MessagePack reader: read runs of scalar array elements in one loop

As for CBOR: when an array opens and the input is contiguous, read the
integers, floats, booleans, and nils that follow in a tight loop instead
of going through parse_msgpack_internal and parse_msgpack_value for every
element. The loop stops at any other kind of value and when fewer than 9
bytes are left; those values are read by the general path as before.
Non-contiguous inputs are not affected.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

include/nlohmann/detail/input/binary_reader.hpp

@@ -22,6 +22,7 @@
     #include <bit>  //byteswap
 #endif
 
+#include <nlohmann/detail/bit_ops.hpp>
 #include <nlohmann/detail/exceptions.hpp>
 #include <nlohmann/detail/input/input_adapters.hpp>
 #include <nlohmann/detail/input/json_sax.hpp>
@@ -741,11 +742,13 @@ class binary_reader
     bool get_cbor_negative_integer()
     {
         NumberType number{};
-        if (JSON_HEDLEY_UNLIKELY(!get_number(number)))
-        {
-            return false;
-        }
+        return get_number(number) && emit_cbor_negative_integer(number);
+    }
 
+    /// pass the CBOR negative integer -1 - @a number to the SAX parser
+    template<typename NumberType>
+    bool emit_cbor_negative_integer(const NumberType number)
+    {
         // the value is -1 - number, which fits into number_integer_t
         // whenever number does; the outer cast undoes the integral promotion
         // for number_integer_t types narrower than int
@@ -970,7 +973,7 @@ class binary_reader
             case 0x95:
             case 0x96:
             case 0x97:
-                return enter_array(conditional_static_cast<std::size_t>(static_cast<unsigned int>(current) & 0x1Fu));
+                return enter_cbor_array(conditional_static_cast<std::size_t>(static_cast<unsigned int>(current) & 0x1Fu));
 
             case 0x98: // array (one-byte uint8_t for n follows)
             case 0x99: // array (two-byte uint16_t for n follow)
@@ -979,7 +982,7 @@ class binary_reader
             {
                 std::uint64_t len{};
                 std::size_t size{};
-                return get_cbor_argument(len) && get_cbor_container_size(len, size, "array") && enter_array(size);
+                return get_cbor_argument(len) && get_cbor_container_size(len, size, "array") && enter_cbor_array(size);
             }
 
             case 0x9F: // array (indefinite length)
@@ -1133,16 +1136,10 @@ class binary_reader
                 return get_half_float(false);
 
             case 0xFA: // Single-Precision Float (four-byte IEEE 754)
-            {
-                float number{};
-                return get_number(number) && emit_float(number);
-            }
+                return get_cbor_float<float>();
 
             case 0xFB: // Double-Precision Float (eight-byte IEEE 754)
-            {
-                double number{};
-                return get_number(number) && emit_float(number);
-            }
+                return get_cbor_float<double>();
 
             default: // anything else (0xFF is handled inside the other types)
             {
@@ -1151,6 +1148,16 @@ class binary_reader
         }
     }
 
+    /// read a single- or double-precision float (kept out of
+    /// parse_cbor_value, which is inlined into parse_cbor_internal)
+    template<typename NumberType>
+    JSON_HEDLEY_NEVER_INLINE
+    bool get_cbor_float()
+    {
+        NumberType number{};
+        return get_number(number) && emit_float(number);
+    }
+
     /*!
     @brief reports a nested indefinite-length CBOR string or byte array
     @param[in] type_name  name of the rejected string type
@@ -1676,6 +1683,157 @@ class binary_reader
         return true;
     }
 
+    /*!
+    @brief read a big-endian number of type @a NumberType at @a data
+    */
+    template<typename NumberType>
+    NumberType load_big_endian(const unsigned char* data) const noexcept
+    {
+        NumberType result{};
+        std::memcpy(&result, data, sizeof(NumberType));
+        if (is_little_endian)
+        {
+            byte_swap(result);
+        }
+        return result;
+    }
+
+    /*!
+    @brief read the integers, floats, booleans, and nulls that follow in a
+           definite-length array, without the per-element bookkeeping of
+           @ref parse_cbor_internal
+
+    Stops at the end of the array, at any other kind of value, and when fewer
+    than 9 bytes (the longest of these values) are left in the input, leaving
+    those to @ref parse_cbor_value, which reads them the same way.
+
+    @param[in,out] remaining  the number of elements of the array not yet read
+    @return whether the SAX parser accepted the values
+    */
+    bool parse_cbor_scalar_run(std::size_t& remaining)
+    {
+        while (remaining != 0 && ia.bulk_remaining() >= 9)
+        {
+            const auto* const data = reinterpret_cast<const unsigned char*>(ia.bulk_data());
+            const unsigned char head = data[0];
+            std::size_t size = 0;
+            bool accepted = true;
+            switch (head)
+            {
+                case 0x18:
+                    size = 2;
+                    accepted = emit_unsigned(data[1]);
+                    break;
+                case 0x19:
+                    size = 3;
+                    accepted = emit_unsigned(load_big_endian<std::uint16_t>(data + 1));
+                    break;
+                case 0x1A:
+                    size = 5;
+                    accepted = emit_unsigned(load_big_endian<std::uint32_t>(data + 1));
+                    break;
+                case 0x1B:
+                    size = 9;
+                    accepted = emit_unsigned(load_big_endian<std::uint64_t>(data + 1));
+                    break;
+                case 0x38:
+                    size = 2;
+                    accepted = emit_cbor_negative_integer(data[1]);
+                    break;
+                case 0x39:
+                    size = 3;
+                    accepted = emit_cbor_negative_integer(load_big_endian<std::uint16_t>(data + 1));
+                    break;
+                case 0x3A:
+                    size = 5;
+                    accepted = emit_cbor_negative_integer(load_big_endian<std::uint32_t>(data + 1));
+                    break;
+                case 0x3B:
+                    size = 9;
+                    accepted = emit_cbor_negative_integer(load_big_endian<std::uint64_t>(data + 1));
+                    break;
+                case 0xF4:
+                    size = 1;
+                    accepted = sax->boolean(false);
+                    break;
+                case 0xF5:
+                    size = 1;
+                    accepted = sax->boolean(true);
+                    break;
+                case 0xF6:
+                    size = 1;
+                    accepted = sax->null();
+                    break;
+                case 0xFA:
+                    size = 5;
+                    accepted = emit_float(load_big_endian<float>(data + 1));
+                    break;
+                case 0xFB:
+                    size = 9;
+                    accepted = emit_float(load_big_endian<double>(data + 1));
+                    break;
+                default:
+                    if (head <= 0x17)
+                    {
+                        size = 1;
+                        accepted = sax->number_unsigned(static_cast<number_unsigned_t>(head));
+                    }
+                    else if (head >= 0x20 && head <= 0x37)
+                    {
+                        size = 1;
+                        accepted = sax->number_integer(static_cast<std::int8_t>(0x20 - 1 - head));
+                    }
+                    else
+                    {
+                        // any other value is read by parse_cbor_value
+                        return true;
+                    }
+                    break;
+            }
+            // as if read by get() and get_number(): current is the head byte
+            ia.bulk_skip(size);
+            chars_read += size;
+            current = head;
+            --remaining;
+            if (JSON_HEDLEY_UNLIKELY(!accepted))
+            {
+                return false;
+            }
+        }
+        return true;
+    }
+
+    /*!
+    @brief open a definite-length array and read the scalars it begins with
+           (see @ref parse_cbor_scalar_run)
+
+    Only the start of an array is tried, which covers arrays of numbers
+    (e.g. coordinates) but costs arrays of other values only one test, not one
+    per element.
+
+    @param[in] len  the number of elements
+    @return whether the SAX parser accepted the array and the values read
+    */
+    bool enter_cbor_array(const std::size_t len)
+    {
+        return enter_array(len) && (len == 0 || parse_cbor_array_start(std::integral_constant<bool, bulk_scan> {}));
+    }
+
+    /// read the scalars the array just opened begins with; out of line, so
+    /// that @ref parse_cbor_value stays small enough to be inlined into
+    /// @ref parse_cbor_internal
+    JSON_HEDLEY_NEVER_INLINE
+    bool parse_cbor_array_start(std::true_type /*bulk*/)
+    {
+        return parse_cbor_scalar_run(container_stack.back().remaining);
+    }
+
+    /// input that is not contiguous: every value is read by parse_cbor_value
+    bool parse_cbor_array_start(std::false_type /*bulk*/) const noexcept
+    {
+        return true;
+    }
+
     /*!
     @brief read a CBOR value and everything nested inside it
 
@@ -1705,6 +1863,7 @@ class binary_reader
                 // pop_back() below, which destroys the container_stack element
                 // it would otherwise alias
                 const container_frame top = container_stack.back();
+                const bool is_object = top.is_object;
                 bool at_end = false;
 
                 if (top.remaining != npos)
@@ -1716,7 +1875,7 @@ class binary_reader
                     {
                         // claim the element about to be read
                         --container_stack.back().remaining;
-                        if (top.is_object)
+                        if (is_object)
                         {
                             get();
                         }
@@ -1729,7 +1888,7 @@ class binary_reader
                     // Testing for it consumes a byte, which is the first byte
                     // of the next element when it is not one.
                     at_end = (get() == 0xFF);
-                    fetch = top.is_object;
+                    fetch = is_object;
                 }
 
                 if (at_end)
@@ -1746,7 +1905,7 @@ class binary_reader
                     continue;
                 }
 
-                if (top.is_object)
+                if (is_object)
                 {
                     if (JSON_HEDLEY_UNLIKELY(!skip_cbor_key_tags(tag_handler)))
                     {
@@ -1971,7 +2130,7 @@ class binary_reader
             case 0x9D:
             case 0x9E:
             case 0x9F:
-                return enter_array(conditional_static_cast<std::size_t>(static_cast<unsigned int>(current) & 0x0Fu));
+                return enter_msgpack_array(conditional_static_cast<std::size_t>(static_cast<unsigned int>(current) & 0x0Fu));
 
             // fixstr
             case 0xA0:
@@ -2102,13 +2261,13 @@ class binary_reader
             case 0xDC: // array 16
             {
                 std::uint16_t len{};
-                return get_number(len) && enter_array(static_cast<std::size_t>(len));
+                return get_number(len) && enter_msgpack_array(static_cast<std::size_t>(len));
             }
 
             case 0xDD: // array 32
             {
                 std::uint32_t len{};
-                return get_number(len) && enter_array(conditional_static_cast<std::size_t>(len));
+                return get_number(len) && enter_msgpack_array(conditional_static_cast<std::size_t>(len));
             }
 
             case 0xDE: // map 16
@@ -2450,6 +2609,138 @@ class binary_reader
         }
     }
 
+    /*!
+    @brief read the integers, floats, booleans, and nils that follow in a
+           MessagePack array, without the per-element bookkeeping of
+           @ref parse_msgpack_internal
+
+    Stops at the end of the array, at any other kind of value, and when fewer
+    than 9 bytes (the longest of these values) are left in the input, leaving
+    those to @ref parse_msgpack_value, which reads them the same way.
+
+    @param[in,out] remaining  the number of elements of the array not yet read
+    @return whether the SAX parser accepted the values
+    */
+    bool parse_msgpack_scalar_run(std::size_t& remaining)
+    {
+        while (remaining != 0 && ia.bulk_remaining() >= 9)
+        {
+            const auto* const data = reinterpret_cast<const unsigned char*>(ia.bulk_data());
+            const unsigned char head = data[0];
+            std::size_t size = 0;
+            bool accepted = true;
+            switch (head)
+            {
+                case 0xC0: // nil
+                    size = 1;
+                    accepted = sax->null();
+                    break;
+                case 0xC2: // false
+                    size = 1;
+                    accepted = sax->boolean(false);
+                    break;
+                case 0xC3: // true
+                    size = 1;
+                    accepted = sax->boolean(true);
+                    break;
+                case 0xCA: // float 32
+                    size = 5;
+                    accepted = emit_float(load_big_endian<float>(data + 1));
+                    break;
+                case 0xCB: // float 64
+                    size = 9;
+                    accepted = emit_float(load_big_endian<double>(data + 1));
+                    break;
+                case 0xCC: // uint 8
+                    size = 2;
+                    accepted = emit_unsigned(data[1]);
+                    break;
+                case 0xCD: // uint 16
+                    size = 3;
+                    accepted = emit_unsigned(load_big_endian<std::uint16_t>(data + 1));
+                    break;
+                case 0xCE: // uint 32
+                    size = 5;
+                    accepted = emit_unsigned(load_big_endian<std::uint32_t>(data + 1));
+                    break;
+                case 0xCF: // uint 64
+                    size = 9;
+                    accepted = emit_unsigned(load_big_endian<std::uint64_t>(data + 1));
+                    break;
+                case 0xD0: // int 8
+                    size = 2;
+                    accepted = emit_signed(static_cast<std::int8_t>(data[1]));
+                    break;
+                case 0xD1: // int 16
+                    size = 3;
+                    accepted = emit_signed(load_big_endian<std::int16_t>(data + 1));
+                    break;
+                case 0xD2: // int 32
+                    size = 5;
+                    accepted = emit_signed(load_big_endian<std::int32_t>(data + 1));
+                    break;
+                case 0xD3: // int 64
+                    size = 9;
+                    accepted = emit_signed(load_big_endian<std::int64_t>(data + 1));
+                    break;
+                default:
+                    if (head <= 0x7F)
+                    {
+                        // positive fixint
+                        size = 1;
+                        accepted = sax->number_unsigned(static_cast<number_unsigned_t>(head));
+                    }
+                    else if (head >= 0xE0)
+                    {
+                        // negative fixint
+                        size = 1;
+                        accepted = sax->number_integer(static_cast<std::int8_t>(head));
+                    }
+                    else
+                    {
+                        // any other value is read by parse_msgpack_value
+                        return true;
+                    }
+                    break;
+            }
+            // as if read by get() and get_number(): current is the last byte
+            ia.bulk_skip(size);
+            chars_read += size;
+            current = data[size - 1];
+            --remaining;
+            if (JSON_HEDLEY_UNLIKELY(!accepted))
+            {
+                return false;
+            }
+        }
+        return true;
+    }
+
+    /*!
+    @brief open a MessagePack array and read the scalars it begins with (see
+           @ref parse_msgpack_scalar_run)
+    @param[in] len  the number of elements
+    @return whether the SAX parser accepted the array and the values read
+    */
+    bool enter_msgpack_array(const std::size_t len)
+    {
+        return enter_array(len) && (len == 0 || parse_msgpack_array_start(std::integral_constant<bool, bulk_scan> {}));
+    }
+
+    /// read the scalars the array just opened begins with; out of line, so
+    /// that @ref parse_msgpack_value stays small enough to be inlined
+    JSON_HEDLEY_NEVER_INLINE
+    bool parse_msgpack_array_start(std::true_type /*bulk*/)
+    {
+        return parse_msgpack_scalar_run(container_stack.back().remaining);
+    }
+
+    /// input that is not contiguous: every value is read by parse_msgpack_value
+    bool parse_msgpack_array_start(std::false_type /*bulk*/) const noexcept
+    {
+        return true;
+    }
+
     /*!
     @brief read a MessagePack value and everything nested inside it
 
@@ -3691,13 +3982,13 @@ class binary_reader
                 return sax->null();
 
             case 0xFB:
-                return sax->number_float(static_cast<number_float_t>(-1.0), "");
+                return sax->number_float(static_cast<number_float_t>(-1.0), no_source);
 
             case 0xFC:
-                return sax->number_float(static_cast<number_float_t>(0.0), "");
+                return sax->number_float(static_cast<number_float_t>(0.0), no_source);
 
             case 0xFD:
-                return sax->number_float(static_cast<number_float_t>(1.0), "");
+                return sax->number_float(static_cast<number_float_t>(1.0), no_source);
 
             case 0xFF: // empty string
             {
@@ -4066,6 +4357,27 @@ class binary_reader
         return current;
     }
 
+    /// reverse the byte order of a double with a single byteswap(): GCC does
+    /// not merge the byte-wise swap of the generic version into one
+    static void byte_swap(double& number) noexcept
+    {
+        static_assert(sizeof(double) == sizeof(std::uint64_t), "double must have 64 bits");
+        std::uint64_t bits{};
+        std::memcpy(&bits, &number, sizeof(bits));
+        bits = byteswap(bits);
+        std::memcpy(&number, &bits, sizeof(bits));
+    }
+
+    /// reverse the byte order of a float with a single byteswap()
+    static void byte_swap(float& number) noexcept
+    {
+        static_assert(sizeof(float) == sizeof(std::uint32_t), "float must have 32 bits");
+        std::uint32_t bits{};
+        std::memcpy(&bits, &number, sizeof(bits));
+        bits = byteswap(bits);
+        std::memcpy(&number, &bits, sizeof(bits));
+    }
+
     template<class NumberType>
     static void byte_swap(NumberType& number)
     {
@@ -4201,7 +4513,7 @@ class binary_reader
         {
             return out_of_range_error(exception_id::number_overflow, "number overflow", "value");
         }
-        return sax->number_float(result, "");
+        return sax->number_float(result, no_source);
     }
 
     /*!
@@ -4261,7 +4573,7 @@ class binary_reader
         }();
         return sax->number_float((half & 0x8000u) != 0
                                  ? static_cast<number_float_t>(-val)
-                                 : static_cast<number_float_t>(val), "");
+                                 : static_cast<number_float_t>(val), no_source);
     }
 
     /*!
@@ -4586,6 +4898,10 @@ class binary_reader
     /// the SAX parser
     json_sax_t* sax = nullptr;
 
+    /// the (empty) source text of every number_float event: binary formats
+    /// have none, and a string_t built for each float would cost more
+    const string_t no_source{};
+
     /// the containers that have been opened and not closed yet; see @ref container_frame
     std::vector<container_frame> container_stack{};
 

single_include/nlohmann/json.hpp

@@ -8852,6 +8852,233 @@ NLOHMANN_JSON_NAMESPACE_END
     #include <bit>  //byteswap
 #endif
 
+// #include <nlohmann/detail/bit_ops.hpp>
+//     __ _____ _____ _____
+//  __|  |   __|     |   | |  JSON for Modern C++
+// |  |  |__   |  |  | | | |  version 3.12.0
+// |_____|_____|_____|_|___|  https://github.com/nlohmann/json
+//
+// SPDX-FileCopyrightText: 2013-2026 Niels Lohmann <https://nlohmann.me>
+// SPDX-License-Identifier: MIT
+
+
+
+#include <cstdint> // uint64_t
+#include <cstring> // memcpy
+#if defined(_MSC_VER) && (defined(_M_X64) || defined(_M_ARM64)) && (!defined(__SIZEOF_INT128__) || (!defined(__GNUC__) && !defined(__clang__)))
+    #include <intrin0.h> // __umulh, _umul128, _BitScanForward64, _BitScanReverse64
+#endif
+
+#ifdef _MSC_VER
+    #include <cstdlib> // _byteswap_ushort, _byteswap_ulong, _byteswap_uint64
+#endif
+#if defined(_MSC_VER) && !defined(__clang__) && (defined(_M_IX86) || defined(_M_ARM))
+    #include <intrin.h> // _BitScanForward, _BitScanReverse
+#endif
+
+// #include <nlohmann/detail/macro_scope.hpp>
+// JSON_HEDLEY_ALWAYS_INLINE, NLOHMANN_JSON_NAMESPACE_BEGIN
+
+#if defined(__cpp_lib_bit_cast) || defined(__cpp_lib_bitops) || defined(__cpp_lib_byteswap)
+    #include <bit> // bit_cast, byteswap, countl_zero, countr_zero
+#endif
+
+// Portable bit-level helpers for the number and string scanners. They use
+// compiler builtins or platform-specific intrinsics where available, the
+// C++20 <bit> functions next, and plain C++ otherwise, so they work regardless
+// of byte order. (The builtins and intrinsics come first: MSVC's
+// std::countl_zero checks for LZCNT at run time unless AVX2 is enabled.)
+
+NLOHMANN_JSON_NAMESPACE_BEGIN
+namespace detail
+{
+
+/// number of leading zero bits of x (x != 0)
+inline int count_leading_zeros(std::uint64_t x) noexcept
+{
+#if defined(__GNUC__) || defined(__clang__)
+    return __builtin_clzll(x);
+#elif defined(_MSC_VER) && (defined(_M_X64) || defined(_M_ARM64))
+    unsigned long index = 0; // NOLINT(runtime/int): the type _BitScan*64 takes
+    _BitScanReverse64(&index, x);
+    return 63 - static_cast<int>(index);
+#elif defined(_MSC_VER) && (defined(_M_IX86) || defined(_M_ARM))
+    unsigned long index = 0; // NOLINT(runtime/int): the type _BitScan* takes
+    if (_BitScanReverse(&index, static_cast<unsigned long>(x >> 32u))) // NOLINT(runtime/int)
+    {
+        return 31 - static_cast<int>(index);
+    }
+    _BitScanReverse(&index, static_cast<unsigned long>(x)); // NOLINT(runtime/int)
+    return 63 - static_cast<int>(index);
+#elif defined(__cpp_lib_bitops)
+    return std::countl_zero(x);
+#else
+    int n = 0;
+    for (int shift = 32; shift != 0; shift >>= 1)
+    {
+        if ((x >> (64 - shift)) == 0)
+        {
+            n += shift;
+            x <<= shift;
+        }
+    }
+    return n;
+#endif
+}
+
+/// number of trailing zero bits of x (x != 0)
+inline int count_trailing_zeros(std::uint64_t x) noexcept
+{
+#if defined(__GNUC__) || defined(__clang__)
+    return __builtin_ctzll(x);
+#elif defined(_MSC_VER) && (defined(_M_X64) || defined(_M_ARM64))
+    unsigned long index = 0; // NOLINT(runtime/int): the type _BitScan*64 takes
+    _BitScanForward64(&index, x);
+    return static_cast<int>(index);
+#elif defined(_MSC_VER) && (defined(_M_IX86) || defined(_M_ARM))
+    unsigned long index = 0; // NOLINT(runtime/int): the type _BitScan* takes
+    if (_BitScanForward(&index, static_cast<unsigned long>(x))) // NOLINT(runtime/int)
+    {
+        return static_cast<int>(index);
+    }
+    _BitScanForward(&index, static_cast<unsigned long>(x >> 32u)); // NOLINT(runtime/int)
+    return 32 + static_cast<int>(index);
+#elif defined(__cpp_lib_bitops)
+    return std::countr_zero(x);
+#else
+    int n = 0;
+    for (int shift = 32; shift != 0; shift >>= 1)
+    {
+        if ((x << (64 - shift)) == 0)
+        {
+            n += shift;
+            x >>= shift;
+        }
+    }
+    return n;
+#endif
+}
+
+/// the object representation of @a from as a @a To (std::bit_cast in C++20)
+template<typename To, typename From>
+JSON_HEDLEY_ALWAYS_INLINE To bit_cast(const From& from) noexcept
+{
+    static_assert(sizeof(To) == sizeof(From), "bit_cast needs types of the same size");
+#if defined(__cpp_lib_bit_cast)
+    return std::bit_cast<To>(from);
+#else
+    To to {};
+    std::memcpy(&to, &from, sizeof(To));
+    return to;
+#endif
+}
+
+/// x with its bytes in reverse order (std::byteswap in C++23); compilers lower
+/// it to a single bswap/rev/movbe
+inline std::uint16_t byteswap(std::uint16_t x) noexcept
+{
+#if defined(__GNUC__) || defined(__clang__)
+    return __builtin_bswap16(x);
+#elif defined(_MSC_VER)
+    return _byteswap_ushort(x);
+#elif defined(__cpp_lib_byteswap)
+    return std::byteswap(x);
+#else
+    return static_cast<std::uint16_t>((x >> 8u) | (x << 8u));
+#endif
+}
+
+/// x with its bytes in reverse order (std::byteswap in C++23)
+inline std::uint32_t byteswap(std::uint32_t x) noexcept
+{
+#if defined(__GNUC__) || defined(__clang__)
+    return __builtin_bswap32(x);
+#elif defined(_MSC_VER)
+    return _byteswap_ulong(x);
+#elif defined(__cpp_lib_byteswap)
+    return std::byteswap(x);
+#else
+    return ((x & 0x000000FFu) << 24u) | ((x & 0x0000FF00u) << 8u)
+           | ((x & 0x00FF0000u) >> 8u) | ((x & 0xFF000000u) >> 24u);
+#endif
+}
+
+/// x with its bytes in reverse order (std::byteswap in C++23)
+inline std::uint64_t byteswap(std::uint64_t x) noexcept
+{
+#if defined(__GNUC__) || defined(__clang__)
+    return __builtin_bswap64(x);
+#elif defined(_MSC_VER)
+    return _byteswap_uint64(x);
+#elif defined(__cpp_lib_byteswap)
+    return std::byteswap(x);
+#else
+    x = ((x & 0x00000000FFFFFFFFull) << 32u) | ((x & 0xFFFFFFFF00000000ull) >> 32u);
+    x = ((x & 0x0000FFFF0000FFFFull) << 16u) | ((x & 0xFFFF0000FFFF0000ull) >> 16u);
+    x = ((x & 0x00FF00FF00FF00FFull) << 8u) | ((x & 0xFF00FF00FF00FF00ull) >> 8u);
+    return x;
+#endif
+}
+
+/// the 128-bit product of two 64-bit numbers
+struct uint128_parts
+{
+    std::uint64_t low;
+    std::uint64_t high;
+};
+
+inline uint128_parts full_multiplication(std::uint64_t a, std::uint64_t b) noexcept
+{
+#if defined(__SIZEOF_INT128__)
+    __extension__ using uint128 = unsigned __int128;
+    const uint128 r = static_cast<uint128>(a) * b;
+    return {static_cast<std::uint64_t>(r), static_cast<std::uint64_t>(r >> 64u)};
+#elif defined(_MSC_VER) && defined(_M_X64)
+    std::uint64_t high = 0;
+    const std::uint64_t low = _umul128(a, b, &high);
+    return {low, high};
+#elif defined(_MSC_VER) && defined(_M_ARM64)
+    return {a * b, __umulh(a, b)};
+#else
+    const std::uint64_t a_lo = a & 0xFFFFFFFFu;
+    const std::uint64_t a_hi = a >> 32u;
+    const std::uint64_t b_lo = b & 0xFFFFFFFFu;
+    const std::uint64_t b_hi = b >> 32u;
+    const std::uint64_t lo_lo = a_lo * b_lo;
+    const std::uint64_t hi_lo = a_hi * b_lo;
+    const std::uint64_t lo_hi = a_lo * b_hi;
+    const std::uint64_t hi_hi = a_hi * b_hi;
+    const std::uint64_t cross = (lo_lo >> 32u) + (hi_lo & 0xFFFFFFFFu) + lo_hi;
+    return {(cross << 32u) | (lo_lo & 0xFFFFFFFFu), (hi_lo >> 32u) + (cross >> 32u) + hi_hi};
+#endif
+}
+
+/// eight bytes as a little-endian word (a single load on little-endian
+/// targets; always inlined, as GCC otherwise calls it in the number loops)
+JSON_HEDLEY_ALWAYS_INLINE std::uint64_t read_eight_bytes(const unsigned char* b) noexcept
+{
+#if JSON_LITTLE_ENDIAN_TARGET
+    // the byte order already matches
+    std::uint64_t result = 0;
+    std::memcpy(&result, b, sizeof(result));
+    return result;
+#else
+    return static_cast<std::uint64_t>(b[0]) | (static_cast<std::uint64_t>(b[1]) << 8u)
+           | (static_cast<std::uint64_t>(b[2]) << 16u) | (static_cast<std::uint64_t>(b[3]) << 24u)
+           | (static_cast<std::uint64_t>(b[4]) << 32u) | (static_cast<std::uint64_t>(b[5]) << 40u)
+           | (static_cast<std::uint64_t>(b[6]) << 48u) | (static_cast<std::uint64_t>(b[7]) << 56u);
+#endif
+}
+
+/// eight bytes as a little-endian word
+JSON_HEDLEY_ALWAYS_INLINE std::uint64_t read_eight_bytes(const char* p) noexcept
+{
+    return read_eight_bytes(reinterpret_cast<const unsigned char*>(p)); // NOLINT(cppcoreguidelines-pro-type-reinterpret-cast)
+}
+
+}  // namespace detail
+NLOHMANN_JSON_NAMESPACE_END
+
 // #include <nlohmann/detail/exceptions.hpp>
 
 // #include <nlohmann/detail/input/input_adapters.hpp>
@@ -10316,231 +10543,6 @@ NLOHMANN_JSON_NAMESPACE_END
 #include <vector> // vector
 
 // #include <nlohmann/detail/bit_ops.hpp>
-//     __ _____ _____ _____
-//  __|  |   __|     |   | |  JSON for Modern C++
-// |  |  |__   |  |  | | | |  version 3.12.0
-// |_____|_____|_____|_|___|  https://github.com/nlohmann/json
-//
-// SPDX-FileCopyrightText: 2013-2026 Niels Lohmann <https://nlohmann.me>
-// SPDX-License-Identifier: MIT
-
-
-
-#include <cstdint> // uint64_t
-#include <cstring> // memcpy
-#if defined(_MSC_VER) && (defined(_M_X64) || defined(_M_ARM64)) && (!defined(__SIZEOF_INT128__) || (!defined(__GNUC__) && !defined(__clang__)))
-    #include <intrin0.h> // __umulh, _umul128, _BitScanForward64, _BitScanReverse64
-#endif
-
-#ifdef _MSC_VER
-    #include <cstdlib> // _byteswap_ushort, _byteswap_ulong, _byteswap_uint64
-#endif
-#if defined(_MSC_VER) && !defined(__clang__) && (defined(_M_IX86) || defined(_M_ARM))
-    #include <intrin.h> // _BitScanForward, _BitScanReverse
-#endif
-
-// #include <nlohmann/detail/macro_scope.hpp>
-// JSON_HEDLEY_ALWAYS_INLINE, NLOHMANN_JSON_NAMESPACE_BEGIN
-
-#if defined(__cpp_lib_bit_cast) || defined(__cpp_lib_bitops) || defined(__cpp_lib_byteswap)
-    #include <bit> // bit_cast, byteswap, countl_zero, countr_zero
-#endif
-
-// Portable bit-level helpers for the number and string scanners. They use
-// compiler builtins or platform-specific intrinsics where available, the
-// C++20 <bit> functions next, and plain C++ otherwise, so they work regardless
-// of byte order. (The builtins and intrinsics come first: MSVC's
-// std::countl_zero checks for LZCNT at run time unless AVX2 is enabled.)
-
-NLOHMANN_JSON_NAMESPACE_BEGIN
-namespace detail
-{
-
-/// number of leading zero bits of x (x != 0)
-inline int count_leading_zeros(std::uint64_t x) noexcept
-{
-#if defined(__GNUC__) || defined(__clang__)
-    return __builtin_clzll(x);
-#elif defined(_MSC_VER) && (defined(_M_X64) || defined(_M_ARM64))
-    unsigned long index = 0; // NOLINT(runtime/int): the type _BitScan*64 takes
-    _BitScanReverse64(&index, x);
-    return 63 - static_cast<int>(index);
-#elif defined(_MSC_VER) && (defined(_M_IX86) || defined(_M_ARM))
-    unsigned long index = 0; // NOLINT(runtime/int): the type _BitScan* takes
-    if (_BitScanReverse(&index, static_cast<unsigned long>(x >> 32u))) // NOLINT(runtime/int)
-    {
-        return 31 - static_cast<int>(index);
-    }
-    _BitScanReverse(&index, static_cast<unsigned long>(x)); // NOLINT(runtime/int)
-    return 63 - static_cast<int>(index);
-#elif defined(__cpp_lib_bitops)
-    return std::countl_zero(x);
-#else
-    int n = 0;
-    for (int shift = 32; shift != 0; shift >>= 1)
-    {
-        if ((x >> (64 - shift)) == 0)
-        {
-            n += shift;
-            x <<= shift;
-        }
-    }
-    return n;
-#endif
-}
-
-/// number of trailing zero bits of x (x != 0)
-inline int count_trailing_zeros(std::uint64_t x) noexcept
-{
-#if defined(__GNUC__) || defined(__clang__)
-    return __builtin_ctzll(x);
-#elif defined(_MSC_VER) && (defined(_M_X64) || defined(_M_ARM64))
-    unsigned long index = 0; // NOLINT(runtime/int): the type _BitScan*64 takes
-    _BitScanForward64(&index, x);
-    return static_cast<int>(index);
-#elif defined(_MSC_VER) && (defined(_M_IX86) || defined(_M_ARM))
-    unsigned long index = 0; // NOLINT(runtime/int): the type _BitScan* takes
-    if (_BitScanForward(&index, static_cast<unsigned long>(x))) // NOLINT(runtime/int)
-    {
-        return static_cast<int>(index);
-    }
-    _BitScanForward(&index, static_cast<unsigned long>(x >> 32u)); // NOLINT(runtime/int)
-    return 32 + static_cast<int>(index);
-#elif defined(__cpp_lib_bitops)
-    return std::countr_zero(x);
-#else
-    int n = 0;
-    for (int shift = 32; shift != 0; shift >>= 1)
-    {
-        if ((x << (64 - shift)) == 0)
-        {
-            n += shift;
-            x >>= shift;
-        }
-    }
-    return n;
-#endif
-}
-
-/// the object representation of @a from as a @a To (std::bit_cast in C++20)
-template<typename To, typename From>
-JSON_HEDLEY_ALWAYS_INLINE To bit_cast(const From& from) noexcept
-{
-    static_assert(sizeof(To) == sizeof(From), "bit_cast needs types of the same size");
-#if defined(__cpp_lib_bit_cast)
-    return std::bit_cast<To>(from);
-#else
-    To to {};
-    std::memcpy(&to, &from, sizeof(To));
-    return to;
-#endif
-}
-
-/// x with its bytes in reverse order (std::byteswap in C++23); compilers lower
-/// it to a single bswap/rev/movbe
-inline std::uint16_t byteswap(std::uint16_t x) noexcept
-{
-#if defined(__GNUC__) || defined(__clang__)
-    return __builtin_bswap16(x);
-#elif defined(_MSC_VER)
-    return _byteswap_ushort(x);
-#elif defined(__cpp_lib_byteswap)
-    return std::byteswap(x);
-#else
-    return static_cast<std::uint16_t>((x >> 8u) | (x << 8u));
-#endif
-}
-
-/// x with its bytes in reverse order (std::byteswap in C++23)
-inline std::uint32_t byteswap(std::uint32_t x) noexcept
-{
-#if defined(__GNUC__) || defined(__clang__)
-    return __builtin_bswap32(x);
-#elif defined(_MSC_VER)
-    return _byteswap_ulong(x);
-#elif defined(__cpp_lib_byteswap)
-    return std::byteswap(x);
-#else
-    return ((x & 0x000000FFu) << 24u) | ((x & 0x0000FF00u) << 8u)
-           | ((x & 0x00FF0000u) >> 8u) | ((x & 0xFF000000u) >> 24u);
-#endif
-}
-
-/// x with its bytes in reverse order (std::byteswap in C++23)
-inline std::uint64_t byteswap(std::uint64_t x) noexcept
-{
-#if defined(__GNUC__) || defined(__clang__)
-    return __builtin_bswap64(x);
-#elif defined(_MSC_VER)
-    return _byteswap_uint64(x);
-#elif defined(__cpp_lib_byteswap)
-    return std::byteswap(x);
-#else
-    x = ((x & 0x00000000FFFFFFFFull) << 32u) | ((x & 0xFFFFFFFF00000000ull) >> 32u);
-    x = ((x & 0x0000FFFF0000FFFFull) << 16u) | ((x & 0xFFFF0000FFFF0000ull) >> 16u);
-    x = ((x & 0x00FF00FF00FF00FFull) << 8u) | ((x & 0xFF00FF00FF00FF00ull) >> 8u);
-    return x;
-#endif
-}
-
-/// the 128-bit product of two 64-bit numbers
-struct uint128_parts
-{
-    std::uint64_t low;
-    std::uint64_t high;
-};
-
-inline uint128_parts full_multiplication(std::uint64_t a, std::uint64_t b) noexcept
-{
-#if defined(__SIZEOF_INT128__)
-    __extension__ using uint128 = unsigned __int128;
-    const uint128 r = static_cast<uint128>(a) * b;
-    return {static_cast<std::uint64_t>(r), static_cast<std::uint64_t>(r >> 64u)};
-#elif defined(_MSC_VER) && defined(_M_X64)
-    std::uint64_t high = 0;
-    const std::uint64_t low = _umul128(a, b, &high);
-    return {low, high};
-#elif defined(_MSC_VER) && defined(_M_ARM64)
-    return {a * b, __umulh(a, b)};
-#else
-    const std::uint64_t a_lo = a & 0xFFFFFFFFu;
-    const std::uint64_t a_hi = a >> 32u;
-    const std::uint64_t b_lo = b & 0xFFFFFFFFu;
-    const std::uint64_t b_hi = b >> 32u;
-    const std::uint64_t lo_lo = a_lo * b_lo;
-    const std::uint64_t hi_lo = a_hi * b_lo;
-    const std::uint64_t lo_hi = a_lo * b_hi;
-    const std::uint64_t hi_hi = a_hi * b_hi;
-    const std::uint64_t cross = (lo_lo >> 32u) + (hi_lo & 0xFFFFFFFFu) + lo_hi;
-    return {(cross << 32u) | (lo_lo & 0xFFFFFFFFu), (hi_lo >> 32u) + (cross >> 32u) + hi_hi};
-#endif
-}
-
-/// eight bytes as a little-endian word (a single load on little-endian
-/// targets; always inlined, as GCC otherwise calls it in the number loops)
-JSON_HEDLEY_ALWAYS_INLINE std::uint64_t read_eight_bytes(const unsigned char* b) noexcept
-{
-#if JSON_LITTLE_ENDIAN_TARGET
-    // the byte order already matches
-    std::uint64_t result = 0;
-    std::memcpy(&result, b, sizeof(result));
-    return result;
-#else
-    return static_cast<std::uint64_t>(b[0]) | (static_cast<std::uint64_t>(b[1]) << 8u)
-           | (static_cast<std::uint64_t>(b[2]) << 16u) | (static_cast<std::uint64_t>(b[3]) << 24u)
-           | (static_cast<std::uint64_t>(b[4]) << 32u) | (static_cast<std::uint64_t>(b[5]) << 40u)
-           | (static_cast<std::uint64_t>(b[6]) << 48u) | (static_cast<std::uint64_t>(b[7]) << 56u);
-#endif
-}
-
-/// eight bytes as a little-endian word
-JSON_HEDLEY_ALWAYS_INLINE std::uint64_t read_eight_bytes(const char* p) noexcept
-{
-    return read_eight_bytes(reinterpret_cast<const unsigned char*>(p)); // NOLINT(cppcoreguidelines-pro-type-reinterpret-cast)
-}
-
-}  // namespace detail
-NLOHMANN_JSON_NAMESPACE_END
 
 // #include <nlohmann/detail/input/input_adapters.hpp>
 
@@ -18517,11 +18519,13 @@ class binary_reader
     bool get_cbor_negative_integer()
     {
         NumberType number{};
-        if (JSON_HEDLEY_UNLIKELY(!get_number(number)))
-        {
-            return false;
-        }
+        return get_number(number) && emit_cbor_negative_integer(number);
+    }
 
+    /// pass the CBOR negative integer -1 - @a number to the SAX parser
+    template<typename NumberType>
+    bool emit_cbor_negative_integer(const NumberType number)
+    {
         // the value is -1 - number, which fits into number_integer_t
         // whenever number does; the outer cast undoes the integral promotion
         // for number_integer_t types narrower than int
@@ -18746,7 +18750,7 @@ class binary_reader
             case 0x95:
             case 0x96:
             case 0x97:
-                return enter_array(conditional_static_cast<std::size_t>(static_cast<unsigned int>(current) & 0x1Fu));
+                return enter_cbor_array(conditional_static_cast<std::size_t>(static_cast<unsigned int>(current) & 0x1Fu));
 
             case 0x98: // array (one-byte uint8_t for n follows)
             case 0x99: // array (two-byte uint16_t for n follow)
@@ -18755,7 +18759,7 @@ class binary_reader
             {
                 std::uint64_t len{};
                 std::size_t size{};
-                return get_cbor_argument(len) && get_cbor_container_size(len, size, "array") && enter_array(size);
+                return get_cbor_argument(len) && get_cbor_container_size(len, size, "array") && enter_cbor_array(size);
             }
 
             case 0x9F: // array (indefinite length)
@@ -18909,16 +18913,10 @@ class binary_reader
                 return get_half_float(false);
 
             case 0xFA: // Single-Precision Float (four-byte IEEE 754)
-            {
-                float number{};
-                return get_number(number) && emit_float(number);
-            }
+                return get_cbor_float<float>();
 
             case 0xFB: // Double-Precision Float (eight-byte IEEE 754)
-            {
-                double number{};
-                return get_number(number) && emit_float(number);
-            }
+                return get_cbor_float<double>();
 
             default: // anything else (0xFF is handled inside the other types)
             {
@@ -18927,6 +18925,16 @@ class binary_reader
         }
     }
 
+    /// read a single- or double-precision float (kept out of
+    /// parse_cbor_value, which is inlined into parse_cbor_internal)
+    template<typename NumberType>
+    JSON_HEDLEY_NEVER_INLINE
+    bool get_cbor_float()
+    {
+        NumberType number{};
+        return get_number(number) && emit_float(number);
+    }
+
     /*!
     @brief reports a nested indefinite-length CBOR string or byte array
     @param[in] type_name  name of the rejected string type
@@ -19452,6 +19460,157 @@ class binary_reader
         return true;
     }
 
+    /*!
+    @brief read a big-endian number of type @a NumberType at @a data
+    */
+    template<typename NumberType>
+    NumberType load_big_endian(const unsigned char* data) const noexcept
+    {
+        NumberType result{};
+        std::memcpy(&result, data, sizeof(NumberType));
+        if (is_little_endian)
+        {
+            byte_swap(result);
+        }
+        return result;
+    }
+
+    /*!
+    @brief read the integers, floats, booleans, and nulls that follow in a
+           definite-length array, without the per-element bookkeeping of
+           @ref parse_cbor_internal
+
+    Stops at the end of the array, at any other kind of value, and when fewer
+    than 9 bytes (the longest of these values) are left in the input, leaving
+    those to @ref parse_cbor_value, which reads them the same way.
+
+    @param[in,out] remaining  the number of elements of the array not yet read
+    @return whether the SAX parser accepted the values
+    */
+    bool parse_cbor_scalar_run(std::size_t& remaining)
+    {
+        while (remaining != 0 && ia.bulk_remaining() >= 9)
+        {
+            const auto* const data = reinterpret_cast<const unsigned char*>(ia.bulk_data());
+            const unsigned char head = data[0];
+            std::size_t size = 0;
+            bool accepted = true;
+            switch (head)
+            {
+                case 0x18:
+                    size = 2;
+                    accepted = emit_unsigned(data[1]);
+                    break;
+                case 0x19:
+                    size = 3;
+                    accepted = emit_unsigned(load_big_endian<std::uint16_t>(data + 1));
+                    break;
+                case 0x1A:
+                    size = 5;
+                    accepted = emit_unsigned(load_big_endian<std::uint32_t>(data + 1));
+                    break;
+                case 0x1B:
+                    size = 9;
+                    accepted = emit_unsigned(load_big_endian<std::uint64_t>(data + 1));
+                    break;
+                case 0x38:
+                    size = 2;
+                    accepted = emit_cbor_negative_integer(data[1]);
+                    break;
+                case 0x39:
+                    size = 3;
+                    accepted = emit_cbor_negative_integer(load_big_endian<std::uint16_t>(data + 1));
+                    break;
+                case 0x3A:
+                    size = 5;
+                    accepted = emit_cbor_negative_integer(load_big_endian<std::uint32_t>(data + 1));
+                    break;
+                case 0x3B:
+                    size = 9;
+                    accepted = emit_cbor_negative_integer(load_big_endian<std::uint64_t>(data + 1));
+                    break;
+                case 0xF4:
+                    size = 1;
+                    accepted = sax->boolean(false);
+                    break;
+                case 0xF5:
+                    size = 1;
+                    accepted = sax->boolean(true);
+                    break;
+                case 0xF6:
+                    size = 1;
+                    accepted = sax->null();
+                    break;
+                case 0xFA:
+                    size = 5;
+                    accepted = emit_float(load_big_endian<float>(data + 1));
+                    break;
+                case 0xFB:
+                    size = 9;
+                    accepted = emit_float(load_big_endian<double>(data + 1));
+                    break;
+                default:
+                    if (head <= 0x17)
+                    {
+                        size = 1;
+                        accepted = sax->number_unsigned(static_cast<number_unsigned_t>(head));
+                    }
+                    else if (head >= 0x20 && head <= 0x37)
+                    {
+                        size = 1;
+                        accepted = sax->number_integer(static_cast<std::int8_t>(0x20 - 1 - head));
+                    }
+                    else
+                    {
+                        // any other value is read by parse_cbor_value
+                        return true;
+                    }
+                    break;
+            }
+            // as if read by get() and get_number(): current is the head byte
+            ia.bulk_skip(size);
+            chars_read += size;
+            current = head;
+            --remaining;
+            if (JSON_HEDLEY_UNLIKELY(!accepted))
+            {
+                return false;
+            }
+        }
+        return true;
+    }
+
+    /*!
+    @brief open a definite-length array and read the scalars it begins with
+           (see @ref parse_cbor_scalar_run)
+
+    Only the start of an array is tried, which covers arrays of numbers
+    (e.g. coordinates) but costs arrays of other values only one test, not one
+    per element.
+
+    @param[in] len  the number of elements
+    @return whether the SAX parser accepted the array and the values read
+    */
+    bool enter_cbor_array(const std::size_t len)
+    {
+        return enter_array(len) && (len == 0 || parse_cbor_array_start(std::integral_constant<bool, bulk_scan> {}));
+    }
+
+    /// read the scalars the array just opened begins with; out of line, so
+    /// that @ref parse_cbor_value stays small enough to be inlined into
+    /// @ref parse_cbor_internal
+    JSON_HEDLEY_NEVER_INLINE
+    bool parse_cbor_array_start(std::true_type /*bulk*/)
+    {
+        return parse_cbor_scalar_run(container_stack.back().remaining);
+    }
+
+    /// input that is not contiguous: every value is read by parse_cbor_value
+    bool parse_cbor_array_start(std::false_type /*bulk*/) const noexcept
+    {
+        return true;
+    }
+
     /*!
     @brief read a CBOR value and everything nested inside it
 
@@ -19481,6 +19640,7 @@ class binary_reader
                 // pop_back() below, which destroys the container_stack element
                 // it would otherwise alias
                 const container_frame top = container_stack.back();
+                const bool is_object = top.is_object;
                 bool at_end = false;
 
                 if (top.remaining != npos)
@@ -19492,7 +19652,7 @@ class binary_reader
                     {
                         // claim the element about to be read
                         --container_stack.back().remaining;
-                        if (top.is_object)
+                        if (is_object)
                         {
                             get();
                         }
@@ -19505,7 +19665,7 @@ class binary_reader
                     // Testing for it consumes a byte, which is the first byte
                     // of the next element when it is not one.
                     at_end = (get() == 0xFF);
-                    fetch = top.is_object;
+                    fetch = is_object;
                 }
 
                 if (at_end)
@@ -19522,7 +19682,7 @@ class binary_reader
                     continue;
                 }
 
-                if (top.is_object)
+                if (is_object)
                 {
                     if (JSON_HEDLEY_UNLIKELY(!skip_cbor_key_tags(tag_handler)))
                     {
@@ -19747,7 +19907,7 @@ class binary_reader
             case 0x9D:
             case 0x9E:
             case 0x9F:
-                return enter_array(conditional_static_cast<std::size_t>(static_cast<unsigned int>(current) & 0x0Fu));
+                return enter_msgpack_array(conditional_static_cast<std::size_t>(static_cast<unsigned int>(current) & 0x0Fu));
 
             // fixstr
             case 0xA0:
@@ -19878,13 +20038,13 @@ class binary_reader
             case 0xDC: // array 16
             {
                 std::uint16_t len{};
-                return get_number(len) && enter_array(static_cast<std::size_t>(len));
+                return get_number(len) && enter_msgpack_array(static_cast<std::size_t>(len));
             }
 
             case 0xDD: // array 32
             {
                 std::uint32_t len{};
-                return get_number(len) && enter_array(conditional_static_cast<std::size_t>(len));
+                return get_number(len) && enter_msgpack_array(conditional_static_cast<std::size_t>(len));
             }
 
             case 0xDE: // map 16
@@ -20226,6 +20386,138 @@ class binary_reader
         }
     }
 
+    /*!
+    @brief read the integers, floats, booleans, and nils that follow in a
+           MessagePack array, without the per-element bookkeeping of
+           @ref parse_msgpack_internal
+
+    Stops at the end of the array, at any other kind of value, and when fewer
+    than 9 bytes (the longest of these values) are left in the input, leaving
+    those to @ref parse_msgpack_value, which reads them the same way.
+
+    @param[in,out] remaining  the number of elements of the array not yet read
+    @return whether the SAX parser accepted the values
+    */
+    bool parse_msgpack_scalar_run(std::size_t& remaining)
+    {
+        while (remaining != 0 && ia.bulk_remaining() >= 9)
+        {
+            const auto* const data = reinterpret_cast<const unsigned char*>(ia.bulk_data());
+            const unsigned char head = data[0];
+            std::size_t size = 0;
+            bool accepted = true;
+            switch (head)
+            {
+                case 0xC0: // nil
+                    size = 1;
+                    accepted = sax->null();
+                    break;
+                case 0xC2: // false
+                    size = 1;
+                    accepted = sax->boolean(false);
+                    break;
+                case 0xC3: // true
+                    size = 1;
+                    accepted = sax->boolean(true);
+                    break;
+                case 0xCA: // float 32
+                    size = 5;
+                    accepted = emit_float(load_big_endian<float>(data + 1));
+                    break;
+                case 0xCB: // float 64
+                    size = 9;
+                    accepted = emit_float(load_big_endian<double>(data + 1));
+                    break;
+                case 0xCC: // uint 8
+                    size = 2;
+                    accepted = emit_unsigned(data[1]);
+                    break;
+                case 0xCD: // uint 16
+                    size = 3;
+                    accepted = emit_unsigned(load_big_endian<std::uint16_t>(data + 1));
+                    break;
+                case 0xCE: // uint 32
+                    size = 5;
+                    accepted = emit_unsigned(load_big_endian<std::uint32_t>(data + 1));
+                    break;
+                case 0xCF: // uint 64
+                    size = 9;
+                    accepted = emit_unsigned(load_big_endian<std::uint64_t>(data + 1));
+                    break;
+                case 0xD0: // int 8
+                    size = 2;
+                    accepted = emit_signed(static_cast<std::int8_t>(data[1]));
+                    break;
+                case 0xD1: // int 16
+                    size = 3;
+                    accepted = emit_signed(load_big_endian<std::int16_t>(data + 1));
+                    break;
+                case 0xD2: // int 32
+                    size = 5;
+                    accepted = emit_signed(load_big_endian<std::int32_t>(data + 1));
+                    break;
+                case 0xD3: // int 64
+                    size = 9;
+                    accepted = emit_signed(load_big_endian<std::int64_t>(data + 1));
+                    break;
+                default:
+                    if (head <= 0x7F)
+                    {
+                        // positive fixint
+                        size = 1;
+                        accepted = sax->number_unsigned(static_cast<number_unsigned_t>(head));
+                    }
+                    else if (head >= 0xE0)
+                    {
+                        // negative fixint
+                        size = 1;
+                        accepted = sax->number_integer(static_cast<std::int8_t>(head));
+                    }
+                    else
+                    {
+                        // any other value is read by parse_msgpack_value
+                        return true;
+                    }
+                    break;
+            }
+            // as if read by get() and get_number(): current is the last byte
+            ia.bulk_skip(size);
+            chars_read += size;
+            current = data[size - 1];
+            --remaining;
+            if (JSON_HEDLEY_UNLIKELY(!accepted))
+            {
+                return false;
+            }
+        }
+        return true;
+    }
+
+    /*!
+    @brief open a MessagePack array and read the scalars it begins with (see
+           @ref parse_msgpack_scalar_run)
+    @param[in] len  the number of elements
+    @return whether the SAX parser accepted the array and the values read
+    */
+    bool enter_msgpack_array(const std::size_t len)
+    {
+        return enter_array(len) && (len == 0 || parse_msgpack_array_start(std::integral_constant<bool, bulk_scan> {}));
+    }
+
+    /// read the scalars the array just opened begins with; out of line, so
+    /// that @ref parse_msgpack_value stays small enough to be inlined
+    JSON_HEDLEY_NEVER_INLINE
+    bool parse_msgpack_array_start(std::true_type /*bulk*/)
+    {
+        return parse_msgpack_scalar_run(container_stack.back().remaining);
+    }
+
+    /// input that is not contiguous: every value is read by parse_msgpack_value
+    bool parse_msgpack_array_start(std::false_type /*bulk*/) const noexcept
+    {
+        return true;
+    }
+
     /*!
     @brief read a MessagePack value and everything nested inside it
 
@@ -21467,13 +21759,13 @@ class binary_reader
                 return sax->null();
 
             case 0xFB:
-                return sax->number_float(static_cast<number_float_t>(-1.0), "");
+                return sax->number_float(static_cast<number_float_t>(-1.0), no_source);
 
             case 0xFC:
-                return sax->number_float(static_cast<number_float_t>(0.0), "");
+                return sax->number_float(static_cast<number_float_t>(0.0), no_source);
 
             case 0xFD:
-                return sax->number_float(static_cast<number_float_t>(1.0), "");
+                return sax->number_float(static_cast<number_float_t>(1.0), no_source);
 
             case 0xFF: // empty string
             {
@@ -21842,6 +22134,27 @@ class binary_reader
         return current;
     }
 
+    /// reverse the byte order of a double with a single byteswap(): GCC does
+    /// not merge the byte-wise swap of the generic version into one
+    static void byte_swap(double& number) noexcept
+    {
+        static_assert(sizeof(double) == sizeof(std::uint64_t), "double must have 64 bits");
+        std::uint64_t bits{};
+        std::memcpy(&bits, &number, sizeof(bits));
+        bits = byteswap(bits);
+        std::memcpy(&number, &bits, sizeof(bits));
+    }
+
+    /// reverse the byte order of a float with a single byteswap()
+    static void byte_swap(float& number) noexcept
+    {
+        static_assert(sizeof(float) == sizeof(std::uint32_t), "float must have 32 bits");
+        std::uint32_t bits{};
+        std::memcpy(&bits, &number, sizeof(bits));
+        bits = byteswap(bits);
+        std::memcpy(&number, &bits, sizeof(bits));
+    }
+
     template<class NumberType>
     static void byte_swap(NumberType& number)
     {
@@ -21977,7 +22290,7 @@ class binary_reader
         {
             return out_of_range_error(exception_id::number_overflow, "number overflow", "value");
         }
-        return sax->number_float(result, "");
+        return sax->number_float(result, no_source);
     }
 
     /*!
@@ -22037,7 +22350,7 @@ class binary_reader
         }();
         return sax->number_float((half & 0x8000u) != 0
                                  ? static_cast<number_float_t>(-val)
-                                 : static_cast<number_float_t>(val), "");
+                                 : static_cast<number_float_t>(val), no_source);
     }
 
     /*!
@@ -22362,6 +22675,10 @@ class binary_reader
     /// the SAX parser
     json_sax_t* sax = nullptr;
 
+    /// the (empty) source text of every number_float event: binary formats
+    /// have none, and a string_t built for each float would cost more
+    const string_t no_source{};
+
     /// the containers that have been opened and not closed yet; see @ref container_frame
     std::vector<container_frame> container_stack{};
 

tests/src/unit-cbor.cpp

@@ -29,7 +29,9 @@ using nlohmann::json;
 #include "round_trip_corpus.hpp"
 #include "test_utils.hpp"
 #include "sax_countdown.hpp"
+#include "sax_event_loggers.hpp"
 using utils::SaxCountdown;
+using utils::SaxEventLogger;
 
 
 TEST_CASE("CBOR")
@@ -2446,6 +2448,78 @@ TEST_CASE("CBOR SAX parsing stops at every event")
     CHECK(count_events(std::vector<std::uint8_t>({0xBF, 0x61, 'a', 0x9F, 0x01, 0xFF, 0xFF})) == 6);
 }
 
+TEST_CASE("CBOR arrays of scalars from contiguous and non-contiguous input")
+{
+    // contiguous input reads the integers, floats, booleans, and nulls an
+    // array begins with in one loop, other input value by value; both must
+    // give the same events, value, or error for the complete document and for
+    // every truncation of it, and stop at every event a SAX handler rejects
+    const std::vector<std::uint8_t> cbor =
+    {
+        0xA3,                                                       // map(3)
+        0x61, 'a',                                                  // "a"
+        0x98, 0x16,                                                 // array(22)
+        0x00, 0x17, 0x18, 0xFF, 0x19, 0x01, 0x00,                   // 0, 23, 255, 256
+        0x1A, 0x00, 0x01, 0x00, 0x00,                               // 65536
+        0x1B, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,       // 18446744073709551615
+        0x20, 0x37, 0x38, 0xFF, 0x39, 0x01, 0x00,                   // -1, -24, -256, -257
+        0x3A, 0x00, 0x01, 0x00, 0x00,                               // -65537
+        0x3B, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,       // -18446744073709551616 (a float)
+        0x3B, 0x7F, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,       // -9223372036854775808
+        0xF4, 0xF5, 0xF6,                                           // false, true, null
+        0xFA, 0x3F, 0xC0, 0x00, 0x00,                               // 1.5 (float32)
+        0xFB, 0xBF, 0xB9, 0x99, 0x99, 0x99, 0x99, 0x99, 0x9A,       // -0.1
+        0xF9, 0x3C, 0x00,                                           // 1.0 (float16, ends the loop)
+        0x01, 0x61, 'a',                                            // 1, "a"
+        0xFB, 0x40, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,       // 2.5
+        0x61, 'b',                                                  // "b"
+        0x83,                                                       // array(3)
+        0x82, 0xFB, 0x3F, 0xF4, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // [1.25,
+        0xFB, 0xC0, 0x0E, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,       //  -3.75]
+        0x82, 0xFB, 0x3F, 0xF4, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // [1.25,
+        0xFB, 0xC0, 0x0E, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,       //  -3.75]
+        0x80,                                                       // []
+        0x61, 'c',                                                  // "c"
+        0x84, 0xA0, 0x01, 0xC1, 0x02, 0x81, 0x03                    // [{}, 1, tag(1) 2, [3]]
+    };
+    const auto tags = json::cbor_tag_handler_t::ignore;
+
+    const json j = json::from_cbor(cbor, true, true, tags);
+    CHECK(j["a"].size() == 22);
+    CHECK(j["a"][5] == 18446744073709551615u);
+    CHECK(j["a"][11].is_number_float());
+    CHECK(j["a"][11] == -18446744073709551616.0);
+    CHECK(j["a"][12] == (std::numeric_limits<std::int64_t>::min)());
+    CHECK(j["a"][17] == -0.1);
+    CHECK(j["b"] == json::parse("[[1.25, -3.75], [1.25, -3.75], []]"));
+    CHECK(j["c"] == json::parse("[{}, 1, 2, [3]]"));
+
+    for (std::size_t length = 0; length <= cbor.size(); ++length)
+    {
+        CAPTURE(length)
+        const std::vector<std::uint8_t> contiguous(cbor.begin(), cbor.begin() + static_cast<std::ptrdiff_t>(length));
+        const std::list<std::uint8_t> bytewise(contiguous.begin(), contiguous.end());
+
+        CHECK(json::from_cbor(contiguous, true, false, tags).dump() == json::from_cbor(bytewise.begin(), bytewise.end(), true, false, tags).dump());
+
+        SaxEventLogger from_contiguous;
+        SaxEventLogger from_bytewise;
+        CHECK(json::sax_parse(contiguous, &from_contiguous, json::input_format_t::cbor, true, false, false, tags)
+              == json::sax_parse(bytewise.begin(), bytewise.end(), &from_bytewise, json::input_format_t::cbor, true, false, false, tags));
+        CHECK(from_contiguous.events == from_bytewise.events);
+    }
+
+    for (int events = 0; events <= 60; ++events)
+    {
+        CAPTURE(events)
+        const std::list<std::uint8_t> bytewise(cbor.begin(), cbor.end());
+        SaxCountdown from_contiguous(events);
+        SaxCountdown from_bytewise(events);
+        CHECK(json::sax_parse(cbor, &from_contiguous, json::input_format_t::cbor, true, false, false, tags)
+              == json::sax_parse(bytewise.begin(), bytewise.end(), &from_bytewise, json::input_format_t::cbor, true, false, false, tags));
+    }
+}
+
 TEST_CASE("CBOR indefinite-length strings do not recurse per chunk")
 {
     // Reading an indefinite-length string or byte array used to call itself

tests/src/unit-msgpack.cpp

@@ -27,12 +27,15 @@ using nlohmann::json;
 #include <sstream>
 #include <iomanip>
 #include <limits>
+#include <list>
 #include <set>
 #include "make_test_data_available.hpp"
 #include "round_trip_corpus.hpp"
 #include "test_utils.hpp"
 #include "sax_countdown.hpp"
+#include "sax_event_loggers.hpp"
 using utils::SaxCountdown;
+using utils::SaxEventLogger;
 
 
 TEST_CASE("MessagePack")
@@ -2212,3 +2215,97 @@ TEST_CASE("MessagePack large strings and binaries (chunked reader)")
         }
     }
 }
+
+TEST_CASE("MessagePack arrays of scalars from contiguous and non-contiguous input")
+{
+    // contiguous input reads the integers, floats, booleans, and nils an
+    // array begins with in one loop, other input value by value; both must
+    // give the same events, value, or error for the complete document and for
+    // every truncation of it, and stop at every event a SAX handler rejects
+    const std::vector<std::uint8_t> msgpack =
+    {
+        0x83,                                                       // map(3)
+        0xA1, 'a',                                                  // "a"
+        0xDC, 0x00, 0x16,                                           // array 16 (22)
+        0x00, 0x7F, 0xCC, 0xFF, 0xCD, 0x01, 0x00,                   // 0, 127, 255, 256
+        0xCE, 0x00, 0x01, 0x00, 0x00,                               // 65536
+        0xCF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,       // 18446744073709551615
+        0xFF, 0xE0, 0xD0, 0x80, 0xD1, 0xFF, 0x00,                   // -1, -32, -128, -256
+        0xD2, 0xFF, 0xFE, 0xFF, 0xFF,                               // -65537
+        0xD3, 0x80, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,       // -9223372036854775808
+        0xC2, 0xC3, 0xC0,                                           // false, true, nil
+        0xCA, 0x3F, 0xC0, 0x00, 0x00,                               // 1.5 (float 32)
+        0xCB, 0xBF, 0xB9, 0x99, 0x99, 0x99, 0x99, 0x99, 0x9A,       // -0.1
+        0xA1, 'x',                                                  // "x" (ends the loop)
+        0x01,                                                       // 1
+        0xCB, 0x40, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,       // 2.5
+        0xD9, 0x01, 'y',                                            // "y" (str 8)
+        0xC0,                                                       // nil
+        0xA1, 'b',                                                  // "b"
+        0x93,                                                       // fixarray(3)
+        0x92, 0xCB, 0x3F, 0xF4, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // [1.25,
+        0xCB, 0xC0, 0x0E, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,       //  -3.75]
+        0x92, 0xCB, 0x3F, 0xF4, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, // [1.25,
+        0xCB, 0xC0, 0x0E, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,       //  -3.75]
+        0x90,                                                       // []
+        0xA1, 'c',                                                  // "c"
+        0x94, 0x80, 0x01, 0xC4, 0x01, 0x05, 0x91, 0x03              // [{}, 1, bin 8, [3]]
+    };
+
+    const json j = json::from_msgpack(msgpack);
+    CHECK(j["a"].size() == 22);
+    CHECK(j["a"][5] == 18446744073709551615u);
+    CHECK(j["a"][7] == -32);
+    CHECK(j["a"][11] == (std::numeric_limits<std::int64_t>::min)());
+    CHECK(j["a"][15] == 1.5);
+    CHECK(j["a"][16] == -0.1);
+    CHECK(j["a"][17] == "x");
+    CHECK(j["b"] == json::parse("[[1.25, -3.75], [1.25, -3.75], []]"));
+    CHECK(j["c"][2].is_binary());
+
+    for (std::size_t length = 0; length <= msgpack.size(); ++length)
+    {
+        CAPTURE(length)
+        const std::vector<std::uint8_t> contiguous(msgpack.begin(), msgpack.begin() + static_cast<std::ptrdiff_t>(length));
+        const std::list<std::uint8_t> bytewise(contiguous.begin(), contiguous.end());
+
+        CHECK(json::from_msgpack(contiguous, true, false).dump() == json::from_msgpack(bytewise.begin(), bytewise.end(), true, false).dump());
+#if !defined(JSON_NOEXCEPTION)
+        std::string contiguous_error;
+        std::string bytewise_error;
+        try
+        {
+            static_cast<void>(json::from_msgpack(contiguous));
+        }
+        catch (const json::parse_error& e)
+        {
+            contiguous_error = e.what();
+        }
+        try
+        {
+            static_cast<void>(json::from_msgpack(bytewise.begin(), bytewise.end()));
+        }
+        catch (const json::parse_error& e)
+        {
+            bytewise_error = e.what();
+        }
+        CHECK(contiguous_error == bytewise_error);
+#endif
+
+        SaxEventLogger from_contiguous;
+        SaxEventLogger from_bytewise;
+        CHECK(json::sax_parse(contiguous, &from_contiguous, json::input_format_t::msgpack)
+              == json::sax_parse(bytewise.begin(), bytewise.end(), &from_bytewise, json::input_format_t::msgpack));
+        CHECK(from_contiguous.events == from_bytewise.events);
+    }
+
+    for (int events = 0; events <= 60; ++events)
+    {
+        CAPTURE(events)
+        const std::list<std::uint8_t> bytewise(msgpack.begin(), msgpack.end());
+        SaxCountdown from_contiguous(events);
+        SaxCountdown from_bytewise(events);
+        CHECK(json::sax_parse(msgpack, &from_contiguous, json::input_format_t::msgpack)
+              == json::sax_parse(bytewise.begin(), bytewise.end(), &from_bytewise, json::input_format_t::msgpack));
+    }
+}