json.hpp

                    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<char>::eof();
            utf8_bytes_filled = 1;
        }
        else
        {
            // get the current character
            const int wc = static_cast<int>(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<std::char_traits<char>::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;
};

class input_adapter
{
  public:
    // native support

    /// input adapter for input stream
    input_adapter(std::istream& i)
        : ia(std::make_shared<input_stream_adapter>(i)) {}

    /// input adapter for input stream
    input_adapter(std::istream&& i)
        : ia(std::make_shared<input_stream_adapter>(i)) {}

    input_adapter(const std::wstring& ws)
        : ia(std::make_shared<wide_string_input_adapter<std::wstring>>(ws)) {}

    input_adapter(const std::u16string& ws)
        : ia(std::make_shared<wide_string_input_adapter<std::u16string>>(ws)) {}

    input_adapter(const std::u32string& ws)
        : ia(std::make_shared<wide_string_input_adapter<std::u32string>>(ws)) {}

    /// input adapter for buffer
    template<typename CharT,
             typename std::enable_if<
                 std::is_pointer<CharT>::value and
                 std::is_integral<typename std::remove_pointer<CharT>::type>::value and
                 sizeof(typename std::remove_pointer<CharT>::type) == 1,
                 int>::type = 0>
    input_adapter(CharT b, std::size_t l)
        : ia(std::make_shared<input_buffer_adapter>(reinterpret_cast<const char*>(b), l)) {}

    // derived support

    /// input adapter for string literal
    template<typename CharT,
             typename std::enable_if<
                 std::is_pointer<CharT>::value and
                 std::is_integral<typename std::remove_pointer<CharT>::type>::value and
                 sizeof(typename std::remove_pointer<CharT>::type) == 1,
                 int>::type = 0>
    input_adapter(CharT b)
        : input_adapter(reinterpret_cast<const char*>(b),
                        std::strlen(reinterpret_cast<const char*>(b))) {}

    /// input adapter for iterator range with contiguous storage
    template<class IteratorType,
             typename std::enable_if<
                 std::is_same<typename std::iterator_traits<IteratorType>::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<bool, int>(true, 0),
                   [&first](std::pair<bool, int> 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<IteratorType>::value_type) == 1,
            "each element in the iterator range must have the size of 1 byte");

        const auto len = static_cast<size_t>(std::distance(first, last));
        if (JSON_LIKELY(len > 0))
        {
            // there is at least one element: use the address of first
            ia = std::make_shared<input_buffer_adapter>(reinterpret_cast<const char*>(&(*first)), len);
        }
        else
        {
            // the address of first cannot be used: use nullptr
            ia = std::make_shared<input_buffer_adapter>(nullptr, len);
        }
    }

    /// input adapter for array
    template<class T, std::size_t N>
    input_adapter(T (&array)[N])
        : input_adapter(std::begin(array), std::end(array)) {}

    /// input adapter for contiguous container
    template<class ContiguousContainer, typename
             std::enable_if<not std::is_pointer<ContiguousContainer>::value and
                            std::is_base_of<std::random_access_iterator_tag, typename std::iterator_traits<decltype(std::begin(std::declval<ContiguousContainer const>()))>::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;
};
}
}

// #include <nlohmann/detail/input/lexer.hpp>


#include <clocale> // localeconv
#include <cstddef> // size_t
#include <cstdlib> // strtof, strtod, strtold, strtoll, strtoull
#include <cstdio> // snprintf
#include <initializer_list> // initializer_list
#include <string> // char_traits, string
#include <vector> // vector

// #include <nlohmann/detail/macro_scope.hpp>

// #include <nlohmann/detail/input/input_adapters.hpp>


namespace nlohmann
{
namespace detail
{
///////////
// lexer //
///////////

/*!
@brief lexical analysis

This class organizes the lexical analysis during JSON deserialization.
*/
template<typename BasicJsonType>
class lexer
{
    using number_integer_t = typename BasicJsonType::number_integer_t;
    using number_unsigned_t = typename BasicJsonType::number_unsigned_t;
    using number_float_t = typename BasicJsonType::number_float_t;
    using string_t = typename BasicJsonType::string_t;

  public:
    /// token types for the parser
    enum class token_type
    {
        uninitialized,    ///< indicating the scanner is uninitialized
        literal_true,     ///< the `true` literal
        literal_false,    ///< the `false` literal
        literal_null,     ///< the `null` literal
        value_string,     ///< a string -- use get_string() for actual value
        value_unsigned,   ///< an unsigned integer -- use get_number_unsigned() for actual value
        value_integer,    ///< a signed integer -- use get_number_integer() for actual value
        value_float,      ///< an floating point number -- use get_number_float() for actual value
        begin_array,      ///< the character for array begin `[`
        begin_object,     ///< the character for object begin `{`
        end_array,        ///< the character for array end `]`
        end_object,       ///< the character for object end `}`
        name_separator,   ///< the name separator `:`
        value_separator,  ///< the value separator `,`
        parse_error,      ///< indicating a parse error
        end_of_input,     ///< indicating the end of the input buffer
        literal_or_value  ///< a literal or the begin of a value (only for diagnostics)
    };

    /// return name of values of type token_type (only used for errors)
    static const char* token_type_name(const token_type t) noexcept
    {
        switch (t)
        {
            case token_type::uninitialized:
                return "<uninitialized>";
            case token_type::literal_true:
                return "true literal";
            case token_type::literal_false:
                return "false literal";
            case token_type::literal_null:
                return "null literal";
            case token_type::value_string:
                return "string literal";
            case lexer::token_type::value_unsigned:
            case lexer::token_type::value_integer:
            case lexer::token_type::value_float:
                return "number literal";
            case token_type::begin_array:
                return "'['";
            case token_type::begin_object:
                return "'{'";
            case token_type::end_array:
                return "']'";
            case token_type::end_object:
                return "'}'";
            case token_type::name_separator:
                return "':'";
            case token_type::value_separator:
                return "','";
            case token_type::parse_error:
                return "<parse error>";
            case token_type::end_of_input:
                return "end of input";
            case token_type::literal_or_value:
                return "'[', '{', or a literal";
            default: // catch non-enum values
                return "unknown token"; // LCOV_EXCL_LINE
        }
    }

    explicit lexer(detail::input_adapter_t&& adapter)
        : ia(std::move(adapter)), decimal_point_char(get_decimal_point()) {}

    // delete because of pointer members
    lexer(const lexer&) = delete;
    lexer& operator=(lexer&) = delete;

  private:
    /////////////////////
    // locales
    /////////////////////

    /// return the locale-dependent decimal point
    static char get_decimal_point() noexcept
    {
        const auto loc = localeconv();
        assert(loc != nullptr);
        return (loc->decimal_point == nullptr) ? '.' : *(loc->decimal_point);
    }

    /////////////////////
    // scan functions
    /////////////////////

    /*!
    @brief get codepoint from 4 hex characters following `\u`

    For input "\u c1 c2 c3 c4" the codepoint is:
      (c1 * 0x1000) + (c2 * 0x0100) + (c3 * 0x0010) + c4
    = (c1 << 12) + (c2 << 8) + (c3 << 4) + (c4 << 0)

    Furthermore, the possible characters '0'..'9', 'A'..'F', and 'a'..'f'
    must be converted to the integers 0x0..0x9, 0xA..0xF, 0xA..0xF, resp. The
    conversion is done by subtracting the offset (0x30, 0x37, and 0x57)
    between the ASCII value of the character and the desired integer value.

    @return codepoint (0x0000..0xFFFF) or -1 in case of an error (e.g. EOF or
            non-hex character)
    */
    int get_codepoint()
    {
        // this function only makes sense after reading `\u`
        assert(current == 'u');
        int codepoint = 0;

        const auto factors = { 12, 8, 4, 0 };
        for (const auto factor : factors)
        {
            get();

            if (current >= '0' and current <= '9')
            {
                codepoint += ((current - 0x30) << factor);
            }
            else if (current >= 'A' and current <= 'F')
            {
                codepoint += ((current - 0x37) << factor);
            }
            else if (current >= 'a' and current <= 'f')
            {
                codepoint += ((current - 0x57) << factor);
            }
            else
            {
                return -1;
            }
        }

        assert(0x0000 <= codepoint and codepoint <= 0xFFFF);
        return codepoint;
    }

    /*!
    @brief check if the next byte(s) are inside a given range

    Adds the current byte and, for each passed range, reads a new byte and
    checks if it is inside the range. If a violation was detected, set up an
    error message and return false. Otherwise, return true.

    @param[in] ranges  list of integers; interpreted as list of pairs of
                       inclusive lower and upper bound, respectively

    @pre The passed list @a ranges must have 2, 4, or 6 elements; that is,
         1, 2, or 3 pairs. This precondition is enforced by an assertion.

    @return true if and only if no range violation was detected
    */
    bool next_byte_in_range(std::initializer_list<int> ranges)
    {
        assert(ranges.size() == 2 or ranges.size() == 4 or ranges.size() == 6);
        add(current);

        for (auto range = ranges.begin(); range != ranges.end(); ++range)
        {
            get();
            if (JSON_LIKELY(*range <= current and current <= *(++range)))
            {
                add(current);
            }
            else
            {
                error_message = "invalid string: ill-formed UTF-8 byte";
                return false;
            }
        }

        return true;
    }

    /*!
    @brief scan a string literal

    This function scans a string according to Sect. 7 of RFC 7159. While
    scanning, bytes are escaped and copied into buffer token_buffer. Then the
    function returns successfully, token_buffer is *not* null-terminated (as it
    may contain \0 bytes), and token_buffer.size() is the number of bytes in the
    string.

    @return token_type::value_string if string could be successfully scanned,
            token_type::parse_error otherwise

    @note In case of errors, variable error_message contains a textual
          description.
    */
    token_type scan_string()
    {
        // reset token_buffer (ignore opening quote)
        reset();

        // we entered the function by reading an open quote
        assert(current == '\"');

        while (true)
        {
            // get next character
            switch (get())
            {
                // end of file while parsing string
                case std::char_traits<char>::eof():
                {
                    error_message = "invalid string: missing closing quote";
                    return token_type::parse_error;
                }

                // closing quote
                case '\"':
                {
                    return token_type::value_string;
                }

                // escapes
                case '\\':
                {
                    switch (get())
                    {
                        // quotation mark
                        case '\"':
                            add('\"');
                            break;
                        // reverse solidus
                        case '\\':
                            add('\\');
                            break;
                        // solidus
                        case '/':
                            add('/');
                            break;
                        // backspace
                        case 'b':
                            add('\b');
                            break;
                        // form feed
                        case 'f':
                            add('\f');
                            break;
                        // line feed
                        case 'n':
                            add('\n');
                            break;
                        // carriage return
                        case 'r':
                            add('\r');
                            break;
                        // tab
                        case 't':
                            add('\t');
                            break;

                        // unicode escapes
                        case 'u':
                        {
                            const int codepoint1 = get_codepoint();
                            int codepoint = codepoint1; // start with codepoint1

                            if (JSON_UNLIKELY(codepoint1 == -1))
                            {
                                error_message = "invalid string: '\\u' must be followed by 4 hex digits";
                                return token_type::parse_error;
                            }

                            // check if code point is a high surrogate
                            if (0xD800 <= codepoint1 and codepoint1 <= 0xDBFF)
                            {
                                // expect next \uxxxx entry
                                if (JSON_LIKELY(get() == '\\' and get() == 'u'))
                                {
                                    const int codepoint2 = get_codepoint();

                                    if (JSON_UNLIKELY(codepoint2 == -1))
                                    {
                                        error_message = "invalid string: '\\u' must be followed by 4 hex digits";
                                        return token_type::parse_error;
                                    }

                                    // check if codepoint2 is a low surrogate
                                    if (JSON_LIKELY(0xDC00 <= codepoint2 and codepoint2 <= 0xDFFF))
                                    {
                                        // overwrite codepoint
                                        codepoint =
                                            // high surrogate occupies the most significant 22 bits
                                            (codepoint1 << 10)
                                            // low surrogate occupies the least significant 15 bits
                                            + codepoint2
                                            // there is still the 0xD800, 0xDC00 and 0x10000 noise
                                            // in the result so we have to subtract with:
                                            // (0xD800 << 10) + DC00 - 0x10000 = 0x35FDC00
                                            - 0x35FDC00;
                                    }
                                    else
                                    {
                                        error_message = "invalid string: surrogate U+DC00..U+DFFF must be followed by U+DC00..U+DFFF";
                                        return token_type::parse_error;
                                    }
                                }
                                else
                                {
                                    error_message = "invalid string: surrogate U+DC00..U+DFFF must be followed by U+DC00..U+DFFF";
                                    return token_type::parse_error;
                                }
                            }
                            else
                            {
                                if (JSON_UNLIKELY(0xDC00 <= codepoint1 and codepoint1 <= 0xDFFF))
                                {
                                    error_message = "invalid string: surrogate U+DC00..U+DFFF must follow U+D800..U+DBFF";
                                    return token_type::parse_error;
                                }
                            }

                            // result of the above calculation yields a proper codepoint
                            assert(0x00 <= codepoint and codepoint <= 0x10FFFF);

                            // translate codepoint into bytes
                            if (codepoint < 0x80)
                            {
                                // 1-byte characters: 0xxxxxxx (ASCII)
                                add(codepoint);
                            }
                            else if (codepoint <= 0x7FF)
                            {
                                // 2-byte characters: 110xxxxx 10xxxxxx
                                add(0xC0 | (codepoint >> 6));
                                add(0x80 | (codepoint & 0x3F));
                            }
                            else if (codepoint <= 0xFFFF)
                            {
                                // 3-byte characters: 1110xxxx 10xxxxxx 10xxxxxx
                                add(0xE0 | (codepoint >> 12));
                                add(0x80 | ((codepoint >> 6) & 0x3F));
                                add(0x80 | (codepoint & 0x3F));
                            }
                            else
                            {
                                // 4-byte characters: 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
                                add(0xF0 | (codepoint >> 18));
                                add(0x80 | ((codepoint >> 12) & 0x3F));
                                add(0x80 | ((codepoint >> 6) & 0x3F));
                                add(0x80 | (codepoint & 0x3F));
                            }

                            break;
                        }

                        // other characters after escape
                        default:
                            error_message = "invalid string: forbidden character after backslash";
                            return token_type::parse_error;
                    }

                    break;
                }

                // invalid control characters
                case 0x00:
                case 0x01:
                case 0x02:
                case 0x03:
                case 0x04:
                case 0x05:
                case 0x06:
                case 0x07:
                case 0x08:
                case 0x09:
                case 0x0A:
                case 0x0B:
                case 0x0C:
                case 0x0D:
                case 0x0E:
                case 0x0F:
                case 0x10:
                case 0x11:
                case 0x12:
                case 0x13:
                case 0x14:
                case 0x15:
                case 0x16:
                case 0x17:
                case 0x18:
                case 0x19:
                case 0x1A:
                case 0x1B:
                case 0x1C:
                case 0x1D:
                case 0x1E:
                case 0x1F:
                {
                    error_message = "invalid string: control character must be escaped";
                    return token_type::parse_error;
                }

                // U+0020..U+007F (except U+0022 (quote) and U+005C (backspace))
                case 0x20:
                case 0x21:
                case 0x23:
                case 0x24:
                case 0x25:
                case 0x26:
                case 0x27:
                case 0x28:
                case 0x29:
                case 0x2A:
                case 0x2B:
                case 0x2C:
                case 0x2D:
                case 0x2E:
                case 0x2F:
                case 0x30:
                case 0x31:
                case 0x32:
                case 0x33:
                case 0x34:
                case 0x35:
                case 0x36:
                case 0x37:
                case 0x38:
                case 0x39:
                case 0x3A:
                case 0x3B:
                case 0x3C:
                case 0x3D:
                case 0x3E:
                case 0x3F:
                case 0x40:
                case 0x41:
                case 0x42:
                case 0x43:
                case 0x44:
                case 0x45:
                case 0x46:
                case 0x47:
                case 0x48:
                case 0x49:
                case 0x4A:
                case 0x4B:
                case 0x4C:
                case 0x4D:
                case 0x4E:
                case 0x4F:
                case 0x50:
                case 0x51:
                case 0x52:
                case 0x53:
                case 0x54:
                case 0x55:
                case 0x56:
                case 0x57:
                case 0x58:
                case 0x59:
                case 0x5A:
                case 0x5B:
                case 0x5D:
                case 0x5E:
                case 0x5F:
                case 0x60:
                case 0x61:
                case 0x62:
                case 0x63:
                case 0x64:
                case 0x65:
                case 0x66:
                case 0x67:
                case 0x68:
                case 0x69:
                case 0x6A:
                case 0x6B:
                case 0x6C:
                case 0x6D:
                case 0x6E:
                case 0x6F:
                case 0x70:
                case 0x71:
                case 0x72:
                case 0x73:
                case 0x74:
                case 0x75:
                case 0x76:
                case 0x77:
                case 0x78:
                case 0x79:
                case 0x7A:
                case 0x7B:
                case 0x7C:
                case 0x7D:
                case 0x7E:
                case 0x7F:
                {
                    add(current);
                    break;
                }

                // U+0080..U+07FF: bytes C2..DF 80..BF
                case 0xC2:
                case 0xC3:
                case 0xC4:
                case 0xC5:
                case 0xC6:
                case 0xC7:
                case 0xC8:
                case 0xC9:
                case 0xCA:
                case 0xCB:
                case 0xCC:
                case 0xCD:
                case 0xCE:
                case 0xCF:
                case 0xD0:
                case 0xD1:
                case 0xD2:
                case 0xD3:
                case 0xD4:
                case 0xD5:
                case 0xD6:
                case 0xD7:
                case 0xD8:
                case 0xD9:
                case 0xDA:
                case 0xDB:
                case 0xDC:
                case 0xDD:
                case 0xDE:
                case 0xDF:
                {
                    if (JSON_UNLIKELY(not next_byte_in_range({0x80, 0xBF})))
                    {
                        return token_type::parse_error;
                    }
                    break;
                }

                // U+0800..U+0FFF: bytes E0 A0..BF 80..BF
                case 0xE0:
                {
                    if (JSON_UNLIKELY(not (next_byte_in_range({0xA0, 0xBF, 0x80, 0xBF}))))
                    {
                        return token_type::parse_error;
                    }
                    break;
                }

                // U+1000..U+CFFF: bytes E1..EC 80..BF 80..BF
                // U+E000..U+FFFF: bytes EE..EF 80..BF 80..BF
                case 0xE1:
                case 0xE2:
                case 0xE3:
                case 0xE4:
                case 0xE5:
                case 0xE6:
                case 0xE7:
                case 0xE8:
                case 0xE9:
                case 0xEA:
                case 0xEB:
                case 0xEC:
                case 0xEE:
                case 0xEF:
                {
                    if (JSON_UNLIKELY(not (next_byte_in_range({0x80, 0xBF, 0x80, 0xBF}))))
                    {
                        return token_type::parse_error;
                    }
                    break;
                }

                // U+D000..U+D7FF: bytes ED 80..9F 80..BF
                case 0xED:
                {
                    if (JSON_UNLIKELY(not (next_byte_in_range({0x80, 0x9F, 0x80, 0xBF}))))
                    {
                        return token_type::parse_error;
                    }
                    break;
                }

                // U+10000..U+3FFFF F0 90..BF 80..BF 80..BF
                case 0xF0:
                {
                    if (JSON_UNLIKELY(not (next_byte_in_range({0x90, 0xBF, 0x80, 0xBF, 0x80, 0xBF}))))
                    {
                        return token_type::parse_error;
                    }
                    break;
                }

                // U+40000..U+FFFFF F1..F3 80..BF 80..BF 80..BF
                case 0xF1:
                case 0xF2:
                case 0xF3:
                {
                    if (JSON_UNLIKELY(not (next_byte_in_range({0x80, 0xBF, 0x80, 0xBF, 0x80, 0xBF}))))
                    {
                        return token_type::parse_error;
                    }
                    break;
                }

                // U+100000..U+10FFFF F4 80..8F 80..BF 80..BF
                case 0xF4:
                {
                    if (JSON_UNLIKELY(not (next_byte_in_range({0x80, 0x8F, 0x80, 0xBF, 0x80, 0xBF}))))
                    {
                        return token_type::parse_error;
                    }
                    break;
                }

                // remaining bytes (80..C1 and F5..FF) are ill-formed
                default:
                {
                    error_message = "invalid string: ill-formed UTF-8 byte";
                    return token_type::parse_error;
                }
            }
        }
    }

    static void strtof(float& f, const char* str, char** endptr) noexcept
    {
        f = std::strtof(str, endptr);
    }

    static void strtof(double& f, const char* str, char** endptr) noexcept
    {
        f = std::strtod(str, endptr);
    }

    static void strtof(long double& f, const char* str, char** endptr) noexcept
    {
        f = std::strtold(str, endptr);
    }

    /*!
    @brief scan a number literal

    This function scans a string according to Sect. 6 of RFC 7159.

    The function is realized with a deterministic finite state machine derived
    from the grammar described in RFC 7159. Starting in state "init", the
    input is read and used to determined the next state. Only state "done"
    accepts the number. State "error" is a trap state to model errors. In the
    table below, "anything" means any character but the ones listed before.

    state    | 0        | 1-9      | e E      | +       | -       | .        | anything
    ---------|----------|----------|----------|---------|---------|----------|-----------
    init     | zero     | any1     | [error]  | [error] | minus   | [error]  | [error]
    minus    | zero     | any1     | [error]  | [error] | [error] | [error]  | [error]
    zero     | done     | done     | exponent | done    | done    | decimal1 | done
    any1     | any1     | any1     | exponent | done    | done    | decimal1 | done
    decimal1 | decimal2 | [error]  | [error]  | [error] | [error] | [error]  | [error]
    decimal2 | decimal2 | decimal2 | exponent | done    | done    | done     | done
    exponent | any2     | any2     | [error]  | sign    | sign    | [error]  | [error]
    sign     | any2     | any2     | [error]  | [error] | [error] | [error]  | [error]
    any2     | any2     | any2     | done     | done    | done    | done     | done

    The state machine is realized with one label per state (prefixed with
    "scan_number_") and `goto` statements between them. The state machine
    contains cycles, but any cycle can be left when EOF is read. Therefore,
    the function is guaranteed to terminate.

    During scanning, the read bytes are stored in token_buffer. This string is
    then converted to a signed integer, an unsigned integer, or a
    floating-point number.

    @return token_type::value_unsigned, token_type::value_integer, or
            token_type::value_float if number could be successfully scanned,
            token_type::parse_error otherwise

    @note The scanner is independent of the current locale. Internally, the
          locale's decimal point is used instead of `.` to work with the
          locale-dependent converters.
    */
    token_type scan_number()
    {
        // reset token_buffer to store the number's bytes
        reset();

        // the type of the parsed number; initially set to unsigned; will be
        // changed if minus sign, decimal point or exponent is read
        token_type number_type = token_type::value_unsigned;

        // state (init): we just found out we need to scan a number
        switch (current)
        {
            case '-':
            {
                add(current);
                goto scan_number_minus;
            }

            case '0':
            {
                add(current);
                goto scan_number_zero;
            }

            case '1':
            case '2':
            case '3':
            case '4':
            case '5':
            case '6':
            case '7':
            case '8':
            case '9':
            {
                add(current);
                goto scan_number_any1;
            }

            default:
            {
                // all other characters are rejected outside scan_number()
                assert(false); // LCOV_EXCL_LINE
            }
        }

scan_number_minus:
        // state: we just parsed a leading minus sign
        number_type = token_type::value_integer;
        switch (get())
        {
            case '0':
            {
                add(current);
                goto scan_number_zero;
            }

            case '1':
            case '2':
            case '3':
            case '4':
            case '5':
            case '6':
            case '7':
            case '8':
            case '9':
            {
                add(current);
                goto scan_number_any1;
            }

            default:
            {
                error_message = "invalid number; expected digit after '-'";
                return token_type::parse_error;
            }
        }

scan_number_zero:
        // state: we just parse a zero (maybe with a leading minus sign)
        switch (get())
        {
            case '.':
            {
                add(decimal_point_char);
                goto scan_number_decimal1;
            }

            case 'e':
            case 'E':
            {
                add(current);
                goto scan_number_exponent;
            }

            default:
                goto scan_number_done;
        }

scan_number_any1:
        // state: we just parsed a number 0-9 (maybe with a leading minus sign)
        switch (get())