#pragma once #include // min #include // array #include // assert #include // size_t #include // strlen #include // streamsize, streamoff, streampos #include // istream #include // begin, end, iterator_traits, random_access_iterator_tag, distance, next #include // shared_ptr, make_shared, addressof #include // accumulate #include // string, char_traits #include // enable_if, is_base_of, is_pointer, is_integral, remove_pointer #include // pair, declval #include namespace nlohmann { namespace detail { //////////////////// // input adapters // //////////////////// /*! @brief abstract input adapter interface Produces a stream of std::char_traits::int_type characters from a std::istream, a buffer, or some other input type. Accepts the return of exactly one non-EOF character for future input. The int_type characters returned consist of all valid char values as positive values (typically unsigned char), plus an EOF value outside that range, specified by the value of the function std::char_traits::eof(). This value is typically -1, but could be any arbitrary value which is not a valid char value. */ struct input_adapter_protocol { /// get a character [0,255] or std::char_traits::eof(). virtual std::char_traits::int_type get_character() = 0; /// restore the last non-eof() character to input virtual void unget_character() = 0; virtual ~input_adapter_protocol() = default; }; /// a type to simplify interfaces using input_adapter_t = std::shared_ptr; /*! Input adapter for a (caching) istream. Ignores a UFT Byte Order Mark at beginning of input. Does not support changing the underlying std::streambuf in mid-input. Maintains underlying std::istream and std::streambuf to support subsequent use of standard std::istream operations to process any input characters following those used in parsing the JSON input. Clears the std::istream flags; any input errors (e.g., EOF) will be detected by the first subsequent call for input from the std::istream. */ class input_stream_adapter : public input_adapter_protocol { public: ~input_stream_adapter() override { // clear stream flags; we use underlying streambuf I/O, do not // maintain ifstream flags is.clear(); } explicit input_stream_adapter(std::istream& i) : is(i), sb(*i.rdbuf()) { // skip byte order mark std::char_traits::int_type c; if ((c = get_character()) == 0xEF) { if ((c = get_character()) == 0xBB) { if ((c = get_character()) == 0xBF) { return; // Ignore BOM } else if (c != std::char_traits::eof()) { is.unget(); } is.putback('\xBB'); } else if (c != std::char_traits::eof()) { is.unget(); } is.putback('\xEF'); } else if (c != std::char_traits::eof()) { is.unget(); // no byte order mark; process as usual } } // delete because of pointer members input_stream_adapter(const input_stream_adapter&) = delete; input_stream_adapter& operator=(input_stream_adapter&) = delete; // std::istream/std::streambuf use std::char_traits::to_int_type, to // ensure that std::char_traits::eof() and the character 0xFF do not // end up as the same value, eg. 0xFFFFFFFF. std::char_traits::int_type get_character() override { return sb.sbumpc(); } void unget_character() override { sb.sungetc(); // is.unget() avoided for performance } private: /// the associated input stream std::istream& is; std::streambuf& sb; }; /// input adapter for buffer input class input_buffer_adapter : public input_adapter_protocol { public: input_buffer_adapter(const char* b, const std::size_t l) : cursor(b), limit(b + l), start(b) { // skip byte order mark if (l >= 3 and b[0] == '\xEF' and b[1] == '\xBB' and b[2] == '\xBF') { cursor += 3; } } // delete because of pointer members input_buffer_adapter(const input_buffer_adapter&) = delete; input_buffer_adapter& operator=(input_buffer_adapter&) = delete; std::char_traits::int_type get_character() noexcept override { if (JSON_LIKELY(cursor < limit)) { return std::char_traits::to_int_type(*(cursor++)); } return std::char_traits::eof(); } void unget_character() noexcept override { if (JSON_LIKELY(cursor > start)) { --cursor; } } private: /// pointer to the current character const char* cursor; /// pointer past the last character const char* limit; /// pointer to the first character const char* start; }; template class wide_string_input_adapter : public input_adapter_protocol { private: using char_t = typename WideStringType::value_type; public: wide_string_input_adapter(const WideStringType& w) : str(w) {} std::char_traits::int_type get_character() noexcept override { // unget_character() was called previously: return the last character if (next_unget) { next_unget = false; return last_char; } // check if buffer needs to be filled if (utf8_bytes_index == utf8_bytes_filled) { if (sizeof(char_t) == 2) { fill_buffer_utf16(); } else { fill_buffer_utf32(); } assert(utf8_bytes_filled > 0); assert(utf8_bytes_index == 0); } // use buffer assert(utf8_bytes_filled > 0); assert(utf8_bytes_index < utf8_bytes_filled); return (last_char = utf8_bytes[utf8_bytes_index++]); } void unget_character() noexcept override { next_unget = true; } private: void fill_buffer_utf16() { utf8_bytes_index = 0; if (current_wchar == str.size()) { utf8_bytes[0] = std::char_traits::eof(); utf8_bytes_filled = 1; } else { // get the current character const char_t wc = str[current_wchar++]; // UTF-16 to UTF-8 encoding if (wc < 0x80) { utf8_bytes[0] = wc; utf8_bytes_filled = 1; } else if (wc <= 0x7FF) { utf8_bytes[0] = 0xC0 | ((wc >> 6)); utf8_bytes[1] = 0x80 | (wc & 0x3F); utf8_bytes_filled = 2; } else if (0xD800 > wc or wc >= 0xE000) { utf8_bytes[0] = 0xE0 | ((wc >> 12)); utf8_bytes[1] = 0x80 | ((wc >> 6) & 0x3F); utf8_bytes[2] = 0x80 | (wc & 0x3F); utf8_bytes_filled = 3; } else { if (current_wchar < str.size()) { const char_t wc2 = str[current_wchar++]; const int charcode = 0x10000 + (((wc & 0x3FF) << 10) | (wc2 & 0x3FF)); utf8_bytes[0] = 0xf0 | (charcode >> 18); utf8_bytes[1] = 0x80 | ((charcode >> 12) & 0x3F); utf8_bytes[2] = 0x80 | ((charcode >> 6) & 0x3F); utf8_bytes[3] = 0x80 | (charcode & 0x3F); utf8_bytes_filled = 4; } else { // unknown character ++current_wchar; utf8_bytes[0] = wc; utf8_bytes_filled = 1; } } } } void fill_buffer_utf32() { utf8_bytes_index = 0; if (current_wchar == str.size()) { utf8_bytes[0] = std::char_traits::eof(); utf8_bytes_filled = 1; } else { // get the current character const char_t wc = str[current_wchar++]; // UTF-32 to UTF-8 encoding if (wc < 0x80) { utf8_bytes[0] = wc; utf8_bytes_filled = 1; } else if (wc <= 0x7FF) { utf8_bytes[0] = 0xC0 | ((wc >> 6) & 0x1F); utf8_bytes[1] = 0x80 | (wc & 0x3F); utf8_bytes_filled = 2; } else if (wc <= 0xFFFF) { utf8_bytes[0] = 0xE0 | ((wc >> 12) & 0x0F); utf8_bytes[1] = 0x80 | ((wc >> 6) & 0x3F); utf8_bytes[2] = 0x80 | (wc & 0x3F); utf8_bytes_filled = 3; } else if (wc <= 0x10FFFF) { utf8_bytes[0] = 0xF0 | ((wc >> 18 ) & 0x07); utf8_bytes[1] = 0x80 | ((wc >> 12) & 0x3F); utf8_bytes[2] = 0x80 | ((wc >> 6) & 0x3F); utf8_bytes[3] = 0x80 | (wc & 0x3F); utf8_bytes_filled = 4; } else { // unknown character utf8_bytes[0] = wc; utf8_bytes_filled = 1; } } } private: /// the wstring to process const WideStringType& str; /// index of the current wchar in str std::size_t current_wchar = 0; /// a buffer for UTF-8 bytes std::array::int_type, 4> utf8_bytes = {{0, 0, 0, 0}}; /// index to the utf8_codes array for the next valid byte std::size_t utf8_bytes_index = 0; /// number of valid bytes in the utf8_codes array std::size_t utf8_bytes_filled = 0; /// the last character (returned after unget_character() is called) std::char_traits::int_type last_char = 0; /// whether get_character() should return last_char bool next_unget = false; }; class input_adapter { public: // native support /// input adapter for input stream input_adapter(std::istream& i) : ia(std::make_shared(i)) {} /// input adapter for input stream input_adapter(std::istream&& i) : ia(std::make_shared(i)) {} input_adapter(const std::wstring& ws) : ia(std::make_shared>(ws)) {} input_adapter(const std::u16string& ws) : ia(std::make_shared>(ws)) {} input_adapter(const std::u32string& ws) : ia(std::make_shared>(ws)) {} /// input adapter for buffer template::value and std::is_integral::type>::value and sizeof(typename std::remove_pointer::type) == 1, int>::type = 0> input_adapter(CharT b, std::size_t l) : ia(std::make_shared(reinterpret_cast(b), l)) {} // derived support /// input adapter for string literal template::value and std::is_integral::type>::value and sizeof(typename std::remove_pointer::type) == 1, int>::type = 0> input_adapter(CharT b) : input_adapter(reinterpret_cast(b), std::strlen(reinterpret_cast(b))) {} /// input adapter for iterator range with contiguous storage template::iterator_category, std::random_access_iterator_tag>::value, int>::type = 0> input_adapter(IteratorType first, IteratorType last) { // assertion to check that the iterator range is indeed contiguous, // see http://stackoverflow.com/a/35008842/266378 for more discussion assert(std::accumulate( first, last, std::pair(true, 0), [&first](std::pair res, decltype(*first) val) { res.first &= (val == *(std::next(std::addressof(*first), res.second++))); return res; }).first); // assertion to check that each element is 1 byte long static_assert( sizeof(typename std::iterator_traits::value_type) == 1, "each element in the iterator range must have the size of 1 byte"); const auto len = static_cast(std::distance(first, last)); if (JSON_LIKELY(len > 0)) { // there is at least one element: use the address of first ia = std::make_shared(reinterpret_cast(&(*first)), len); } else { // the address of first cannot be used: use nullptr ia = std::make_shared(nullptr, len); } } /// input adapter for array template input_adapter(T (&array)[N]) : input_adapter(std::begin(array), std::end(array)) {} /// input adapter for contiguous container template::value and std::is_base_of()))>::iterator_category>::value, int>::type = 0> input_adapter(const ContiguousContainer& c) : input_adapter(std::begin(c), std::end(c)) {} operator input_adapter_t() { return ia; } private: /// the actual adapter input_adapter_t ia = nullptr; }; } }