node_element); - return xml_node(); + result.set_name(name_); + + return result; } - xml_node xml_node::previous_sibling() const + PUGI__FN xml_node xml_node::prepend_child(const char_t* name_) { - if (!_root) return xml_node(); - - if (_root->prev_sibling_c->next_sibling) return xml_node(_root->prev_sibling_c); - else return xml_node(); + xml_node result = prepend_child(node_element); + + result.set_name(name_); + + return result; } - xml_node xml_node::parent() const + PUGI__FN xml_node xml_node::insert_child_after(const char_t* name_, const xml_node& node) { - return _root ? xml_node(_root->parent) : xml_node(); + xml_node result = insert_child_after(node_element, node); + + result.set_name(name_); + + return result; } - xml_node xml_node::root() const + PUGI__FN xml_node xml_node::insert_child_before(const char_t* name_, const xml_node& node) { - if (!_root) return xml_node(); + xml_node result = insert_child_before(node_element, node); - xml_memory_page* page = reinterpret_cast(_root->header & xml_memory_page_pointer_mask); + result.set_name(name_); - return xml_node(static_cast(page->allocator)); + return result; } - const char_t* xml_node::child_value() const + PUGI__FN xml_node xml_node::append_copy(const xml_node& proto) { - if (!_root) return PUGIXML_TEXT(""); - - for (xml_node_struct* i = _root->first_child; i; i = i->next_sibling) - { - xml_node_type type = static_cast((i->header & xml_memory_page_type_mask) + 1); + xml_node_type type_ = proto.type(); + if (!impl::allow_insert_child(type(), type_)) return xml_node(); - if (i->value && (type == node_pcdata || type == node_cdata)) - return i->value; - } + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_node(); - return PUGIXML_TEXT(""); + xml_node n(impl::allocate_node(alloc, type_)); + if (!n) return xml_node(); + + impl::append_node(n._root, _root); + impl::node_copy_tree(n._root, proto._root); + + return n; } - const char_t* xml_node::child_value(const char_t* name) const + PUGI__FN xml_node xml_node::prepend_copy(const xml_node& proto) { - return child(name).child_value(); + xml_node_type type_ = proto.type(); + if (!impl::allow_insert_child(type(), type_)) return xml_node(); + + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_node(); + + xml_node n(impl::allocate_node(alloc, type_)); + if (!n) return xml_node(); + + impl::prepend_node(n._root, _root); + impl::node_copy_tree(n._root, proto._root); + + return n; } - xml_attribute xml_node::first_attribute() const + PUGI__FN xml_node xml_node::insert_copy_after(const xml_node& proto, const xml_node& node) { - return _root ? xml_attribute(_root->first_attribute) : xml_attribute(); + xml_node_type type_ = proto.type(); + if (!impl::allow_insert_child(type(), type_)) return xml_node(); + if (!node._root || node._root->parent != _root) return xml_node(); + + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_node(); + + xml_node n(impl::allocate_node(alloc, type_)); + if (!n) return xml_node(); + + impl::insert_node_after(n._root, node._root); + impl::node_copy_tree(n._root, proto._root); + + return n; } - xml_attribute xml_node::last_attribute() const + PUGI__FN xml_node xml_node::insert_copy_before(const xml_node& proto, const xml_node& node) { - return _root && _root->first_attribute ? xml_attribute(_root->first_attribute->prev_attribute_c) : xml_attribute(); + xml_node_type type_ = proto.type(); + if (!impl::allow_insert_child(type(), type_)) return xml_node(); + if (!node._root || node._root->parent != _root) return xml_node(); + + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_node(); + + xml_node n(impl::allocate_node(alloc, type_)); + if (!n) return xml_node(); + + impl::insert_node_before(n._root, node._root); + impl::node_copy_tree(n._root, proto._root); + + return n; } - xml_node xml_node::first_child() const + PUGI__FN xml_node xml_node::append_move(const xml_node& moved) { - return _root ? xml_node(_root->first_child) : xml_node(); + if (!impl::allow_move(*this, moved)) return xml_node(); + + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_node(); + + // disable document_buffer_order optimization since moving nodes around changes document order without changing buffer pointers + impl::get_document(_root).header |= impl::xml_memory_page_contents_shared_mask; + + impl::remove_node(moved._root); + impl::append_node(moved._root, _root); + + return moved; } - xml_node xml_node::last_child() const + PUGI__FN xml_node xml_node::prepend_move(const xml_node& moved) { - return _root && _root->first_child ? xml_node(_root->first_child->prev_sibling_c) : xml_node(); + if (!impl::allow_move(*this, moved)) return xml_node(); + + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_node(); + + // disable document_buffer_order optimization since moving nodes around changes document order without changing buffer pointers + impl::get_document(_root).header |= impl::xml_memory_page_contents_shared_mask; + + impl::remove_node(moved._root); + impl::prepend_node(moved._root, _root); + + return moved; } - bool xml_node::set_name(const char_t* rhs) + PUGI__FN xml_node xml_node::insert_move_after(const xml_node& moved, const xml_node& node) { - switch (type()) - { - case node_pi: - case node_declaration: - case node_element: - return strcpy_insitu(_root->name, _root->header, xml_memory_page_name_allocated_mask, rhs); + if (!impl::allow_move(*this, moved)) return xml_node(); + if (!node._root || node._root->parent != _root) return xml_node(); + if (moved._root == node._root) return xml_node(); - default: - return false; - } + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_node(); + + // disable document_buffer_order optimization since moving nodes around changes document order without changing buffer pointers + impl::get_document(_root).header |= impl::xml_memory_page_contents_shared_mask; + + impl::remove_node(moved._root); + impl::insert_node_after(moved._root, node._root); + + return moved; } - - bool xml_node::set_value(const char_t* rhs) + + PUGI__FN xml_node xml_node::insert_move_before(const xml_node& moved, const xml_node& node) { - switch (type()) - { - case node_pi: - case node_cdata: - case node_pcdata: - case node_comment: - case node_doctype: - return strcpy_insitu(_root->value, _root->header, xml_memory_page_value_allocated_mask, rhs); + if (!impl::allow_move(*this, moved)) return xml_node(); + if (!node._root || node._root->parent != _root) return xml_node(); + if (moved._root == node._root) return xml_node(); - default: - return false; - } + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return xml_node(); + + // disable document_buffer_order optimization since moving nodes around changes document order without changing buffer pointers + impl::get_document(_root).header |= impl::xml_memory_page_contents_shared_mask; + + impl::remove_node(moved._root); + impl::insert_node_before(moved._root, node._root); + + return moved; } - xml_attribute xml_node::append_attribute(const char_t* name) + PUGI__FN bool xml_node::remove_attribute(const char_t* name_) { - if (type() != node_element && type() != node_declaration) return xml_attribute(); - - xml_attribute a(append_attribute_ll(_root, get_allocator(_root))); - a.set_name(name); - - return a; + return remove_attribute(attribute(name_)); } - xml_attribute xml_node::prepend_attribute(const char_t* name) + PUGI__FN bool xml_node::remove_attribute(const xml_attribute& a) { - if (type() != node_element && type() != node_declaration) return xml_attribute(); - - xml_attribute a(allocate_attribute(get_allocator(_root))); - if (!a) return xml_attribute(); + if (!_root || !a._attr) return false; + if (!impl::is_attribute_of(a._attr, _root)) return false; - a.set_name(name); - - xml_attribute_struct* head = _root->first_attribute; + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return false; - if (head) - { - a._attr->prev_attribute_c = head->prev_attribute_c; - head->prev_attribute_c = a._attr; - } - else - a._attr->prev_attribute_c = a._attr; - - a._attr->next_attribute = head; - _root->first_attribute = a._attr; - - return a; + impl::remove_attribute(a._attr, _root); + impl::destroy_attribute(a._attr, alloc); + + return true; } - xml_attribute xml_node::insert_attribute_before(const char_t* name, const xml_attribute& attr) + PUGI__FN bool xml_node::remove_child(const char_t* name_) { - if ((type() != node_element && type() != node_declaration) || attr.empty()) return xml_attribute(); - - // check that attribute belongs to *this - xml_attribute_struct* cur = attr._attr; - - while (cur->prev_attribute_c->next_attribute) cur = cur->prev_attribute_c; + return remove_child(child(name_)); + } - if (cur != _root->first_attribute) return xml_attribute(); + PUGI__FN bool xml_node::remove_child(const xml_node& n) + { + if (!_root || !n._root || n._root->parent != _root) return false; - xml_attribute a(allocate_attribute(get_allocator(_root))); - if (!a) return xml_attribute(); + impl::xml_allocator& alloc = impl::get_allocator(_root); + if (!alloc.reserve()) return false; - a.set_name(name); + impl::remove_node(n._root); + impl::destroy_node(n._root, alloc); - if (attr._attr->prev_attribute_c->next_attribute) - attr._attr->prev_attribute_c->next_attribute = a._attr; - else - _root->first_attribute = a._attr; - - a._attr->prev_attribute_c = attr._attr->prev_attribute_c; - a._attr->next_attribute = attr._attr; - attr._attr->prev_attribute_c = a._attr; - - return a; + return true; } - xml_attribute xml_node::insert_attribute_after(const char_t* name, const xml_attribute& attr) + PUGI__FN xml_parse_result xml_node::append_buffer(const void* contents, size_t size, unsigned int options, xml_encoding encoding) { - if ((type() != node_element && type() != node_declaration) || attr.empty()) return xml_attribute(); - - // check that attribute belongs to *this - xml_attribute_struct* cur = attr._attr; + // append_buffer is only valid for elements/documents + if (!impl::allow_insert_child(type(), node_element)) return impl::make_parse_result(status_append_invalid_root); - while (cur->prev_attribute_c->next_attribute) cur = cur->prev_attribute_c; + // get document node + impl::xml_document_struct* doc = &impl::get_document(_root); - if (cur != _root->first_attribute) return xml_attribute(); + // disable document_buffer_order optimization since in a document with multiple buffers comparing buffer pointers does not make sense + doc->header |= impl::xml_memory_page_contents_shared_mask; - xml_attribute a(allocate_attribute(get_allocator(_root))); - if (!a) return xml_attribute(); + // get extra buffer element (we'll store the document fragment buffer there so that we can deallocate it later) + impl::xml_memory_page* page = 0; + impl::xml_extra_buffer* extra = static_cast(doc->allocate_memory(sizeof(impl::xml_extra_buffer) + sizeof(void*), page)); + (void)page; - a.set_name(name); + if (!extra) return impl::make_parse_result(status_out_of_memory); - if (attr._attr->next_attribute) - attr._attr->next_attribute->prev_attribute_c = a._attr; - else - _root->first_attribute->prev_attribute_c = a._attr; - - a._attr->next_attribute = attr._attr->next_attribute; - a._attr->prev_attribute_c = attr._attr; - attr._attr->next_attribute = a._attr; + #ifdef PUGIXML_COMPACT + // align the memory block to a pointer boundary; this is required for compact mode where memory allocations are only 4b aligned + // note that this requires up to sizeof(void*)-1 additional memory, which the allocation above takes into account + extra = reinterpret_cast((reinterpret_cast(extra) + (sizeof(void*) - 1)) & ~(sizeof(void*) - 1)); + #endif - return a; + // add extra buffer to the list + extra->buffer = 0; + extra->next = doc->extra_buffers; + doc->extra_buffers = extra; + + // name of the root has to be NULL before parsing - otherwise closing node mismatches will not be detected at the top level + impl::name_null_sentry sentry(_root); + + return impl::load_buffer_impl(doc, _root, const_cast(contents), size, options, encoding, false, false, &extra->buffer); } - xml_attribute xml_node::append_copy(const xml_attribute& proto) + PUGI__FN xml_node xml_node::find_child_by_attribute(const char_t* name_, const char_t* attr_name, const char_t* attr_value) const { - if (!proto) return xml_attribute(); + if (!_root) return xml_node(); - xml_attribute result = append_attribute(proto.name()); - result.set_value(proto.value()); + for (xml_node_struct* i = _root->first_child; i; i = i->next_sibling) + if (i->name && impl::strequal(name_, i->name)) + { + for (xml_attribute_struct* a = i->first_attribute; a; a = a->next_attribute) + if (a->name && impl::strequal(attr_name, a->name) && impl::strequal(attr_value, a->value ? a->value + 0 : PUGIXML_TEXT(""))) + return xml_node(i); + } - return result; + return xml_node(); } - xml_attribute xml_node::prepend_copy(const xml_attribute& proto) + PUGI__FN xml_node xml_node::find_child_by_attribute(const char_t* attr_name, const char_t* attr_value) const { - if (!proto) return xml_attribute(); + if (!_root) return xml_node(); - xml_attribute result = prepend_attribute(proto.name()); - result.set_value(proto.value()); + for (xml_node_struct* i = _root->first_child; i; i = i->next_sibling) + for (xml_attribute_struct* a = i->first_attribute; a; a = a->next_attribute) + if (a->name && impl::strequal(attr_name, a->name) && impl::strequal(attr_value, a->value ? a->value + 0 : PUGIXML_TEXT(""))) + return xml_node(i); - return result; + return xml_node(); } - xml_attribute xml_node::insert_copy_after(const xml_attribute& proto, const xml_attribute& attr) +#ifndef PUGIXML_NO_STL + PUGI__FN string_t xml_node::path(char_t delimiter) const { - if (!proto) return xml_attribute(); + if (!_root) return string_t(); - xml_attribute result = insert_attribute_after(proto.name(), attr); - result.set_value(proto.value()); + size_t offset = 0; - return result; - } + for (xml_node_struct* i = _root; i; i = i->parent) + { + offset += (i != _root); + offset += i->name ? impl::strlength(i->name) : 0; + } - xml_attribute xml_node::insert_copy_before(const xml_attribute& proto, const xml_attribute& attr) - { - if (!proto) return xml_attribute(); + string_t result; + result.resize(offset); + + for (xml_node_struct* j = _root; j; j = j->parent) + { + if (j != _root) + result[--offset] = delimiter; + + if (j->name) + { + size_t length = impl::strlength(j->name); + + offset -= length; + memcpy(&result[offset], j->name, length * sizeof(char_t)); + } + } - xml_attribute result = insert_attribute_before(proto.name(), attr); - result.set_value(proto.value()); + assert(offset == 0); return result; } +#endif - xml_node xml_node::append_child(xml_node_type type) + PUGI__FN xml_node xml_node::first_element_by_path(const char_t* path_, char_t delimiter) const { - if (!allow_insert_child(this->type(), type)) return xml_node(); - - xml_node n(append_node(_root, get_allocator(_root), type)); + xml_node found = *this; // Current search context. - if (type == node_declaration) n.set_name(PUGIXML_TEXT("xml")); + if (!_root || !path_[0]) return found; - return n; - } + if (path_[0] == delimiter) + { + // Absolute path; e.g. '/foo/bar' + found = found.root(); + ++path_; + } - xml_node xml_node::prepend_child(xml_node_type type) - { - if (!allow_insert_child(this->type(), type)) return xml_node(); - - xml_node n(allocate_node(get_allocator(_root), type)); - if (!n) return xml_node(); + const char_t* path_segment = path_; - n._root->parent = _root; + while (*path_segment == delimiter) ++path_segment; - xml_node_struct* head = _root->first_child; + const char_t* path_segment_end = path_segment; - if (head) - { - n._root->prev_sibling_c = head->prev_sibling_c; - head->prev_sibling_c = n._root; - } - else - n._root->prev_sibling_c = n._root; - - n._root->next_sibling = head; - _root->first_child = n._root; - - if (type == node_declaration) n.set_name(PUGIXML_TEXT("xml")); + while (*path_segment_end && *path_segment_end != delimiter) ++path_segment_end; - return n; - } + if (path_segment == path_segment_end) return found; - xml_node xml_node::insert_child_before(xml_node_type type, const xml_node& node) - { - if (!allow_insert_child(this->type(), type)) return xml_node(); - if (!node._root || node._root->parent != _root) return xml_node(); - - xml_node n(allocate_node(get_allocator(_root), type)); - if (!n) return xml_node(); + const char_t* next_segment = path_segment_end; + + while (*next_segment == delimiter) ++next_segment; - n._root->parent = _root; - - if (node._root->prev_sibling_c->next_sibling) - node._root->prev_sibling_c->next_sibling = n._root; + if (*path_segment == '.' && path_segment + 1 == path_segment_end) + return found.first_element_by_path(next_segment, delimiter); + else if (*path_segment == '.' && *(path_segment+1) == '.' && path_segment + 2 == path_segment_end) + return found.parent().first_element_by_path(next_segment, delimiter); else - _root->first_child = n._root; - - n._root->prev_sibling_c = node._root->prev_sibling_c; - n._root->next_sibling = node._root; - node._root->prev_sibling_c = n._root; + { + for (xml_node_struct* j = found._root->first_child; j; j = j->next_sibling) + { + if (j->name && impl::strequalrange(j->name, path_segment, static_cast(path_segment_end - path_segment))) + { + xml_node subsearch = xml_node(j).first_element_by_path(next_segment, delimiter); - if (type == node_declaration) n.set_name(PUGIXML_TEXT("xml")); + if (subsearch) return subsearch; + } + } - return n; + return xml_node(); + } } - xml_node xml_node::insert_child_after(xml_node_type type, const xml_node& node) + PUGI__FN bool xml_node::traverse(xml_tree_walker& walker) { - if (!allow_insert_child(this->type(), type)) return xml_node(); - if (!node._root || node._root->parent != _root) return xml_node(); - - xml_node n(allocate_node(get_allocator(_root), type)); - if (!n) return xml_node(); + walker._depth = -1; - n._root->parent = _root; - - if (node._root->next_sibling) - node._root->next_sibling->prev_sibling_c = n._root; - else - _root->first_child->prev_sibling_c = n._root; - - n._root->next_sibling = node._root->next_sibling; - n._root->prev_sibling_c = node._root; - node._root->next_sibling = n._root; + xml_node arg_begin(_root); + if (!walker.begin(arg_begin)) return false; - if (type == node_declaration) n.set_name(PUGIXML_TEXT("xml")); + xml_node_struct* cur = _root ? _root->first_child + 0 : 0; - return n; - } + if (cur) + { + ++walker._depth; + + do + { + xml_node arg_for_each(cur); + if (!walker.for_each(arg_for_each)) + return false; + + if (cur->first_child) + { + ++walker._depth; + cur = cur->first_child; + } + else if (cur->next_sibling) + cur = cur->next_sibling; + else + { + while (!cur->next_sibling && cur != _root && cur->parent) + { + --walker._depth; + cur = cur->parent; + } + + if (cur != _root) + cur = cur->next_sibling; + } + } + while (cur && cur != _root); + } - xml_node xml_node::append_child(const char_t* name) - { - xml_node result = append_child(node_element); + assert(walker._depth == -1); - result.set_name(name); + xml_node arg_end(_root); + return walker.end(arg_end); + } - return result; - } + PUGI__FN size_t xml_node::hash_value() const + { + return static_cast(reinterpret_cast(_root) / sizeof(xml_node_struct)); + } - xml_node xml_node::prepend_child(const char_t* name) - { - xml_node result = prepend_child(node_element); + PUGI__FN xml_node_struct* xml_node::internal_object() const + { + return _root; + } - result.set_name(name); + PUGI__FN void xml_node::print(xml_writer& writer, const char_t* indent, unsigned int flags, xml_encoding encoding, unsigned int depth) const + { + if (!_root) return; - return result; - } + impl::xml_buffered_writer buffered_writer(writer, encoding); - xml_node xml_node::insert_child_after(const char_t* name, const xml_node& node) - { - xml_node result = insert_child_after(node_element, node); + impl::node_output(buffered_writer, _root, indent, flags, depth); - result.set_name(name); + buffered_writer.flush(); + } - return result; - } +#ifndef PUGIXML_NO_STL + PUGI__FN void xml_node::print(std::basic_ostream >& stream, const char_t* indent, unsigned int flags, xml_encoding encoding, unsigned int depth) const + { + xml_writer_stream writer(stream); - xml_node xml_node::insert_child_before(const char_t* name, const xml_node& node) - { - xml_node result = insert_child_before(node_element, node); + print(writer, indent, flags, encoding, depth); + } - result.set_name(name); + PUGI__FN void xml_node::print(std::basic_ostream >& stream, const char_t* indent, unsigned int flags, unsigned int depth) const + { + xml_writer_stream writer(stream); - return result; - } + print(writer, indent, flags, encoding_wchar, depth); + } +#endif - xml_node xml_node::append_copy(const xml_node& proto) + PUGI__FN ptrdiff_t xml_node::offset_debug() const { - xml_node result = append_child(proto.type()); + if (!_root) return -1; - if (result) recursive_copy_skip(result, proto, result); + impl::xml_document_struct& doc = impl::get_document(_root); - return result; + // we can determine the offset reliably only if there is exactly once parse buffer + if (!doc.buffer || doc.extra_buffers) return -1; + + switch (type()) + { + case node_document: + return 0; + + case node_element: + case node_declaration: + case node_pi: + return _root->name && (_root->header & impl::xml_memory_page_name_allocated_or_shared_mask) == 0 ? _root->name - doc.buffer : -1; + + case node_pcdata: + case node_cdata: + case node_comment: + case node_doctype: + return _root->value && (_root->header & impl::xml_memory_page_value_allocated_or_shared_mask) == 0 ? _root->value - doc.buffer : -1; + + default: + assert(false && "Invalid node type"); // unreachable + return -1; + } } - xml_node xml_node::prepend_copy(const xml_node& proto) +#ifdef __BORLANDC__ + PUGI__FN bool operator&&(const xml_node& lhs, bool rhs) { - xml_node result = prepend_child(proto.type()); + return (bool)lhs && rhs; + } - if (result) recursive_copy_skip(result, proto, result); + PUGI__FN bool operator||(const xml_node& lhs, bool rhs) + { + return (bool)lhs || rhs; + } +#endif - return result; + PUGI__FN xml_text::xml_text(xml_node_struct* root): _root(root) + { } - xml_node xml_node::insert_copy_after(const xml_node& proto, const xml_node& node) + PUGI__FN xml_node_struct* xml_text::_data() const { - xml_node result = insert_child_after(proto.type(), node); + if (!_root || impl::is_text_node(_root)) return _root; - if (result) recursive_copy_skip(result, proto, result); + // element nodes can have value if parse_embed_pcdata was used + if (PUGI__NODETYPE(_root) == node_element && _root->value) + return _root; - return result; + for (xml_node_struct* node = _root->first_child; node; node = node->next_sibling) + if (impl::is_text_node(node)) + return node; + + return 0; } - xml_node xml_node::insert_copy_before(const xml_node& proto, const xml_node& node) + PUGI__FN xml_node_struct* xml_text::_data_new() { - xml_node result = insert_child_before(proto.type(), node); + xml_node_struct* d = _data(); + if (d) return d; - if (result) recursive_copy_skip(result, proto, result); + return xml_node(_root).append_child(node_pcdata).internal_object(); + } - return result; + PUGI__FN xml_text::xml_text(): _root(0) + { } - bool xml_node::remove_attribute(const char_t* name) + PUGI__FN static void unspecified_bool_xml_text(xml_text***) { - return remove_attribute(attribute(name)); } - bool xml_node::remove_attribute(const xml_attribute& a) + PUGI__FN xml_text::operator xml_text::unspecified_bool_type() const { - if (!_root || !a._attr) return false; + return _data() ? unspecified_bool_xml_text : 0; + } + + PUGI__FN bool xml_text::operator!() const + { + return !_data(); + } + + PUGI__FN bool xml_text::empty() const + { + return _data() == 0; + } - // check that attribute belongs to *this - xml_attribute_struct* attr = a._attr; + PUGI__FN const char_t* xml_text::get() const + { + xml_node_struct* d = _data(); - while (attr->prev_attribute_c->next_attribute) attr = attr->prev_attribute_c; + return (d && d->value) ? d->value + 0 : PUGIXML_TEXT(""); + } - if (attr != _root->first_attribute) return false; + PUGI__FN const char_t* xml_text::as_string(const char_t* def) const + { + xml_node_struct* d = _data(); - if (a._attr->next_attribute) a._attr->next_attribute->prev_attribute_c = a._attr->prev_attribute_c; - else if (_root->first_attribute) _root->first_attribute->prev_attribute_c = a._attr->prev_attribute_c; - - if (a._attr->prev_attribute_c->next_attribute) a._attr->prev_attribute_c->next_attribute = a._attr->next_attribute; - else _root->first_attribute = a._attr->next_attribute; + return (d && d->value) ? d->value + 0 : def; + } - destroy_attribute(a._attr, get_allocator(_root)); + PUGI__FN int xml_text::as_int(int def) const + { + xml_node_struct* d = _data(); - return true; + return (d && d->value) ? impl::get_value_int(d->value) : def; } - bool xml_node::remove_child(const char_t* name) + PUGI__FN unsigned int xml_text::as_uint(unsigned int def) const { - return remove_child(child(name)); + xml_node_struct* d = _data(); + + return (d && d->value) ? impl::get_value_uint(d->value) : def; } - bool xml_node::remove_child(const xml_node& n) + PUGI__FN double xml_text::as_double(double def) const { - if (!_root || !n._root || n._root->parent != _root) return false; - - if (n._root->next_sibling) n._root->next_sibling->prev_sibling_c = n._root->prev_sibling_c; - else if (_root->first_child) _root->first_child->prev_sibling_c = n._root->prev_sibling_c; - - if (n._root->prev_sibling_c->next_sibling) n._root->prev_sibling_c->next_sibling = n._root->next_sibling; - else _root->first_child = n._root->next_sibling; - - destroy_node(n._root, get_allocator(_root)); + xml_node_struct* d = _data(); - return true; + return (d && d->value) ? impl::get_value_double(d->value) : def; } - xml_node xml_node::find_child_by_attribute(const char_t* name, const char_t* attr_name, const char_t* attr_value) const + PUGI__FN float xml_text::as_float(float def) const { - if (!_root) return xml_node(); - - for (xml_node_struct* i = _root->first_child; i; i = i->next_sibling) - if (i->name && strequal(name, i->name)) - { - for (xml_attribute_struct* a = i->first_attribute; a; a = a->next_attribute) - if (strequal(attr_name, a->name) && strequal(attr_value, a->value)) - return xml_node(i); - } + xml_node_struct* d = _data(); - return xml_node(); + return (d && d->value) ? impl::get_value_float(d->value) : def; } - xml_node xml_node::find_child_by_attribute(const char_t* attr_name, const char_t* attr_value) const + PUGI__FN bool xml_text::as_bool(bool def) const { - if (!_root) return xml_node(); - - for (xml_node_struct* i = _root->first_child; i; i = i->next_sibling) - for (xml_attribute_struct* a = i->first_attribute; a; a = a->next_attribute) - if (strequal(attr_name, a->name) && strequal(attr_value, a->value)) - return xml_node(i); + xml_node_struct* d = _data(); - return xml_node(); + return (d && d->value) ? impl::get_value_bool(d->value) : def; } -#ifndef PUGIXML_NO_STL - string_t xml_node::path(char_t delimiter) const +#ifdef PUGIXML_HAS_LONG_LONG + PUGI__FN long long xml_text::as_llong(long long def) const { - string_t path; + xml_node_struct* d = _data(); - xml_node cursor = *this; // Make a copy. - - path = cursor.name(); + return (d && d->value) ? impl::get_value_llong(d->value) : def; + } - while (cursor.parent()) - { - cursor = cursor.parent(); - - string_t temp = cursor.name(); - temp += delimiter; - temp += path; - path.swap(temp); - } + PUGI__FN unsigned long long xml_text::as_ullong(unsigned long long def) const + { + xml_node_struct* d = _data(); - return path; + return (d && d->value) ? impl::get_value_ullong(d->value) : def; } #endif - xml_node xml_node::first_element_by_path(const char_t* path, char_t delimiter) const + PUGI__FN bool xml_text::set(const char_t* rhs) { - xml_node found = *this; // Current search context. - - if (!_root || !path || !path[0]) return found; + xml_node_struct* dn = _data_new(); - if (path[0] == delimiter) - { - // Absolute path; e.g. '/foo/bar' - found = found.root(); - ++path; - } + return dn ? impl::strcpy_insitu(dn->value, dn->header, impl::xml_memory_page_value_allocated_mask, rhs, impl::strlength(rhs)) : false; + } - const char_t* path_segment = path; + PUGI__FN bool xml_text::set(int rhs) + { + xml_node_struct* dn = _data_new(); - while (*path_segment == delimiter) ++path_segment; + return dn ? impl::set_value_integer(dn->value, dn->header, impl::xml_memory_page_value_allocated_mask, rhs, rhs < 0) : false; + } - const char_t* path_segment_end = path_segment; + PUGI__FN bool xml_text::set(unsigned int rhs) + { + xml_node_struct* dn = _data_new(); - while (*path_segment_end && *path_segment_end != delimiter) ++path_segment_end; + return dn ? impl::set_value_integer(dn->value, dn->header, impl::xml_memory_page_value_allocated_mask, rhs, false) : false; + } - if (path_segment == path_segment_end) return found; + PUGI__FN bool xml_text::set(long rhs) + { + xml_node_struct* dn = _data_new(); - const char_t* next_segment = path_segment_end; + return dn ? impl::set_value_integer(dn->value, dn->header, impl::xml_memory_page_value_allocated_mask, rhs, rhs < 0) : false; + } - while (*next_segment == delimiter) ++next_segment; + PUGI__FN bool xml_text::set(unsigned long rhs) + { + xml_node_struct* dn = _data_new(); - if (*path_segment == '.' && path_segment + 1 == path_segment_end) - return found.first_element_by_path(next_segment, delimiter); - else if (*path_segment == '.' && *(path_segment+1) == '.' && path_segment + 2 == path_segment_end) - return found.parent().first_element_by_path(next_segment, delimiter); - else - { - for (xml_node_struct* j = found._root->first_child; j; j = j->next_sibling) - { - if (j->name && strequalrange(j->name, path_segment, static_cast(path_segment_end - path_segment))) - { - xml_node subsearch = xml_node(j).first_element_by_path(next_segment, delimiter); + return dn ? impl::set_value_integer(dn->value, dn->header, impl::xml_memory_page_value_allocated_mask, rhs, false) : false; + } - if (subsearch) return subsearch; - } - } + PUGI__FN bool xml_text::set(float rhs) + { + xml_node_struct* dn = _data_new(); - return xml_node(); - } + return dn ? impl::set_value_convert(dn->value, dn->header, impl::xml_memory_page_value_allocated_mask, rhs) : false; } - bool xml_node::traverse(xml_tree_walker& walker) + PUGI__FN bool xml_text::set(double rhs) { - walker._depth = -1; - - xml_node arg_begin = *this; - if (!walker.begin(arg_begin)) return false; - - xml_node cur = first_child(); - - if (cur) - { - ++walker._depth; + xml_node_struct* dn = _data_new(); - do - { - xml_node arg_for_each = cur; - if (!walker.for_each(arg_for_each)) - return false; - - if (cur.first_child()) - { - ++walker._depth; - cur = cur.first_child(); - } - else if (cur.next_sibling()) - cur = cur.next_sibling(); - else - { - // Borland C++ workaround - while (!cur.next_sibling() && cur != *this && (bool)cur.parent()) - { - --walker._depth; - cur = cur.parent(); - } - - if (cur != *this) - cur = cur.next_sibling(); - } - } - while (cur && cur != *this); - } + return dn ? impl::set_value_convert(dn->value, dn->header, impl::xml_memory_page_value_allocated_mask, rhs) : false; + } - assert(walker._depth == -1); + PUGI__FN bool xml_text::set(bool rhs) + { + xml_node_struct* dn = _data_new(); - xml_node arg_end = *this; - return walker.end(arg_end); + return dn ? impl::set_value_bool(dn->value, dn->header, impl::xml_memory_page_value_allocated_mask, rhs) : false; } - size_t xml_node::hash_value() const - { - return static_cast(reinterpret_cast(_root) / sizeof(xml_node_struct)); - } - - xml_node_struct* xml_node::internal_object() const +#ifdef PUGIXML_HAS_LONG_LONG + PUGI__FN bool xml_text::set(long long rhs) { - return _root; + xml_node_struct* dn = _data_new(); + + return dn ? impl::set_value_integer(dn->value, dn->header, impl::xml_memory_page_value_allocated_mask, rhs, rhs < 0) : false; } - void xml_node::print(xml_writer& writer, const char_t* indent, unsigned int flags, xml_encoding encoding, unsigned int depth) const + PUGI__FN bool xml_text::set(unsigned long long rhs) { - if (!_root) return; - - xml_buffered_writer buffered_writer(writer, encoding); + xml_node_struct* dn = _data_new(); - node_output(buffered_writer, *this, indent, flags, depth); + return dn ? impl::set_value_integer(dn->value, dn->header, impl::xml_memory_page_value_allocated_mask, rhs, false) : false; } +#endif -#ifndef PUGIXML_NO_STL - void xml_node::print(std::basic_ostream >& stream, const char_t* indent, unsigned int flags, xml_encoding encoding, unsigned int depth) const + PUGI__FN xml_text& xml_text::operator=(const char_t* rhs) { - xml_writer_stream writer(stream); - - print(writer, indent, flags, encoding, depth); + set(rhs); + return *this; } - void xml_node::print(std::basic_ostream >& stream, const char_t* indent, unsigned int flags, unsigned int depth) const + PUGI__FN xml_text& xml_text::operator=(int rhs) { - xml_writer_stream writer(stream); + set(rhs); + return *this; + } - print(writer, indent, flags, encoding_wchar, depth); + PUGI__FN xml_text& xml_text::operator=(unsigned int rhs) + { + set(rhs); + return *this; } -#endif - ptrdiff_t xml_node::offset_debug() const + PUGI__FN xml_text& xml_text::operator=(long rhs) { - xml_node_struct* r = root()._root; + set(rhs); + return *this; + } - if (!r) return -1; + PUGI__FN xml_text& xml_text::operator=(unsigned long rhs) + { + set(rhs); + return *this; + } - const char_t* buffer = static_cast(r)->buffer; + PUGI__FN xml_text& xml_text::operator=(double rhs) + { + set(rhs); + return *this; + } - if (!buffer) return -1; + PUGI__FN xml_text& xml_text::operator=(float rhs) + { + set(rhs); + return *this; + } - switch (type()) - { - case node_document: - return 0; + PUGI__FN xml_text& xml_text::operator=(bool rhs) + { + set(rhs); + return *this; + } - case node_element: - case node_declaration: - case node_pi: - return (_root->header & xml_memory_page_name_allocated_mask) ? -1 : _root->name - buffer; +#ifdef PUGIXML_HAS_LONG_LONG + PUGI__FN xml_text& xml_text::operator=(long long rhs) + { + set(rhs); + return *this; + } - case node_pcdata: - case node_cdata: - case node_comment: - case node_doctype: - return (_root->header & xml_memory_page_value_allocated_mask) ? -1 : _root->value - buffer; + PUGI__FN xml_text& xml_text::operator=(unsigned long long rhs) + { + set(rhs); + return *this; + } +#endif - default: - return -1; - } + PUGI__FN xml_node xml_text::data() const + { + return xml_node(_data()); } #ifdef __BORLANDC__ - bool operator&&(const xml_node& lhs, bool rhs) + PUGI__FN bool operator&&(const xml_text& lhs, bool rhs) { return (bool)lhs && rhs; } - bool operator||(const xml_node& lhs, bool rhs) + PUGI__FN bool operator||(const xml_text& lhs, bool rhs) { return (bool)lhs || rhs; } #endif - xml_node_iterator::xml_node_iterator() + PUGI__FN xml_node_iterator::xml_node_iterator() { } - xml_node_iterator::xml_node_iterator(const xml_node& node): _wrap(node), _parent(node.parent()) + PUGI__FN xml_node_iterator::xml_node_iterator(const xml_node& node): _wrap(node), _parent(node.parent()) { } - xml_node_iterator::xml_node_iterator(xml_node_struct* ref, xml_node_struct* parent): _wrap(ref), _parent(parent) + PUGI__FN xml_node_iterator::xml_node_iterator(xml_node_struct* ref, xml_node_struct* parent): _wrap(ref), _parent(parent) { } - bool xml_node_iterator::operator==(const xml_node_iterator& rhs) const + PUGI__FN bool xml_node_iterator::operator==(const xml_node_iterator& rhs) const { return _wrap._root == rhs._wrap._root && _parent._root == rhs._parent._root; } - - bool xml_node_iterator::operator!=(const xml_node_iterator& rhs) const + + PUGI__FN bool xml_node_iterator::operator!=(const xml_node_iterator& rhs) const { return _wrap._root != rhs._wrap._root || _parent._root != rhs._parent._root; } - xml_node& xml_node_iterator::operator*() + PUGI__FN xml_node& xml_node_iterator::operator*() const { assert(_wrap._root); return _wrap; } - xml_node* xml_node_iterator::operator->() + PUGI__FN xml_node* xml_node_iterator::operator->() const { assert(_wrap._root); - return &_wrap; + return const_cast(&_wrap); // BCC5 workaround } - const xml_node_iterator& xml_node_iterator::operator++() + PUGI__FN const xml_node_iterator& xml_node_iterator::operator++() { assert(_wrap._root); _wrap._root = _wrap._root->next_sibling; return *this; } - xml_node_iterator xml_node_iterator::operator++(int) + PUGI__FN xml_node_iterator xml_node_iterator::operator++(int) { xml_node_iterator temp = *this; ++*this; return temp; } - const xml_node_iterator& xml_node_iterator::operator--() + PUGI__FN const xml_node_iterator& xml_node_iterator::operator--() { _wrap = _wrap._root ? _wrap.previous_sibling() : _parent.last_child(); return *this; } - xml_node_iterator xml_node_iterator::operator--(int) + PUGI__FN xml_node_iterator xml_node_iterator::operator--(int) { xml_node_iterator temp = *this; --*this; return temp; } - xml_attribute_iterator::xml_attribute_iterator() + PUGI__FN xml_attribute_iterator::xml_attribute_iterator() { } - xml_attribute_iterator::xml_attribute_iterator(const xml_attribute& attr, const xml_node& parent): _wrap(attr), _parent(parent) + PUGI__FN xml_attribute_iterator::xml_attribute_iterator(const xml_attribute& attr, const xml_node& parent): _wrap(attr), _parent(parent) { } - xml_attribute_iterator::xml_attribute_iterator(xml_attribute_struct* ref, xml_node_struct* parent): _wrap(ref), _parent(parent) + PUGI__FN xml_attribute_iterator::xml_attribute_iterator(xml_attribute_struct* ref, xml_node_struct* parent): _wrap(ref), _parent(parent) { } - bool xml_attribute_iterator::operator==(const xml_attribute_iterator& rhs) const + PUGI__FN bool xml_attribute_iterator::operator==(const xml_attribute_iterator& rhs) const { return _wrap._attr == rhs._wrap._attr && _parent._root == rhs._parent._root; } - - bool xml_attribute_iterator::operator!=(const xml_attribute_iterator& rhs) const + + PUGI__FN bool xml_attribute_iterator::operator!=(const xml_attribute_iterator& rhs) const { return _wrap._attr != rhs._wrap._attr || _parent._root != rhs._parent._root; } - xml_attribute& xml_attribute_iterator::operator*() + PUGI__FN xml_attribute& xml_attribute_iterator::operator*() const { assert(_wrap._attr); return _wrap; } - xml_attribute* xml_attribute_iterator::operator->() + PUGI__FN xml_attribute* xml_attribute_iterator::operator->() const { assert(_wrap._attr); - return &_wrap; + return const_cast(&_wrap); // BCC5 workaround } - const xml_attribute_iterator& xml_attribute_iterator::operator++() + PUGI__FN const xml_attribute_iterator& xml_attribute_iterator::operator++() { assert(_wrap._attr); _wrap._attr = _wrap._attr->next_attribute; return *this; } - xml_attribute_iterator xml_attribute_iterator::operator++(int) + PUGI__FN xml_attribute_iterator xml_attribute_iterator::operator++(int) { xml_attribute_iterator temp = *this; ++*this; return temp; } - const xml_attribute_iterator& xml_attribute_iterator::operator--() + PUGI__FN const xml_attribute_iterator& xml_attribute_iterator::operator--() { _wrap = _wrap._attr ? _wrap.previous_attribute() : _parent.last_attribute(); return *this; } - xml_attribute_iterator xml_attribute_iterator::operator--(int) + PUGI__FN xml_attribute_iterator xml_attribute_iterator::operator--(int) { xml_attribute_iterator temp = *this; --*this; return temp; } - xml_parse_result::xml_parse_result(): status(status_internal_error), offset(0), encoding(encoding_auto) - { - } + PUGI__FN xml_named_node_iterator::xml_named_node_iterator(): _name(0) + { + } + + PUGI__FN xml_named_node_iterator::xml_named_node_iterator(const xml_node& node, const char_t* name): _wrap(node), _parent(node.parent()), _name(name) + { + } + + PUGI__FN xml_named_node_iterator::xml_named_node_iterator(xml_node_struct* ref, xml_node_struct* parent, const char_t* name): _wrap(ref), _parent(parent), _name(name) + { + } + + PUGI__FN bool xml_named_node_iterator::operator==(const xml_named_node_iterator& rhs) const + { + return _wrap._root == rhs._wrap._root && _parent._root == rhs._parent._root; + } + + PUGI__FN bool xml_named_node_iterator::operator!=(const xml_named_node_iterator& rhs) const + { + return _wrap._root != rhs._wrap._root || _parent._root != rhs._parent._root; + } + + PUGI__FN xml_node& xml_named_node_iterator::operator*() const + { + assert(_wrap._root); + return _wrap; + } + + PUGI__FN xml_node* xml_named_node_iterator::operator->() const + { + assert(_wrap._root); + return const_cast(&_wrap); // BCC5 workaround + } + + PUGI__FN const xml_named_node_iterator& xml_named_node_iterator::operator++() + { + assert(_wrap._root); + _wrap = _wrap.next_sibling(_name); + return *this; + } + + PUGI__FN xml_named_node_iterator xml_named_node_iterator::operator++(int) + { + xml_named_node_iterator temp = *this; + ++*this; + return temp; + } + + PUGI__FN const xml_named_node_iterator& xml_named_node_iterator::operator--() + { + if (_wrap._root) + _wrap = _wrap.previous_sibling(_name); + else + { + _wrap = _parent.last_child(); + + if (!impl::strequal(_wrap.name(), _name)) + _wrap = _wrap.previous_sibling(_name); + } + + return *this; + } + + PUGI__FN xml_named_node_iterator xml_named_node_iterator::operator--(int) + { + xml_named_node_iterator temp = *this; + --*this; + return temp; + } + + PUGI__FN xml_parse_result::xml_parse_result(): status(status_internal_error), offset(0), encoding(encoding_auto) + { + } - xml_parse_result::operator bool() const - { - return status == status_ok; - } + PUGI__FN xml_parse_result::operator bool() const + { + return status == status_ok; + } - const char* xml_parse_result::description() const + PUGI__FN const char* xml_parse_result::description() const { switch (status) { @@ -4485,106 +6826,263 @@ namespace pugi case status_bad_end_element: return "Error parsing end element tag"; case status_end_element_mismatch: return "Start-end tags mismatch"; - default: return "Unknown error"; + case status_append_invalid_root: return "Unable to append nodes: root is not an element or document"; + + case status_no_document_element: return "No document element found"; + + default: return "Unknown error"; + } + } + + PUGI__FN xml_document::xml_document(): _buffer(0) + { + _create(); + } + + PUGI__FN xml_document::~xml_document() + { + _destroy(); + } + +#ifdef PUGIXML_HAS_MOVE + PUGI__FN xml_document::xml_document(xml_document&& rhs) PUGIXML_NOEXCEPT_IF_NOT_COMPACT: _buffer(0) + { + _create(); + _move(rhs); + } + + PUGI__FN xml_document& xml_document::operator=(xml_document&& rhs) PUGIXML_NOEXCEPT_IF_NOT_COMPACT + { + if (this == &rhs) return *this; + + _destroy(); + _create(); + _move(rhs); + + return *this; + } +#endif + + PUGI__FN void xml_document::reset() + { + _destroy(); + _create(); + } + + PUGI__FN void xml_document::reset(const xml_document& proto) + { + reset(); + + for (xml_node cur = proto.first_child(); cur; cur = cur.next_sibling()) + append_copy(cur); + } + + PUGI__FN void xml_document::_create() + { + assert(!_root); + + #ifdef PUGIXML_COMPACT + // space for page marker for the first page (uint32_t), rounded up to pointer size; assumes pointers are at least 32-bit + const size_t page_offset = sizeof(void*); + #else + const size_t page_offset = 0; + #endif + + // initialize sentinel page + PUGI__STATIC_ASSERT(sizeof(impl::xml_memory_page) + sizeof(impl::xml_document_struct) + page_offset <= sizeof(_memory)); + + // prepare page structure + impl::xml_memory_page* page = impl::xml_memory_page::construct(_memory); + assert(page); + + page->busy_size = impl::xml_memory_page_size; + + // setup first page marker + #ifdef PUGIXML_COMPACT + // round-trip through void* to avoid 'cast increases required alignment of target type' warning + page->compact_page_marker = reinterpret_cast(static_cast(reinterpret_cast(page) + sizeof(impl::xml_memory_page))); + *page->compact_page_marker = sizeof(impl::xml_memory_page); + #endif + + // allocate new root + _root = new (reinterpret_cast(page) + sizeof(impl::xml_memory_page) + page_offset) impl::xml_document_struct(page); + _root->prev_sibling_c = _root; + + // setup sentinel page + page->allocator = static_cast(_root); + + // setup hash table pointer in allocator + #ifdef PUGIXML_COMPACT + page->allocator->_hash = &static_cast(_root)->hash; + #endif + + // verify the document allocation + assert(reinterpret_cast(_root) + sizeof(impl::xml_document_struct) <= _memory + sizeof(_memory)); + } + + PUGI__FN void xml_document::_destroy() + { + assert(_root); + + // destroy static storage + if (_buffer) + { + impl::xml_memory::deallocate(_buffer); + _buffer = 0; + } + + // destroy extra buffers (note: no need to destroy linked list nodes, they're allocated using document allocator) + for (impl::xml_extra_buffer* extra = static_cast(_root)->extra_buffers; extra; extra = extra->next) + { + if (extra->buffer) impl::xml_memory::deallocate(extra->buffer); + } + + // destroy dynamic storage, leave sentinel page (it's in static memory) + impl::xml_memory_page* root_page = PUGI__GETPAGE(_root); + assert(root_page && !root_page->prev); + assert(reinterpret_cast(root_page) >= _memory && reinterpret_cast(root_page) < _memory + sizeof(_memory)); + + for (impl::xml_memory_page* page = root_page->next; page; ) + { + impl::xml_memory_page* next = page->next; + + impl::xml_allocator::deallocate_page(page); + + page = next; } - } - xml_document::xml_document(): _buffer(0) - { - create(); - } + #ifdef PUGIXML_COMPACT + // destroy hash table + static_cast(_root)->hash.clear(); + #endif - xml_document::~xml_document() - { - destroy(); + _root = 0; } - void xml_document::reset() +#ifdef PUGIXML_HAS_MOVE + PUGI__FN void xml_document::_move(xml_document& rhs) PUGIXML_NOEXCEPT_IF_NOT_COMPACT { - destroy(); - create(); - } + impl::xml_document_struct* doc = static_cast(_root); + impl::xml_document_struct* other = static_cast(rhs._root); - void xml_document::reset(const xml_document& proto) - { - reset(); + // save first child pointer for later; this needs hash access + xml_node_struct* other_first_child = other->first_child; - for (xml_node cur = proto.first_child(); cur; cur = cur.next_sibling()) - append_copy(cur); - } + #ifdef PUGIXML_COMPACT + // reserve space for the hash table up front; this is the only operation that can fail + // if it does, we have no choice but to throw (if we have exceptions) + if (other_first_child) + { + size_t other_children = 0; + for (xml_node_struct* node = other_first_child; node; node = node->next_sibling) + other_children++; - void xml_document::create() - { - // initialize sentinel page - STATIC_ASSERT(offsetof(xml_memory_page, data) + sizeof(xml_document_struct) + xml_memory_page_alignment <= sizeof(_memory)); + // in compact mode, each pointer assignment could result in a hash table request + // during move, we have to relocate document first_child and parents of all children + // normally there's just one child and its parent has a pointerless encoding but + // we assume the worst here + if (!other->_hash->reserve(other_children + 1)) + { + #ifdef PUGIXML_NO_EXCEPTIONS + return; + #else + throw std::bad_alloc(); + #endif + } + } + #endif - // align upwards to page boundary - void* page_memory = reinterpret_cast((reinterpret_cast(_memory) + (xml_memory_page_alignment - 1)) & ~(xml_memory_page_alignment - 1)); + // move allocation state + doc->_root = other->_root; + doc->_busy_size = other->_busy_size; - // prepare page structure - xml_memory_page* page = xml_memory_page::construct(page_memory); + // move buffer state + doc->buffer = other->buffer; + doc->extra_buffers = other->extra_buffers; + _buffer = rhs._buffer; - page->busy_size = xml_memory_page_size; + #ifdef PUGIXML_COMPACT + // move compact hash; note that the hash table can have pointers to other but they will be "inactive", similarly to nodes removed with remove_child + doc->hash = other->hash; + doc->_hash = &doc->hash; - // allocate new root - _root = new (page->data) xml_document_struct(page); - _root->prev_sibling_c = _root; + // make sure we don't access other hash up until the end when we reinitialize other document + other->_hash = 0; + #endif - // setup sentinel page - page->allocator = static_cast(_root); - } + // move page structure + impl::xml_memory_page* doc_page = PUGI__GETPAGE(doc); + assert(doc_page && !doc_page->prev && !doc_page->next); - void xml_document::destroy() - { - // destroy static storage - if (_buffer) + impl::xml_memory_page* other_page = PUGI__GETPAGE(other); + assert(other_page && !other_page->prev); + + // relink pages since root page is embedded into xml_document + if (impl::xml_memory_page* page = other_page->next) { - global_deallocate(_buffer); - _buffer = 0; + assert(page->prev == other_page); + + page->prev = doc_page; + + doc_page->next = page; + other_page->next = 0; } - // destroy dynamic storage, leave sentinel page (it's in static memory) - if (_root) + // make sure pages point to the correct document state + for (impl::xml_memory_page* page = doc_page->next; page; page = page->next) { - xml_memory_page* root_page = reinterpret_cast(_root->header & xml_memory_page_pointer_mask); - assert(root_page && !root_page->prev && !root_page->memory); + assert(page->allocator == other); - // destroy all pages - for (xml_memory_page* page = root_page->next; page; ) - { - xml_memory_page* next = page->next; + page->allocator = doc; - xml_allocator::deallocate_page(page); + #ifdef PUGIXML_COMPACT + // this automatically migrates most children between documents and prevents ->parent assignment from allocating + if (page->compact_shared_parent == other) + page->compact_shared_parent = doc; + #endif + } - page = next; - } + // move tree structure + assert(!doc->first_child); + + doc->first_child = other_first_child; - // cleanup root page - root_page->allocator = 0; - root_page->next = 0; - root_page->busy_size = root_page->freed_size = 0; + for (xml_node_struct* node = other_first_child; node; node = node->next_sibling) + { + #ifdef PUGIXML_COMPACT + // most children will have migrated when we reassigned compact_shared_parent + assert(node->parent == other || node->parent == doc); - _root = 0; + node->parent = doc; + #else + assert(node->parent == other); + node->parent = doc; + #endif } + + // reset other document + new (other) impl::xml_document_struct(PUGI__GETPAGE(other)); + rhs._buffer = 0; } +#endif #ifndef PUGIXML_NO_STL - xml_parse_result xml_document::load(std::basic_istream >& stream, unsigned int options, xml_encoding encoding) + PUGI__FN xml_parse_result xml_document::load(std::basic_istream >& stream, unsigned int options, xml_encoding encoding) { reset(); - return load_stream_impl(*this, stream, options, encoding); + return impl::load_stream_impl(static_cast(_root), stream, options, encoding, &_buffer); } - xml_parse_result xml_document::load(std::basic_istream >& stream, unsigned int options) + PUGI__FN xml_parse_result xml_document::load(std::basic_istream >& stream, unsigned int options) { reset(); - return load_stream_impl(*this, stream, options, encoding_wchar); + return impl::load_stream_impl(static_cast(_root), stream, options, encoding_wchar, &_buffer); } #endif - xml_parse_result xml_document::load(const char_t* contents, unsigned int options) + PUGI__FN xml_parse_result xml_document::load_string(const char_t* contents, unsigned int options) { // Force native encoding (skip autodetection) #ifdef PUGIXML_WCHAR_MODE @@ -4593,97 +7091,92 @@ namespace pugi xml_encoding encoding = encoding_utf8; #endif - return load_buffer(contents, strlength(contents) * sizeof(char_t), options, encoding); + return load_buffer(contents, impl::strlength(contents) * sizeof(char_t), options, encoding); } - xml_parse_result xml_document::load_file(const char* path, unsigned int options, xml_encoding encoding) + PUGI__FN xml_parse_result xml_document::load(const char_t* contents, unsigned int options) { - reset(); - - FILE* file = fopen(path, "rb"); - - return load_file_impl(*this, file, options, encoding); + return load_string(contents, options); } - xml_parse_result xml_document::load_file(const wchar_t* path, unsigned int options, xml_encoding encoding) + PUGI__FN xml_parse_result xml_document::load_file(const char* path_, unsigned int options, xml_encoding encoding) { reset(); - FILE* file = open_file_wide(path, L"rb"); + using impl::auto_deleter; // MSVC7 workaround + auto_deleter file(fopen(path_, "rb"), impl::close_file); - return load_file_impl(*this, file, options, encoding); + return impl::load_file_impl(static_cast(_root), file.data, options, encoding, &_buffer); } - xml_parse_result xml_document::load_buffer_impl(void* contents, size_t size, unsigned int options, xml_encoding encoding, bool is_mutable, bool own) + PUGI__FN xml_parse_result xml_document::load_file(const wchar_t* path_, unsigned int options, xml_encoding encoding) { reset(); - // check input buffer - assert(contents || size == 0); - - // get actual encoding - xml_encoding buffer_encoding = get_buffer_encoding(encoding, contents, size); - - // get private buffer - char_t* buffer = 0; - size_t length = 0; - - if (!convert_buffer(buffer, length, buffer_encoding, contents, size, is_mutable)) return make_parse_result(status_out_of_memory); - - // delete original buffer if we performed a conversion - if (own && buffer != contents && contents) global_deallocate(contents); - - // parse - xml_parse_result res = xml_parser::parse(buffer, length, _root, options); - - // remember encoding - res.encoding = buffer_encoding; + using impl::auto_deleter; // MSVC7 workaround + auto_deleter file(impl::open_file_wide(path_, L"rb"), impl::close_file); - // grab onto buffer if it's our buffer, user is responsible for deallocating contens himself - if (own || buffer != contents) _buffer = buffer; - - return res; + return impl::load_file_impl(static_cast(_root), file.data, options, encoding, &_buffer); } - xml_parse_result xml_document::load_buffer(const void* contents, size_t size, unsigned int options, xml_encoding encoding) + PUGI__FN xml_parse_result xml_document::load_buffer(const void* contents, size_t size, unsigned int options, xml_encoding encoding) { - return load_buffer_impl(const_cast(contents), size, options, encoding, false, false); + reset(); + + return impl::load_buffer_impl(static_cast(_root), _root, const_cast(contents), size, options, encoding, false, false, &_buffer); } - xml_parse_result xml_document::load_buffer_inplace(void* contents, size_t size, unsigned int options, xml_encoding encoding) + PUGI__FN xml_parse_result xml_document::load_buffer_inplace(void* contents, size_t size, unsigned int options, xml_encoding encoding) { - return load_buffer_impl(contents, size, options, encoding, true, false); + reset(); + + return impl::load_buffer_impl(static_cast(_root), _root, contents, size, options, encoding, true, false, &_buffer); } - - xml_parse_result xml_document::load_buffer_inplace_own(void* contents, size_t size, unsigned int options, xml_encoding encoding) + + PUGI__FN xml_parse_result xml_document::load_buffer_inplace_own(void* contents, size_t size, unsigned int options, xml_encoding encoding) { - return load_buffer_impl(contents, size, options, encoding, true, true); + reset(); + + return impl::load_buffer_impl(static_cast(_root), _root, contents, size, options, encoding, true, true, &_buffer); } - void xml_document::save(xml_writer& writer, const char_t* indent, unsigned int flags, xml_encoding encoding) const + PUGI__FN void xml_document::save(xml_writer& writer, const char_t* indent, unsigned int flags, xml_encoding encoding) const { - if (flags & format_write_bom) write_bom(writer, get_write_encoding(encoding)); + impl::xml_buffered_writer buffered_writer(writer, encoding); - xml_buffered_writer buffered_writer(writer, encoding); + if ((flags & format_write_bom) && encoding != encoding_latin1) + { + // BOM always represents the codepoint U+FEFF, so just write it in native encoding + #ifdef PUGIXML_WCHAR_MODE + unsigned int bom = 0xfeff; + buffered_writer.write(static_cast(bom)); + #else + buffered_writer.write('\xef', '\xbb', '\xbf'); + #endif + } - if (!(flags & format_no_declaration) && !has_declaration(*this)) + if (!(flags & format_no_declaration) && !impl::has_declaration(_root)) { - buffered_writer.write(PUGIXML_TEXT("")); + buffered_writer.write_string(PUGIXML_TEXT("'); if (!(flags & format_raw)) buffered_writer.write('\n'); } - node_output(buffered_writer, *this, indent, flags, 0); + impl::node_output(buffered_writer, _root, indent, flags, 0); + + buffered_writer.flush(); } #ifndef PUGIXML_NO_STL - void xml_document::save(std::basic_ostream >& stream, const char_t* indent, unsigned int flags, xml_encoding encoding) const + PUGI__FN void xml_document::save(std::basic_ostream >& stream, const char_t* indent, unsigned int flags, xml_encoding encoding) const { xml_writer_stream writer(stream); save(writer, indent, flags, encoding); } - void xml_document::save(std::basic_ostream >& stream, const char_t* indent, unsigned int flags) const + PUGI__FN void xml_document::save(std::basic_ostream >& stream, const char_t* indent, unsigned int flags) const { xml_writer_stream writer(stream); @@ -4691,94 +7184,91 @@ namespace pugi } #endif - bool xml_document::save_file(const char* path, const char_t* indent, unsigned int flags, xml_encoding encoding) const + PUGI__FN bool xml_document::save_file(const char* path_, const char_t* indent, unsigned int flags, xml_encoding encoding) const { - FILE* file = fopen(path, "wb"); - if (!file) return false; - - xml_writer_file writer(file); - save(writer, indent, flags, encoding); + using impl::auto_deleter; // MSVC7 workaround + auto_deleter file(fopen(path_, (flags & format_save_file_text) ? "w" : "wb"), impl::close_file); - fclose(file); - - return true; + return impl::save_file_impl(*this, file.data, indent, flags, encoding); } - bool xml_document::save_file(const wchar_t* path, const char_t* indent, unsigned int flags, xml_encoding encoding) const + PUGI__FN bool xml_document::save_file(const wchar_t* path_, const char_t* indent, unsigned int flags, xml_encoding encoding) const { - FILE* file = open_file_wide(path, L"wb"); - if (!file) return false; + using impl::auto_deleter; // MSVC7 workaround + auto_deleter file(impl::open_file_wide(path_, (flags & format_save_file_text) ? L"w" : L"wb"), impl::close_file); - xml_writer_file writer(file); - save(writer, indent, flags, encoding); - - fclose(file); - - return true; + return impl::save_file_impl(*this, file.data, indent, flags, encoding); } - xml_node xml_document::document_element() const - { + PUGI__FN xml_node xml_document::document_element() const + { + assert(_root); + for (xml_node_struct* i = _root->first_child; i; i = i->next_sibling) - if ((i->header & xml_memory_page_type_mask) + 1 == node_element) - return xml_node(i); + if (PUGI__NODETYPE(i) == node_element) + return xml_node(i); - return xml_node(); - } + return xml_node(); + } #ifndef PUGIXML_NO_STL - std::string PUGIXML_FUNCTION as_utf8(const wchar_t* str) + PUGI__FN std::string PUGIXML_FUNCTION as_utf8(const wchar_t* str) { assert(str); - return as_utf8_impl(str, wcslen(str)); + return impl::as_utf8_impl(str, impl::strlength_wide(str)); } - std::string PUGIXML_FUNCTION as_utf8(const std::wstring& str) + PUGI__FN std::string PUGIXML_FUNCTION as_utf8(const std::basic_string& str) { - return as_utf8_impl(str.c_str(), str.size()); + return impl::as_utf8_impl(str.c_str(), str.size()); } - - std::wstring PUGIXML_FUNCTION as_wide(const char* str) + + PUGI__FN std::basic_string PUGIXML_FUNCTION as_wide(const char* str) { assert(str); - return as_wide_impl(str, strlen(str)); + return impl::as_wide_impl(str, strlen(str)); } - - std::wstring PUGIXML_FUNCTION as_wide(const std::string& str) + + PUGI__FN std::basic_string PUGIXML_FUNCTION as_wide(const std::string& str) { - return as_wide_impl(str.c_str(), str.size()); + return impl::as_wide_impl(str.c_str(), str.size()); } #endif - void PUGIXML_FUNCTION set_memory_management_functions(allocation_function allocate, deallocation_function deallocate) - { - global_allocate = allocate; - global_deallocate = deallocate; - } - - allocation_function PUGIXML_FUNCTION get_memory_allocation_function() - { - return global_allocate; - } - - deallocation_function PUGIXML_FUNCTION get_memory_deallocation_function() - { - return global_deallocate; - } + PUGI__FN void PUGIXML_FUNCTION set_memory_management_functions(allocation_function allocate, deallocation_function deallocate) + { + impl::xml_memory::allocate = allocate; + impl::xml_memory::deallocate = deallocate; + } + + PUGI__FN allocation_function PUGIXML_FUNCTION get_memory_allocation_function() + { + return impl::xml_memory::allocate; + } + + PUGI__FN deallocation_function PUGIXML_FUNCTION get_memory_deallocation_function() + { + return impl::xml_memory::deallocate; + } } #if !defined(PUGIXML_NO_STL) && (defined(_MSC_VER) || defined(__ICC)) namespace std { // Workarounds for (non-standard) iterator category detection for older versions (MSVC7/IC8 and earlier) - std::bidirectional_iterator_tag _Iter_cat(const xml_node_iterator&) + PUGI__FN std::bidirectional_iterator_tag _Iter_cat(const pugi::xml_node_iterator&) + { + return std::bidirectional_iterator_tag(); + } + + PUGI__FN std::bidirectional_iterator_tag _Iter_cat(const pugi::xml_attribute_iterator&) { return std::bidirectional_iterator_tag(); } - std::bidirectional_iterator_tag _Iter_cat(const xml_attribute_iterator&) + PUGI__FN std::bidirectional_iterator_tag _Iter_cat(const pugi::xml_named_node_iterator&) { return std::bidirectional_iterator_tag(); } @@ -4789,12 +7279,17 @@ namespace std namespace std { // Workarounds for (non-standard) iterator category detection - std::bidirectional_iterator_tag __iterator_category(const xml_node_iterator&) + PUGI__FN std::bidirectional_iterator_tag __iterator_category(const pugi::xml_node_iterator&) { return std::bidirectional_iterator_tag(); } - std::bidirectional_iterator_tag __iterator_category(const xml_attribute_iterator&) + PUGI__FN std::bidirectional_iterator_tag __iterator_category(const pugi::xml_attribute_iterator&) + { + return std::bidirectional_iterator_tag(); + } + + PUGI__FN std::bidirectional_iterator_tag __iterator_category(const pugi::xml_named_node_iterator&) { return std::bidirectional_iterator_tag(); } @@ -4802,10 +7297,8 @@ namespace std #endif #ifndef PUGIXML_NO_XPATH - // STL replacements -namespace -{ +PUGI__NS_BEGIN struct equal_to { template bool operator()(const T& lhs, const T& rhs) const @@ -4858,18 +7351,18 @@ namespace template void reverse(I begin, I end) { - while (begin + 1 < end) swap(*begin++, *--end); + while (end - begin > 1) swap(*begin++, *--end); } template I unique(I begin, I end) { // fast skip head - while (begin + 1 < end && *begin != *(begin + 1)) begin++; + while (end - begin > 1 && *begin != *(begin + 1)) begin++; if (begin == end) return begin; // last written element - I write = begin++; + I write = begin++; // merge unique elements while (begin != end) @@ -4884,134 +7377,76 @@ namespace return write + 1; } - template void copy_backwards(I begin, I end, I target) - { - while (begin != end) *--target = *--end; - } - - template void insertion_sort(I begin, I end, const Pred& pred, T*) + template void insertion_sort(T* begin, T* end, const Pred& pred) { - assert(begin != end); + if (begin == end) + return; - for (I it = begin + 1; it != end; ++it) + for (T* it = begin + 1; it != end; ++it) { T val = *it; + T* hole = it; - if (pred(val, *begin)) + // move hole backwards + while (hole > begin && pred(val, *(hole - 1))) { - // move to front - copy_backwards(begin, it, it + 1); - *begin = val; + *hole = *(hole - 1); + hole--; } - else - { - I hole = it; - // move hole backwards - while (pred(val, *(hole - 1))) - { - *hole = *(hole - 1); - hole--; - } - - // fill hole with element - *hole = val; - } + // fill hole with element + *hole = val; } } - // std variant for elements with == - template void partition(I begin, I middle, I end, const Pred& pred, I* out_eqbeg, I* out_eqend) + template I median3(I first, I middle, I last, const Pred& pred) { - I eqbeg = middle, eqend = middle + 1; + if (pred(*middle, *first)) swap(middle, first); + if (pred(*last, *middle)) swap(last, middle); + if (pred(*middle, *first)) swap(middle, first); - // expand equal range - while (eqbeg != begin && *(eqbeg - 1) == *eqbeg) --eqbeg; - while (eqend != end && *eqend == *eqbeg) ++eqend; - - // process outer elements - I ltend = eqbeg, gtbeg = eqend; - - for (;;) - { - // find the element from the right side that belongs to the left one - for (; gtbeg != end; ++gtbeg) - if (!pred(*eqbeg, *gtbeg)) - { - if (*gtbeg == *eqbeg) swap(*gtbeg, *eqend++); - else break; - } - - // find the element from the left side that belongs to the right one - for (; ltend != begin; --ltend) - if (!pred(*(ltend - 1), *eqbeg)) - { - if (*eqbeg == *(ltend - 1)) swap(*(ltend - 1), *--eqbeg); - else break; - } - - // scanned all elements - if (gtbeg == end && ltend == begin) - { - *out_eqbeg = eqbeg; - *out_eqend = eqend; - return; - } - - // make room for elements by moving equal area - if (gtbeg == end) - { - if (--ltend != --eqbeg) swap(*ltend, *eqbeg); - swap(*eqbeg, *--eqend); - } - else if (ltend == begin) - { - if (eqend != gtbeg) swap(*eqbeg, *eqend); - ++eqend; - swap(*gtbeg++, *eqbeg++); - } - else swap(*gtbeg++, *--ltend); - } + return middle; } - template void median3(I first, I middle, I last, const Pred& pred) + template void partition3(T* begin, T* end, T pivot, const Pred& pred, T** out_eqbeg, T** out_eqend) { - if (pred(*middle, *first)) swap(*middle, *first); - if (pred(*last, *middle)) swap(*last, *middle); - if (pred(*middle, *first)) swap(*middle, *first); - } + // invariant: array is split into 4 groups: = < ? > (each variable denotes the boundary between the groups) + T* eq = begin; + T* lt = begin; + T* gt = end; - template void median(I first, I middle, I last, const Pred& pred) - { - if (last - first <= 40) + while (lt < gt) { - // median of three for small chunks - median3(first, middle, last, pred); + if (pred(*lt, pivot)) + lt++; + else if (*lt == pivot) + swap(*eq++, *lt++); + else + swap(*lt, *--gt); } - else - { - // median of nine - size_t step = (last - first + 1) / 8; - median3(first, first + step, first + 2 * step, pred); - median3(middle - step, middle, middle + step, pred); - median3(last - 2 * step, last - step, last, pred); - median3(first + step, middle, last - step, pred); - } + // we now have just 4 groups: = < >; move equal elements to the middle + T* eqbeg = gt; + + for (T* it = begin; it != eq; ++it) + swap(*it, *--eqbeg); + + *out_eqbeg = eqbeg; + *out_eqend = gt; } template void sort(I begin, I end, const Pred& pred) { // sort large chunks - while (end - begin > 32) + while (end - begin > 16) { // find median element I middle = begin + (end - begin) / 2; - median(begin, middle, end - 1, pred); + I median = median3(begin, middle, end - 1, pred); // partition in three chunks (< = >) I eqbeg, eqend; - partition(begin, middle, end, pred, &eqbeg, &eqend); + partition3(begin, end, *median, pred, &eqbeg, &eqend); // loop on larger half if (eqbeg - begin > end - eqend) @@ -5027,121 +7462,120 @@ namespace } // insertion sort small chunk - if (begin != end) insertion_sort(begin, end, pred, &*begin); + insertion_sort(begin, end, pred); } -} +PUGI__NS_END // Allocator used for AST and evaluation stacks -namespace -{ +PUGI__NS_BEGIN + static const size_t xpath_memory_page_size = + #ifdef PUGIXML_MEMORY_XPATH_PAGE_SIZE + PUGIXML_MEMORY_XPATH_PAGE_SIZE + #else + 4096 + #endif + ; + + static const uintptr_t xpath_memory_block_alignment = sizeof(double) > sizeof(void*) ? sizeof(double) : sizeof(void*); + struct xpath_memory_block - { + { xpath_memory_block* next; + size_t capacity; - char data[4096]; + union + { + char data[xpath_memory_page_size]; + double alignment; + }; }; - - class xpath_allocator + + struct xpath_allocator { xpath_memory_block* _root; size_t _root_size; + bool* _error; - public: - #ifdef PUGIXML_NO_EXCEPTIONS - jmp_buf* error_handler; - #endif - - xpath_allocator(xpath_memory_block* root, size_t root_size = 0): _root(root), _root_size(root_size) + xpath_allocator(xpath_memory_block* root, bool* error = 0): _root(root), _root_size(0), _error(error) { - #ifdef PUGIXML_NO_EXCEPTIONS - error_handler = 0; - #endif } - - void* allocate_nothrow(size_t size) - { - const size_t block_capacity = sizeof(_root->data); - // align size so that we're able to store pointers in subsequent blocks - size = (size + sizeof(void*) - 1) & ~(sizeof(void*) - 1); + void* allocate(size_t size) + { + // round size up to block alignment boundary + size = (size + xpath_memory_block_alignment - 1) & ~(xpath_memory_block_alignment - 1); - if (_root_size + size <= block_capacity) + if (_root_size + size <= _root->capacity) { - void* buf = _root->data + _root_size; + void* buf = &_root->data[0] + _root_size; _root_size += size; return buf; } else { - size_t block_data_size = (size > block_capacity) ? size : block_capacity; - size_t block_size = block_data_size + offsetof(xpath_memory_block, data); + // make sure we have at least 1/4th of the page free after allocation to satisfy subsequent allocation requests + size_t block_capacity_base = sizeof(_root->data); + size_t block_capacity_req = size + block_capacity_base / 4; + size_t block_capacity = (block_capacity_base > block_capacity_req) ? block_capacity_base : block_capacity_req; + + size_t block_size = block_capacity + offsetof(xpath_memory_block, data); + + xpath_memory_block* block = static_cast(xml_memory::allocate(block_size)); + if (!block) + { + if (_error) *_error = true; + return 0; + } - xpath_memory_block* block = static_cast(global_allocate(block_size)); - if (!block) return 0; - block->next = _root; - + block->capacity = block_capacity; + _root = block; _root_size = size; - - return block->data; - } - } - - void* allocate(size_t size) - { - void* result = allocate_nothrow(size); - if (!result) - { - #ifdef PUGIXML_NO_EXCEPTIONS - assert(error_handler); - longjmp(*error_handler, 1); - #else - throw std::bad_alloc(); - #endif + return block->data; } - - return result; } void* reallocate(void* ptr, size_t old_size, size_t new_size) { - // align size so that we're able to store pointers in subsequent blocks - old_size = (old_size + sizeof(void*) - 1) & ~(sizeof(void*) - 1); - new_size = (new_size + sizeof(void*) - 1) & ~(sizeof(void*) - 1); + // round size up to block alignment boundary + old_size = (old_size + xpath_memory_block_alignment - 1) & ~(xpath_memory_block_alignment - 1); + new_size = (new_size + xpath_memory_block_alignment - 1) & ~(xpath_memory_block_alignment - 1); // we can only reallocate the last object - assert(ptr == 0 || static_cast(ptr) + old_size == _root->data + _root_size); + assert(ptr == 0 || static_cast(ptr) + old_size == &_root->data[0] + _root_size); - // adjust root size so that we have not allocated the object at all - bool only_object = (_root_size == old_size); - - if (ptr) _root_size -= old_size; + // try to reallocate the object inplace + if (ptr && _root_size - old_size + new_size <= _root->capacity) + { + _root_size = _root_size - old_size + new_size; + return ptr; + } - // allocate a new version (this will obviously reuse the memory if possible) + // allocate a new block void* result = allocate(new_size); - assert(result); + if (!result) return 0; // we have a new block - if (result != ptr && ptr) + if (ptr) { - // copy old data - assert(new_size > old_size); + // copy old data (we only support growing) + assert(new_size >= old_size); memcpy(result, ptr, old_size); // free the previous page if it had no other objects - if (only_object) - { - assert(_root->data == result); - assert(_root->next); + assert(_root->data == result); + assert(_root->next); + if (_root->next->data == ptr) + { + // deallocate the whole page, unless it was the first one xpath_memory_block* next = _root->next->next; if (next) { - // deallocate the whole page, unless it was the first one - global_deallocate(_root->next); + xml_memory::deallocate(_root->next); _root->next = next; } } @@ -5159,7 +7593,7 @@ namespace { xpath_memory_block* next = cur->next; - global_deallocate(cur); + xml_memory::deallocate(cur); cur = next; } @@ -5178,7 +7612,7 @@ namespace { xpath_memory_block* next = cur->next; - global_deallocate(cur); + xml_memory::deallocate(cur); cur = next; } @@ -5212,21 +7646,15 @@ namespace xpath_allocator result; xpath_allocator temp; xpath_stack stack; + bool oom; - #ifdef PUGIXML_NO_EXCEPTIONS - jmp_buf error_handler; - #endif - - xpath_stack_data(): result(blocks + 0), temp(blocks + 1) + xpath_stack_data(): result(blocks + 0, &oom), temp(blocks + 1, &oom), oom(false) { blocks[0].next = blocks[1].next = 0; + blocks[0].capacity = blocks[1].capacity = sizeof(blocks[0].data); stack.result = &result; stack.temp = &temp; - - #ifdef PUGIXML_NO_EXCEPTIONS - result.error_handler = temp.error_handler = &error_handler; - #endif } ~xpath_stack_data() @@ -5235,20 +7663,20 @@ namespace temp.release(); } }; -} +PUGI__NS_END // String class -namespace -{ +PUGI__NS_BEGIN class xpath_string { const char_t* _buffer; bool _uses_heap; + size_t _length_heap; static char_t* duplicate_string(const char_t* string, size_t length, xpath_allocator* alloc) { char_t* result = static_cast(alloc->allocate((length + 1) * sizeof(char_t))); - assert(result); + if (!result) return 0; memcpy(result, string, length * sizeof(char_t)); result[length] = 0; @@ -5256,36 +7684,38 @@ namespace return result; } - static char_t* duplicate_string(const char_t* string, xpath_allocator* alloc) + xpath_string(const char_t* buffer, bool uses_heap_, size_t length_heap): _buffer(buffer), _uses_heap(uses_heap_), _length_heap(length_heap) { - return duplicate_string(string, strlength(string), alloc); } public: - xpath_string(): _buffer(PUGIXML_TEXT("")), _uses_heap(false) + static xpath_string from_const(const char_t* str) { + return xpath_string(str, false, 0); } - explicit xpath_string(const char_t* str, xpath_allocator* alloc) + static xpath_string from_heap_preallocated(const char_t* begin, const char_t* end) { - bool empty = (*str == 0); + assert(begin <= end && *end == 0); - _buffer = empty ? PUGIXML_TEXT("") : duplicate_string(str, alloc); - _uses_heap = !empty; + return xpath_string(begin, true, static_cast(end - begin)); } - explicit xpath_string(const char_t* str, bool use_heap): _buffer(str), _uses_heap(use_heap) - { - } - - xpath_string(const char_t* begin, const char_t* end, xpath_allocator* alloc) + static xpath_string from_heap(const char_t* begin, const char_t* end, xpath_allocator* alloc) { assert(begin <= end); - bool empty = (begin == end); + if (begin == end) + return xpath_string(); + + size_t length = static_cast(end - begin); + const char_t* data = duplicate_string(begin, length, alloc); + + return data ? xpath_string(data, true, length) : xpath_string(); + } - _buffer = empty ? PUGIXML_TEXT("") : duplicate_string(begin, static_cast(end - begin), alloc); - _uses_heap = !empty; + xpath_string(): _buffer(PUGIXML_TEXT("")), _uses_heap(false), _length_heap(0) + { } void append(const xpath_string& o, xpath_allocator* alloc) @@ -5301,24 +7731,25 @@ namespace else { // need to make heap copy - size_t target_length = strlength(_buffer); - size_t source_length = strlength(o._buffer); - size_t length = target_length + source_length; + size_t target_length = length(); + size_t source_length = o.length(); + size_t result_length = target_length + source_length; // allocate new buffer - char_t* result = static_cast(alloc->reallocate(_uses_heap ? const_cast(_buffer) : 0, (target_length + 1) * sizeof(char_t), (length + 1) * sizeof(char_t))); - assert(result); + char_t* result = static_cast(alloc->reallocate(_uses_heap ? const_cast(_buffer) : 0, (target_length + 1) * sizeof(char_t), (result_length + 1) * sizeof(char_t))); + if (!result) return; // append first string to the new buffer in case there was no reallocation if (!_uses_heap) memcpy(result, _buffer, target_length * sizeof(char_t)); // append second string to the new buffer memcpy(result + target_length, o._buffer, source_length * sizeof(char_t)); - result[length] = 0; + result[result_length] = 0; // finalize _buffer = result; _uses_heap = true; + _length_heap = result_length; } } @@ -5329,16 +7760,22 @@ namespace size_t length() const { - return strlength(_buffer); + return _uses_heap ? _length_heap : strlength(_buffer); } - + char_t* data(xpath_allocator* alloc) { // make private heap copy if (!_uses_heap) { - _buffer = duplicate_string(_buffer, alloc); + size_t length_ = strlength(_buffer); + const char_t* data_ = duplicate_string(_buffer, length_, alloc); + + if (!data_) return 0; + + _buffer = data_; _uses_heap = true; + _length_heap = length_; } return const_cast(_buffer); @@ -5364,16 +7801,10 @@ namespace return _uses_heap; } }; +PUGI__NS_END - xpath_string xpath_string_const(const char_t* str) - { - return xpath_string(str, false); - } -} - -namespace -{ - bool starts_with(const char_t* string, const char_t* pattern) +PUGI__NS_BEGIN + PUGI__FN bool starts_with(const char_t* string, const char_t* pattern) { while (*pattern && *string == *pattern) { @@ -5384,7 +7815,7 @@ namespace return *pattern == 0; } - const char_t* find_char(const char_t* s, char_t c) + PUGI__FN const char_t* find_char(const char_t* s, char_t c) { #ifdef PUGIXML_WCHAR_MODE return wcschr(s, c); @@ -5393,7 +7824,7 @@ namespace #endif } - const char_t* find_substring(const char_t* s, const char_t* p) + PUGI__FN const char_t* find_substring(const char_t* s, const char_t* p) { #ifdef PUGIXML_WCHAR_MODE // MSVC6 wcsstr bug workaround (if s is empty it always returns 0) @@ -5404,18 +7835,18 @@ namespace } // Converts symbol to lower case, if it is an ASCII one - char_t tolower_ascii(char_t ch) + PUGI__FN char_t tolower_ascii(char_t ch) { return static_cast(ch - 'A') < 26 ? static_cast(ch | ' ') : ch; } - xpath_string string_value(const xpath_node& na, xpath_allocator* alloc) + PUGI__FN xpath_string string_value(const xpath_node& na, xpath_allocator* alloc) { if (na.attribute()) - return xpath_string_const(na.attribute().value()); + return xpath_string::from_const(na.attribute().value()); else { - const xml_node& n = na.node(); + xml_node n = na.node(); switch (n.type()) { @@ -5423,19 +7854,23 @@ namespace case node_cdata: case node_comment: case node_pi: - return xpath_string_const(n.value()); - + return xpath_string::from_const(n.value()); + case node_document: case node_element: { xpath_string result; + // element nodes can have value if parse_embed_pcdata was used + if (n.value()[0]) + result.append(xpath_string::from_const(n.value()), alloc); + xml_node cur = n.first_child(); - + while (cur && cur != n) { if (cur.type() == node_pcdata || cur.type() == node_cdata) - result.append(xpath_string_const(cur.value()), alloc); + result.append(xpath_string::from_const(cur.value()), alloc); if (cur.first_child()) cur = cur.first_child(); @@ -5446,100 +7881,135 @@ namespace while (!cur.next_sibling() && cur != n) cur = cur.parent(); - if (cur != n) cur = cur.next_sibling(); - } - } - - return result; - } - - default: - return xpath_string(); - } + if (cur != n) cur = cur.next_sibling(); + } + } + + return result; + } + + default: + return xpath_string(); + } + } + } + + PUGI__FN bool node_is_before_sibling(xml_node_struct* ln, xml_node_struct* rn) + { + assert(ln->parent == rn->parent); + + // there is no common ancestor (the shared parent is null), nodes are from different documents + if (!ln->parent) return ln < rn; + + // determine sibling order + xml_node_struct* ls = ln; + xml_node_struct* rs = rn; + + while (ls && rs) + { + if (ls == rn) return true; + if (rs == ln) return false; + + ls = ls->next_sibling; + rs = rs->next_sibling; + } + + // if rn sibling chain ended ln must be before rn + return !rs; + } + + PUGI__FN bool node_is_before(xml_node_struct* ln, xml_node_struct* rn) + { + // find common ancestor at the same depth, if any + xml_node_struct* lp = ln; + xml_node_struct* rp = rn; + + while (lp && rp && lp->parent != rp->parent) + { + lp = lp->parent; + rp = rp->parent; + } + + // parents are the same! + if (lp && rp) return node_is_before_sibling(lp, rp); + + // nodes are at different depths, need to normalize heights + bool left_higher = !lp; + + while (lp) + { + lp = lp->parent; + ln = ln->parent; + } + + while (rp) + { + rp = rp->parent; + rn = rn->parent; + } + + // one node is the ancestor of the other + if (ln == rn) return left_higher; + + // find common ancestor... again + while (ln->parent != rn->parent) + { + ln = ln->parent; + rn = rn->parent; } + + return node_is_before_sibling(ln, rn); } - - unsigned int node_height(xml_node n) + + PUGI__FN bool node_is_ancestor(xml_node_struct* parent, xml_node_struct* node) { - unsigned int result = 0; - - while (n) - { - ++result; - n = n.parent(); - } - - return result; + while (node && node != parent) node = node->parent; + + return parent && node == parent; } - - bool node_is_before(xml_node ln, unsigned int lh, xml_node rn, unsigned int rh) + + PUGI__FN const void* document_buffer_order(const xpath_node& xnode) { - // normalize heights - for (unsigned int i = rh; i < lh; i++) ln = ln.parent(); - for (unsigned int j = lh; j < rh; j++) rn = rn.parent(); - - // one node is the ancestor of the other - if (ln == rn) return lh < rh; - - // find common ancestor - while (ln.parent() != rn.parent()) - { - ln = ln.parent(); - rn = rn.parent(); - } + xml_node_struct* node = xnode.node().internal_object(); - // there is no common ancestor (the shared parent is null), nodes are from different documents - if (!ln.parent()) return ln < rn; + if (node) + { + if ((get_document(node).header & xml_memory_page_contents_shared_mask) == 0) + { + if (node->name && (node->header & impl::xml_memory_page_name_allocated_or_shared_mask) == 0) return node->name; + if (node->value && (node->header & impl::xml_memory_page_value_allocated_or_shared_mask) == 0) return node->value; + } - // determine sibling order - for (; ln; ln = ln.next_sibling()) - if (ln == rn) - return true; - - return false; - } - - bool node_is_ancestor(xml_node parent, xml_node node) - { - while (node && node != parent) node = node.parent(); - - return parent && node == parent; - } - - const void* document_order(const xpath_node& xnode) - { - xml_node_struct* node = xnode.node().internal_object(); - - if (node) - { - if (node->name && (node->header & xml_memory_page_name_allocated_mask) == 0) return node->name; - if (node->value && (node->header & xml_memory_page_value_allocated_mask) == 0) return node->value; - return 0; - } - - xml_attribute_struct* attr = xnode.attribute().internal_object(); - - if (attr) - { - if ((attr->header & xml_memory_page_name_allocated_mask) == 0) return attr->name; - if ((attr->header & xml_memory_page_value_allocated_mask) == 0) return attr->value; - return 0; - } + return 0; + } + + xml_attribute_struct* attr = xnode.attribute().internal_object(); + + if (attr) + { + if ((get_document(attr).header & xml_memory_page_contents_shared_mask) == 0) + { + if ((attr->header & impl::xml_memory_page_name_allocated_or_shared_mask) == 0) return attr->name; + if ((attr->header & impl::xml_memory_page_value_allocated_or_shared_mask) == 0) return attr->value; + } + + return 0; + } return 0; - } - + } + struct document_order_comparator { bool operator()(const xpath_node& lhs, const xpath_node& rhs) const { // optimized document order based check - const void* lo = document_order(lhs); - const void* ro = document_order(rhs); + const void* lo = document_buffer_order(lhs); + const void* ro = document_buffer_order(rhs); if (lo && ro) return lo < ro; - // slow comparison + // slow comparison xml_node ln = lhs.node(), rn = rhs.node(); // compare attributes @@ -5549,13 +8019,13 @@ namespace if (lhs.parent() == rhs.parent()) { // determine sibling order - for (xml_attribute a = lhs.attribute(); a; a = a.next_attribute()) - if (a == rhs.attribute()) - return true; - - return false; + for (xml_attribute a = lhs.attribute(); a; a = a.next_attribute()) + if (a == rhs.attribute()) + return true; + + return false; } - + // compare attribute parents ln = lhs.parent(); rn = rhs.parent(); @@ -5564,23 +8034,22 @@ namespace { // attributes go after the parent element if (lhs.parent() == rhs.node()) return false; - + ln = lhs.parent(); } else if (rhs.attribute()) { // attributes go after the parent element if (rhs.parent() == lhs.node()) return true; - + rn = rhs.parent(); } if (ln == rn) return false; - - unsigned int lh = node_height(ln); - unsigned int rh = node_height(rn); - - return node_is_before(ln, lh, rn, rh); + + if (!ln || !rn) return ln < rn; + + return node_is_before(ln.internal_object(), rn.internal_object()); } }; @@ -5592,23 +8061,25 @@ namespace else return rhs.attribute() ? false : lhs.node() < rhs.node(); } }; - - double gen_nan() + + PUGI__FN double gen_nan() { #if defined(__STDC_IEC_559__) || ((FLT_RADIX - 0 == 2) && (FLT_MAX_EXP - 0 == 128) && (FLT_MANT_DIG - 0 == 24)) - union { float f; int32_t i; } u[sizeof(float) == sizeof(int32_t) ? 1 : -1]; - u[0].i = 0x7fc00000; - return u[0].f; + PUGI__STATIC_ASSERT(sizeof(float) == sizeof(uint32_t)); + typedef uint32_t UI; // BCC5 workaround + union { float f; UI i; } u; + u.i = 0x7fc00000; + return u.f; #else // fallback const volatile double zero = 0.0; return zero / zero; #endif } - - bool is_nan(double value) + + PUGI__FN bool is_nan(double value) { - #if defined(_MSC_VER) || defined(__BORLANDC__) + #if defined(PUGI__MSVC_CRT_VERSION) || defined(__BORLANDC__) return !!_isnan(value); #elif defined(fpclassify) && defined(FP_NAN) return fpclassify(value) == FP_NAN; @@ -5618,13 +8089,13 @@ namespace return v != v; #endif } - - const char_t* convert_number_to_string_special(double value) + + PUGI__FN const char_t* convert_number_to_string_special(double value) { - #if defined(_MSC_VER) || defined(__BORLANDC__) + #if defined(PUGI__MSVC_CRT_VERSION) || defined(__BORLANDC__) if (_finite(value)) return (value == 0) ? PUGIXML_TEXT("0") : 0; if (_isnan(value)) return PUGIXML_TEXT("NaN"); - return PUGIXML_TEXT("-Infinity") + (value > 0); + return value > 0 ? PUGIXML_TEXT("Infinity") : PUGIXML_TEXT("-Infinity"); #elif defined(fpclassify) && defined(FP_NAN) && defined(FP_INFINITE) && defined(FP_ZERO) switch (fpclassify(value)) { @@ -5632,7 +8103,7 @@ namespace return PUGIXML_TEXT("NaN"); case FP_INFINITE: - return PUGIXML_TEXT("-Infinity") + (value > 0); + return value > 0 ? PUGIXML_TEXT("Infinity") : PUGIXML_TEXT("-Infinity"); case FP_ZERO: return PUGIXML_TEXT("0"); @@ -5646,17 +8117,17 @@ namespace if (v == 0) return PUGIXML_TEXT("0"); if (v != v) return PUGIXML_TEXT("NaN"); - if (v * 2 == v) return PUGIXML_TEXT("-Infinity") + (value > 0); + if (v * 2 == v) return value > 0 ? PUGIXML_TEXT("Infinity") : PUGIXML_TEXT("-Infinity"); return 0; #endif } - - bool convert_number_to_boolean(double value) + + PUGI__FN bool convert_number_to_boolean(double value) { return (value != 0 && !is_nan(value)); } - - void truncate_zeros(char* begin, char* end) + + PUGI__FN void truncate_zeros(char* begin, char* end) { while (begin != end && end[-1] == '0') end--; @@ -5664,12 +8135,12 @@ namespace } // gets mantissa digits in the form of 0.xxxxx with 0. implied and the exponent -#if defined(_MSC_VER) && _MSC_VER >= 1400 - void convert_number_to_mantissa_exponent(double value, char* buffer, size_t buffer_size, char** out_mantissa, int* out_exponent) +#if defined(PUGI__MSVC_CRT_VERSION) && PUGI__MSVC_CRT_VERSION >= 1400 && !defined(_WIN32_WCE) + PUGI__FN void convert_number_to_mantissa_exponent(double value, char (&buffer)[32], char** out_mantissa, int* out_exponent) { // get base values int sign, exponent; - _ecvt_s(buffer, buffer_size, value, DBL_DIG + 1, &exponent, &sign); + _ecvt_s(buffer, sizeof(buffer), value, DBL_DIG + 1, &exponent, &sign); // truncate redundant zeros truncate_zeros(buffer, buffer + strlen(buffer)); @@ -5679,12 +8150,10 @@ namespace *out_exponent = exponent; } #else - void convert_number_to_mantissa_exponent(double value, char* buffer, size_t buffer_size, char** out_mantissa, int* out_exponent) + PUGI__FN void convert_number_to_mantissa_exponent(double value, char (&buffer)[32], char** out_mantissa, int* out_exponent) { // get a scientific notation value with IEEE DBL_DIG decimals - sprintf(buffer, "%.*e", DBL_DIG, value); - assert(strlen(buffer) < buffer_size); - (void)!buffer_size; + PUGI__SNPRINTF(buffer, "%.*e", DBL_DIG, value); // get the exponent (possibly negative) char* exponent_string = strchr(buffer, 'e'); @@ -5710,21 +8179,25 @@ namespace } #endif - xpath_string convert_number_to_string(double value, xpath_allocator* alloc) + PUGI__FN xpath_string convert_number_to_string(double value, xpath_allocator* alloc) { // try special number conversion const char_t* special = convert_number_to_string_special(value); - if (special) return xpath_string_const(special); + if (special) return xpath_string::from_const(special); // get mantissa + exponent form - char mantissa_buffer[64]; + char mantissa_buffer[32]; char* mantissa; int exponent; - convert_number_to_mantissa_exponent(value, mantissa_buffer, sizeof(mantissa_buffer), &mantissa, &exponent); + convert_number_to_mantissa_exponent(value, mantissa_buffer, &mantissa, &exponent); + + // allocate a buffer of suitable length for the number + size_t result_size = strlen(mantissa_buffer) + (exponent > 0 ? exponent : -exponent) + 4; + char_t* result = static_cast(alloc->allocate(sizeof(char_t) * result_size)); + if (!result) return xpath_string(); // make the number! - char_t result[512]; char_t* s = result; // sign @@ -5739,7 +8212,7 @@ namespace { while (exponent > 0) { - assert(*mantissa == 0 || (unsigned)(*mantissa - '0') <= 9); + assert(*mantissa == 0 || static_cast(*mantissa - '0') <= 9); *s++ = *mantissa ? *mantissa++ : '0'; exponent--; } @@ -5761,22 +8234,22 @@ namespace // extra mantissa digits while (*mantissa) { - assert((unsigned)(*mantissa - '0') <= 9); + assert(static_cast(*mantissa - '0') <= 9); *s++ = *mantissa++; } } // zero-terminate - assert(s < result + sizeof(result) / sizeof(result[0])); + assert(s < result + result_size); *s = 0; - return xpath_string(result, alloc); + return xpath_string::from_heap_preallocated(result, s); } - - bool check_string_to_number_format(const char_t* string) + + PUGI__FN bool check_string_to_number_format(const char_t* string) { // parse leading whitespace - while (IS_CHARTYPE(*string, ct_space)) ++string; + while (PUGI__IS_CHARTYPE(*string, ct_space)) ++string; // parse sign if (*string == '-') ++string; @@ -5784,26 +8257,26 @@ namespace if (!*string) return false; // if there is no integer part, there should be a decimal part with at least one digit - if (!IS_CHARTYPEX(string[0], ctx_digit) && (string[0] != '.' || !IS_CHARTYPEX(string[1], ctx_digit))) return false; + if (!PUGI__IS_CHARTYPEX(string[0], ctx_digit) && (string[0] != '.' || !PUGI__IS_CHARTYPEX(string[1], ctx_digit))) return false; // parse integer part - while (IS_CHARTYPEX(*string, ctx_digit)) ++string; + while (PUGI__IS_CHARTYPEX(*string, ctx_digit)) ++string; // parse decimal part if (*string == '.') { ++string; - while (IS_CHARTYPEX(*string, ctx_digit)) ++string; + while (PUGI__IS_CHARTYPEX(*string, ctx_digit)) ++string; } // parse trailing whitespace - while (IS_CHARTYPE(*string, ct_space)) ++string; + while (PUGI__IS_CHARTYPE(*string, ct_space)) ++string; return *string == 0; } - double convert_string_to_number(const char_t* string) + PUGI__FN double convert_string_to_number(const char_t* string) { // check string format if (!check_string_to_number_format(string)) return gen_nan(); @@ -5812,21 +8285,19 @@ namespace #ifdef PUGIXML_WCHAR_MODE return wcstod(string, 0); #else - return atof(string); + return strtod(string, 0); #endif } - bool convert_string_to_number(const char_t* begin, const char_t* end, double* out_result) + PUGI__FN bool convert_string_to_number_scratch(char_t (&buffer)[32], const char_t* begin, const char_t* end, double* out_result) { - char_t buffer[32]; - size_t length = static_cast(end - begin); char_t* scratch = buffer; if (length >= sizeof(buffer) / sizeof(buffer[0])) { // need to make dummy on-heap copy - scratch = static_cast(global_allocate((length + 1) * sizeof(char_t))); + scratch = static_cast(xml_memory::allocate((length + 1) * sizeof(char_t))); if (!scratch) return false; } @@ -5837,33 +8308,33 @@ namespace *out_result = convert_string_to_number(scratch); // free dummy buffer - if (scratch != buffer) global_deallocate(scratch); + if (scratch != buffer) xml_memory::deallocate(scratch); return true; } - - double round_nearest(double value) + + PUGI__FN double round_nearest(double value) { return floor(value + 0.5); } - double round_nearest_nzero(double value) + PUGI__FN double round_nearest_nzero(double value) { // same as round_nearest, but returns -0 for [-0.5, -0] // ceil is used to differentiate between +0 and -0 (we return -0 for [-0.5, -0] and +0 for +0) return (value >= -0.5 && value <= 0) ? ceil(value) : floor(value + 0.5); } - - const char_t* qualified_name(const xpath_node& node) + + PUGI__FN const char_t* qualified_name(const xpath_node& node) { return node.attribute() ? node.attribute().name() : node.node().name(); } - - const char_t* local_name(const xpath_node& node) + + PUGI__FN const char_t* local_name(const xpath_node& node) { const char_t* name = qualified_name(node); const char_t* p = find_char(name, ':'); - + return p ? p + 1 : name; } @@ -5880,7 +8351,7 @@ namespace prefix_length = pos ? static_cast(pos - name) : 0; } - bool operator()(const xml_attribute& a) const + bool operator()(xml_attribute a) const { const char_t* name = a.name(); @@ -5890,51 +8361,51 @@ namespace } }; - const char_t* namespace_uri(const xml_node& node) + PUGI__FN const char_t* namespace_uri(xml_node node) { namespace_uri_predicate pred = node.name(); - + xml_node p = node; - + while (p) { xml_attribute a = p.find_attribute(pred); - + if (a) return a.value(); - + p = p.parent(); } - + return PUGIXML_TEXT(""); } - const char_t* namespace_uri(const xml_attribute& attr, const xml_node& parent) + PUGI__FN const char_t* namespace_uri(xml_attribute attr, xml_node parent) { namespace_uri_predicate pred = attr.name(); - + // Default namespace does not apply to attributes if (!pred.prefix) return PUGIXML_TEXT(""); - + xml_node p = parent; - + while (p) { xml_attribute a = p.find_attribute(pred); - + if (a) return a.value(); - + p = p.parent(); } - + return PUGIXML_TEXT(""); } - const char_t* namespace_uri(const xpath_node& node) + PUGI__FN const char_t* namespace_uri(const xpath_node& node) { return node.attribute() ? namespace_uri(node.attribute(), node.parent()) : namespace_uri(node.node()); } - void normalize_space(char_t* buffer) + PUGI__FN char_t* normalize_space(char_t* buffer) { char_t* write = buffer; @@ -5942,10 +8413,10 @@ namespace { char_t ch = *it++; - if (IS_CHARTYPE(ch, ct_space)) + if (PUGI__IS_CHARTYPE(ch, ct_space)) { // replace whitespace sequence with single space - while (IS_CHARTYPE(*it, ct_space)) it++; + while (PUGI__IS_CHARTYPE(*it, ct_space)) it++; // avoid leading spaces if (write != buffer) *write++ = ' '; @@ -5954,21 +8425,21 @@ namespace } // remove trailing space - if (write != buffer && IS_CHARTYPE(write[-1], ct_space)) write--; + if (write != buffer && PUGI__IS_CHARTYPE(write[-1], ct_space)) write--; // zero-terminate *write = 0; + + return write; } - void translate(char_t* buffer, const char_t* from, const char_t* to) + PUGI__FN char_t* translate(char_t* buffer, const char_t* from, const char_t* to, size_t to_length) { - size_t to_length = strlength(to); - char_t* write = buffer; while (*buffer) { - DMC_VOLATILE char_t ch = *buffer++; + PUGI__DMC_VOLATILE char_t ch = *buffer++; const char_t* pos = find_char(from, ch); @@ -5980,11 +8451,80 @@ namespace // zero-terminate *write = 0; + + return write; + } + + PUGI__FN unsigned char* translate_table_generate(xpath_allocator* alloc, const char_t* from, const char_t* to) + { + unsigned char table[128] = {0}; + + while (*from) + { + unsigned int fc = static_cast(*from); + unsigned int tc = static_cast(*to); + + if (fc >= 128 || tc >= 128) + return 0; + + // code=128 means "skip character" + if (!table[fc]) + table[fc] = static_cast(tc ? tc : 128); + + from++; + if (tc) to++; + } + + for (int i = 0; i < 128; ++i) + if (!table[i]) + table[i] = static_cast(i); + + void* result = alloc->allocate(sizeof(table)); + if (!result) return 0; + + memcpy(result, table, sizeof(table)); + + return static_cast(result); + } + + PUGI__FN char_t* translate_table(char_t* buffer, const unsigned char* table) + { + char_t* write = buffer; + + while (*buffer) + { + char_t ch = *buffer++; + unsigned int index = static_cast(ch); + + if (index < 128) + { + unsigned char code = table[index]; + + // code=128 means "skip character" (table size is 128 so 128 can be a special value) + // this code skips these characters without extra branches + *write = static_cast(code); + write += 1 - (code >> 7); + } + else + { + *write++ = ch; + } + } + + // zero-terminate + *write = 0; + + return write; + } + + inline bool is_xpath_attribute(const char_t* name) + { + return !(starts_with(name, PUGIXML_TEXT("xmlns")) && (name[5] == 0 || name[5] == ':')); } struct xpath_variable_boolean: xpath_variable { - xpath_variable_boolean(): value(false) + xpath_variable_boolean(): xpath_variable(xpath_type_boolean), value(false) { } @@ -5994,7 +8534,7 @@ namespace struct xpath_variable_number: xpath_variable { - xpath_variable_number(): value(0) + xpath_variable_number(): xpath_variable(xpath_type_number), value(0) { } @@ -6004,13 +8544,13 @@ namespace struct xpath_variable_string: xpath_variable { - xpath_variable_string(): value(0) + xpath_variable_string(): xpath_variable(xpath_type_string), value(0) { } ~xpath_variable_string() { - if (value) global_deallocate(value); + if (value) xml_memory::deallocate(value); } char_t* value; @@ -6019,13 +8559,17 @@ namespace struct xpath_variable_node_set: xpath_variable { + xpath_variable_node_set(): xpath_variable(xpath_type_node_set) + { + } + xpath_node_set value; char_t name[1]; }; - const xpath_node_set dummy_node_set; + static const xpath_node_set dummy_node_set; - unsigned int hash_string(const char_t* str) + PUGI__FN PUGI__UNSIGNED_OVERFLOW unsigned int hash_string(const char_t* str) { // Jenkins one-at-a-time hash (http://en.wikipedia.org/wiki/Jenkins_hash_function#one-at-a-time) unsigned int result = 0; @@ -6036,21 +8580,21 @@ namespace result += result << 10; result ^= result >> 6; } - + result += result << 3; result ^= result >> 11; result += result << 15; - + return result; } - template T* new_xpath_variable(const char_t* name) + template PUGI__FN T* new_xpath_variable(const char_t* name) { size_t length = strlength(name); if (length == 0) return 0; // empty variable names are invalid // $$ we can't use offsetof(T, name) because T is non-POD, so we just allocate additional length characters - void* memory = global_allocate(sizeof(T) + length * sizeof(char_t)); + void* memory = xml_memory::allocate(sizeof(T) + length * sizeof(char_t)); if (!memory) return 0; T* result = new (memory) T(); @@ -6060,7 +8604,7 @@ namespace return result; } - xpath_variable* new_xpath_variable(xpath_value_type type, const char_t* name) + PUGI__FN xpath_variable* new_xpath_variable(xpath_value_type type, const char_t* name) { switch (type) { @@ -6081,13 +8625,13 @@ namespace } } - template void delete_xpath_variable(T* var) + template PUGI__FN void delete_xpath_variable(T* var) { var->~T(); - global_deallocate(var); + xml_memory::deallocate(var); } - void delete_xpath_variable(xpath_value_type type, xpath_variable* var) + PUGI__FN void delete_xpath_variable(xpath_value_type type, xpath_variable* var) { switch (type) { @@ -6108,57 +8652,99 @@ namespace break; default: - assert(!"Invalid variable type"); + assert(false && "Invalid variable type"); // unreachable } } - xpath_variable* get_variable(xpath_variable_set* set, const char_t* begin, const char_t* end) + PUGI__FN bool copy_xpath_variable(xpath_variable* lhs, const xpath_variable* rhs) { - char_t buffer[32]; + switch (rhs->type()) + { + case xpath_type_node_set: + return lhs->set(static_cast(rhs)->value); + + case xpath_type_number: + return lhs->set(static_cast(rhs)->value); + + case xpath_type_string: + return lhs->set(static_cast(rhs)->value); + + case xpath_type_boolean: + return lhs->set(static_cast(rhs)->value); + + default: + assert(false && "Invalid variable type"); // unreachable + return false; + } + } + PUGI__FN bool get_variable_scratch(char_t (&buffer)[32], xpath_variable_set* set, const char_t* begin, const char_t* end, xpath_variable** out_result) + { size_t length = static_cast(end - begin); char_t* scratch = buffer; if (length >= sizeof(buffer) / sizeof(buffer[0])) { // need to make dummy on-heap copy - scratch = static_cast(global_allocate((length + 1) * sizeof(char_t))); - if (!scratch) return 0; + scratch = static_cast(xml_memory::allocate((length + 1) * sizeof(char_t))); + if (!scratch) return false; } // copy string to zero-terminated buffer and perform lookup memcpy(scratch, begin, length * sizeof(char_t)); scratch[length] = 0; - xpath_variable* result = set->get(scratch); + *out_result = set->get(scratch); // free dummy buffer - if (scratch != buffer) global_deallocate(scratch); + if (scratch != buffer) xml_memory::deallocate(scratch); - return result; + return true; } -} +PUGI__NS_END // Internal node set class -namespace -{ - xpath_node_set::type_t xpath_sort(xpath_node* begin, xpath_node* end, xpath_node_set::type_t type, bool rev) +PUGI__NS_BEGIN + PUGI__FN xpath_node_set::type_t xpath_get_order(const xpath_node* begin, const xpath_node* end) + { + if (end - begin < 2) + return xpath_node_set::type_sorted; + + document_order_comparator cmp; + + bool first = cmp(begin[0], begin[1]); + + for (const xpath_node* it = begin + 1; it + 1 < end; ++it) + if (cmp(it[0], it[1]) != first) + return xpath_node_set::type_unsorted; + + return first ? xpath_node_set::type_sorted : xpath_node_set::type_sorted_reverse; + } + + PUGI__FN xpath_node_set::type_t xpath_sort(xpath_node* begin, xpath_node* end, xpath_node_set::type_t type, bool rev) { xpath_node_set::type_t order = rev ? xpath_node_set::type_sorted_reverse : xpath_node_set::type_sorted; if (type == xpath_node_set::type_unsorted) { - sort(begin, end, document_order_comparator()); + xpath_node_set::type_t sorted = xpath_get_order(begin, end); + + if (sorted == xpath_node_set::type_unsorted) + { + sort(begin, end, document_order_comparator()); - type = xpath_node_set::type_sorted; + type = xpath_node_set::type_sorted; + } + else + type = sorted; } - + if (type != order) reverse(begin, end); - + return order; } - xpath_node xpath_first(const xpath_node* begin, const xpath_node* end, xpath_node_set::type_t type) + PUGI__FN xpath_node xpath_first(const xpath_node* begin, const xpath_node* end, xpath_node_set::type_t type) { if (begin == end) return xpath_node(); @@ -6174,10 +8760,11 @@ namespace return *min_element(begin, end, document_order_comparator()); default: - assert(!"Invalid node set type"); + assert(false && "Invalid node set type"); // unreachable return xpath_node(); } } + class xpath_node_set_raw { xpath_node_set::type_t _type; @@ -6216,47 +8803,37 @@ namespace return xpath_first(_begin, _end, _type); } + void push_back_grow(const xpath_node& node, xpath_allocator* alloc); + void push_back(const xpath_node& node, xpath_allocator* alloc) { - if (_end == _eos) - { - size_t capacity = static_cast(_eos - _begin); - - // get new capacity (1.5x rule) - size_t new_capacity = capacity + capacity / 2 + 1; - - // reallocate the old array or allocate a new one - xpath_node* data = static_cast(alloc->reallocate(_begin, capacity * sizeof(xpath_node), new_capacity * sizeof(xpath_node))); - assert(data); - - // finalize - _begin = data; - _end = data + capacity; - _eos = data + new_capacity; - } - - *_end++ = node; + if (_end != _eos) + *_end++ = node; + else + push_back_grow(node, alloc); } - void append(const xpath_node* begin, const xpath_node* end, xpath_allocator* alloc) + void append(const xpath_node* begin_, const xpath_node* end_, xpath_allocator* alloc) { - size_t size = static_cast(_end - _begin); + if (begin_ == end_) return; + + size_t size_ = static_cast(_end - _begin); size_t capacity = static_cast(_eos - _begin); - size_t count = static_cast(end - begin); + size_t count = static_cast(end_ - begin_); - if (size + count > capacity) + if (size_ + count > capacity) { // reallocate the old array or allocate a new one - xpath_node* data = static_cast(alloc->reallocate(_begin, capacity * sizeof(xpath_node), (size + count) * sizeof(xpath_node))); - assert(data); + xpath_node* data = static_cast(alloc->reallocate(_begin, capacity * sizeof(xpath_node), (size_ + count) * sizeof(xpath_node))); + if (!data) return; // finalize _begin = data; - _end = data + size; - _eos = data + size + count; + _end = data + size_; + _eos = data + size_ + count; } - memcpy(_end, begin, count * sizeof(xpath_node)); + memcpy(_end, begin_, count * sizeof(xpath_node)); _end += count; } @@ -6276,7 +8853,7 @@ namespace { if (_type == xpath_node_set::type_unsorted) sort(_begin, _end, duplicate_comparator()); - + _end = unique(_begin, _end); } @@ -6285,21 +8862,40 @@ namespace return _type; } - void set_type(xpath_node_set::type_t type) + void set_type(xpath_node_set::type_t value) { - _type = type; + _type = value; } }; -} -namespace -{ + PUGI__FN_NO_INLINE void xpath_node_set_raw::push_back_grow(const xpath_node& node, xpath_allocator* alloc) + { + size_t capacity = static_cast(_eos - _begin); + + // get new capacity (1.5x rule) + size_t new_capacity = capacity + capacity / 2 + 1; + + // reallocate the old array or allocate a new one + xpath_node* data = static_cast(alloc->reallocate(_begin, capacity * sizeof(xpath_node), new_capacity * sizeof(xpath_node))); + if (!data) return; + + // finalize + _begin = data; + _end = data + capacity; + _eos = data + new_capacity; + + // push + *_end++ = node; + } +PUGI__NS_END + +PUGI__NS_BEGIN struct xpath_context { xpath_node n; size_t position, size; - xpath_context(const xpath_node& n, size_t position, size_t size): n(n), position(position), size(size) + xpath_context(const xpath_node& n_, size_t position_, size_t size_): n(n_), position(position_), size(size_) { } }; @@ -6365,17 +8961,17 @@ namespace { next(); } - + const char_t* state() const { return _cur; } - + void next() { const char_t* cur = _cur; - while (IS_CHARTYPE(*cur, ct_space)) ++cur; + while (PUGI__IS_CHARTYPE(*cur, ct_space)) ++cur; // save lexeme position for error reporting _cur_lexeme_pos = cur; @@ -6385,7 +8981,7 @@ namespace case 0: _cur_lexeme = lex_eof; break; - + case '>': if (*(cur+1) == '=') { @@ -6429,7 +9025,7 @@ namespace _cur_lexeme = lex_equal; break; - + case '+': cur += 1; _cur_lexeme = lex_plus; @@ -6453,25 +9049,25 @@ namespace _cur_lexeme = lex_union; break; - + case '$': cur += 1; - if (IS_CHARTYPEX(*cur, ctx_start_symbol)) + if (PUGI__IS_CHARTYPEX(*cur, ctx_start_symbol)) { _cur_lexeme_contents.begin = cur; - while (IS_CHARTYPEX(*cur, ctx_symbol)) cur++; + while (PUGI__IS_CHARTYPEX(*cur, ctx_symbol)) cur++; - if (cur[0] == ':' && IS_CHARTYPEX(cur[1], ctx_symbol)) // qname + if (cur[0] == ':' && PUGI__IS_CHARTYPEX(cur[1], ctx_symbol)) // qname { cur++; // : - while (IS_CHARTYPEX(*cur, ctx_symbol)) cur++; + while (PUGI__IS_CHARTYPEX(*cur, ctx_symbol)) cur++; } _cur_lexeme_contents.end = cur; - + _cur_lexeme = lex_var_ref; } else @@ -6492,7 +9088,7 @@ namespace _cur_lexeme = lex_close_brace; break; - + case '[': cur += 1; _cur_lexeme = lex_open_square_brace; @@ -6523,23 +9119,23 @@ namespace _cur_lexeme = lex_slash; } break; - + case '.': if (*(cur+1) == '.') { cur += 2; _cur_lexeme = lex_double_dot; } - else if (IS_CHARTYPEX(*(cur+1), ctx_digit)) + else if (PUGI__IS_CHARTYPEX(*(cur+1), ctx_digit)) { _cur_lexeme_contents.begin = cur; // . ++cur; - while (IS_CHARTYPEX(*cur, ctx_digit)) cur++; + while (PUGI__IS_CHARTYPEX(*cur, ctx_digit)) cur++; _cur_lexeme_contents.end = cur; - + _cur_lexeme = lex_number; } else @@ -6565,7 +9161,7 @@ namespace _cur_lexeme_contents.begin = cur; while (*cur && *cur != terminator) cur++; _cur_lexeme_contents.end = cur; - + if (!*cur) _cur_lexeme = lex_none; else @@ -6590,28 +9186,28 @@ namespace break; default: - if (IS_CHARTYPEX(*cur, ctx_digit)) + if (PUGI__IS_CHARTYPEX(*cur, ctx_digit)) { _cur_lexeme_contents.begin = cur; - while (IS_CHARTYPEX(*cur, ctx_digit)) cur++; - + while (PUGI__IS_CHARTYPEX(*cur, ctx_digit)) cur++; + if (*cur == '.') { cur++; - while (IS_CHARTYPEX(*cur, ctx_digit)) cur++; + while (PUGI__IS_CHARTYPEX(*cur, ctx_digit)) cur++; } _cur_lexeme_contents.end = cur; _cur_lexeme = lex_number; } - else if (IS_CHARTYPEX(*cur, ctx_start_symbol)) + else if (PUGI__IS_CHARTYPEX(*cur, ctx_start_symbol)) { _cur_lexeme_contents.begin = cur; - while (IS_CHARTYPEX(*cur, ctx_symbol)) cur++; + while (PUGI__IS_CHARTYPEX(*cur, ctx_symbol)) cur++; if (cur[0] == ':') { @@ -6619,16 +9215,16 @@ namespace { cur += 2; // :* } - else if (IS_CHARTYPEX(cur[1], ctx_symbol)) // namespace test qname + else if (PUGI__IS_CHARTYPEX(cur[1], ctx_symbol)) // namespace test qname { cur++; // : - while (IS_CHARTYPEX(*cur, ctx_symbol)) cur++; + while (PUGI__IS_CHARTYPEX(*cur, ctx_symbol)) cur++; } } _cur_lexeme_contents.end = cur; - + _cur_lexeme = lex_string; } else @@ -6660,10 +9256,11 @@ namespace enum ast_type_t { + ast_unknown, ast_op_or, // left or right ast_op_and, // left and right ast_op_equal, // left = right - ast_op_not_equal, // left != right + ast_op_not_equal, // left != right ast_op_less, // left < right ast_op_greater, // left > right ast_op_less_or_equal, // left <= right @@ -6677,7 +9274,6 @@ namespace ast_op_union, // left | right ast_predicate, // apply predicate to set; next points to next predicate ast_filter, // select * from left where right - ast_filter_posinv, // select * from left where right; proximity position invariant ast_string_constant, // string constant ast_number_constant, // number constant ast_variable, // variable @@ -6717,7 +9313,10 @@ namespace ast_func_ceiling, // ceiling(left) ast_func_round, // round(left) ast_step, // process set left with step - ast_step_root // select root node + ast_step_root, // select root node + + ast_opt_translate_table, // translate(left, right, third) where right/third are constants + ast_opt_compare_attribute // @name = 'string' }; enum axis_t @@ -6736,7 +9335,7 @@ namespace axis_preceding_sibling, axis_self }; - + enum nodetest_t { nodetest_none, @@ -6750,13 +9349,28 @@ namespace nodetest_all_in_namespace }; + enum predicate_t + { + predicate_default, + predicate_posinv, + predicate_constant, + predicate_constant_one + }; + + enum nodeset_eval_t + { + nodeset_eval_all, + nodeset_eval_any, + nodeset_eval_first + }; + template struct axis_to_type { static const axis_t axis; }; template const axis_t axis_to_type::axis = N; - + class xpath_ast_node { private: @@ -6764,8 +9378,10 @@ namespace char _type; char _rettype; - // for ast_step / ast_predicate + // for ast_step char _axis; + + // for ast_step/ast_predicate/ast_filter char _test; // tree node structure @@ -6783,6 +9399,8 @@ namespace xpath_variable* variable; // node test for ast_step (node name/namespace/node type/pi target) const char_t* nodetest; + // table for ast_opt_translate_table + const unsigned char* table; } _data; xpath_ast_node(const xpath_ast_node&); @@ -6812,8 +9430,8 @@ namespace { xpath_allocator_capture cr(stack.result); - xpath_node_set_raw ls = lhs->eval_node_set(c, stack); - xpath_node_set_raw rs = rhs->eval_node_set(c, stack); + xpath_node_set_raw ls = lhs->eval_node_set(c, stack, nodeset_eval_all); + xpath_node_set_raw rs = rhs->eval_node_set(c, stack, nodeset_eval_all); for (const xpath_node* li = ls.begin(); li != ls.end(); ++li) for (const xpath_node* ri = rs.begin(); ri != rs.end(); ++ri) @@ -6841,7 +9459,7 @@ namespace xpath_allocator_capture cr(stack.result); double l = lhs->eval_number(c, stack); - xpath_node_set_raw rs = rhs->eval_node_set(c, stack); + xpath_node_set_raw rs = rhs->eval_node_set(c, stack, nodeset_eval_all); for (const xpath_node* ri = rs.begin(); ri != rs.end(); ++ri) { @@ -6858,7 +9476,7 @@ namespace xpath_allocator_capture cr(stack.result); xpath_string l = lhs->eval_string(c, stack); - xpath_node_set_raw rs = rhs->eval_node_set(c, stack); + xpath_node_set_raw rs = rhs->eval_node_set(c, stack, nodeset_eval_all); for (const xpath_node* ri = rs.begin(); ri != rs.end(); ++ri) { @@ -6872,10 +9490,15 @@ namespace } } - assert(!"Wrong types"); + assert(false && "Wrong types"); // unreachable return false; } + static bool eval_once(xpath_node_set::type_t type, nodeset_eval_t eval) + { + return type == xpath_node_set::type_sorted ? eval != nodeset_eval_all : eval == nodeset_eval_any; + } + template static bool compare_rel(xpath_ast_node* lhs, xpath_ast_node* rhs, const xpath_context& c, const xpath_stack& stack, const Comp& comp) { xpath_value_type lt = lhs->rettype(), rt = rhs->rettype(); @@ -6886,8 +9509,8 @@ namespace { xpath_allocator_capture cr(stack.result); - xpath_node_set_raw ls = lhs->eval_node_set(c, stack); - xpath_node_set_raw rs = rhs->eval_node_set(c, stack); + xpath_node_set_raw ls = lhs->eval_node_set(c, stack, nodeset_eval_all); + xpath_node_set_raw rs = rhs->eval_node_set(c, stack, nodeset_eval_all); for (const xpath_node* li = ls.begin(); li != ls.end(); ++li) { @@ -6911,7 +9534,7 @@ namespace xpath_allocator_capture cr(stack.result); double l = lhs->eval_number(c, stack); - xpath_node_set_raw rs = rhs->eval_node_set(c, stack); + xpath_node_set_raw rs = rhs->eval_node_set(c, stack, nodeset_eval_all); for (const xpath_node* ri = rs.begin(); ri != rs.end(); ++ri) { @@ -6927,7 +9550,7 @@ namespace { xpath_allocator_capture cr(stack.result); - xpath_node_set_raw ls = lhs->eval_node_set(c, stack); + xpath_node_set_raw ls = lhs->eval_node_set(c, stack, nodeset_eval_all); double r = rhs->eval_number(c, stack); for (const xpath_node* li = ls.begin(); li != ls.end(); ++li) @@ -6942,128 +9565,231 @@ namespace } else { - assert(!"Wrong types"); + assert(false && "Wrong types"); // unreachable return false; } } - void apply_predicate(xpath_node_set_raw& ns, size_t first, xpath_ast_node* expr, const xpath_stack& stack) + static void apply_predicate_boolean(xpath_node_set_raw& ns, size_t first, xpath_ast_node* expr, const xpath_stack& stack, bool once) { assert(ns.size() >= first); + assert(expr->rettype() != xpath_type_number); size_t i = 1; size_t size = ns.size() - first; - + xpath_node* last = ns.begin() + first; - + // remove_if... or well, sort of for (xpath_node* it = last; it != ns.end(); ++it, ++i) { xpath_context c(*it, i, size); - - if (expr->rettype() == xpath_type_number) + + if (expr->eval_boolean(c, stack)) { - if (expr->eval_number(c, stack) == i) - *last++ = *it; + *last++ = *it; + + if (once) break; } - else if (expr->eval_boolean(c, stack)) + } + + ns.truncate(last); + } + + static void apply_predicate_number(xpath_node_set_raw& ns, size_t first, xpath_ast_node* expr, const xpath_stack& stack, bool once) + { + assert(ns.size() >= first); + assert(expr->rettype() == xpath_type_number); + + size_t i = 1; + size_t size = ns.size() - first; + + xpath_node* last = ns.begin() + first; + + // remove_if... or well, sort of + for (xpath_node* it = last; it != ns.end(); ++it, ++i) + { + xpath_context c(*it, i, size); + + if (expr->eval_number(c, stack) == i) + { *last++ = *it; + + if (once) break; + } } - + ns.truncate(last); } - void apply_predicates(xpath_node_set_raw& ns, size_t first, const xpath_stack& stack) + static void apply_predicate_number_const(xpath_node_set_raw& ns, size_t first, xpath_ast_node* expr, const xpath_stack& stack) { - if (ns.size() == first) return; - - for (xpath_ast_node* pred = _right; pred; pred = pred->_next) + assert(ns.size() >= first); + assert(expr->rettype() == xpath_type_number); + + size_t size = ns.size() - first; + + xpath_node* last = ns.begin() + first; + + xpath_context c(xpath_node(), 1, size); + + double er = expr->eval_number(c, stack); + + if (er >= 1.0 && er <= size) { - apply_predicate(ns, first, pred->_left, stack); + size_t eri = static_cast(er); + + if (er == eri) + { + xpath_node r = last[eri - 1]; + + *last++ = r; + } } + + ns.truncate(last); + } + + void apply_predicate(xpath_node_set_raw& ns, size_t first, const xpath_stack& stack, bool once) + { + if (ns.size() == first) return; + + assert(_type == ast_filter || _type == ast_predicate); + + if (_test == predicate_constant || _test == predicate_constant_one) + apply_predicate_number_const(ns, first, _right, stack); + else if (_right->rettype() == xpath_type_number) + apply_predicate_number(ns, first, _right, stack, once); + else + apply_predicate_boolean(ns, first, _right, stack, once); } - void step_push(xpath_node_set_raw& ns, const xml_attribute& a, const xml_node& parent, xpath_allocator* alloc) + void apply_predicates(xpath_node_set_raw& ns, size_t first, const xpath_stack& stack, nodeset_eval_t eval) { - if (!a) return; + if (ns.size() == first) return; - const char_t* name = a.name(); + bool last_once = eval_once(ns.type(), eval); + + for (xpath_ast_node* pred = _right; pred; pred = pred->_next) + pred->apply_predicate(ns, first, stack, !pred->_next && last_once); + } + + bool step_push(xpath_node_set_raw& ns, xml_attribute_struct* a, xml_node_struct* parent, xpath_allocator* alloc) + { + assert(a); + + const char_t* name = a->name ? a->name + 0 : PUGIXML_TEXT(""); - // There are no attribute nodes corresponding to attributes that declare namespaces - // That is, "xmlns:..." or "xmlns" - if (starts_with(name, PUGIXML_TEXT("xmlns")) && (name[5] == 0 || name[5] == ':')) return; - switch (_test) { case nodetest_name: - if (strequal(name, _data.nodetest)) ns.push_back(xpath_node(a, parent), alloc); + if (strequal(name, _data.nodetest) && is_xpath_attribute(name)) + { + ns.push_back(xpath_node(xml_attribute(a), xml_node(parent)), alloc); + return true; + } break; - + case nodetest_type_node: case nodetest_all: - ns.push_back(xpath_node(a, parent), alloc); + if (is_xpath_attribute(name)) + { + ns.push_back(xpath_node(xml_attribute(a), xml_node(parent)), alloc); + return true; + } break; - + case nodetest_all_in_namespace: - if (starts_with(name, _data.nodetest)) - ns.push_back(xpath_node(a, parent), alloc); + if (starts_with(name, _data.nodetest) && is_xpath_attribute(name)) + { + ns.push_back(xpath_node(xml_attribute(a), xml_node(parent)), alloc); + return true; + } break; - + default: ; } + + return false; } - - void step_push(xpath_node_set_raw& ns, const xml_node& n, xpath_allocator* alloc) + + bool step_push(xpath_node_set_raw& ns, xml_node_struct* n, xpath_allocator* alloc) { - if (!n) return; + assert(n); + + xml_node_type type = PUGI__NODETYPE(n); switch (_test) { case nodetest_name: - if (n.type() == node_element && strequal(n.name(), _data.nodetest)) ns.push_back(n, alloc); + if (type == node_element && n->name && strequal(n->name, _data.nodetest)) + { + ns.push_back(xml_node(n), alloc); + return true; + } break; - + case nodetest_type_node: - ns.push_back(n, alloc); - break; - + ns.push_back(xml_node(n), alloc); + return true; + case nodetest_type_comment: - if (n.type() == node_comment) - ns.push_back(n, alloc); + if (type == node_comment) + { + ns.push_back(xml_node(n), alloc); + return true; + } break; - + case nodetest_type_text: - if (n.type() == node_pcdata || n.type() == node_cdata) - ns.push_back(n, alloc); + if (type == node_pcdata || type == node_cdata) + { + ns.push_back(xml_node(n), alloc); + return true; + } break; - + case nodetest_type_pi: - if (n.type() == node_pi) - ns.push_back(n, alloc); + if (type == node_pi) + { + ns.push_back(xml_node(n), alloc); + return true; + } break; - + case nodetest_pi: - if (n.type() == node_pi && strequal(n.name(), _data.nodetest)) - ns.push_back(n, alloc); + if (type == node_pi && n->name && strequal(n->name, _data.nodetest)) + { + ns.push_back(xml_node(n), alloc); + return true; + } break; - + case nodetest_all: - if (n.type() == node_element) - ns.push_back(n, alloc); + if (type == node_element) + { + ns.push_back(xml_node(n), alloc); + return true; + } break; - + case nodetest_all_in_namespace: - if (n.type() == node_element && starts_with(n.name(), _data.nodetest)) - ns.push_back(n, alloc); + if (type == node_element && n->name && starts_with(n->name, _data.nodetest)) + { + ns.push_back(xml_node(n), alloc); + return true; + } break; default: - assert(!"Unknown axis"); - } + assert(false && "Unknown axis"); // unreachable + } + + return false; } - template void step_fill(xpath_node_set_raw& ns, const xml_node& n, xpath_allocator* alloc, T) + template void step_fill(xpath_node_set_raw& ns, xml_node_struct* n, xpath_allocator* alloc, bool once, T) { const axis_t axis = T::axis; @@ -7071,86 +9797,103 @@ namespace { case axis_attribute: { - for (xml_attribute a = n.first_attribute(); a; a = a.next_attribute()) - step_push(ns, a, n, alloc); - + for (xml_attribute_struct* a = n->first_attribute; a; a = a->next_attribute) + if (step_push(ns, a, n, alloc) & once) + return; + break; } - + case axis_child: { - for (xml_node c = n.first_child(); c; c = c.next_sibling()) - step_push(ns, c, alloc); - + for (xml_node_struct* c = n->first_child; c; c = c->next_sibling) + if (step_push(ns, c, alloc) & once) + return; + break; } - + case axis_descendant: case axis_descendant_or_self: { if (axis == axis_descendant_or_self) - step_push(ns, n, alloc); - - xml_node cur = n.first_child(); - - while (cur && cur != n) + if (step_push(ns, n, alloc) & once) + return; + + xml_node_struct* cur = n->first_child; + + while (cur) { - step_push(ns, cur, alloc); - - if (cur.first_child()) - cur = cur.first_child(); - else if (cur.next_sibling()) - cur = cur.next_sibling(); + if (step_push(ns, cur, alloc) & once) + return; + + if (cur->first_child) + cur = cur->first_child; else { - while (!cur.next_sibling() && cur != n) - cur = cur.parent(); - - if (cur != n) cur = cur.next_sibling(); + while (!cur->next_sibling) + { + cur = cur->parent; + + if (cur == n) return; + } + + cur = cur->next_sibling; } } - + break; } - + case axis_following_sibling: { - for (xml_node c = n.next_sibling(); c; c = c.next_sibling()) - step_push(ns, c, alloc); - + for (xml_node_struct* c = n->next_sibling; c; c = c->next_sibling) + if (step_push(ns, c, alloc) & once) + return; + break; } - + case axis_preceding_sibling: { - for (xml_node c = n.previous_sibling(); c; c = c.previous_sibling()) - step_push(ns, c, alloc); - + for (xml_node_struct* c = n->prev_sibling_c; c->next_sibling; c = c->prev_sibling_c) + if (step_push(ns, c, alloc) & once) + return; + break; } - + case axis_following: { - xml_node cur = n; + xml_node_struct* cur = n; // exit from this node so that we don't include descendants - while (cur && !cur.next_sibling()) cur = cur.parent(); - cur = cur.next_sibling(); + while (!cur->next_sibling) + { + cur = cur->parent; + + if (!cur) return; + } + + cur = cur->next_sibling; - for (;;) + while (cur) { - step_push(ns, cur, alloc); + if (step_push(ns, cur, alloc) & once) + return; - if (cur.first_child()) - cur = cur.first_child(); - else if (cur.next_sibling()) - cur = cur.next_sibling(); + if (cur->first_child) + cur = cur->first_child; else { - while (cur && !cur.next_sibling()) cur = cur.parent(); - cur = cur.next_sibling(); + while (!cur->next_sibling) + { + cur = cur->parent; + + if (!cur) return; + } - if (!cur) break; + cur = cur->next_sibling; } } @@ -7159,58 +9902,63 @@ namespace case axis_preceding: { - xml_node cur = n; + xml_node_struct* cur = n; + + // exit from this node so that we don't include descendants + while (!cur->prev_sibling_c->next_sibling) + { + cur = cur->parent; + + if (!cur) return; + } - while (cur && !cur.previous_sibling()) cur = cur.parent(); - cur = cur.previous_sibling(); + cur = cur->prev_sibling_c; - for (;;) + while (cur) { - if (cur.last_child()) - cur = cur.last_child(); + if (cur->first_child) + cur = cur->first_child->prev_sibling_c; else { // leaf node, can't be ancestor - step_push(ns, cur, alloc); + if (step_push(ns, cur, alloc) & once) + return; - if (cur.previous_sibling()) - cur = cur.previous_sibling(); - else + while (!cur->prev_sibling_c->next_sibling) { - do - { - cur = cur.parent(); - if (!cur) break; + cur = cur->parent; - if (!node_is_ancestor(cur, n)) step_push(ns, cur, alloc); - } - while (!cur.previous_sibling()); - - cur = cur.previous_sibling(); + if (!cur) return; - if (!cur) break; + if (!node_is_ancestor(cur, n)) + if (step_push(ns, cur, alloc) & once) + return; } + + cur = cur->prev_sibling_c; } } break; } - + case axis_ancestor: case axis_ancestor_or_self: { if (axis == axis_ancestor_or_self) - step_push(ns, n, alloc); + if (step_push(ns, n, alloc) & once) + return; + + xml_node_struct* cur = n->parent; - xml_node cur = n.parent(); - while (cur) { - step_push(ns, cur, alloc); - - cur = cur.parent(); + if (step_push(ns, cur, alloc) & once) + return; + + cur = cur->parent; } - + break; } @@ -7223,17 +9971,18 @@ namespace case axis_parent: { - if (n.parent()) step_push(ns, n.parent(), alloc); + if (n->parent) + step_push(ns, n->parent, alloc); break; } - + default: - assert(!"Unimplemented axis"); + assert(false && "Unimplemented axis"); // unreachable } } - - template void step_fill(xpath_node_set_raw& ns, const xml_attribute& a, const xml_node& p, xpath_allocator* alloc, T v) + + template void step_fill(xpath_node_set_raw& ns, xml_attribute_struct* a, xml_node_struct* p, xpath_allocator* alloc, bool once, T v) { const axis_t axis = T::axis; @@ -7243,17 +9992,19 @@ namespace case axis_ancestor_or_self: { if (axis == axis_ancestor_or_self && _test == nodetest_type_node) // reject attributes based on principal node type test - step_push(ns, a, p, alloc); + if (step_push(ns, a, p, alloc) & once) + return; + + xml_node_struct* cur = p; - xml_node cur = p; - while (cur) { - step_push(ns, cur, alloc); - - cur = cur.parent(); + if (step_push(ns, cur, alloc) & once) + return; + + cur = cur->parent; } - + break; } @@ -7268,23 +10019,26 @@ namespace case axis_following: { - xml_node cur = p; - - for (;;) + xml_node_struct* cur = p; + + while (cur) { - if (cur.first_child()) - cur = cur.first_child(); - else if (cur.next_sibling()) - cur = cur.next_sibling(); + if (cur->first_child) + cur = cur->first_child; else { - while (cur && !cur.next_sibling()) cur = cur.parent(); - cur = cur.next_sibling(); - - if (!cur) break; + while (!cur->next_sibling) + { + cur = cur->parent; + + if (!cur) return; + } + + cur = cur->next_sibling; } - step_push(ns, cur, alloc); + if (step_push(ns, cur, alloc) & once) + return; } break; @@ -7300,26 +10054,43 @@ namespace case axis_preceding: { // preceding:: axis does not include attribute nodes and attribute ancestors (they are the same as parent's ancestors), so we can reuse node preceding - step_fill(ns, p, alloc, v); + step_fill(ns, p, alloc, once, v); break; } - + default: - assert(!"Unimplemented axis"); + assert(false && "Unimplemented axis"); // unreachable } } - - template xpath_node_set_raw step_do(const xpath_context& c, const xpath_stack& stack, T v) + + template void step_fill(xpath_node_set_raw& ns, const xpath_node& xn, xpath_allocator* alloc, bool once, T v) + { + const axis_t axis = T::axis; + const bool axis_has_attributes = (axis == axis_ancestor || axis == axis_ancestor_or_self || axis == axis_descendant_or_self || axis == axis_following || axis == axis_parent || axis == axis_preceding || axis == axis_self); + + if (xn.node()) + step_fill(ns, xn.node().internal_object(), alloc, once, v); + else if (axis_has_attributes && xn.attribute() && xn.parent()) + step_fill(ns, xn.attribute().internal_object(), xn.parent().internal_object(), alloc, once, v); + } + + template xpath_node_set_raw step_do(const xpath_context& c, const xpath_stack& stack, nodeset_eval_t eval, T v) { const axis_t axis = T::axis; - bool attributes = (axis == axis_ancestor || axis == axis_ancestor_or_self || axis == axis_descendant_or_self || axis == axis_following || axis == axis_parent || axis == axis_preceding || axis == axis_self); + const bool axis_reverse = (axis == axis_ancestor || axis == axis_ancestor_or_self || axis == axis_preceding || axis == axis_preceding_sibling); + const xpath_node_set::type_t axis_type = axis_reverse ? xpath_node_set::type_sorted_reverse : xpath_node_set::type_sorted; + + bool once = + (axis == axis_attribute && _test == nodetest_name) || + (!_right && eval_once(axis_type, eval)) || + (_right && !_right->_next && _right->_test == predicate_constant_one); xpath_node_set_raw ns; - ns.set_type((axis == axis_ancestor || axis == axis_ancestor_or_self || axis == axis_preceding || axis == axis_preceding_sibling) ? xpath_node_set::type_sorted_reverse : xpath_node_set::type_sorted); + ns.set_type(axis_type); if (_left) { - xpath_node_set_raw s = _left->eval_node_set(c, stack); + xpath_node_set_raw s = _left->eval_node_set(c, stack, nodeset_eval_all); // self axis preserves the original order if (axis == axis_self) ns.set_type(s.type()); @@ -7330,23 +10101,15 @@ namespace // in general, all axes generate elements in a particular order, but there is no order guarantee if axis is applied to two nodes if (axis != axis_self && size != 0) ns.set_type(xpath_node_set::type_unsorted); - - if (it->node()) - step_fill(ns, it->node(), stack.result, v); - else if (attributes) - step_fill(ns, it->attribute(), it->parent(), stack.result, v); - - apply_predicates(ns, size, stack); + + step_fill(ns, *it, stack.result, once, v); + if (_right) apply_predicates(ns, size, stack, eval); } } else { - if (c.n.node()) - step_fill(ns, c.n.node(), stack.result, v); - else if (attributes) - step_fill(ns, c.n.attribute(), c.n.parent(), stack.result, v); - - apply_predicates(ns, 0, stack); + step_fill(ns, c.n, stack.result, once, v); + if (_right) apply_predicates(ns, 0, stack, eval); } // child, attribute and self axes always generate unique set of nodes @@ -7356,40 +10119,47 @@ namespace return ns; } - + public: - xpath_ast_node(ast_type_t type, xpath_value_type rettype, const char_t* value): - _type((char)type), _rettype((char)rettype), _axis(0), _test(0), _left(0), _right(0), _next(0) + xpath_ast_node(ast_type_t type, xpath_value_type rettype_, const char_t* value): + _type(static_cast(type)), _rettype(static_cast(rettype_)), _axis(0), _test(0), _left(0), _right(0), _next(0) { assert(type == ast_string_constant); _data.string = value; } - xpath_ast_node(ast_type_t type, xpath_value_type rettype, double value): - _type((char)type), _rettype((char)rettype), _axis(0), _test(0), _left(0), _right(0), _next(0) + xpath_ast_node(ast_type_t type, xpath_value_type rettype_, double value): + _type(static_cast(type)), _rettype(static_cast(rettype_)), _axis(0), _test(0), _left(0), _right(0), _next(0) { assert(type == ast_number_constant); _data.number = value; } - - xpath_ast_node(ast_type_t type, xpath_value_type rettype, xpath_variable* value): - _type((char)type), _rettype((char)rettype), _axis(0), _test(0), _left(0), _right(0), _next(0) + + xpath_ast_node(ast_type_t type, xpath_value_type rettype_, xpath_variable* value): + _type(static_cast(type)), _rettype(static_cast(rettype_)), _axis(0), _test(0), _left(0), _right(0), _next(0) { assert(type == ast_variable); _data.variable = value; } - - xpath_ast_node(ast_type_t type, xpath_value_type rettype, xpath_ast_node* left = 0, xpath_ast_node* right = 0): - _type((char)type), _rettype((char)rettype), _axis(0), _test(0), _left(left), _right(right), _next(0) + + xpath_ast_node(ast_type_t type, xpath_value_type rettype_, xpath_ast_node* left = 0, xpath_ast_node* right = 0): + _type(static_cast(type)), _rettype(static_cast(rettype_)), _axis(0), _test(0), _left(left), _right(right), _next(0) { } xpath_ast_node(ast_type_t type, xpath_ast_node* left, axis_t axis, nodetest_t test, const char_t* contents): - _type((char)type), _rettype(xpath_type_node_set), _axis((char)axis), _test((char)test), _left(left), _right(0), _next(0) + _type(static_cast(type)), _rettype(xpath_type_node_set), _axis(static_cast(axis)), _test(static_cast(test)), _left(left), _right(0), _next(0) { + assert(type == ast_step); _data.nodetest = contents; } + xpath_ast_node(ast_type_t type, xpath_ast_node* left, xpath_ast_node* right, predicate_t test): + _type(static_cast(type)), _rettype(xpath_type_node_set), _axis(0), _test(static_cast(test)), _left(left), _right(right), _next(0) + { + assert(type == ast_filter || type == ast_predicate); + } + void set_next(xpath_ast_node* value) { _next = value; @@ -7406,25 +10176,25 @@ namespace { case ast_op_or: return _left->eval_boolean(c, stack) || _right->eval_boolean(c, stack); - + case ast_op_and: return _left->eval_boolean(c, stack) && _right->eval_boolean(c, stack); - + case ast_op_equal: return compare_eq(_left, _right, c, stack, equal_to()); case ast_op_not_equal: return compare_eq(_left, _right, c, stack, not_equal_to()); - + case ast_op_less: return compare_rel(_left, _right, c, stack, less()); - + case ast_op_greater: return compare_rel(_right, _left, c, stack, less()); case ast_op_less_or_equal: return compare_rel(_left, _right, c, stack, less_equal()); - + case ast_op_greater_or_equal: return compare_rel(_right, _left, c, stack, less_equal()); @@ -7450,79 +10220,87 @@ namespace case ast_func_boolean: return _left->eval_boolean(c, stack); - + case ast_func_not: return !_left->eval_boolean(c, stack); - + case ast_func_true: return true; - + case ast_func_false: return false; case ast_func_lang: { if (c.n.attribute()) return false; - + xpath_allocator_capture cr(stack.result); xpath_string lang = _left->eval_string(c, stack); - + for (xml_node n = c.n.node(); n; n = n.parent()) { xml_attribute a = n.attribute(PUGIXML_TEXT("xml:lang")); - + if (a) { const char_t* value = a.value(); - + // strnicmp / strncasecmp is not portable for (const char_t* lit = lang.c_str(); *lit; ++lit) { if (tolower_ascii(*lit) != tolower_ascii(*value)) return false; ++value; } - + return *value == 0 || *value == '-'; } } - + return false; } + case ast_opt_compare_attribute: + { + const char_t* value = (_right->_type == ast_string_constant) ? _right->_data.string : _right->_data.variable->get_string(); + + xml_attribute attr = c.n.node().attribute(_left->_data.nodetest); + + return attr && strequal(attr.value(), value) && is_xpath_attribute(attr.name()); + } + case ast_variable: { assert(_rettype == _data.variable->type()); if (_rettype == xpath_type_boolean) return _data.variable->get_boolean(); - - // fallthrough to type conversion } + // fallthrough default: { switch (_rettype) { case xpath_type_number: return convert_number_to_boolean(eval_number(c, stack)); - + case xpath_type_string: { xpath_allocator_capture cr(stack.result); return !eval_string(c, stack).empty(); } - - case xpath_type_node_set: + + case xpath_type_node_set: { xpath_allocator_capture cr(stack.result); - return !eval_node_set(c, stack).empty(); + return !eval_node_set(c, stack, nodeset_eval_any).empty(); } default: - assert(!"Wrong expression for return type boolean"); + assert(false && "Wrong expression for return type boolean"); // unreachable return false; } } @@ -7535,7 +10313,7 @@ namespace { case ast_op_add: return _left->eval_number(c, stack) + _right->eval_number(c, stack); - + case ast_op_subtract: return _left->eval_number(c, stack) - _right->eval_number(c, stack); @@ -7555,39 +10333,39 @@ namespace return _data.number; case ast_func_last: - return (double)c.size; - + return static_cast(c.size); + case ast_func_position: - return (double)c.position; + return static_cast(c.position); case ast_func_count: { xpath_allocator_capture cr(stack.result); - return (double)_left->eval_node_set(c, stack).size(); + return static_cast(_left->eval_node_set(c, stack, nodeset_eval_all).size()); } - + case ast_func_string_length_0: { xpath_allocator_capture cr(stack.result); - return (double)string_value(c.n, stack.result).length(); + return static_cast(string_value(c.n, stack.result).length()); } - + case ast_func_string_length_1: { xpath_allocator_capture cr(stack.result); - return (double)_left->eval_string(c, stack).length(); + return static_cast(_left->eval_string(c, stack).length()); } - + case ast_func_number_0: { xpath_allocator_capture cr(stack.result); return convert_string_to_number(string_value(c.n, stack.result).c_str()); } - + case ast_func_number_1: return _left->eval_number(c, stack); @@ -7596,76 +10374,75 @@ namespace xpath_allocator_capture cr(stack.result); double r = 0; - - xpath_node_set_raw ns = _left->eval_node_set(c, stack); - + + xpath_node_set_raw ns = _left->eval_node_set(c, stack, nodeset_eval_all); + for (const xpath_node* it = ns.begin(); it != ns.end(); ++it) { xpath_allocator_capture cri(stack.result); r += convert_string_to_number(string_value(*it, stack.result).c_str()); } - + return r; } case ast_func_floor: { double r = _left->eval_number(c, stack); - + return r == r ? floor(r) : r; } case ast_func_ceiling: { double r = _left->eval_number(c, stack); - + return r == r ? ceil(r) : r; } case ast_func_round: return round_nearest_nzero(_left->eval_number(c, stack)); - + case ast_variable: { assert(_rettype == _data.variable->type()); if (_rettype == xpath_type_number) return _data.variable->get_number(); - - // fallthrough to type conversion } + // fallthrough default: { switch (_rettype) { case xpath_type_boolean: return eval_boolean(c, stack) ? 1 : 0; - + case xpath_type_string: { xpath_allocator_capture cr(stack.result); return convert_string_to_number(eval_string(c, stack).c_str()); } - + case xpath_type_node_set: { xpath_allocator_capture cr(stack.result); return convert_string_to_number(eval_string(c, stack).c_str()); } - + default: - assert(!"Wrong expression for return type number"); + assert(false && "Wrong expression for return type number"); // unreachable return 0; } - + } } } - + xpath_string eval_string_concat(const xpath_context& c, const xpath_stack& stack) { assert(_type == ast_func_concat); @@ -7676,16 +10453,9 @@ namespace size_t count = 1; for (xpath_ast_node* nc = _right; nc; nc = nc->_next) count++; - // gather all strings - xpath_string static_buffer[4]; - xpath_string* buffer = static_buffer; - - // allocate on-heap for large concats - if (count > sizeof(static_buffer) / sizeof(static_buffer[0])) - { - buffer = static_cast(stack.temp->allocate(count * sizeof(xpath_string))); - assert(buffer); - } + // allocate a buffer for temporary string objects + xpath_string* buffer = static_cast(stack.temp->allocate(count * sizeof(xpath_string))); + if (!buffer) return xpath_string(); // evaluate all strings to temporary stack xpath_stack swapped_stack = {stack.temp, stack.result}; @@ -7702,7 +10472,7 @@ namespace // create final string char_t* result = static_cast(stack.result->allocate((length + 1) * sizeof(char_t))); - assert(result); + if (!result) return xpath_string(); char_t* ri = result; @@ -7712,7 +10482,7 @@ namespace *ri = 0; - return xpath_string(result, true); + return xpath_string::from_heap_preallocated(result, ri); } xpath_string eval_string(const xpath_context& c, const xpath_stack& stack) @@ -7720,57 +10490,57 @@ namespace switch (_type) { case ast_string_constant: - return xpath_string_const(_data.string); - + return xpath_string::from_const(_data.string); + case ast_func_local_name_0: { xpath_node na = c.n; - - return xpath_string_const(local_name(na)); + + return xpath_string::from_const(local_name(na)); } case ast_func_local_name_1: { xpath_allocator_capture cr(stack.result); - xpath_node_set_raw ns = _left->eval_node_set(c, stack); + xpath_node_set_raw ns = _left->eval_node_set(c, stack, nodeset_eval_first); xpath_node na = ns.first(); - - return xpath_string_const(local_name(na)); + + return xpath_string::from_const(local_name(na)); } case ast_func_name_0: { xpath_node na = c.n; - - return xpath_string_const(qualified_name(na)); + + return xpath_string::from_const(qualified_name(na)); } case ast_func_name_1: { xpath_allocator_capture cr(stack.result); - xpath_node_set_raw ns = _left->eval_node_set(c, stack); + xpath_node_set_raw ns = _left->eval_node_set(c, stack, nodeset_eval_first); xpath_node na = ns.first(); - - return xpath_string_const(qualified_name(na)); + + return xpath_string::from_const(qualified_name(na)); } case ast_func_namespace_uri_0: { xpath_node na = c.n; - - return xpath_string_const(namespace_uri(na)); + + return xpath_string::from_const(namespace_uri(na)); } case ast_func_namespace_uri_1: { xpath_allocator_capture cr(stack.result); - xpath_node_set_raw ns = _left->eval_node_set(c, stack); + xpath_node_set_raw ns = _left->eval_node_set(c, stack, nodeset_eval_first); xpath_node na = ns.first(); - - return xpath_string_const(namespace_uri(na)); + + return xpath_string::from_const(namespace_uri(na)); } case ast_func_string_0: @@ -7792,10 +10562,10 @@ namespace xpath_string p = _right->eval_string(c, swapped_stack); const char_t* pos = find_substring(s.c_str(), p.c_str()); - - return pos ? xpath_string(s.c_str(), pos, stack.result) : xpath_string(); + + return pos ? xpath_string::from_heap(s.c_str(), pos, stack.result) : xpath_string(); } - + case ast_func_substring_after: { xpath_allocator_capture cr(stack.temp); @@ -7804,13 +10574,14 @@ namespace xpath_string s = _left->eval_string(c, swapped_stack); xpath_string p = _right->eval_string(c, swapped_stack); - + const char_t* pos = find_substring(s.c_str(), p.c_str()); if (!pos) return xpath_string(); - const char_t* result = pos + p.length(); + const char_t* rbegin = pos + p.length(); + const char_t* rend = s.c_str() + s.length(); - return s.uses_heap() ? xpath_string(result, stack.result) : xpath_string_const(result); + return s.uses_heap() ? xpath_string::from_heap(rbegin, rend, stack.result) : xpath_string::from_const(rbegin); } case ast_func_substring_2: @@ -7823,18 +10594,19 @@ namespace size_t s_length = s.length(); double first = round_nearest(_right->eval_number(c, stack)); - + if (is_nan(first)) return xpath_string(); // NaN else if (first >= s_length + 1) return xpath_string(); - - size_t pos = first < 1 ? 1 : (size_t)first; + + size_t pos = first < 1 ? 1 : static_cast(first); assert(1 <= pos && pos <= s_length + 1); const char_t* rbegin = s.c_str() + (pos - 1); - - return s.uses_heap() ? xpath_string(rbegin, stack.result) : xpath_string_const(rbegin); + const char_t* rend = s.c_str() + s.length(); + + return s.uses_heap() ? xpath_string::from_heap(rbegin, rend, stack.result) : xpath_string::from_const(rbegin); } - + case ast_func_substring_3: { xpath_allocator_capture cr(stack.temp); @@ -7846,38 +10618,44 @@ namespace double first = round_nearest(_right->eval_number(c, stack)); double last = first + round_nearest(_right->_next->eval_number(c, stack)); - + if (is_nan(first) || is_nan(last)) return xpath_string(); else if (first >= s_length + 1) return xpath_string(); else if (first >= last) return xpath_string(); else if (last < 1) return xpath_string(); - - size_t pos = first < 1 ? 1 : (size_t)first; - size_t end = last >= s_length + 1 ? s_length + 1 : (size_t)last; + + size_t pos = first < 1 ? 1 : static_cast(first); + size_t end = last >= s_length + 1 ? s_length + 1 : static_cast(last); assert(1 <= pos && pos <= end && end <= s_length + 1); const char_t* rbegin = s.c_str() + (pos - 1); const char_t* rend = s.c_str() + (end - 1); - return (end == s_length + 1 && !s.uses_heap()) ? xpath_string_const(rbegin) : xpath_string(rbegin, rend, stack.result); + return (end == s_length + 1 && !s.uses_heap()) ? xpath_string::from_const(rbegin) : xpath_string::from_heap(rbegin, rend, stack.result); } case ast_func_normalize_space_0: { xpath_string s = string_value(c.n, stack.result); - normalize_space(s.data(stack.result)); + char_t* begin = s.data(stack.result); + if (!begin) return xpath_string(); - return s; + char_t* end = normalize_space(begin); + + return xpath_string::from_heap_preallocated(begin, end); } case ast_func_normalize_space_1: { xpath_string s = _left->eval_string(c, stack); - normalize_space(s.data(stack.result)); - - return s; + char_t* begin = s.data(stack.result); + if (!begin) return xpath_string(); + + char_t* end = normalize_space(begin); + + return xpath_string::from_heap_preallocated(begin, end); } case ast_func_translate: @@ -7890,9 +10668,24 @@ namespace xpath_string from = _right->eval_string(c, swapped_stack); xpath_string to = _right->_next->eval_string(c, swapped_stack); - translate(s.data(stack.result), from.c_str(), to.c_str()); + char_t* begin = s.data(stack.result); + if (!begin) return xpath_string(); + + char_t* end = translate(begin, from.c_str(), to.c_str(), to.length()); - return s; + return xpath_string::from_heap_preallocated(begin, end); + } + + case ast_opt_translate_table: + { + xpath_string s = _left->eval_string(c, stack); + + char_t* begin = s.data(stack.result); + if (!begin) return xpath_string(); + + char_t* end = translate_table(begin, _data.table); + + return xpath_string::from_heap_preallocated(begin, end); } case ast_variable: @@ -7900,40 +10693,39 @@ namespace assert(_rettype == _data.variable->type()); if (_rettype == xpath_type_string) - return xpath_string_const(_data.variable->get_string()); - - // fallthrough to type conversion + return xpath_string::from_const(_data.variable->get_string()); } + // fallthrough default: { switch (_rettype) { case xpath_type_boolean: - return xpath_string_const(eval_boolean(c, stack) ? PUGIXML_TEXT("true") : PUGIXML_TEXT("false")); - + return xpath_string::from_const(eval_boolean(c, stack) ? PUGIXML_TEXT("true") : PUGIXML_TEXT("false")); + case xpath_type_number: return convert_number_to_string(eval_number(c, stack), stack.result); - + case xpath_type_node_set: { xpath_allocator_capture cr(stack.temp); xpath_stack swapped_stack = {stack.temp, stack.result}; - xpath_node_set_raw ns = eval_node_set(c, swapped_stack); + xpath_node_set_raw ns = eval_node_set(c, swapped_stack, nodeset_eval_first); return ns.empty() ? xpath_string() : string_value(ns.first(), stack.result); } - + default: - assert(!"Wrong expression for return type string"); + assert(false && "Wrong expression for return type string"); // unreachable return xpath_string(); } } } } - xpath_node_set_raw eval_node_set(const xpath_context& c, const xpath_stack& stack) + xpath_node_set_raw eval_node_set(const xpath_context& c, const xpath_stack& stack, nodeset_eval_t eval) { switch (_type) { @@ -7943,77 +10735,82 @@ namespace xpath_stack swapped_stack = {stack.temp, stack.result}; - xpath_node_set_raw ls = _left->eval_node_set(c, swapped_stack); - xpath_node_set_raw rs = _right->eval_node_set(c, stack); - + xpath_node_set_raw ls = _left->eval_node_set(c, swapped_stack, eval); + xpath_node_set_raw rs = _right->eval_node_set(c, stack, eval); + // we can optimize merging two sorted sets, but this is a very rare operation, so don't bother - rs.set_type(xpath_node_set::type_unsorted); + rs.set_type(xpath_node_set::type_unsorted); rs.append(ls.begin(), ls.end(), stack.result); rs.remove_duplicates(); - + return rs; } case ast_filter: - case ast_filter_posinv: { - xpath_node_set_raw set = _left->eval_node_set(c, stack); + xpath_node_set_raw set = _left->eval_node_set(c, stack, _test == predicate_constant_one ? nodeset_eval_first : nodeset_eval_all); // either expression is a number or it contains position() call; sort by document order - if (_type == ast_filter) set.sort_do(); + if (_test != predicate_posinv) set.sort_do(); + + bool once = eval_once(set.type(), eval); + + apply_predicate(set, 0, stack, once); - apply_predicate(set, 0, _right, stack); - return set; } - + case ast_func_id: return xpath_node_set_raw(); - + case ast_step: { switch (_axis) { case axis_ancestor: - return step_do(c, stack, axis_to_type()); - + return step_do(c, stack, eval, axis_to_type()); + case axis_ancestor_or_self: - return step_do(c, stack, axis_to_type()); + return step_do(c, stack, eval, axis_to_type()); case axis_attribute: - return step_do(c, stack, axis_to_type()); + return step_do(c, stack, eval, axis_to_type()); case axis_child: - return step_do(c, stack, axis_to_type()); - + return step_do(c, stack, eval, axis_to_type()); + case axis_descendant: - return step_do(c, stack, axis_to_type()); + return step_do(c, stack, eval, axis_to_type()); case axis_descendant_or_self: - return step_do(c, stack, axis_to_type()); + return step_do(c, stack, eval, axis_to_type()); case axis_following: - return step_do(c, stack, axis_to_type()); - + return step_do(c, stack, eval, axis_to_type()); + case axis_following_sibling: - return step_do(c, stack, axis_to_type()); - + return step_do(c, stack, eval, axis_to_type()); + case axis_namespace: // namespaced axis is not supported return xpath_node_set_raw(); - + case axis_parent: - return step_do(c, stack, axis_to_type()); - + return step_do(c, stack, eval, axis_to_type()); + case axis_preceding: - return step_do(c, stack, axis_to_type()); + return step_do(c, stack, eval, axis_to_type()); case axis_preceding_sibling: - return step_do(c, stack, axis_to_type()); - + return step_do(c, stack, eval, axis_to_type()); + case axis_self: - return step_do(c, stack, axis_to_type()); + return step_do(c, stack, eval, axis_to_type()); + + default: + assert(false && "Unknown axis"); // unreachable + return xpath_node_set_raw(); } } @@ -8044,23 +10841,97 @@ namespace ns.set_type(s.type()); ns.append(s.begin(), s.end(), stack.result); - return ns; - } + return ns; + } + } + + // fallthrough + default: + assert(false && "Wrong expression for return type node set"); // unreachable + return xpath_node_set_raw(); + } + } + + void optimize(xpath_allocator* alloc) + { + if (_left) + _left->optimize(alloc); + + if (_right) + _right->optimize(alloc); + + if (_next) + _next->optimize(alloc); + + optimize_self(alloc); + } + + void optimize_self(xpath_allocator* alloc) + { + // Rewrite [position()=expr] with [expr] + // Note that this step has to go before classification to recognize [position()=1] + if ((_type == ast_filter || _type == ast_predicate) && + _right->_type == ast_op_equal && _right->_left->_type == ast_func_position && _right->_right->_rettype == xpath_type_number) + { + _right = _right->_right; + } + + // Classify filter/predicate ops to perform various optimizations during evaluation + if (_type == ast_filter || _type == ast_predicate) + { + assert(_test == predicate_default); + + if (_right->_type == ast_number_constant && _right->_data.number == 1.0) + _test = predicate_constant_one; + else if (_right->_rettype == xpath_type_number && (_right->_type == ast_number_constant || _right->_type == ast_variable || _right->_type == ast_func_last)) + _test = predicate_constant; + else if (_right->_rettype != xpath_type_number && _right->is_posinv_expr()) + _test = predicate_posinv; + } + + // Rewrite descendant-or-self::node()/child::foo with descendant::foo + // The former is a full form of //foo, the latter is much faster since it executes the node test immediately + // Do a similar kind of rewrite for self/descendant/descendant-or-self axes + // Note that we only rewrite positionally invariant steps (//foo[1] != /descendant::foo[1]) + if (_type == ast_step && (_axis == axis_child || _axis == axis_self || _axis == axis_descendant || _axis == axis_descendant_or_self) && _left && + _left->_type == ast_step && _left->_axis == axis_descendant_or_self && _left->_test == nodetest_type_node && !_left->_right && + is_posinv_step()) + { + if (_axis == axis_child || _axis == axis_descendant) + _axis = axis_descendant; + else + _axis = axis_descendant_or_self; + + _left = _left->_left; + } + + // Use optimized lookup table implementation for translate() with constant arguments + if (_type == ast_func_translate && _right->_type == ast_string_constant && _right->_next->_type == ast_string_constant) + { + unsigned char* table = translate_table_generate(alloc, _right->_data.string, _right->_next->_data.string); - // fallthrough to type conversion + if (table) + { + _type = ast_opt_translate_table; + _data.table = table; + } } - default: - assert(!"Wrong expression for return type node set"); - return xpath_node_set_raw(); + // Use optimized path for @attr = 'value' or @attr = $value + if (_type == ast_op_equal && + _left->_type == ast_step && _left->_axis == axis_attribute && _left->_test == nodetest_name && !_left->_left && !_left->_right && + (_right->_type == ast_string_constant || (_right->_type == ast_variable && _right->_rettype == xpath_type_string))) + { + _type = ast_opt_compare_attribute; } } - - bool is_posinv() + + bool is_posinv_expr() const { switch (_type) { case ast_func_position: + case ast_func_last: return false; case ast_string_constant: @@ -8074,19 +10945,33 @@ namespace case ast_predicate: case ast_filter: - case ast_filter_posinv: return true; default: - if (_left && !_left->is_posinv()) return false; - + if (_left && !_left->is_posinv_expr()) return false; + for (xpath_ast_node* n = _right; n; n = n->_next) - if (!n->is_posinv()) return false; - + if (!n->is_posinv_expr()) return false; + return true; } } + bool is_posinv_step() const + { + assert(_type == ast_step); + + for (xpath_ast_node* n = _right; n; n = n->_next) + { + assert(n->_type == ast_predicate); + + if (n->_test != predicate_posinv) + return false; + } + + return true; + } + xpath_value_type rettype() const { return static_cast(_rettype); @@ -8095,72 +10980,87 @@ namespace struct xpath_parser { - xpath_allocator* _alloc; - xpath_lexer _lexer; + xpath_allocator* _alloc; + xpath_lexer _lexer; const char_t* _query; xpath_variable_set* _variables; xpath_parse_result* _result; - #ifdef PUGIXML_NO_EXCEPTIONS - jmp_buf _error_handler; - #endif + char_t _scratch[32]; - void throw_error(const char* message) + xpath_ast_node* error(const char* message) { _result->error = message; _result->offset = _lexer.current_pos() - _query; - #ifdef PUGIXML_NO_EXCEPTIONS - longjmp(_error_handler, 1); - #else - throw xpath_exception(*_result); - #endif + return 0; } - void throw_error_oom() - { - #ifdef PUGIXML_NO_EXCEPTIONS - throw_error("Out of memory"); - #else - throw std::bad_alloc(); - #endif - } + xpath_ast_node* error_oom() + { + assert(_alloc->_error); + *_alloc->_error = true; + + return 0; + } void* alloc_node() { - void* result = _alloc->allocate_nothrow(sizeof(xpath_ast_node)); + return _alloc->allocate(sizeof(xpath_ast_node)); + } - if (!result) throw_error_oom(); + xpath_ast_node* alloc_node(ast_type_t type, xpath_value_type rettype, const char_t* value) + { + void* memory = alloc_node(); + return memory ? new (memory) xpath_ast_node(type, rettype, value) : 0; + } - return result; + xpath_ast_node* alloc_node(ast_type_t type, xpath_value_type rettype, double value) + { + void* memory = alloc_node(); + return memory ? new (memory) xpath_ast_node(type, rettype, value) : 0; } - const char_t* alloc_string(const xpath_lexer_string& value) + xpath_ast_node* alloc_node(ast_type_t type, xpath_value_type rettype, xpath_variable* value) { - if (value.begin) - { - size_t length = static_cast(value.end - value.begin); + void* memory = alloc_node(); + return memory ? new (memory) xpath_ast_node(type, rettype, value) : 0; + } - char_t* c = static_cast(_alloc->allocate_nothrow((length + 1) * sizeof(char_t))); - if (!c) throw_error_oom(); + xpath_ast_node* alloc_node(ast_type_t type, xpath_value_type rettype, xpath_ast_node* left = 0, xpath_ast_node* right = 0) + { + void* memory = alloc_node(); + return memory ? new (memory) xpath_ast_node(type, rettype, left, right) : 0; + } - memcpy(c, value.begin, length * sizeof(char_t)); - c[length] = 0; + xpath_ast_node* alloc_node(ast_type_t type, xpath_ast_node* left, axis_t axis, nodetest_t test, const char_t* contents) + { + void* memory = alloc_node(); + return memory ? new (memory) xpath_ast_node(type, left, axis, test, contents) : 0; + } - return c; - } - else return 0; + xpath_ast_node* alloc_node(ast_type_t type, xpath_ast_node* left, xpath_ast_node* right, predicate_t test) + { + void* memory = alloc_node(); + return memory ? new (memory) xpath_ast_node(type, left, right, test) : 0; } - xpath_ast_node* parse_function_helper(ast_type_t type0, ast_type_t type1, size_t argc, xpath_ast_node* args[2]) + const char_t* alloc_string(const xpath_lexer_string& value) { - assert(argc <= 1); + if (!value.begin) + return PUGIXML_TEXT(""); + + size_t length = static_cast(value.end - value.begin); + + char_t* c = static_cast(_alloc->allocate((length + 1) * sizeof(char_t))); + if (!c) return 0; - if (argc == 1 && args[0]->rettype() != xpath_type_node_set) throw_error("Function has to be applied to node set"); + memcpy(c, value.begin, length * sizeof(char_t)); + c[length] = 0; - return new (alloc_node()) xpath_ast_node(argc == 0 ? type0 : type1, xpath_type_string, args[0]); + return c; } xpath_ast_node* parse_function(const xpath_lexer_string& name, size_t argc, xpath_ast_node* args[2]) @@ -8169,108 +11069,118 @@ namespace { case 'b': if (name == PUGIXML_TEXT("boolean") && argc == 1) - return new (alloc_node()) xpath_ast_node(ast_func_boolean, xpath_type_boolean, args[0]); - + return alloc_node(ast_func_boolean, xpath_type_boolean, args[0]); + break; - + case 'c': if (name == PUGIXML_TEXT("count") && argc == 1) { - if (args[0]->rettype() != xpath_type_node_set) throw_error("Function has to be applied to node set"); - return new (alloc_node()) xpath_ast_node(ast_func_count, xpath_type_number, args[0]); + if (args[0]->rettype() != xpath_type_node_set) return error("Function has to be applied to node set"); + return alloc_node(ast_func_count, xpath_type_number, args[0]); } else if (name == PUGIXML_TEXT("contains") && argc == 2) - return new (alloc_node()) xpath_ast_node(ast_func_contains, xpath_type_string, args[0], args[1]); + return alloc_node(ast_func_contains, xpath_type_boolean, args[0], args[1]); else if (name == PUGIXML_TEXT("concat") && argc >= 2) - return new (alloc_node()) xpath_ast_node(ast_func_concat, xpath_type_string, args[0], args[1]); + return alloc_node(ast_func_concat, xpath_type_string, args[0], args[1]); else if (name == PUGIXML_TEXT("ceiling") && argc == 1) - return new (alloc_node()) xpath_ast_node(ast_func_ceiling, xpath_type_number, args[0]); - + return alloc_node(ast_func_ceiling, xpath_type_number, args[0]); + break; - + case 'f': if (name == PUGIXML_TEXT("false") && argc == 0) - return new (alloc_node()) xpath_ast_node(ast_func_false, xpath_type_boolean); + return alloc_node(ast_func_false, xpath_type_boolean); else if (name == PUGIXML_TEXT("floor") && argc == 1) - return new (alloc_node()) xpath_ast_node(ast_func_floor, xpath_type_number, args[0]); - + return alloc_node(ast_func_floor, xpath_type_number, args[0]); + break; - + case 'i': if (name == PUGIXML_TEXT("id") && argc == 1) - return new (alloc_node()) xpath_ast_node(ast_func_id, xpath_type_node_set, args[0]); - + return alloc_node(ast_func_id, xpath_type_node_set, args[0]); + break; - + case 'l': if (name == PUGIXML_TEXT("last") && argc == 0) - return new (alloc_node()) xpath_ast_node(ast_func_last, xpath_type_number); + return alloc_node(ast_func_last, xpath_type_number); else if (name == PUGIXML_TEXT("lang") && argc == 1) - return new (alloc_node()) xpath_ast_node(ast_func_lang, xpath_type_boolean, args[0]); + return alloc_node(ast_func_lang, xpath_type_boolean, args[0]); else if (name == PUGIXML_TEXT("local-name") && argc <= 1) - return parse_function_helper(ast_func_local_name_0, ast_func_local_name_1, argc, args); - + { + if (argc == 1 && args[0]->rettype() != xpath_type_node_set) return error("Function has to be applied to node set"); + return alloc_node(argc == 0 ? ast_func_local_name_0 : ast_func_local_name_1, xpath_type_string, args[0]); + } + break; - + case 'n': if (name == PUGIXML_TEXT("name") && argc <= 1) - return parse_function_helper(ast_func_name_0, ast_func_name_1, argc, args); + { + if (argc == 1 && args[0]->rettype() != xpath_type_node_set) return error("Function has to be applied to node set"); + return alloc_node(argc == 0 ? ast_func_name_0 : ast_func_name_1, xpath_type_string, args[0]); + } else if (name == PUGIXML_TEXT("namespace-uri") && argc <= 1) - return parse_function_helper(ast_func_namespace_uri_0, ast_func_namespace_uri_1, argc, args); + { + if (argc == 1 && args[0]->rettype() != xpath_type_node_set) return error("Function has to be applied to node set"); + return alloc_node(argc == 0 ? ast_func_namespace_uri_0 : ast_func_namespace_uri_1, xpath_type_string, args[0]); + } else if (name == PUGIXML_TEXT("normalize-space") && argc <= 1) - return new (alloc_node()) xpath_ast_node(argc == 0 ? ast_func_normalize_space_0 : ast_func_normalize_space_1, xpath_type_string, args[0], args[1]); + return alloc_node(argc == 0 ? ast_func_normalize_space_0 : ast_func_normalize_space_1, xpath_type_string, args[0], args[1]); else if (name == PUGIXML_TEXT("not") && argc == 1) - return new (alloc_node()) xpath_ast_node(ast_func_not, xpath_type_boolean, args[0]); + return alloc_node(ast_func_not, xpath_type_boolean, args[0]); else if (name == PUGIXML_TEXT("number") && argc <= 1) - return new (alloc_node()) xpath_ast_node(argc == 0 ? ast_func_number_0 : ast_func_number_1, xpath_type_number, args[0]); - + return alloc_node(argc == 0 ? ast_func_number_0 : ast_func_number_1, xpath_type_number, args[0]); + break; - + case 'p': if (name == PUGIXML_TEXT("position") && argc == 0) - return new (alloc_node()) xpath_ast_node(ast_func_position, xpath_type_number); - + return alloc_node(ast_func_position, xpath_type_number); + break; - + case 'r': if (name == PUGIXML_TEXT("round") && argc == 1) - return new (alloc_node()) xpath_ast_node(ast_func_round, xpath_type_number, args[0]); + return alloc_node(ast_func_round, xpath_type_number, args[0]); break; - + case 's': if (name == PUGIXML_TEXT("string") && argc <= 1) - return new (alloc_node()) xpath_ast_node(argc == 0 ? ast_func_string_0 : ast_func_string_1, xpath_type_string, args[0]); + return alloc_node(argc == 0 ? ast_func_string_0 : ast_func_string_1, xpath_type_string, args[0]); else if (name == PUGIXML_TEXT("string-length") && argc <= 1) - return new (alloc_node()) xpath_ast_node(argc == 0 ? ast_func_string_length_0 : ast_func_string_length_1, xpath_type_string, args[0]); + return alloc_node(argc == 0 ? ast_func_string_length_0 : ast_func_string_length_1, xpath_type_number, args[0]); else if (name == PUGIXML_TEXT("starts-with") && argc == 2) - return new (alloc_node()) xpath_ast_node(ast_func_starts_with, xpath_type_boolean, args[0], args[1]); + return alloc_node(ast_func_starts_with, xpath_type_boolean, args[0], args[1]); else if (name == PUGIXML_TEXT("substring-before") && argc == 2) - return new (alloc_node()) xpath_ast_node(ast_func_substring_before, xpath_type_string, args[0], args[1]); + return alloc_node(ast_func_substring_before, xpath_type_string, args[0], args[1]); else if (name == PUGIXML_TEXT("substring-after") && argc == 2) - return new (alloc_node()) xpath_ast_node(ast_func_substring_after, xpath_type_string, args[0], args[1]); + return alloc_node(ast_func_substring_after, xpath_type_string, args[0], args[1]); else if (name == PUGIXML_TEXT("substring") && (argc == 2 || argc == 3)) - return new (alloc_node()) xpath_ast_node(argc == 2 ? ast_func_substring_2 : ast_func_substring_3, xpath_type_string, args[0], args[1]); + return alloc_node(argc == 2 ? ast_func_substring_2 : ast_func_substring_3, xpath_type_string, args[0], args[1]); else if (name == PUGIXML_TEXT("sum") && argc == 1) { - if (args[0]->rettype() != xpath_type_node_set) throw_error("Function has to be applied to node set"); - return new (alloc_node()) xpath_ast_node(ast_func_sum, xpath_type_number, args[0]); + if (args[0]->rettype() != xpath_type_node_set) return error("Function has to be applied to node set"); + return alloc_node(ast_func_sum, xpath_type_number, args[0]); } break; - + case 't': if (name == PUGIXML_TEXT("translate") && argc == 3) - return new (alloc_node()) xpath_ast_node(ast_func_translate, xpath_type_string, args[0], args[1]); + return alloc_node(ast_func_translate, xpath_type_string, args[0], args[1]); else if (name == PUGIXML_TEXT("true") && argc == 0) - return new (alloc_node()) xpath_ast_node(ast_func_true, xpath_type_boolean); - + return alloc_node(ast_func_true, xpath_type_boolean); + break; - } - throw_error("Unrecognized function or wrong parameter count"); + default: + break; + } - return 0; + return error("Unrecognized function or wrong parameter count"); } axis_t parse_axis_name(const xpath_lexer_string& name, bool& specified) @@ -8286,37 +11196,37 @@ namespace return axis_ancestor_or_self; else if (name == PUGIXML_TEXT("attribute")) return axis_attribute; - + break; - + case 'c': if (name == PUGIXML_TEXT("child")) return axis_child; - + break; - + case 'd': if (name == PUGIXML_TEXT("descendant")) return axis_descendant; else if (name == PUGIXML_TEXT("descendant-or-self")) return axis_descendant_or_self; - + break; - + case 'f': if (name == PUGIXML_TEXT("following")) return axis_following; else if (name == PUGIXML_TEXT("following-sibling")) return axis_following_sibling; - + break; - + case 'n': if (name == PUGIXML_TEXT("namespace")) return axis_namespace; - + break; - + case 'p': if (name == PUGIXML_TEXT("parent")) return axis_parent; @@ -8324,13 +11234,16 @@ namespace return axis_preceding; else if (name == PUGIXML_TEXT("preceding-sibling")) return axis_preceding_sibling; - + break; - + case 's': if (name == PUGIXML_TEXT("self")) return axis_self; - + + break; + + default: break; } @@ -8365,31 +11278,36 @@ namespace return nodetest_type_text; break; + + default: + break; } return nodetest_none; } - // PrimaryExpr ::= VariableReference | '(' Expr ')' | Literal | Number | FunctionCall - xpath_ast_node* parse_primary_expression() - { - switch (_lexer.current()) - { - case lex_var_ref: - { + // PrimaryExpr ::= VariableReference | '(' Expr ')' | Literal | Number | FunctionCall + xpath_ast_node* parse_primary_expression() + { + switch (_lexer.current()) + { + case lex_var_ref: + { xpath_lexer_string name = _lexer.contents(); if (!_variables) - throw_error("Unknown variable: variable set is not provided"); + return error("Unknown variable: variable set is not provided"); - xpath_variable* var = get_variable(_variables, name.begin, name.end); + xpath_variable* var = 0; + if (!get_variable_scratch(_scratch, _variables, name.begin, name.end, &var)) + return error_oom(); if (!var) - throw_error("Unknown variable: variable set does not contain the given name"); + return error("Unknown variable: variable set does not contain the given name"); _lexer.next(); - return new (alloc_node()) xpath_ast_node(ast_variable, var->type(), var); + return alloc_node(ast_variable, var->type(), var); } case lex_open_brace: @@ -8397,9 +11315,10 @@ namespace _lexer.next(); xpath_ast_node* n = parse_expression(); + if (!n) return 0; if (_lexer.current() != lex_close_brace) - throw_error("Unmatched braces"); + return error("Expected ')' to match an opening '('"); _lexer.next(); @@ -8409,107 +11328,107 @@ namespace case lex_quoted_string: { const char_t* value = alloc_string(_lexer.contents()); + if (!value) return 0; - xpath_ast_node* n = new (alloc_node()) xpath_ast_node(ast_string_constant, xpath_type_string, value); _lexer.next(); - return n; + return alloc_node(ast_string_constant, xpath_type_string, value); } case lex_number: { double value = 0; - if (!convert_string_to_number(_lexer.contents().begin, _lexer.contents().end, &value)) - throw_error_oom(); + if (!convert_string_to_number_scratch(_scratch, _lexer.contents().begin, _lexer.contents().end, &value)) + return error_oom(); - xpath_ast_node* n = new (alloc_node()) xpath_ast_node(ast_number_constant, xpath_type_number, value); _lexer.next(); - return n; + return alloc_node(ast_number_constant, xpath_type_number, value); } case lex_string: { xpath_ast_node* args[2] = {0}; size_t argc = 0; - + xpath_lexer_string function = _lexer.contents(); _lexer.next(); - + xpath_ast_node* last_arg = 0; - + if (_lexer.current() != lex_open_brace) - throw_error("Unrecognized function call"); + return error("Unrecognized function call"); _lexer.next(); - if (_lexer.current() != lex_close_brace) - args[argc++] = parse_expression(); - while (_lexer.current() != lex_close_brace) { - if (_lexer.current() != lex_comma) - throw_error("No comma between function arguments"); - _lexer.next(); - + if (argc > 0) + { + if (_lexer.current() != lex_comma) + return error("No comma between function arguments"); + _lexer.next(); + } + xpath_ast_node* n = parse_expression(); - + if (!n) return 0; + if (argc < 2) args[argc] = n; else last_arg->set_next(n); argc++; last_arg = n; } - + _lexer.next(); return parse_function(function, argc, args); } - default: - throw_error("Unrecognizable primary expression"); + default: + return error("Unrecognizable primary expression"); + } + } + + // FilterExpr ::= PrimaryExpr | FilterExpr Predicate + // Predicate ::= '[' PredicateExpr ']' + // PredicateExpr ::= Expr + xpath_ast_node* parse_filter_expression() + { + xpath_ast_node* n = parse_primary_expression(); + if (!n) return 0; - return 0; - } - } - - // FilterExpr ::= PrimaryExpr | FilterExpr Predicate - // Predicate ::= '[' PredicateExpr ']' - // PredicateExpr ::= Expr - xpath_ast_node* parse_filter_expression() - { - xpath_ast_node* n = parse_primary_expression(); + while (_lexer.current() == lex_open_square_brace) + { + _lexer.next(); - while (_lexer.current() == lex_open_square_brace) - { - _lexer.next(); + if (n->rettype() != xpath_type_node_set) + return error("Predicate has to be applied to node set"); xpath_ast_node* expr = parse_expression(); + if (!expr) return 0; + + n = alloc_node(ast_filter, n, expr, predicate_default); + if (!n) return 0; + + if (_lexer.current() != lex_close_square_brace) + return error("Expected ']' to match an opening '['"); + + _lexer.next(); + } + + return n; + } - if (n->rettype() != xpath_type_node_set) throw_error("Predicate has to be applied to node set"); - - bool posinv = expr->rettype() != xpath_type_number && expr->is_posinv(); - - n = new (alloc_node()) xpath_ast_node(posinv ? ast_filter_posinv : ast_filter, xpath_type_node_set, n, expr); - - if (_lexer.current() != lex_close_square_brace) - throw_error("Unmatched square brace"); - - _lexer.next(); - } - - return n; - } - - // Step ::= AxisSpecifier NodeTest Predicate* | AbbreviatedStep - // AxisSpecifier ::= AxisName '::' | '@'? - // NodeTest ::= NameTest | NodeType '(' ')' | 'processing-instruction' '(' Literal ')' - // NameTest ::= '*' | NCName ':' '*' | QName - // AbbreviatedStep ::= '.' | '..' - xpath_ast_node* parse_step(xpath_ast_node* set) - { + // Step ::= AxisSpecifier NodeTest Predicate* | AbbreviatedStep + // AxisSpecifier ::= AxisName '::' | '@'? + // NodeTest ::= NameTest | NodeType '(' ')' | 'processing-instruction' '(' Literal ')' + // NameTest ::= '*' | NCName ':' '*' | QName + // AbbreviatedStep ::= '.' | '..' + xpath_ast_node* parse_step(xpath_ast_node* set) + { if (set && set->rettype() != xpath_type_node_set) - throw_error("Step has to be applied to node set"); + return error("Step has to be applied to node set"); bool axis_specified = false; axis_t axis = axis_child; // implied child axis @@ -8518,25 +11437,31 @@ namespace { axis = axis_attribute; axis_specified = true; - + _lexer.next(); } else if (_lexer.current() == lex_dot) { _lexer.next(); - - return new (alloc_node()) xpath_ast_node(ast_step, set, axis_self, nodetest_type_node, 0); + + if (_lexer.current() == lex_open_square_brace) + return error("Predicates are not allowed after an abbreviated step"); + + return alloc_node(ast_step, set, axis_self, nodetest_type_node, 0); } else if (_lexer.current() == lex_double_dot) { _lexer.next(); - - return new (alloc_node()) xpath_ast_node(ast_step, set, axis_parent, nodetest_type_node, 0); + + if (_lexer.current() == lex_open_square_brace) + return error("Predicates are not allowed after an abbreviated step"); + + return alloc_node(ast_step, set, axis_parent, nodetest_type_node, 0); } - + nodetest_t nt_type = nodetest_none; xpath_lexer_string nt_name; - + if (_lexer.current() == lex_string) { // node name test @@ -8547,11 +11472,13 @@ namespace if (_lexer.current() == lex_double_colon) { // parse axis name - if (axis_specified) throw_error("Two axis specifiers in one step"); + if (axis_specified) + return error("Two axis specifiers in one step"); axis = parse_axis_name(nt_name, axis_specified); - if (!axis_specified) throw_error("Unknown axis"); + if (!axis_specified) + return error("Unknown axis"); // read actual node test _lexer.next(); @@ -8567,42 +11494,47 @@ namespace nt_name = _lexer.contents(); _lexer.next(); } - else throw_error("Unrecognized node test"); + else + { + return error("Unrecognized node test"); + } } - + if (nt_type == nodetest_none) { // node type test or processing-instruction if (_lexer.current() == lex_open_brace) { _lexer.next(); - + if (_lexer.current() == lex_close_brace) { _lexer.next(); nt_type = parse_node_test_type(nt_name); - if (nt_type == nodetest_none) throw_error("Unrecognized node type"); - + if (nt_type == nodetest_none) + return error("Unrecognized node type"); + nt_name = xpath_lexer_string(); } else if (nt_name == PUGIXML_TEXT("processing-instruction")) { if (_lexer.current() != lex_quoted_string) - throw_error("Only literals are allowed as arguments to processing-instruction()"); - + return error("Only literals are allowed as arguments to processing-instruction()"); + nt_type = nodetest_pi; nt_name = _lexer.contents(); _lexer.next(); - + if (_lexer.current() != lex_close_brace) - throw_error("Unmatched brace near processing-instruction()"); + return error("Unmatched brace near processing-instruction()"); _lexer.next(); } else - throw_error("Unmatched brace near node type test"); - + { + return error("Unmatched brace near node type test"); + } } // QName or NCName:* else @@ -8610,10 +11542,13 @@ namespace if (nt_name.end - nt_name.begin > 2 && nt_name.end[-2] == ':' && nt_name.end[-1] == '*') // NCName:* { nt_name.end--; // erase * - + nt_type = nodetest_all_in_namespace; } - else nt_type = nodetest_name; + else + { + nt_type = nodetest_name; + } } } } @@ -8622,61 +11557,76 @@ namespace nt_type = nodetest_all; _lexer.next(); } - else throw_error("Unrecognized node test"); - - xpath_ast_node* n = new (alloc_node()) xpath_ast_node(ast_step, set, axis, nt_type, alloc_string(nt_name)); - + else + { + return error("Unrecognized node test"); + } + + const char_t* nt_name_copy = alloc_string(nt_name); + if (!nt_name_copy) return 0; + + xpath_ast_node* n = alloc_node(ast_step, set, axis, nt_type, nt_name_copy); + if (!n) return 0; + xpath_ast_node* last = 0; - + while (_lexer.current() == lex_open_square_brace) { _lexer.next(); - + xpath_ast_node* expr = parse_expression(); + if (!expr) return 0; + + xpath_ast_node* pred = alloc_node(ast_predicate, 0, expr, predicate_default); + if (!pred) return 0; - xpath_ast_node* pred = new (alloc_node()) xpath_ast_node(ast_predicate, xpath_type_node_set, expr); - if (_lexer.current() != lex_close_square_brace) - throw_error("Unmatched square brace"); + return error("Expected ']' to match an opening '['"); _lexer.next(); - + if (last) last->set_next(pred); else n->set_right(pred); - + last = pred; } - + return n; - } - - // RelativeLocationPath ::= Step | RelativeLocationPath '/' Step | RelativeLocationPath '//' Step - xpath_ast_node* parse_relative_location_path(xpath_ast_node* set) - { + } + + // RelativeLocationPath ::= Step | RelativeLocationPath '/' Step | RelativeLocationPath '//' Step + xpath_ast_node* parse_relative_location_path(xpath_ast_node* set) + { xpath_ast_node* n = parse_step(set); - + if (!n) return 0; + while (_lexer.current() == lex_slash || _lexer.current() == lex_double_slash) { lexeme_t l = _lexer.current(); _lexer.next(); if (l == lex_double_slash) - n = new (alloc_node()) xpath_ast_node(ast_step, n, axis_descendant_or_self, nodetest_type_node, 0); - + { + n = alloc_node(ast_step, n, axis_descendant_or_self, nodetest_type_node, 0); + if (!n) return 0; + } + n = parse_step(n); + if (!n) return 0; } - + return n; - } - - // LocationPath ::= RelativeLocationPath | AbsoluteLocationPath - // AbsoluteLocationPath ::= '/' RelativeLocationPath? | '//' RelativeLocationPath - xpath_ast_node* parse_location_path() - { + } + + // LocationPath ::= RelativeLocationPath | AbsoluteLocationPath + // AbsoluteLocationPath ::= '/' RelativeLocationPath? | '//' RelativeLocationPath + xpath_ast_node* parse_location_path() + { if (_lexer.current() == lex_slash) { _lexer.next(); - - xpath_ast_node* n = new (alloc_node()) xpath_ast_node(ast_step_root, xpath_type_node_set); + + xpath_ast_node* n = alloc_node(ast_step_root, xpath_type_node_set); + if (!n) return 0; // relative location path can start from axis_attribute, dot, double_dot, multiply and string lexemes; any other lexeme means standalone root path lexeme_t l = _lexer.current(); @@ -8686,234 +11636,218 @@ namespace else return n; } - else if (_lexer.current() == lex_double_slash) + else if (_lexer.current() == lex_double_slash) + { + _lexer.next(); + + xpath_ast_node* n = alloc_node(ast_step_root, xpath_type_node_set); + if (!n) return 0; + + n = alloc_node(ast_step, n, axis_descendant_or_self, nodetest_type_node, 0); + if (!n) return 0; + + return parse_relative_location_path(n); + } + + // else clause moved outside of if because of bogus warning 'control may reach end of non-void function being inlined' in gcc 4.0.1 + return parse_relative_location_path(0); + } + + // PathExpr ::= LocationPath + // | FilterExpr + // | FilterExpr '/' RelativeLocationPath + // | FilterExpr '//' RelativeLocationPath + // UnionExpr ::= PathExpr | UnionExpr '|' PathExpr + // UnaryExpr ::= UnionExpr | '-' UnaryExpr + xpath_ast_node* parse_path_or_unary_expression() + { + // Clarification. + // PathExpr begins with either LocationPath or FilterExpr. + // FilterExpr begins with PrimaryExpr + // PrimaryExpr begins with '$' in case of it being a variable reference, + // '(' in case of it being an expression, string literal, number constant or + // function call. + if (_lexer.current() == lex_var_ref || _lexer.current() == lex_open_brace || + _lexer.current() == lex_quoted_string || _lexer.current() == lex_number || + _lexer.current() == lex_string) + { + if (_lexer.current() == lex_string) + { + // This is either a function call, or not - if not, we shall proceed with location path + const char_t* state = _lexer.state(); + + while (PUGI__IS_CHARTYPE(*state, ct_space)) ++state; + + if (*state != '(') + return parse_location_path(); + + // This looks like a function call; however this still can be a node-test. Check it. + if (parse_node_test_type(_lexer.contents()) != nodetest_none) + return parse_location_path(); + } + + xpath_ast_node* n = parse_filter_expression(); + if (!n) return 0; + + if (_lexer.current() == lex_slash || _lexer.current() == lex_double_slash) + { + lexeme_t l = _lexer.current(); + _lexer.next(); + + if (l == lex_double_slash) + { + if (n->rettype() != xpath_type_node_set) + return error("Step has to be applied to node set"); + + n = alloc_node(ast_step, n, axis_descendant_or_self, nodetest_type_node, 0); + if (!n) return 0; + } + + // select from location path + return parse_relative_location_path(n); + } + + return n; + } + else if (_lexer.current() == lex_minus) + { + _lexer.next(); + + // precedence 7+ - only parses union expressions + xpath_ast_node* n = parse_expression(7); + if (!n) return 0; + + return alloc_node(ast_op_negate, xpath_type_number, n); + } + else + { + return parse_location_path(); + } + } + + struct binary_op_t + { + ast_type_t asttype; + xpath_value_type rettype; + int precedence; + + binary_op_t(): asttype(ast_unknown), rettype(xpath_type_none), precedence(0) + { + } + + binary_op_t(ast_type_t asttype_, xpath_value_type rettype_, int precedence_): asttype(asttype_), rettype(rettype_), precedence(precedence_) + { + } + + static binary_op_t parse(xpath_lexer& lexer) + { + switch (lexer.current()) + { + case lex_string: + if (lexer.contents() == PUGIXML_TEXT("or")) + return binary_op_t(ast_op_or, xpath_type_boolean, 1); + else if (lexer.contents() == PUGIXML_TEXT("and")) + return binary_op_t(ast_op_and, xpath_type_boolean, 2); + else if (lexer.contents() == PUGIXML_TEXT("div")) + return binary_op_t(ast_op_divide, xpath_type_number, 6); + else if (lexer.contents() == PUGIXML_TEXT("mod")) + return binary_op_t(ast_op_mod, xpath_type_number, 6); + else + return binary_op_t(); + + case lex_equal: + return binary_op_t(ast_op_equal, xpath_type_boolean, 3); + + case lex_not_equal: + return binary_op_t(ast_op_not_equal, xpath_type_boolean, 3); + + case lex_less: + return binary_op_t(ast_op_less, xpath_type_boolean, 4); + + case lex_greater: + return binary_op_t(ast_op_greater, xpath_type_boolean, 4); + + case lex_less_or_equal: + return binary_op_t(ast_op_less_or_equal, xpath_type_boolean, 4); + + case lex_greater_or_equal: + return binary_op_t(ast_op_greater_or_equal, xpath_type_boolean, 4); + + case lex_plus: + return binary_op_t(ast_op_add, xpath_type_number, 5); + + case lex_minus: + return binary_op_t(ast_op_subtract, xpath_type_number, 5); + + case lex_multiply: + return binary_op_t(ast_op_multiply, xpath_type_number, 6); + + case lex_union: + return binary_op_t(ast_op_union, xpath_type_node_set, 7); + + default: + return binary_op_t(); + } + } + }; + + xpath_ast_node* parse_expression_rec(xpath_ast_node* lhs, int limit) + { + binary_op_t op = binary_op_t::parse(_lexer); + + while (op.asttype != ast_unknown && op.precedence >= limit) { _lexer.next(); - - xpath_ast_node* n = new (alloc_node()) xpath_ast_node(ast_step_root, xpath_type_node_set); - n = new (alloc_node()) xpath_ast_node(ast_step, n, axis_descendant_or_self, nodetest_type_node, 0); - - return parse_relative_location_path(n); - } - - // else clause moved outside of if because of bogus warning 'control may reach end of non-void function being inlined' in gcc 4.0.1 - return parse_relative_location_path(0); - } - - // PathExpr ::= LocationPath - // | FilterExpr - // | FilterExpr '/' RelativeLocationPath - // | FilterExpr '//' RelativeLocationPath - xpath_ast_node* parse_path_expression() - { - // Clarification. - // PathExpr begins with either LocationPath or FilterExpr. - // FilterExpr begins with PrimaryExpr - // PrimaryExpr begins with '$' in case of it being a variable reference, - // '(' in case of it being an expression, string literal, number constant or - // function call. - if (_lexer.current() == lex_var_ref || _lexer.current() == lex_open_brace || - _lexer.current() == lex_quoted_string || _lexer.current() == lex_number || - _lexer.current() == lex_string) - { - if (_lexer.current() == lex_string) - { - // This is either a function call, or not - if not, we shall proceed with location path - const char_t* state = _lexer.state(); - - while (IS_CHARTYPE(*state, ct_space)) ++state; - - if (*state != '(') return parse_location_path(); + xpath_ast_node* rhs = parse_path_or_unary_expression(); + if (!rhs) return 0; - // This looks like a function call; however this still can be a node-test. Check it. - if (parse_node_test_type(_lexer.contents()) != nodetest_none) return parse_location_path(); - } - - xpath_ast_node* n = parse_filter_expression(); + binary_op_t nextop = binary_op_t::parse(_lexer); - if (_lexer.current() == lex_slash || _lexer.current() == lex_double_slash) - { - lexeme_t l = _lexer.current(); - _lexer.next(); - - if (l == lex_double_slash) - { - if (n->rettype() != xpath_type_node_set) throw_error("Step has to be applied to node set"); + while (nextop.asttype != ast_unknown && nextop.precedence > op.precedence) + { + rhs = parse_expression_rec(rhs, nextop.precedence); + if (!rhs) return 0; - n = new (alloc_node()) xpath_ast_node(ast_step, n, axis_descendant_or_self, nodetest_type_node, 0); - } - - // select from location path - return parse_relative_location_path(n); - } - - return n; - } - else return parse_location_path(); - } - - // UnionExpr ::= PathExpr | UnionExpr '|' PathExpr - xpath_ast_node* parse_union_expression() - { - xpath_ast_node* n = parse_path_expression(); - - while (_lexer.current() == lex_union) - { - _lexer.next(); - - xpath_ast_node* expr = parse_union_expression(); - - if (n->rettype() != xpath_type_node_set || expr->rettype() != xpath_type_node_set) - throw_error("Union operator has to be applied to node sets"); - - n = new (alloc_node()) xpath_ast_node(ast_op_union, xpath_type_node_set, n, expr); - } - - return n; - } - - // UnaryExpr ::= UnionExpr | '-' UnaryExpr - xpath_ast_node* parse_unary_expression() - { - if (_lexer.current() == lex_minus) - { - _lexer.next(); - - xpath_ast_node* expr = parse_unary_expression(); - - return new (alloc_node()) xpath_ast_node(ast_op_negate, xpath_type_number, expr); - } - else return parse_union_expression(); - } - - // MultiplicativeExpr ::= UnaryExpr - // | MultiplicativeExpr '*' UnaryExpr - // | MultiplicativeExpr 'div' UnaryExpr - // | MultiplicativeExpr 'mod' UnaryExpr - xpath_ast_node* parse_multiplicative_expression() - { - xpath_ast_node* n = parse_unary_expression(); - - while (_lexer.current() == lex_multiply || (_lexer.current() == lex_string && - (_lexer.contents() == PUGIXML_TEXT("mod") || _lexer.contents() == PUGIXML_TEXT("div")))) - { - ast_type_t op = _lexer.current() == lex_multiply ? ast_op_multiply : - _lexer.contents().begin[0] == 'd' ? ast_op_divide : ast_op_mod; - _lexer.next(); - - xpath_ast_node* expr = parse_unary_expression(); - - n = new (alloc_node()) xpath_ast_node(op, xpath_type_number, n, expr); - } - - return n; - } - - // AdditiveExpr ::= MultiplicativeExpr - // | AdditiveExpr '+' MultiplicativeExpr - // | AdditiveExpr '-' MultiplicativeExpr - xpath_ast_node* parse_additive_expression() - { - xpath_ast_node* n = parse_multiplicative_expression(); - - while (_lexer.current() == lex_plus || _lexer.current() == lex_minus) - { - lexeme_t l = _lexer.current(); - - _lexer.next(); - - xpath_ast_node* expr = parse_multiplicative_expression(); - - n = new (alloc_node()) xpath_ast_node(l == lex_plus ? ast_op_add : ast_op_subtract, xpath_type_number, n, expr); - } - - return n; - } - - // RelationalExpr ::= AdditiveExpr - // | RelationalExpr '<' AdditiveExpr - // | RelationalExpr '>' AdditiveExpr - // | RelationalExpr '<=' AdditiveExpr - // | RelationalExpr '>=' AdditiveExpr - xpath_ast_node* parse_relational_expression() - { - xpath_ast_node* n = parse_additive_expression(); - - while (_lexer.current() == lex_less || _lexer.current() == lex_less_or_equal || - _lexer.current() == lex_greater || _lexer.current() == lex_greater_or_equal) - { - lexeme_t l = _lexer.current(); - _lexer.next(); - - xpath_ast_node* expr = parse_additive_expression(); - - n = new (alloc_node()) xpath_ast_node(l == lex_less ? ast_op_less : l == lex_greater ? ast_op_greater : - l == lex_less_or_equal ? ast_op_less_or_equal : ast_op_greater_or_equal, xpath_type_boolean, n, expr); - } - - return n; - } - - // EqualityExpr ::= RelationalExpr - // | EqualityExpr '=' RelationalExpr - // | EqualityExpr '!=' RelationalExpr - xpath_ast_node* parse_equality_expression() - { - xpath_ast_node* n = parse_relational_expression(); - - while (_lexer.current() == lex_equal || _lexer.current() == lex_not_equal) - { - lexeme_t l = _lexer.current(); - - _lexer.next(); - - xpath_ast_node* expr = parse_relational_expression(); - - n = new (alloc_node()) xpath_ast_node(l == lex_equal ? ast_op_equal : ast_op_not_equal, xpath_type_boolean, n, expr); - } - - return n; - } - - // AndExpr ::= EqualityExpr | AndExpr 'and' EqualityExpr - xpath_ast_node* parse_and_expression() - { - xpath_ast_node* n = parse_equality_expression(); - - while (_lexer.current() == lex_string && _lexer.contents() == PUGIXML_TEXT("and")) - { - _lexer.next(); - - xpath_ast_node* expr = parse_equality_expression(); - - n = new (alloc_node()) xpath_ast_node(ast_op_and, xpath_type_boolean, n, expr); - } - - return n; - } + nextop = binary_op_t::parse(_lexer); + } - // OrExpr ::= AndExpr | OrExpr 'or' AndExpr - xpath_ast_node* parse_or_expression() - { - xpath_ast_node* n = parse_and_expression(); + if (op.asttype == ast_op_union && (lhs->rettype() != xpath_type_node_set || rhs->rettype() != xpath_type_node_set)) + return error("Union operator has to be applied to node sets"); - while (_lexer.current() == lex_string && _lexer.contents() == PUGIXML_TEXT("or")) - { - _lexer.next(); + lhs = alloc_node(op.asttype, op.rettype, lhs, rhs); + if (!lhs) return 0; - xpath_ast_node* expr = parse_and_expression(); + op = binary_op_t::parse(_lexer); + } - n = new (alloc_node()) xpath_ast_node(ast_op_or, xpath_type_boolean, n, expr); - } + return lhs; + } - return n; - } - // Expr ::= OrExpr - xpath_ast_node* parse_expression() - { - return parse_or_expression(); + // OrExpr ::= AndExpr | OrExpr 'or' AndExpr + // AndExpr ::= EqualityExpr | AndExpr 'and' EqualityExpr + // EqualityExpr ::= RelationalExpr + // | EqualityExpr '=' RelationalExpr + // | EqualityExpr '!=' RelationalExpr + // RelationalExpr ::= AdditiveExpr + // | RelationalExpr '<' AdditiveExpr + // | RelationalExpr '>' AdditiveExpr + // | RelationalExpr '<=' AdditiveExpr + // | RelationalExpr '>=' AdditiveExpr + // AdditiveExpr ::= MultiplicativeExpr + // | AdditiveExpr '+' MultiplicativeExpr + // | AdditiveExpr '-' MultiplicativeExpr + // MultiplicativeExpr ::= UnaryExpr + // | MultiplicativeExpr '*' UnaryExpr + // | MultiplicativeExpr 'div' UnaryExpr + // | MultiplicativeExpr 'mod' UnaryExpr + xpath_ast_node* parse_expression(int limit = 0) + { + xpath_ast_node* n = parse_path_or_unary_expression(); + if (!n) return 0; + + return parse_expression_rec(n, limit); } xpath_parser(const char_t* query, xpath_variable_set* variables, xpath_allocator* alloc, xpath_parse_result* result): _alloc(alloc), _lexer(query), _query(query), _variables(variables), _result(result) @@ -8922,174 +11856,179 @@ namespace xpath_ast_node* parse() { - xpath_ast_node* result = parse_expression(); - + xpath_ast_node* n = parse_expression(); + if (!n) return 0; + + // check if there are unparsed tokens left if (_lexer.current() != lex_eof) - { - // there are still unparsed tokens left, error - throw_error("Incorrect query"); - } - - return result; + return error("Incorrect query"); + + return n; } static xpath_ast_node* parse(const char_t* query, xpath_variable_set* variables, xpath_allocator* alloc, xpath_parse_result* result) { xpath_parser parser(query, variables, alloc, result); - #ifdef PUGIXML_NO_EXCEPTIONS - int error = setjmp(parser._error_handler); - - return (error == 0) ? parser.parse() : 0; - #else return parser.parse(); - #endif } }; - struct xpath_query_impl - { + struct xpath_query_impl + { static xpath_query_impl* create() { - void* memory = global_allocate(sizeof(xpath_query_impl)); + void* memory = xml_memory::allocate(sizeof(xpath_query_impl)); + if (!memory) return 0; - return new (memory) xpath_query_impl(); + return new (memory) xpath_query_impl(); } - static void destroy(void* ptr) + static void destroy(xpath_query_impl* impl) { - if (!ptr) return; - // free all allocated pages - static_cast(ptr)->alloc.release(); + impl->alloc.release(); // free allocator memory (with the first page) - global_deallocate(ptr); + xml_memory::deallocate(impl); } - xpath_query_impl(): root(0), alloc(&block) - { - block.next = 0; - } + xpath_query_impl(): root(0), alloc(&block, &oom), oom(false) + { + block.next = 0; + block.capacity = sizeof(block.data); + } - xpath_ast_node* root; - xpath_allocator alloc; - xpath_memory_block block; - }; + xpath_ast_node* root; + xpath_allocator alloc; + xpath_memory_block block; + bool oom; + }; - xpath_string evaluate_string_impl(xpath_query_impl* impl, const xpath_node& n, xpath_stack_data& sd) + PUGI__FN impl::xpath_ast_node* evaluate_node_set_prepare(xpath_query_impl* impl) { - if (!impl) return xpath_string(); + if (!impl) return 0; - #ifdef PUGIXML_NO_EXCEPTIONS - if (setjmp(sd.error_handler)) return xpath_string(); - #endif + if (impl->root->rettype() != xpath_type_node_set) + { + #ifdef PUGIXML_NO_EXCEPTIONS + return 0; + #else + xpath_parse_result res; + res.error = "Expression does not evaluate to node set"; - xpath_context c(n, 1, 1); + throw xpath_exception(res); + #endif + } - return impl->root->eval_string(c, sd.stack); + return impl->root; } -} +PUGI__NS_END namespace pugi { #ifndef PUGIXML_NO_EXCEPTIONS - xpath_exception::xpath_exception(const xpath_parse_result& result): _result(result) + PUGI__FN xpath_exception::xpath_exception(const xpath_parse_result& result_): _result(result_) { - assert(result.error); + assert(_result.error); } - - const char* xpath_exception::what() const throw() + + PUGI__FN const char* xpath_exception::what() const throw() { return _result.error; } - const xpath_parse_result& xpath_exception::result() const + PUGI__FN const xpath_parse_result& xpath_exception::result() const { return _result; } #endif - - xpath_node::xpath_node() + + PUGI__FN xpath_node::xpath_node() { } - - xpath_node::xpath_node(const xml_node& node): _node(node) + + PUGI__FN xpath_node::xpath_node(const xml_node& node_): _node(node_) { } - - xpath_node::xpath_node(const xml_attribute& attribute, const xml_node& parent): _node(attribute ? parent : xml_node()), _attribute(attribute) + + PUGI__FN xpath_node::xpath_node(const xml_attribute& attribute_, const xml_node& parent_): _node(attribute_ ? parent_ : xml_node()), _attribute(attribute_) { } - xml_node xpath_node::node() const + PUGI__FN xml_node xpath_node::node() const { return _attribute ? xml_node() : _node; } - - xml_attribute xpath_node::attribute() const + + PUGI__FN xml_attribute xpath_node::attribute() const { return _attribute; } - - xml_node xpath_node::parent() const + + PUGI__FN xml_node xpath_node::parent() const { return _attribute ? _node : _node.parent(); } - xpath_node::operator xpath_node::unspecified_bool_type() const + PUGI__FN static void unspecified_bool_xpath_node(xpath_node***) { - return (_node || _attribute) ? &xpath_node::_node : 0; } - - bool xpath_node::operator!() const + + PUGI__FN xpath_node::operator xpath_node::unspecified_bool_type() const + { + return (_node || _attribute) ? unspecified_bool_xpath_node : 0; + } + + PUGI__FN bool xpath_node::operator!() const { return !(_node || _attribute); } - bool xpath_node::operator==(const xpath_node& n) const + PUGI__FN bool xpath_node::operator==(const xpath_node& n) const { return _node == n._node && _attribute == n._attribute; } - - bool xpath_node::operator!=(const xpath_node& n) const + + PUGI__FN bool xpath_node::operator!=(const xpath_node& n) const { return _node != n._node || _attribute != n._attribute; } #ifdef __BORLANDC__ - bool operator&&(const xpath_node& lhs, bool rhs) + PUGI__FN bool operator&&(const xpath_node& lhs, bool rhs) { return (bool)lhs && rhs; } - bool operator||(const xpath_node& lhs, bool rhs) + PUGI__FN bool operator||(const xpath_node& lhs, bool rhs) { return (bool)lhs || rhs; } #endif - void xpath_node_set::_assign(const_iterator begin, const_iterator end) + PUGI__FN void xpath_node_set::_assign(const_iterator begin_, const_iterator end_, type_t type_) { - assert(begin <= end); + assert(begin_ <= end_); - size_t size = static_cast(end - begin); + size_t size_ = static_cast(end_ - begin_); - if (size <= 1) + if (size_ <= 1) { // deallocate old buffer - if (_begin != &_storage) global_deallocate(_begin); + if (_begin != &_storage) impl::xml_memory::deallocate(_begin); // use internal buffer - if (begin != end) _storage = *begin; + if (begin_ != end_) _storage = *begin_; _begin = &_storage; - _end = &_storage + size; + _end = &_storage + size_; + _type = type_; } else { // make heap copy - xpath_node* storage = static_cast(global_allocate(size * sizeof(xpath_node))); + xpath_node* storage = static_cast(impl::xml_memory::allocate(size_ * sizeof(xpath_node))); if (!storage) { @@ -9100,248 +12039,374 @@ namespace pugi #endif } - memcpy(storage, begin, size * sizeof(xpath_node)); - + memcpy(storage, begin_, size_ * sizeof(xpath_node)); + // deallocate old buffer - if (_begin != &_storage) global_deallocate(_begin); + if (_begin != &_storage) impl::xml_memory::deallocate(_begin); // finalize _begin = storage; - _end = storage + size; + _end = storage + size_; + _type = type_; } } - xpath_node_set::xpath_node_set(): _type(type_unsorted), _begin(&_storage), _end(&_storage) +#ifdef PUGIXML_HAS_MOVE + PUGI__FN void xpath_node_set::_move(xpath_node_set& rhs) PUGIXML_NOEXCEPT + { + _type = rhs._type; + _storage = rhs._storage; + _begin = (rhs._begin == &rhs._storage) ? &_storage : rhs._begin; + _end = _begin + (rhs._end - rhs._begin); + + rhs._type = type_unsorted; + rhs._begin = &rhs._storage; + rhs._end = rhs._begin; + } +#endif + + PUGI__FN xpath_node_set::xpath_node_set(): _type(type_unsorted), _begin(&_storage), _end(&_storage) { } - xpath_node_set::xpath_node_set(const_iterator begin, const_iterator end, type_t type): _type(type), _begin(&_storage), _end(&_storage) + PUGI__FN xpath_node_set::xpath_node_set(const_iterator begin_, const_iterator end_, type_t type_): _type(type_unsorted), _begin(&_storage), _end(&_storage) { - _assign(begin, end); + _assign(begin_, end_, type_); } - xpath_node_set::~xpath_node_set() + PUGI__FN xpath_node_set::~xpath_node_set() { - if (_begin != &_storage) global_deallocate(_begin); + if (_begin != &_storage) + impl::xml_memory::deallocate(_begin); } - - xpath_node_set::xpath_node_set(const xpath_node_set& ns): _type(ns._type), _begin(&_storage), _end(&_storage) + + PUGI__FN xpath_node_set::xpath_node_set(const xpath_node_set& ns): _type(type_unsorted), _begin(&_storage), _end(&_storage) { - _assign(ns._begin, ns._end); + _assign(ns._begin, ns._end, ns._type); } - - xpath_node_set& xpath_node_set::operator=(const xpath_node_set& ns) + + PUGI__FN xpath_node_set& xpath_node_set::operator=(const xpath_node_set& ns) { if (this == &ns) return *this; - - _type = ns._type; - _assign(ns._begin, ns._end); + + _assign(ns._begin, ns._end, ns._type); + + return *this; + } + +#ifdef PUGIXML_HAS_MOVE + PUGI__FN xpath_node_set::xpath_node_set(xpath_node_set&& rhs) PUGIXML_NOEXCEPT: _type(type_unsorted), _begin(&_storage), _end(&_storage) + { + _move(rhs); + } + + PUGI__FN xpath_node_set& xpath_node_set::operator=(xpath_node_set&& rhs) PUGIXML_NOEXCEPT + { + if (this == &rhs) return *this; + + if (_begin != &_storage) + impl::xml_memory::deallocate(_begin); + + _move(rhs); return *this; } +#endif - xpath_node_set::type_t xpath_node_set::type() const + PUGI__FN xpath_node_set::type_t xpath_node_set::type() const { return _type; } - - size_t xpath_node_set::size() const + + PUGI__FN size_t xpath_node_set::size() const { return _end - _begin; } - - bool xpath_node_set::empty() const + + PUGI__FN bool xpath_node_set::empty() const { return _begin == _end; } - - const xpath_node& xpath_node_set::operator[](size_t index) const + + PUGI__FN const xpath_node& xpath_node_set::operator[](size_t index) const { assert(index < size()); return _begin[index]; } - xpath_node_set::const_iterator xpath_node_set::begin() const + PUGI__FN xpath_node_set::const_iterator xpath_node_set::begin() const { return _begin; } - - xpath_node_set::const_iterator xpath_node_set::end() const + + PUGI__FN xpath_node_set::const_iterator xpath_node_set::end() const { return _end; } - - void xpath_node_set::sort(bool reverse) + + PUGI__FN void xpath_node_set::sort(bool reverse) + { + _type = impl::xpath_sort(_begin, _end, _type, reverse); + } + + PUGI__FN xpath_node xpath_node_set::first() const { - _type = xpath_sort(_begin, _end, _type, reverse); + return impl::xpath_first(_begin, _end, _type); } - xpath_node xpath_node_set::first() const + PUGI__FN xpath_parse_result::xpath_parse_result(): error("Internal error"), offset(0) { - return xpath_first(_begin, _end, _type); } - xpath_parse_result::xpath_parse_result(): error("Internal error"), offset(0) - { - } + PUGI__FN xpath_parse_result::operator bool() const + { + return error == 0; + } - xpath_parse_result::operator bool() const - { - return error == 0; - } - const char* xpath_parse_result::description() const + PUGI__FN const char* xpath_parse_result::description() const { return error ? error : "No error"; } - xpath_variable::xpath_variable() - { - } + PUGI__FN xpath_variable::xpath_variable(xpath_value_type type_): _type(type_), _next(0) + { + } - const char_t* xpath_variable::name() const + PUGI__FN const char_t* xpath_variable::name() const { switch (_type) { case xpath_type_node_set: - return static_cast(this)->name; + return static_cast(this)->name; case xpath_type_number: - return static_cast(this)->name; + return static_cast(this)->name; case xpath_type_string: - return static_cast(this)->name; + return static_cast(this)->name; case xpath_type_boolean: - return static_cast(this)->name; + return static_cast(this)->name; default: - assert(!"Invalid variable type"); + assert(false && "Invalid variable type"); // unreachable return 0; } } - xpath_value_type xpath_variable::type() const + PUGI__FN xpath_value_type xpath_variable::type() const { return _type; } - bool xpath_variable::get_boolean() const + PUGI__FN bool xpath_variable::get_boolean() const { - return (_type == xpath_type_boolean) ? static_cast(this)->value : false; + return (_type == xpath_type_boolean) ? static_cast(this)->value : false; } - double xpath_variable::get_number() const + PUGI__FN double xpath_variable::get_number() const { - return (_type == xpath_type_number) ? static_cast(this)->value : gen_nan(); + return (_type == xpath_type_number) ? static_cast(this)->value : impl::gen_nan(); } - const char_t* xpath_variable::get_string() const + PUGI__FN const char_t* xpath_variable::get_string() const { - const char_t* value = (_type == xpath_type_string) ? static_cast(this)->value : 0; + const char_t* value = (_type == xpath_type_string) ? static_cast(this)->value : 0; return value ? value : PUGIXML_TEXT(""); } - const xpath_node_set& xpath_variable::get_node_set() const + PUGI__FN const xpath_node_set& xpath_variable::get_node_set() const { - return (_type == xpath_type_node_set) ? static_cast(this)->value : dummy_node_set; + return (_type == xpath_type_node_set) ? static_cast(this)->value : impl::dummy_node_set; } - bool xpath_variable::set(bool value) + PUGI__FN bool xpath_variable::set(bool value) { if (_type != xpath_type_boolean) return false; - static_cast(this)->value = value; + static_cast(this)->value = value; return true; } - bool xpath_variable::set(double value) + PUGI__FN bool xpath_variable::set(double value) { if (_type != xpath_type_number) return false; - static_cast(this)->value = value; + static_cast(this)->value = value; return true; } - bool xpath_variable::set(const char_t* value) + PUGI__FN bool xpath_variable::set(const char_t* value) { if (_type != xpath_type_string) return false; - xpath_variable_string* var = static_cast(this); + impl::xpath_variable_string* var = static_cast(this); // duplicate string - size_t size = (strlength(value) + 1) * sizeof(char_t); + size_t size = (impl::strlength(value) + 1) * sizeof(char_t); - char_t* copy = static_cast(global_allocate(size)); + char_t* copy = static_cast(impl::xml_memory::allocate(size)); if (!copy) return false; memcpy(copy, value, size); // replace old string - if (var->value) global_deallocate(var->value); + if (var->value) impl::xml_memory::deallocate(var->value); var->value = copy; return true; } - bool xpath_variable::set(const xpath_node_set& value) + PUGI__FN bool xpath_variable::set(const xpath_node_set& value) { if (_type != xpath_type_node_set) return false; - static_cast(this)->value = value; + static_cast(this)->value = value; return true; } - xpath_variable_set::xpath_variable_set() + PUGI__FN xpath_variable_set::xpath_variable_set() + { + for (size_t i = 0; i < sizeof(_data) / sizeof(_data[0]); ++i) + _data[i] = 0; + } + + PUGI__FN xpath_variable_set::~xpath_variable_set() + { + for (size_t i = 0; i < sizeof(_data) / sizeof(_data[0]); ++i) + _destroy(_data[i]); + } + + PUGI__FN xpath_variable_set::xpath_variable_set(const xpath_variable_set& rhs) + { + for (size_t i = 0; i < sizeof(_data) / sizeof(_data[0]); ++i) + _data[i] = 0; + + _assign(rhs); + } + + PUGI__FN xpath_variable_set& xpath_variable_set::operator=(const xpath_variable_set& rhs) { - for (size_t i = 0; i < sizeof(_data) / sizeof(_data[0]); ++i) _data[i] = 0; + if (this == &rhs) return *this; + + _assign(rhs); + + return *this; } - xpath_variable_set::~xpath_variable_set() +#ifdef PUGIXML_HAS_MOVE + PUGI__FN xpath_variable_set::xpath_variable_set(xpath_variable_set&& rhs) PUGIXML_NOEXCEPT { for (size_t i = 0; i < sizeof(_data) / sizeof(_data[0]); ++i) { - xpath_variable* var = _data[i]; + _data[i] = rhs._data[i]; + rhs._data[i] = 0; + } + } - while (var) - { - xpath_variable* next = var->_next; + PUGI__FN xpath_variable_set& xpath_variable_set::operator=(xpath_variable_set&& rhs) PUGIXML_NOEXCEPT + { + for (size_t i = 0; i < sizeof(_data) / sizeof(_data[0]); ++i) + { + _destroy(_data[i]); - delete_xpath_variable(var->_type, var); + _data[i] = rhs._data[i]; + rhs._data[i] = 0; + } - var = next; - } + return *this; + } +#endif + + PUGI__FN void xpath_variable_set::_assign(const xpath_variable_set& rhs) + { + xpath_variable_set temp; + + for (size_t i = 0; i < sizeof(_data) / sizeof(_data[0]); ++i) + if (rhs._data[i] && !_clone(rhs._data[i], &temp._data[i])) + return; + + _swap(temp); + } + + PUGI__FN void xpath_variable_set::_swap(xpath_variable_set& rhs) + { + for (size_t i = 0; i < sizeof(_data) / sizeof(_data[0]); ++i) + { + xpath_variable* chain = _data[i]; + + _data[i] = rhs._data[i]; + rhs._data[i] = chain; } } - xpath_variable* xpath_variable_set::find(const char_t* name) const + PUGI__FN xpath_variable* xpath_variable_set::_find(const char_t* name) const { const size_t hash_size = sizeof(_data) / sizeof(_data[0]); - size_t hash = hash_string(name) % hash_size; + size_t hash = impl::hash_string(name) % hash_size; // look for existing variable for (xpath_variable* var = _data[hash]; var; var = var->_next) - if (strequal(var->name(), name)) + if (impl::strequal(var->name(), name)) return var; return 0; } - xpath_variable* xpath_variable_set::add(const char_t* name, xpath_value_type type) + PUGI__FN bool xpath_variable_set::_clone(xpath_variable* var, xpath_variable** out_result) + { + xpath_variable* last = 0; + + while (var) + { + // allocate storage for new variable + xpath_variable* nvar = impl::new_xpath_variable(var->_type, var->name()); + if (!nvar) return false; + + // link the variable to the result immediately to handle failures gracefully + if (last) + last->_next = nvar; + else + *out_result = nvar; + + last = nvar; + + // copy the value; this can fail due to out-of-memory conditions + if (!impl::copy_xpath_variable(nvar, var)) return false; + + var = var->_next; + } + + return true; + } + + PUGI__FN void xpath_variable_set::_destroy(xpath_variable* var) + { + while (var) + { + xpath_variable* next = var->_next; + + impl::delete_xpath_variable(var->_type, var); + + var = next; + } + } + + PUGI__FN xpath_variable* xpath_variable_set::add(const char_t* name, xpath_value_type type) { const size_t hash_size = sizeof(_data) / sizeof(_data[0]); - size_t hash = hash_string(name) % hash_size; + size_t hash = impl::hash_string(name) % hash_size; // look for existing variable for (xpath_variable* var = _data[hash]; var; var = var->_next) - if (strequal(var->name(), name)) + if (impl::strequal(var->name(), name)) return var->type() == type ? var : 0; // add new variable - xpath_variable* result = new_xpath_variable(type, name); + xpath_variable* result = impl::new_xpath_variable(type, name); if (result) { - result->_type = type; result->_next = _data[hash]; _data[hash] = result; @@ -9350,208 +12415,363 @@ namespace pugi return result; } - bool xpath_variable_set::set(const char_t* name, bool value) + PUGI__FN bool xpath_variable_set::set(const char_t* name, bool value) { xpath_variable* var = add(name, xpath_type_boolean); return var ? var->set(value) : false; } - bool xpath_variable_set::set(const char_t* name, double value) + PUGI__FN bool xpath_variable_set::set(const char_t* name, double value) { xpath_variable* var = add(name, xpath_type_number); return var ? var->set(value) : false; } - bool xpath_variable_set::set(const char_t* name, const char_t* value) + PUGI__FN bool xpath_variable_set::set(const char_t* name, const char_t* value) { xpath_variable* var = add(name, xpath_type_string); return var ? var->set(value) : false; } - bool xpath_variable_set::set(const char_t* name, const xpath_node_set& value) + PUGI__FN bool xpath_variable_set::set(const char_t* name, const xpath_node_set& value) { xpath_variable* var = add(name, xpath_type_node_set); return var ? var->set(value) : false; } - xpath_variable* xpath_variable_set::get(const char_t* name) + PUGI__FN xpath_variable* xpath_variable_set::get(const char_t* name) { - return find(name); + return _find(name); } - const xpath_variable* xpath_variable_set::get(const char_t* name) const + PUGI__FN const xpath_variable* xpath_variable_set::get(const char_t* name) const { - return find(name); + return _find(name); } - xpath_query::xpath_query(const char_t* query, xpath_variable_set* variables): _impl(0) + PUGI__FN xpath_query::xpath_query(const char_t* query, xpath_variable_set* variables): _impl(0) { - xpath_query_impl* impl = xpath_query_impl::create(); + impl::xpath_query_impl* qimpl = impl::xpath_query_impl::create(); - if (!impl) + if (!qimpl) { #ifdef PUGIXML_NO_EXCEPTIONS _result.error = "Out of memory"; - #else + #else throw std::bad_alloc(); #endif } else { - buffer_holder impl_holder(impl, xpath_query_impl::destroy); + using impl::auto_deleter; // MSVC7 workaround + auto_deleter impl(qimpl, impl::xpath_query_impl::destroy); - impl->root = xpath_parser::parse(query, variables, &impl->alloc, &_result); + qimpl->root = impl::xpath_parser::parse(query, variables, &qimpl->alloc, &_result); - if (impl->root) + if (qimpl->root) { - _impl = static_cast(impl_holder.release()); + qimpl->root->optimize(&qimpl->alloc); + + _impl = impl.release(); _result.error = 0; } + else + { + #ifdef PUGIXML_NO_EXCEPTIONS + if (qimpl->oom) _result.error = "Out of memory"; + #else + if (qimpl->oom) throw std::bad_alloc(); + throw xpath_exception(_result); + #endif + } } } - xpath_query::~xpath_query() + PUGI__FN xpath_query::xpath_query(): _impl(0) { - xpath_query_impl::destroy(_impl); } - xpath_value_type xpath_query::return_type() const + PUGI__FN xpath_query::~xpath_query() + { + if (_impl) + impl::xpath_query_impl::destroy(static_cast(_impl)); + } + +#ifdef PUGIXML_HAS_MOVE + PUGI__FN xpath_query::xpath_query(xpath_query&& rhs) PUGIXML_NOEXCEPT + { + _impl = rhs._impl; + _result = rhs._result; + rhs._impl = 0; + rhs._result = xpath_parse_result(); + } + + PUGI__FN xpath_query& xpath_query::operator=(xpath_query&& rhs) PUGIXML_NOEXCEPT + { + if (this == &rhs) return *this; + + if (_impl) + impl::xpath_query_impl::destroy(static_cast(_impl)); + + _impl = rhs._impl; + _result = rhs._result; + rhs._impl = 0; + rhs._result = xpath_parse_result(); + + return *this; + } +#endif + + PUGI__FN xpath_value_type xpath_query::return_type() const { if (!_impl) return xpath_type_none; - return static_cast(_impl)->root->rettype(); + return static_cast(_impl)->root->rettype(); } - bool xpath_query::evaluate_boolean(const xpath_node& n) const + PUGI__FN bool xpath_query::evaluate_boolean(const xpath_node& n) const { if (!_impl) return false; - - xpath_context c(n, 1, 1); - xpath_stack_data sd; - #ifdef PUGIXML_NO_EXCEPTIONS - if (setjmp(sd.error_handler)) return false; - #endif - - return static_cast(_impl)->root->eval_boolean(c, sd.stack); + impl::xpath_context c(n, 1, 1); + impl::xpath_stack_data sd; + + bool r = static_cast(_impl)->root->eval_boolean(c, sd.stack); + + if (sd.oom) + { + #ifdef PUGIXML_NO_EXCEPTIONS + return false; + #else + throw std::bad_alloc(); + #endif + } + + return r; } - - double xpath_query::evaluate_number(const xpath_node& n) const + + PUGI__FN double xpath_query::evaluate_number(const xpath_node& n) const { - if (!_impl) return gen_nan(); - - xpath_context c(n, 1, 1); - xpath_stack_data sd; + if (!_impl) return impl::gen_nan(); - #ifdef PUGIXML_NO_EXCEPTIONS - if (setjmp(sd.error_handler)) return gen_nan(); - #endif + impl::xpath_context c(n, 1, 1); + impl::xpath_stack_data sd; + + double r = static_cast(_impl)->root->eval_number(c, sd.stack); + + if (sd.oom) + { + #ifdef PUGIXML_NO_EXCEPTIONS + return impl::gen_nan(); + #else + throw std::bad_alloc(); + #endif + } - return static_cast(_impl)->root->eval_number(c, sd.stack); + return r; } #ifndef PUGIXML_NO_STL - string_t xpath_query::evaluate_string(const xpath_node& n) const + PUGI__FN string_t xpath_query::evaluate_string(const xpath_node& n) const { - xpath_stack_data sd; + if (!_impl) return string_t(); + + impl::xpath_context c(n, 1, 1); + impl::xpath_stack_data sd; + + impl::xpath_string r = static_cast(_impl)->root->eval_string(c, sd.stack); + + if (sd.oom) + { + #ifdef PUGIXML_NO_EXCEPTIONS + return string_t(); + #else + throw std::bad_alloc(); + #endif + } - return evaluate_string_impl(static_cast(_impl), n, sd).c_str(); + return string_t(r.c_str(), r.length()); } #endif - size_t xpath_query::evaluate_string(char_t* buffer, size_t capacity, const xpath_node& n) const + PUGI__FN size_t xpath_query::evaluate_string(char_t* buffer, size_t capacity, const xpath_node& n) const { - xpath_stack_data sd; + impl::xpath_context c(n, 1, 1); + impl::xpath_stack_data sd; - xpath_string r = evaluate_string_impl(static_cast(_impl), n, sd); + impl::xpath_string r = _impl ? static_cast(_impl)->root->eval_string(c, sd.stack) : impl::xpath_string(); + + if (sd.oom) + { + #ifdef PUGIXML_NO_EXCEPTIONS + r = impl::xpath_string(); + #else + throw std::bad_alloc(); + #endif + } size_t full_size = r.length() + 1; - + if (capacity > 0) - { - size_t size = (full_size < capacity) ? full_size : capacity; - assert(size > 0); - - memcpy(buffer, r.c_str(), (size - 1) * sizeof(char_t)); - buffer[size - 1] = 0; - } - + { + size_t size = (full_size < capacity) ? full_size : capacity; + assert(size > 0); + + memcpy(buffer, r.c_str(), (size - 1) * sizeof(char_t)); + buffer[size - 1] = 0; + } + return full_size; } - xpath_node_set xpath_query::evaluate_node_set(const xpath_node& n) const + PUGI__FN xpath_node_set xpath_query::evaluate_node_set(const xpath_node& n) const { - if (!_impl) return xpath_node_set(); + impl::xpath_ast_node* root = impl::evaluate_node_set_prepare(static_cast(_impl)); + if (!root) return xpath_node_set(); - xpath_ast_node* root = static_cast(_impl)->root; + impl::xpath_context c(n, 1, 1); + impl::xpath_stack_data sd; - if (root->rettype() != xpath_type_node_set) + impl::xpath_node_set_raw r = root->eval_node_set(c, sd.stack, impl::nodeset_eval_all); + + if (sd.oom) { #ifdef PUGIXML_NO_EXCEPTIONS return xpath_node_set(); #else - xpath_parse_result result; - result.error = "Expression does not evaluate to node set"; - - throw xpath_exception(result); + throw std::bad_alloc(); #endif } - - xpath_context c(n, 1, 1); - xpath_stack_data sd; - #ifdef PUGIXML_NO_EXCEPTIONS - if (setjmp(sd.error_handler)) return xpath_node_set(); - #endif + return xpath_node_set(r.begin(), r.end(), r.type()); + } - xpath_node_set_raw r = root->eval_node_set(c, sd.stack); + PUGI__FN xpath_node xpath_query::evaluate_node(const xpath_node& n) const + { + impl::xpath_ast_node* root = impl::evaluate_node_set_prepare(static_cast(_impl)); + if (!root) return xpath_node(); - return xpath_node_set(r.begin(), r.end(), r.type()); + impl::xpath_context c(n, 1, 1); + impl::xpath_stack_data sd; + + impl::xpath_node_set_raw r = root->eval_node_set(c, sd.stack, impl::nodeset_eval_first); + + if (sd.oom) + { + #ifdef PUGIXML_NO_EXCEPTIONS + return xpath_node(); + #else + throw std::bad_alloc(); + #endif + } + + return r.first(); } - const xpath_parse_result& xpath_query::result() const + PUGI__FN const xpath_parse_result& xpath_query::result() const { return _result; } - xpath_query::operator xpath_query::unspecified_bool_type() const + PUGI__FN static void unspecified_bool_xpath_query(xpath_query***) + { + } + + PUGI__FN xpath_query::operator xpath_query::unspecified_bool_type() const { - return _impl ? &xpath_query::_impl : 0; + return _impl ? unspecified_bool_xpath_query : 0; } - bool xpath_query::operator!() const + PUGI__FN bool xpath_query::operator!() const { return !_impl; } - xpath_node xml_node::select_single_node(const char_t* query, xpath_variable_set* variables) const + PUGI__FN xpath_node xml_node::select_node(const char_t* query, xpath_variable_set* variables) const { xpath_query q(query, variables); - return select_single_node(q); + return q.evaluate_node(*this); } - xpath_node xml_node::select_single_node(const xpath_query& query) const + PUGI__FN xpath_node xml_node::select_node(const xpath_query& query) const { - xpath_node_set s = query.evaluate_node_set(*this); - return s.empty() ? xpath_node() : s.first(); + return query.evaluate_node(*this); } - xpath_node_set xml_node::select_nodes(const char_t* query, xpath_variable_set* variables) const + PUGI__FN xpath_node_set xml_node::select_nodes(const char_t* query, xpath_variable_set* variables) const { xpath_query q(query, variables); - return select_nodes(q); + return q.evaluate_node_set(*this); } - xpath_node_set xml_node::select_nodes(const xpath_query& query) const + PUGI__FN xpath_node_set xml_node::select_nodes(const xpath_query& query) const { return query.evaluate_node_set(*this); } + + PUGI__FN xpath_node xml_node::select_single_node(const char_t* query, xpath_variable_set* variables) const + { + xpath_query q(query, variables); + return q.evaluate_node(*this); + } + + PUGI__FN xpath_node xml_node::select_single_node(const xpath_query& query) const + { + return query.evaluate_node(*this); + } } #endif +#ifdef __BORLANDC__ +# pragma option pop +#endif + +// Intel C++ does not properly keep warning state for function templates, +// so popping warning state at the end of translation unit leads to warnings in the middle. +#if defined(_MSC_VER) && !defined(__INTEL_COMPILER) +# pragma warning(pop) +#endif + +#if defined(_MSC_VER) && defined(__c2__) +# pragma clang diagnostic pop +#endif + +// Undefine all local macros (makes sure we're not leaking macros in header-only mode) +#undef PUGI__NO_INLINE +#undef PUGI__UNLIKELY +#undef PUGI__STATIC_ASSERT +#undef PUGI__DMC_VOLATILE +#undef PUGI__UNSIGNED_OVERFLOW +#undef PUGI__MSVC_CRT_VERSION +#undef PUGI__SNPRINTF +#undef PUGI__NS_BEGIN +#undef PUGI__NS_END +#undef PUGI__FN +#undef PUGI__FN_NO_INLINE +#undef PUGI__GETHEADER_IMPL +#undef PUGI__GETPAGE_IMPL +#undef PUGI__GETPAGE +#undef PUGI__NODETYPE +#undef PUGI__IS_CHARTYPE_IMPL +#undef PUGI__IS_CHARTYPE +#undef PUGI__IS_CHARTYPEX +#undef PUGI__ENDSWITH +#undef PUGI__SKIPWS +#undef PUGI__OPTSET +#undef PUGI__PUSHNODE +#undef PUGI__POPNODE +#undef PUGI__SCANFOR +#undef PUGI__SCANWHILE +#undef PUGI__SCANWHILE_UNROLL +#undef PUGI__ENDSEG +#undef PUGI__THROW_ERROR +#undef PUGI__CHECK_ERROR + +#endif + /** - * Copyright (c) 2006-2010 Arseny Kapoulkine + * Copyright (c) 2006-2018 Arseny Kapoulkine * * Permission is hereby granted, free of charge, to any person * obtaining a copy of this software and associated documentation @@ -9564,7 +12784,7 @@ namespace pugi * * The above copyright notice and this permission notice shall be * included in all copies or substantial portions of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND @@ -9574,3 +12794,4 @@ namespace pugi * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR * OTHER DEALINGS IN THE SOFTWARE. */ + diff --git a/libpugixml/pugixml.hpp b/libpugixml/pugixml.hpp index 4360996..696cad7 100755 --- a/libpugixml/pugixml.hpp +++ b/libpugixml/pugixml.hpp @@ -1,7 +1,7 @@ /** - * pugixml parser - version 1.0 + * pugixml parser - version 1.9 * -------------------------------------------------------- - * Copyright (C) 2006-2010, by Arseny Kapoulkine (arseny.kapoulkine@gmail.com) + * Copyright (C) 2006-2018, by Arseny Kapoulkine (arseny.kapoulkine@gmail.com) * Report bugs and download new versions at http://pugixml.org/ * * This library is distributed under the MIT License. See notice at the end @@ -11,38 +11,30 @@ * Copyright (C) 2003, by Kristen Wegner (kristen@tima.net) */ +#ifndef PUGIXML_VERSION +// Define version macro; evaluates to major * 100 + minor so that it's safe to use in less-than comparisons +# define PUGIXML_VERSION 190 +#endif + +// Include user configuration file (this can define various configuration macros) +#include "pugiconfig.hpp" + #ifndef HEADER_PUGIXML_HPP #define HEADER_PUGIXML_HPP -#include "pugiconfig.hpp" +// Include stddef.h for size_t and ptrdiff_t +#include -#ifndef PUGIXML_NO_STL -namespace std -{ - struct bidirectional_iterator_tag; - -#ifdef __SUNPRO_CC - // Sun C++ compiler has a bug which forces template argument names in forward declarations to be the same as in actual definitions - template class allocator; - template struct char_traits; - template class basic_istream; - template class basic_ostream; - template class basic_string; -#else - // Borland C++ compiler has a bug which forces template argument names in forward declarations to be the same as in actual definitions - template class allocator; - template struct char_traits; - template class basic_istream; - template class basic_ostream; - template class basic_string; +// Include exception header for XPath +#if !defined(PUGIXML_NO_XPATH) && !defined(PUGIXML_NO_EXCEPTIONS) +# include #endif - // Digital Mars compiler has a bug which requires a forward declaration for explicit instantiation (otherwise type selection is messed up later, producing link errors) - // Also note that we have to declare char_traits as a class here, since it's defined that way -#ifdef __DMC__ - template <> class char_traits; -#endif -} +// Include STL headers +#ifndef PUGIXML_NO_STL +# include +# include +# include #endif // Macro for deprecated features @@ -56,27 +48,67 @@ namespace std # endif #endif -// Include exception header for XPath -#if !defined(PUGIXML_NO_XPATH) && !defined(PUGIXML_NO_EXCEPTIONS) -# include -#endif - // If no API is defined, assume default #ifndef PUGIXML_API -# define PUGIXML_API +# define PUGIXML_API #endif // If no API for classes is defined, assume default #ifndef PUGIXML_CLASS -# define PUGIXML_CLASS PUGIXML_API +# define PUGIXML_CLASS PUGIXML_API #endif // If no API for functions is defined, assume default #ifndef PUGIXML_FUNCTION -# define PUGIXML_FUNCTION PUGIXML_API +# define PUGIXML_FUNCTION PUGIXML_API #endif -#include +// If the platform is known to have long long support, enable long long functions +#ifndef PUGIXML_HAS_LONG_LONG +# if __cplusplus >= 201103 +# define PUGIXML_HAS_LONG_LONG +# elif defined(_MSC_VER) && _MSC_VER >= 1400 +# define PUGIXML_HAS_LONG_LONG +# endif +#endif + +// If the platform is known to have move semantics support, compile move ctor/operator implementation +#ifndef PUGIXML_HAS_MOVE +# if __cplusplus >= 201103 +# define PUGIXML_HAS_MOVE +# elif defined(_MSC_VER) && _MSC_VER >= 1600 +# define PUGIXML_HAS_MOVE +# endif +#endif + +// If C++ is 2011 or higher, add 'noexcept' specifiers +#ifndef PUGIXML_NOEXCEPT +# if __cplusplus >= 201103 +# define PUGIXML_NOEXCEPT noexcept +# elif defined(_MSC_VER) && _MSC_VER >= 1900 +# define PUGIXML_NOEXCEPT noexcept +# else +# define PUGIXML_NOEXCEPT +# endif +#endif + +// Some functions can not be noexcept in compact mode +#ifdef PUGIXML_COMPACT +# define PUGIXML_NOEXCEPT_IF_NOT_COMPACT +#else +# define PUGIXML_NOEXCEPT_IF_NOT_COMPACT PUGIXML_NOEXCEPT +#endif + +// If C++ is 2011 or higher, add 'override' qualifiers +#ifndef PUGIXML_OVERRIDE +# if __cplusplus >= 201103 +# define PUGIXML_OVERRIDE override +# elif defined(_MSC_VER) && _MSC_VER >= 1700 +# define PUGIXML_OVERRIDE override +# else +# define PUGIXML_OVERRIDE +# endif +#endif // Character interface macros #ifdef PUGIXML_WCHAR_MODE @@ -104,7 +136,7 @@ namespace pugi // Tree node types enum xml_node_type { - node_null, // Empty (null) node handle + node_null, // Empty (null) node handle node_document, // A document tree's absolute root node_element, // Element tag, i.e. '' node_pcdata, // Plain character data, i.e. 'text' @@ -112,13 +144,13 @@ namespace pugi node_comment, // Comment tag, i.e. '' node_pi, // Processing instruction, i.e. '' node_declaration, // Document declaration, i.e. '' - node_doctype // Document type declaration, i.e. '' + node_doctype // Document type declaration, i.e. '' }; // Parsing options // Minimal parsing mode (equivalent to turning all other flags off). - // Only elements and PCDATA sections are added to the DOM tree, no text conversions are performed. + // Only elements and PCDATA sections are added to the DOM tree, no text conversions are performed. const unsigned int parse_minimal = 0x0000; // This flag determines if processing instructions (node_pi) are added to the DOM tree. This flag is off by default. @@ -131,7 +163,7 @@ namespace pugi const unsigned int parse_cdata = 0x0004; // This flag determines if plain character data (node_pcdata) that consist only of whitespace are added to the DOM tree. - // This flag is off by default; turning it on usually results in slower parsing and more memory consumption. + // This flag is off by default; turning it on usually results in slower parsing and more memory consumption. const unsigned int parse_ws_pcdata = 0x0008; // This flag determines if character and entity references are expanded during parsing. This flag is on by default. @@ -139,72 +171,107 @@ namespace pugi // This flag determines if EOL characters are normalized (converted to #xA) during parsing. This flag is on by default. const unsigned int parse_eol = 0x0020; - - // This flag determines if attribute values are normalized using CDATA normalization rules during parsing. This flag is on by default. - const unsigned int parse_wconv_attribute = 0x0040; - - // This flag determines if attribute values are normalized using NMTOKENS normalization rules during parsing. This flag is off by default. - const unsigned int parse_wnorm_attribute = 0x0080; - - // This flag determines if document declaration (node_declaration) is added to the DOM tree. This flag is off by default. + + // This flag determines if attribute values are normalized using CDATA normalization rules during parsing. This flag is on by default. + const unsigned int parse_wconv_attribute = 0x0040; + + // This flag determines if attribute values are normalized using NMTOKENS normalization rules during parsing. This flag is off by default. + const unsigned int parse_wnorm_attribute = 0x0080; + + // This flag determines if document declaration (node_declaration) is added to the DOM tree. This flag is off by default. const unsigned int parse_declaration = 0x0100; - // This flag determines if document type declaration (node_doctype) is added to the DOM tree. This flag is off by default. + // This flag determines if document type declaration (node_doctype) is added to the DOM tree. This flag is off by default. const unsigned int parse_doctype = 0x0200; + // This flag determines if plain character data (node_pcdata) that is the only child of the parent node and that consists only + // of whitespace is added to the DOM tree. + // This flag is off by default; turning it on may result in slower parsing and more memory consumption. + const unsigned int parse_ws_pcdata_single = 0x0400; + + // This flag determines if leading and trailing whitespace is to be removed from plain character data. This flag is off by default. + const unsigned int parse_trim_pcdata = 0x0800; + + // This flag determines if plain character data that does not have a parent node is added to the DOM tree, and if an empty document + // is a valid document. This flag is off by default. + const unsigned int parse_fragment = 0x1000; + + // This flag determines if plain character data is be stored in the parent element's value. This significantly changes the structure of + // the document; this flag is only recommended for parsing documents with many PCDATA nodes in memory-constrained environments. + // This flag is off by default. + const unsigned int parse_embed_pcdata = 0x2000; + // The default parsing mode. - // Elements, PCDATA and CDATA sections are added to the DOM tree, character/reference entities are expanded, - // End-of-Line characters are normalized, attribute values are normalized using CDATA normalization rules. + // Elements, PCDATA and CDATA sections are added to the DOM tree, character/reference entities are expanded, + // End-of-Line characters are normalized, attribute values are normalized using CDATA normalization rules. const unsigned int parse_default = parse_cdata | parse_escapes | parse_wconv_attribute | parse_eol; - // The full parsing mode. - // Nodes of all types are added to the DOM tree, character/reference entities are expanded, - // End-of-Line characters are normalized, attribute values are normalized using CDATA normalization rules. - const unsigned int parse_full = parse_default | parse_pi | parse_comments | parse_declaration | parse_doctype; + // The full parsing mode. + // Nodes of all types are added to the DOM tree, character/reference entities are expanded, + // End-of-Line characters are normalized, attribute values are normalized using CDATA normalization rules. + const unsigned int parse_full = parse_default | parse_pi | parse_comments | parse_declaration | parse_doctype; // These flags determine the encoding of input data for XML document enum xml_encoding { - encoding_auto, // Auto-detect input encoding using BOM or < / class xml_object_range + { + public: + typedef It const_iterator; + typedef It iterator; + + xml_object_range(It b, It e): _begin(b), _end(e) + { + } + + It begin() const { return _begin; } + It end() const { return _end; } + + private: + It _begin, _end; + }; + // Writer interface for node printing (see xml_node::print) class PUGIXML_CLASS xml_writer { @@ -226,10 +311,10 @@ namespace pugi class PUGIXML_CLASS xml_writer_file: public xml_writer { public: - // Construct writer from a FILE* object; void* is used to avoid header dependencies on stdio + // Construct writer from a FILE* object; void* is used to avoid header dependencies on stdio xml_writer_file(void* file); - virtual void write(const void* data, size_t size); + virtual void write(const void* data, size_t size) PUGIXML_OVERRIDE; private: void* file; @@ -240,11 +325,11 @@ namespace pugi class PUGIXML_CLASS xml_writer_stream: public xml_writer { public: - // Construct writer from an output stream object + // Construct writer from an output stream object xml_writer_stream(std::basic_ostream >& stream); xml_writer_stream(std::basic_ostream >& stream); - virtual void write(const void* data, size_t size); + virtual void write(const void* data, size_t size) PUGIXML_OVERRIDE; private: std::basic_ostream >* narrow_stream; @@ -260,21 +345,21 @@ namespace pugi private: xml_attribute_struct* _attr; - - typedef xml_attribute_struct* xml_attribute::*unspecified_bool_type; + + typedef void (*unspecified_bool_type)(xml_attribute***); public: - // Default constructor. Constructs an empty attribute. + // Default constructor. Constructs an empty attribute. xml_attribute(); - - // Constructs attribute from internal pointer + + // Constructs attribute from internal pointer explicit xml_attribute(xml_attribute_struct* attr); - // Safe bool conversion operator - operator unspecified_bool_type() const; + // Safe bool conversion operator + operator unspecified_bool_type() const; - // Borland C++ workaround - bool operator!() const; + // Borland C++ workaround + bool operator!() const; // Comparison operators (compares wrapped attribute pointers) bool operator==(const xml_attribute& r) const; @@ -291,38 +376,62 @@ namespace pugi const char_t* name() const; const char_t* value() const; - // Get attribute value as a number, or 0 if conversion did not succeed or attribute is empty - int as_int() const; - unsigned int as_uint() const; - double as_double() const; - float as_float() const; + // Get attribute value, or the default value if attribute is empty + const char_t* as_string(const char_t* def = PUGIXML_TEXT("")) const; - // Get attribute value as bool (returns true if first character is in '1tTyY' set), or false if attribute is empty - bool as_bool() const; + // Get attribute value as a number, or the default value if conversion did not succeed or attribute is empty + int as_int(int def = 0) const; + unsigned int as_uint(unsigned int def = 0) const; + double as_double(double def = 0) const; + float as_float(float def = 0) const; + + #ifdef PUGIXML_HAS_LONG_LONG + long long as_llong(long long def = 0) const; + unsigned long long as_ullong(unsigned long long def = 0) const; + #endif - // Set attribute name/value (returns false if attribute is empty or there is not enough memory) + // Get attribute value as bool (returns true if first character is in '1tTyY' set), or the default value if attribute is empty + bool as_bool(bool def = false) const; + + // Set attribute name/value (returns false if attribute is empty or there is not enough memory) bool set_name(const char_t* rhs); bool set_value(const char_t* rhs); - // Set attribute value with type conversion (numbers are converted to strings, boolean is converted to "true"/"false") + // Set attribute value with type conversion (numbers are converted to strings, boolean is converted to "true"/"false") bool set_value(int rhs); bool set_value(unsigned int rhs); + bool set_value(long rhs); + bool set_value(unsigned long rhs); bool set_value(double rhs); + bool set_value(float rhs); bool set_value(bool rhs); + #ifdef PUGIXML_HAS_LONG_LONG + bool set_value(long long rhs); + bool set_value(unsigned long long rhs); + #endif + // Set attribute value (equivalent to set_value without error checking) xml_attribute& operator=(const char_t* rhs); xml_attribute& operator=(int rhs); xml_attribute& operator=(unsigned int rhs); + xml_attribute& operator=(long rhs); + xml_attribute& operator=(unsigned long rhs); xml_attribute& operator=(double rhs); + xml_attribute& operator=(float rhs); xml_attribute& operator=(bool rhs); - // Get next/previous attribute in the attribute list of the parent node - xml_attribute next_attribute() const; - xml_attribute previous_attribute() const; + #ifdef PUGIXML_HAS_LONG_LONG + xml_attribute& operator=(long long rhs); + xml_attribute& operator=(unsigned long long rhs); + #endif + + // Get next/previous attribute in the attribute list of the parent node + xml_attribute next_attribute() const; + xml_attribute previous_attribute() const; - // Get hash value (unique for handles to the same object) - size_t hash_value() const; + // Get hash value (unique for handles to the same object) + size_t hash_value() const; // Get internal pointer xml_attribute_struct* internal_object() const; @@ -339,25 +448,26 @@ namespace pugi { friend class xml_attribute_iterator; friend class xml_node_iterator; + friend class xml_named_node_iterator; protected: xml_node_struct* _root; - typedef xml_node_struct* xml_node::*unspecified_bool_type; + typedef void (*unspecified_bool_type)(xml_node***); public: // Default constructor. Constructs an empty node. xml_node(); - // Constructs node from internal pointer + // Constructs node from internal pointer explicit xml_node(xml_node_struct* p); - // Safe bool conversion operator + // Safe bool conversion operator operator unspecified_bool_type() const; // Borland C++ workaround bool operator!() const; - + // Comparison operators (compares wrapped node pointers) bool operator==(const xml_node& r) const; bool operator!=(const xml_node& r) const; @@ -372,34 +482,43 @@ namespace pugi // Get node type xml_node_type type() const; - // Get node name/value, or "" if node is empty or it has no name/value + // Get node name, or "" if node is empty or it has no name const char_t* name() const; + + // Get node value, or "" if node is empty or it has no value + // Note: For text node.value() does not return "text"! Use child_value() or text() methods to access text inside nodes. const char_t* value() const; - + // Get attribute list xml_attribute first_attribute() const; - xml_attribute last_attribute() const; + xml_attribute last_attribute() const; - // Get children list + // Get children list xml_node first_child() const; - xml_node last_child() const; + xml_node last_child() const; - // Get next/previous sibling in the children list of the parent node + // Get next/previous sibling in the children list of the parent node xml_node next_sibling() const; xml_node previous_sibling() const; - - // Get parent node + + // Get parent node xml_node parent() const; // Get root of DOM tree this node belongs to xml_node root() const; + // Get text object for the current node + xml_text text() const; + // Get child, attribute or next/previous sibling with the specified name xml_node child(const char_t* name) const; xml_attribute attribute(const char_t* name) const; xml_node next_sibling(const char_t* name) const; xml_node previous_sibling(const char_t* name) const; + // Get attribute, starting the search from a hint (and updating hint so that searching for a sequence of attributes is fast) + xml_attribute attribute(const char_t* name, xml_attribute& hint) const; + // Get child value of current node; that is, value of the first child node of type PCDATA/CDATA const char_t* child_value() const; @@ -409,7 +528,7 @@ namespace pugi // Set node name/value (returns false if node is empty, there is not enough memory, or node can not have name/value) bool set_name(const char_t* rhs); bool set_value(const char_t* rhs); - + // Add attribute with specified name. Returns added attribute, or empty attribute on errors. xml_attribute append_attribute(const char_t* name); xml_attribute prepend_attribute(const char_t* name); @@ -440,6 +559,12 @@ namespace pugi xml_node insert_copy_after(const xml_node& proto, const xml_node& node); xml_node insert_copy_before(const xml_node& proto, const xml_node& node); + // Move the specified node to become a child of this node. Returns moved node, or empty node on errors. + xml_node append_move(const xml_node& moved); + xml_node prepend_move(const xml_node& moved); + xml_node insert_move_after(const xml_node& moved, const xml_node& node); + xml_node insert_move_before(const xml_node& moved, const xml_node& node); + // Remove specified attribute bool remove_attribute(const xml_attribute& a); bool remove_attribute(const char_t* name); @@ -448,15 +573,20 @@ namespace pugi bool remove_child(const xml_node& n); bool remove_child(const char_t* name); + // Parses buffer as an XML document fragment and appends all nodes as children of the current node. + // Copies/converts the buffer, so it may be deleted or changed after the function returns. + // Note: append_buffer allocates memory that has the lifetime of the owning document; removing the appended nodes does not immediately reclaim that memory. + xml_parse_result append_buffer(const void* contents, size_t size, unsigned int options = parse_default, xml_encoding encoding = encoding_auto); + // Find attribute using predicate. Returns first attribute for which predicate returned true. template xml_attribute find_attribute(Predicate pred) const { if (!_root) return xml_attribute(); - + for (xml_attribute attrib = first_attribute(); attrib; attrib = attrib.next_attribute()) if (pred(attrib)) return attrib; - + return xml_attribute(); } @@ -464,12 +594,12 @@ namespace pugi template xml_node find_child(Predicate pred) const { if (!_root) return xml_node(); - + for (xml_node node = first_child(); node; node = node.next_sibling()) if (pred(node)) return node; - - return xml_node(); + + return xml_node(); } // Find node from subtree using predicate. Returns first node from subtree (depth-first), for which predicate returned true. @@ -478,7 +608,7 @@ namespace pugi if (!_root) return xml_node(); xml_node cur = first_child(); - + while (cur._root && cur._root != _root) { if (pred(cur)) return cur; @@ -510,17 +640,22 @@ namespace pugi // Recursively traverse subtree with xml_tree_walker bool traverse(xml_tree_walker& walker); - + #ifndef PUGIXML_NO_XPATH // Select single node by evaluating XPath query. Returns first node from the resulting node set. - xpath_node select_single_node(const char_t* query, xpath_variable_set* variables = 0) const; - xpath_node select_single_node(const xpath_query& query) const; + xpath_node select_node(const char_t* query, xpath_variable_set* variables = 0) const; + xpath_node select_node(const xpath_query& query) const; // Select node set by evaluating XPath query xpath_node_set select_nodes(const char_t* query, xpath_variable_set* variables = 0) const; xpath_node_set select_nodes(const xpath_query& query) const; + + // (deprecated: use select_node instead) Select single node by evaluating XPath query. + PUGIXML_DEPRECATED xpath_node select_single_node(const char_t* query, xpath_variable_set* variables = 0) const; + PUGIXML_DEPRECATED xpath_node select_single_node(const xpath_query& query) const; + #endif - + // Print subtree using a writer object void print(xml_writer& writer, const char_t* indent = PUGIXML_TEXT("\t"), unsigned int flags = format_default, xml_encoding encoding = encoding_auto, unsigned int depth = 0) const; @@ -542,11 +677,16 @@ namespace pugi attribute_iterator attributes_begin() const; attribute_iterator attributes_end() const; + // Range-based for support + xml_object_range children() const; + xml_object_range children(const char_t* name) const; + xml_object_range attributes() const; + // Get node offset in parsed file/string (in char_t units) for debugging purposes ptrdiff_t offset_debug() const; - // Get hash value (unique for handles to the same object) - size_t hash_value() const; + // Get hash value (unique for handles to the same object) + size_t hash_value() const; // Get internal pointer xml_node_struct* internal_object() const; @@ -558,13 +698,102 @@ namespace pugi bool PUGIXML_FUNCTION operator||(const xml_node& lhs, bool rhs); #endif + // A helper for working with text inside PCDATA nodes + class PUGIXML_CLASS xml_text + { + friend class xml_node; + + xml_node_struct* _root; + + typedef void (*unspecified_bool_type)(xml_text***); + + explicit xml_text(xml_node_struct* root); + + xml_node_struct* _data_new(); + xml_node_struct* _data() const; + + public: + // Default constructor. Constructs an empty object. + xml_text(); + + // Safe bool conversion operator + operator unspecified_bool_type() const; + + // Borland C++ workaround + bool operator!() const; + + // Check if text object is empty + bool empty() const; + + // Get text, or "" if object is empty + const char_t* get() const; + + // Get text, or the default value if object is empty + const char_t* as_string(const char_t* def = PUGIXML_TEXT("")) const; + + // Get text as a number, or the default value if conversion did not succeed or object is empty + int as_int(int def = 0) const; + unsigned int as_uint(unsigned int def = 0) const; + double as_double(double def = 0) const; + float as_float(float def = 0) const; + + #ifdef PUGIXML_HAS_LONG_LONG + long long as_llong(long long def = 0) const; + unsigned long long as_ullong(unsigned long long def = 0) const; + #endif + + // Get text as bool (returns true if first character is in '1tTyY' set), or the default value if object is empty + bool as_bool(bool def = false) const; + + // Set text (returns false if object is empty or there is not enough memory) + bool set(const char_t* rhs); + + // Set text with type conversion (numbers are converted to strings, boolean is converted to "true"/"false") + bool set(int rhs); + bool set(unsigned int rhs); + bool set(long rhs); + bool set(unsigned long rhs); + bool set(double rhs); + bool set(float rhs); + bool set(bool rhs); + + #ifdef PUGIXML_HAS_LONG_LONG + bool set(long long rhs); + bool set(unsigned long long rhs); + #endif + + // Set text (equivalent to set without error checking) + xml_text& operator=(const char_t* rhs); + xml_text& operator=(int rhs); + xml_text& operator=(unsigned int rhs); + xml_text& operator=(long rhs); + xml_text& operator=(unsigned long rhs); + xml_text& operator=(double rhs); + xml_text& operator=(float rhs); + xml_text& operator=(bool rhs); + + #ifdef PUGIXML_HAS_LONG_LONG + xml_text& operator=(long long rhs); + xml_text& operator=(unsigned long long rhs); + #endif + + // Get the data node (node_pcdata or node_cdata) for this object + xml_node data() const; + }; + +#ifdef __BORLANDC__ + // Borland C++ workaround + bool PUGIXML_FUNCTION operator&&(const xml_text& lhs, bool rhs); + bool PUGIXML_FUNCTION operator||(const xml_text& lhs, bool rhs); +#endif + // Child node iterator (a bidirectional iterator over a collection of xml_node) class PUGIXML_CLASS xml_node_iterator { friend class xml_node; private: - xml_node _wrap; + mutable xml_node _wrap; xml_node _parent; xml_node_iterator(xml_node_struct* ref, xml_node_struct* parent); @@ -580,18 +809,18 @@ namespace pugi typedef std::bidirectional_iterator_tag iterator_category; #endif - // Default constructor + // Default constructor xml_node_iterator(); - // Construct an iterator which points to the specified node + // Construct an iterator which points to the specified node xml_node_iterator(const xml_node& node); - // Iterator operators + // Iterator operators bool operator==(const xml_node_iterator& rhs) const; bool operator!=(const xml_node_iterator& rhs) const; - xml_node& operator*(); - xml_node* operator->(); + xml_node& operator*() const; + xml_node* operator->() const; const xml_node_iterator& operator++(); xml_node_iterator operator++(int); @@ -606,7 +835,7 @@ namespace pugi friend class xml_node; private: - xml_attribute _wrap; + mutable xml_attribute _wrap; xml_node _parent; xml_attribute_iterator(xml_attribute_struct* ref, xml_node_struct* parent); @@ -622,18 +851,18 @@ namespace pugi typedef std::bidirectional_iterator_tag iterator_category; #endif - // Default constructor + // Default constructor xml_attribute_iterator(); - // Construct an iterator which points to the specified attribute + // Construct an iterator which points to the specified attribute xml_attribute_iterator(const xml_attribute& attr, const xml_node& parent); // Iterator operators bool operator==(const xml_attribute_iterator& rhs) const; bool operator!=(const xml_attribute_iterator& rhs) const; - xml_attribute& operator*(); - xml_attribute* operator->(); + xml_attribute& operator*() const; + xml_attribute* operator->() const; const xml_attribute_iterator& operator++(); xml_attribute_iterator operator++(int); @@ -642,6 +871,49 @@ namespace pugi xml_attribute_iterator operator--(int); }; + // Named node range helper + class PUGIXML_CLASS xml_named_node_iterator + { + friend class xml_node; + + public: + // Iterator traits + typedef ptrdiff_t difference_type; + typedef xml_node value_type; + typedef xml_node* pointer; + typedef xml_node& reference; + + #ifndef PUGIXML_NO_STL + typedef std::bidirectional_iterator_tag iterator_category; + #endif + + // Default constructor + xml_named_node_iterator(); + + // Construct an iterator which points to the specified node + xml_named_node_iterator(const xml_node& node, const char_t* name); + + // Iterator operators + bool operator==(const xml_named_node_iterator& rhs) const; + bool operator!=(const xml_named_node_iterator& rhs) const; + + xml_node& operator*() const; + xml_node* operator->() const; + + const xml_named_node_iterator& operator++(); + xml_named_node_iterator operator++(int); + + const xml_named_node_iterator& operator--(); + xml_named_node_iterator operator--(int); + + private: + mutable xml_node _wrap; + xml_node _parent; + const char_t* _name; + + xml_named_node_iterator(xml_node_struct* ref, xml_node_struct* parent, const char_t* name); + }; + // Abstract tree walker class (see xml_node::traverse) class PUGIXML_CLASS xml_tree_walker { @@ -649,11 +921,11 @@ namespace pugi private: int _depth; - + protected: // Get current traversal depth int depth() const; - + public: xml_tree_walker(); virtual ~xml_tree_walker(); @@ -671,24 +943,28 @@ namespace pugi // Parsing status, returned as part of xml_parse_result object enum xml_parse_status { - status_ok = 0, // No error - - status_file_not_found, // File was not found during load_file() - status_io_error, // Error reading from file/stream - status_out_of_memory, // Could not allocate memory - status_internal_error, // Internal error occurred - - status_unrecognized_tag, // Parser could not determine tag type - - status_bad_pi, // Parsing error occurred while parsing document declaration/processing instruction - status_bad_comment, // Parsing error occurred while parsing comment - status_bad_cdata, // Parsing error occurred while parsing CDATA section - status_bad_doctype, // Parsing error occurred while parsing document type declaration - status_bad_pcdata, // Parsing error occurred while parsing PCDATA section - status_bad_start_element, // Parsing error occurred while parsing start element tag - status_bad_attribute, // Parsing error occurred while parsing element attribute - status_bad_end_element, // Parsing error occurred while parsing end element tag - status_end_element_mismatch // There was a mismatch of start-end tags (closing tag had incorrect name, some tag was not closed or there was an excessive closing tag) + status_ok = 0, // No error + + status_file_not_found, // File was not found during load_file() + status_io_error, // Error reading from file/stream + status_out_of_memory, // Could not allocate memory + status_internal_error, // Internal error occurred + + status_unrecognized_tag, // Parser could not determine tag type + + status_bad_pi, // Parsing error occurred while parsing document declaration/processing instruction + status_bad_comment, // Parsing error occurred while parsing comment + status_bad_cdata, // Parsing error occurred while parsing CDATA section + status_bad_doctype, // Parsing error occurred while parsing document type declaration + status_bad_pcdata, // Parsing error occurred while parsing PCDATA section + status_bad_start_element, // Parsing error occurred while parsing start element tag + status_bad_attribute, // Parsing error occurred while parsing element attribute + status_bad_end_element, // Parsing error occurred while parsing end element tag + status_end_element_mismatch,// There was a mismatch of start-end tags (closing tag had incorrect name, some tag was not closed or there was an excessive closing tag) + + status_append_invalid_root, // Unable to append nodes since root type is not node_element or node_document (exclusive to xml_node::append_buffer) + + status_no_document_element // Parsing resulted in a document without element nodes }; // Parsing result @@ -703,7 +979,7 @@ namespace pugi // Source document encoding xml_encoding encoding; - // Default constructor, initializes object to failed state + // Default constructor, initializes object to failed state xml_parse_result(); // Cast to bool operator @@ -720,15 +996,14 @@ namespace pugi char_t* _buffer; char _memory[192]; - + // Non-copyable semantics xml_document(const xml_document&); - const xml_document& operator=(const xml_document&); + xml_document& operator=(const xml_document&); - void create(); - void destroy(); - - xml_parse_result load_buffer_impl(void* contents, size_t size, unsigned int options, xml_encoding encoding, bool is_mutable, bool own); + void _create(); + void _destroy(); + void _move(xml_document& rhs) PUGIXML_NOEXCEPT_IF_NOT_COMPACT; public: // Default constructor, makes empty document @@ -737,10 +1012,16 @@ namespace pugi // Destructor, invalidates all node/attribute handles to this document ~xml_document(); - // Removes all nodes, leaving the empty document + #ifdef PUGIXML_HAS_MOVE + // Move semantics support + xml_document(xml_document&& rhs) PUGIXML_NOEXCEPT_IF_NOT_COMPACT; + xml_document& operator=(xml_document&& rhs) PUGIXML_NOEXCEPT_IF_NOT_COMPACT; + #endif + + // Removes all nodes, leaving the empty document void reset(); - // Removes all nodes, then copies the entire contents of the specified document + // Removes all nodes, then copies the entire contents of the specified document void reset(const xml_document& proto); #ifndef PUGIXML_NO_STL @@ -749,8 +1030,11 @@ namespace pugi xml_parse_result load(std::basic_istream >& stream, unsigned int options = parse_default); #endif + // (deprecated: use load_string instead) Load document from zero-terminated string. No encoding conversions are applied. + PUGIXML_DEPRECATED xml_parse_result load(const char_t* contents, unsigned int options = parse_default); + // Load document from zero-terminated string. No encoding conversions are applied. - xml_parse_result load(const char_t* contents, unsigned int options = parse_default); + xml_parse_result load_string(const char_t* contents, unsigned int options = parse_default); // Load document from file xml_parse_result load_file(const char* path, unsigned int options = parse_default, xml_encoding encoding = encoding_auto); @@ -760,11 +1044,11 @@ namespace pugi xml_parse_result load_buffer(const void* contents, size_t size, unsigned int options = parse_default, xml_encoding encoding = encoding_auto); // Load document from buffer, using the buffer for in-place parsing (the buffer is modified and used for storage of document data). - // You should ensure that buffer data will persist throughout the document's lifetime, and free the buffer memory manually once document is destroyed. + // You should ensure that buffer data will persist throughout the document's lifetime, and free the buffer memory manually once document is destroyed. xml_parse_result load_buffer_inplace(void* contents, size_t size, unsigned int options = parse_default, xml_encoding encoding = encoding_auto); // Load document from buffer, using the buffer for in-place parsing (the buffer is modified and used for storage of document data). - // You should allocate the buffer with pugixml allocation function; document will free the buffer when it is no longer needed (you can't use it anymore). + // You should allocate the buffer with pugixml allocation function; document will free the buffer when it is no longer needed (you can't use it anymore). xml_parse_result load_buffer_inplace_own(void* contents, size_t size, unsigned int options = parse_default, xml_encoding encoding = encoding_auto); // Save XML document to writer (semantics is slightly different from xml_node::print, see documentation for details). @@ -780,22 +1064,22 @@ namespace pugi bool save_file(const char* path, const char_t* indent = PUGIXML_TEXT("\t"), unsigned int flags = format_default, xml_encoding encoding = encoding_auto) const; bool save_file(const wchar_t* path, const char_t* indent = PUGIXML_TEXT("\t"), unsigned int flags = format_default, xml_encoding encoding = encoding_auto) const; - // Get document element - xml_node document_element() const; + // Get document element + xml_node document_element() const; }; #ifndef PUGIXML_NO_XPATH // XPath query return type enum xpath_value_type { - xpath_type_none, // Unknown type (query failed to compile) + xpath_type_none, // Unknown type (query failed to compile) xpath_type_node_set, // Node set (xpath_node_set) - xpath_type_number, // Number - xpath_type_string, // String - xpath_type_boolean // Boolean + xpath_type_number, // Number + xpath_type_string, // String + xpath_type_boolean // Boolean }; - // XPath parsing result + // XPath parsing result struct PUGIXML_CLASS xpath_parse_result { // Error message (0 if no error) @@ -804,7 +1088,7 @@ namespace pugi // Last parsed offset (in char_t units from string start) ptrdiff_t offset; - // Default constructor, initializes object to failed state + // Default constructor, initializes object to failed state xpath_parse_result(); // Cast to bool operator @@ -823,26 +1107,26 @@ namespace pugi xpath_value_type _type; xpath_variable* _next; - xpath_variable(); + xpath_variable(xpath_value_type type); // Non-copyable semantics xpath_variable(const xpath_variable&); xpath_variable& operator=(const xpath_variable&); - + public: - // Get variable name + // Get variable name const char_t* name() const; - // Get variable type + // Get variable type xpath_value_type type() const; - // Get variable value; no type conversion is performed, default value (false, NaN, empty string, empty node set) is returned on type mismatch error + // Get variable value; no type conversion is performed, default value (false, NaN, empty string, empty node set) is returned on type mismatch error bool get_boolean() const; double get_number() const; const char_t* get_string() const; const xpath_node_set& get_node_set() const; - // Set variable value; no type conversion is performed, false is returned on type mismatch error + // Set variable value; no type conversion is performed, false is returned on type mismatch error bool set(bool value); bool set(double value); bool set(const char_t* value); @@ -855,27 +1139,39 @@ namespace pugi private: xpath_variable* _data[64]; - // Non-copyable semantics - xpath_variable_set(const xpath_variable_set&); - xpath_variable_set& operator=(const xpath_variable_set&); + void _assign(const xpath_variable_set& rhs); + void _swap(xpath_variable_set& rhs); + + xpath_variable* _find(const char_t* name) const; - xpath_variable* find(const char_t* name) const; + static bool _clone(xpath_variable* var, xpath_variable** out_result); + static void _destroy(xpath_variable* var); public: - // Default constructor/destructor + // Default constructor/destructor xpath_variable_set(); ~xpath_variable_set(); - // Add a new variable or get the existing one, if the types match + // Copy constructor/assignment operator + xpath_variable_set(const xpath_variable_set& rhs); + xpath_variable_set& operator=(const xpath_variable_set& rhs); + + #ifdef PUGIXML_HAS_MOVE + // Move semantics support + xpath_variable_set(xpath_variable_set&& rhs) PUGIXML_NOEXCEPT; + xpath_variable_set& operator=(xpath_variable_set&& rhs) PUGIXML_NOEXCEPT; + #endif + + // Add a new variable or get the existing one, if the types match xpath_variable* add(const char_t* name, xpath_value_type type); - // Set value of an existing variable; no type conversion is performed, false is returned if there is no such variable or if types mismatch + // Set value of an existing variable; no type conversion is performed, false is returned if there is no such variable or if types mismatch bool set(const char_t* name, bool value); bool set(const char_t* name, double value); bool set(const char_t* name, const char_t* value); bool set(const char_t* name, const xpath_node_set& value); - // Get existing variable by name + // Get existing variable by name xpath_variable* get(const char_t* name); const xpath_variable* get(const char_t* name) const; }; @@ -887,58 +1183,73 @@ namespace pugi void* _impl; xpath_parse_result _result; - typedef void* xpath_query::*unspecified_bool_type; + typedef void (*unspecified_bool_type)(xpath_query***); // Non-copyable semantics xpath_query(const xpath_query&); xpath_query& operator=(const xpath_query&); public: - // Construct a compiled object from XPath expression. - // If PUGIXML_NO_EXCEPTIONS is not defined, throws xpath_exception on compilation errors. + // Construct a compiled object from XPath expression. + // If PUGIXML_NO_EXCEPTIONS is not defined, throws xpath_exception on compilation errors. explicit xpath_query(const char_t* query, xpath_variable_set* variables = 0); + // Constructor + xpath_query(); + // Destructor ~xpath_query(); + #ifdef PUGIXML_HAS_MOVE + // Move semantics support + xpath_query(xpath_query&& rhs) PUGIXML_NOEXCEPT; + xpath_query& operator=(xpath_query&& rhs) PUGIXML_NOEXCEPT; + #endif + // Get query expression return type xpath_value_type return_type() const; - + // Evaluate expression as boolean value in the specified context; performs type conversion if necessary. - // If PUGIXML_NO_EXCEPTIONS is not defined, throws std::bad_alloc on out of memory errors. + // If PUGIXML_NO_EXCEPTIONS is not defined, throws std::bad_alloc on out of memory errors. bool evaluate_boolean(const xpath_node& n) const; - + // Evaluate expression as double value in the specified context; performs type conversion if necessary. - // If PUGIXML_NO_EXCEPTIONS is not defined, throws std::bad_alloc on out of memory errors. + // If PUGIXML_NO_EXCEPTIONS is not defined, throws std::bad_alloc on out of memory errors. double evaluate_number(const xpath_node& n) const; - + #ifndef PUGIXML_NO_STL // Evaluate expression as string value in the specified context; performs type conversion if necessary. - // If PUGIXML_NO_EXCEPTIONS is not defined, throws std::bad_alloc on out of memory errors. + // If PUGIXML_NO_EXCEPTIONS is not defined, throws std::bad_alloc on out of memory errors. string_t evaluate_string(const xpath_node& n) const; #endif - + // Evaluate expression as string value in the specified context; performs type conversion if necessary. - // At most capacity characters are written to the destination buffer, full result size is returned (includes terminating zero). - // If PUGIXML_NO_EXCEPTIONS is not defined, throws std::bad_alloc on out of memory errors. - // If PUGIXML_NO_EXCEPTIONS is defined, returns empty set instead. + // At most capacity characters are written to the destination buffer, full result size is returned (includes terminating zero). + // If PUGIXML_NO_EXCEPTIONS is not defined, throws std::bad_alloc on out of memory errors. + // If PUGIXML_NO_EXCEPTIONS is defined, returns empty set instead. size_t evaluate_string(char_t* buffer, size_t capacity, const xpath_node& n) const; // Evaluate expression as node set in the specified context. - // If PUGIXML_NO_EXCEPTIONS is not defined, throws xpath_exception on type mismatch and std::bad_alloc on out of memory errors. - // If PUGIXML_NO_EXCEPTIONS is defined, returns empty node set instead. + // If PUGIXML_NO_EXCEPTIONS is not defined, throws xpath_exception on type mismatch and std::bad_alloc on out of memory errors. + // If PUGIXML_NO_EXCEPTIONS is defined, returns empty node set instead. xpath_node_set evaluate_node_set(const xpath_node& n) const; + // Evaluate expression as node set in the specified context. + // Return first node in document order, or empty node if node set is empty. + // If PUGIXML_NO_EXCEPTIONS is not defined, throws xpath_exception on type mismatch and std::bad_alloc on out of memory errors. + // If PUGIXML_NO_EXCEPTIONS is defined, returns empty node instead. + xpath_node evaluate_node(const xpath_node& n) const; + // Get parsing result (used to get compilation errors in PUGIXML_NO_EXCEPTIONS mode) const xpath_parse_result& result() const; // Safe bool conversion operator operator unspecified_bool_type() const; - // Borland C++ workaround + // Borland C++ workaround bool operator!() const; }; - + #ifndef PUGIXML_NO_EXCEPTIONS // XPath exception class class PUGIXML_CLASS xpath_exception: public std::exception @@ -951,26 +1262,26 @@ namespace pugi explicit xpath_exception(const xpath_parse_result& result); // Get error message - virtual const char* what() const throw(); + virtual const char* what() const throw() PUGIXML_OVERRIDE; - // Get parse result + // Get parse result const xpath_parse_result& result() const; }; #endif - + // XPath node class (either xml_node or xml_attribute) class PUGIXML_CLASS xpath_node { private: xml_node _node; xml_attribute _attribute; - - typedef xml_node xpath_node::*unspecified_bool_type; + + typedef void (*unspecified_bool_type)(xpath_node***); public: // Default constructor; constructs empty XPath node xpath_node(); - + // Construct XPath node from XML node/attribute xpath_node(const xml_node& node); xpath_node(const xml_attribute& attribute, const xml_node& parent); @@ -978,15 +1289,15 @@ namespace pugi // Get node/attribute, if any xml_node node() const; xml_attribute attribute() const; - + // Get parent of contained node/attribute xml_node parent() const; - // Safe bool conversion operator + // Safe bool conversion operator operator unspecified_bool_type() const; - - // Borland C++ workaround - bool operator!() const; + + // Borland C++ workaround + bool operator!() const; // Comparison operators bool operator==(const xpath_node& n) const; @@ -1010,10 +1321,13 @@ namespace pugi type_sorted, // Sorted by document order (ascending) type_sorted_reverse // Sorted by document order (descending) }; - + // Constant iterator type typedef const xpath_node* const_iterator; - + + // We define non-constant iterator to be the same as constant iterator so that various generic algorithms (i.e. boost foreach) work + typedef const xpath_node* iterator; + // Default constructor. Constructs empty set. xpath_node_set(); @@ -1022,42 +1336,49 @@ namespace pugi // Destructor ~xpath_node_set(); - + // Copy constructor/assignment operator xpath_node_set(const xpath_node_set& ns); xpath_node_set& operator=(const xpath_node_set& ns); + #ifdef PUGIXML_HAS_MOVE + // Move semantics support + xpath_node_set(xpath_node_set&& rhs) PUGIXML_NOEXCEPT; + xpath_node_set& operator=(xpath_node_set&& rhs) PUGIXML_NOEXCEPT; + #endif + // Get collection type type_t type() const; - + // Get collection size size_t size() const; - // Indexing operator + // Indexing operator const xpath_node& operator[](size_t index) const; - + // Collection iterators const_iterator begin() const; const_iterator end() const; // Sort the collection in ascending/descending order by document order void sort(bool reverse = false); - + // Get first node in the collection by document order xpath_node first() const; - + // Check if collection is empty bool empty() const; - + private: type_t _type; - + xpath_node _storage; - + xpath_node* _begin; xpath_node* _end; - void _assign(const_iterator begin, const_iterator end); + void _assign(const_iterator begin, const_iterator end, type_t type); + void _move(xpath_node_set& rhs) PUGIXML_NOEXCEPT; }; #endif @@ -1065,7 +1386,7 @@ namespace pugi // Convert wide string to UTF8 std::basic_string, std::allocator > PUGIXML_FUNCTION as_utf8(const wchar_t* str); std::basic_string, std::allocator > PUGIXML_FUNCTION as_utf8(const std::basic_string, std::allocator >& str); - + // Convert UTF8 to wide string std::basic_string, std::allocator > PUGIXML_FUNCTION as_wide(const char* str); std::basic_string, std::allocator > PUGIXML_FUNCTION as_wide(const std::basic_string, std::allocator >& str); @@ -1073,16 +1394,16 @@ namespace pugi // Memory allocation function interface; returns pointer to allocated memory or NULL on failure typedef void* (*allocation_function)(size_t size); - + // Memory deallocation function interface - typedef void (*deallocation_function)(void* ptr); - - // Override default memory management functions. All subsequent allocations/deallocations will be performed via supplied functions. - void PUGIXML_FUNCTION set_memory_management_functions(allocation_function allocate, deallocation_function deallocate); - - // Get current memory management functions - allocation_function PUGIXML_FUNCTION get_memory_allocation_function(); - deallocation_function PUGIXML_FUNCTION get_memory_deallocation_function(); + typedef void (*deallocation_function)(void* ptr); + + // Override default memory management functions. All subsequent allocations/deallocations will be performed via supplied functions. + void PUGIXML_FUNCTION set_memory_management_functions(allocation_function allocate, deallocation_function deallocate); + + // Get current memory management functions + allocation_function PUGIXML_FUNCTION get_memory_allocation_function(); + deallocation_function PUGIXML_FUNCTION get_memory_deallocation_function(); } #if !defined(PUGIXML_NO_STL) && (defined(_MSC_VER) || defined(__ICC)) @@ -1091,6 +1412,7 @@ namespace std // Workarounds for (non-standard) iterator category detection for older versions (MSVC7/IC8 and earlier) std::bidirectional_iterator_tag PUGIXML_FUNCTION _Iter_cat(const pugi::xml_node_iterator&); std::bidirectional_iterator_tag PUGIXML_FUNCTION _Iter_cat(const pugi::xml_attribute_iterator&); + std::bidirectional_iterator_tag PUGIXML_FUNCTION _Iter_cat(const pugi::xml_named_node_iterator&); } #endif @@ -1100,13 +1422,21 @@ namespace std // Workarounds for (non-standard) iterator category detection std::bidirectional_iterator_tag PUGIXML_FUNCTION __iterator_category(const pugi::xml_node_iterator&); std::bidirectional_iterator_tag PUGIXML_FUNCTION __iterator_category(const pugi::xml_attribute_iterator&); + std::bidirectional_iterator_tag PUGIXML_FUNCTION __iterator_category(const pugi::xml_named_node_iterator&); } #endif #endif +// Make sure implementation is included in header-only mode +// Use macro expansion in #include to work around QMake (QTBUG-11923) +#if defined(PUGIXML_HEADER_ONLY) && !defined(PUGIXML_SOURCE) +# define PUGIXML_SOURCE "pugixml.cpp" +# include PUGIXML_SOURCE +#endif + /** - * Copyright (c) 2006-2010 Arseny Kapoulkine + * Copyright (c) 2006-2018 Arseny Kapoulkine * * Permission is hereby granted, free of charge, to any person * obtaining a copy of this software and associated documentation @@ -1119,7 +1449,7 @@ namespace std * * The above copyright notice and this permission notice shall be * included in all copies or substantial portions of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND @@ -1129,3 +1459,4 @@ namespace std * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR * OTHER DEALINGS IN THE SOFTWARE. */ +