Fix CI: useless cast in the growth of the view's node index

GCC -Werror=useless-cast (ci_test_gcc on Linux x86-64) rejected
static_cast<std::size_t>(guess + (guess / 4) + 64): the sum is a
std::uint64_t prvalue, the same type as std::size_t there, while the cast
is needed where std::size_t is 32 bits wide. Cast a named variable
instead, which GCC does not report. The build stopped at an earlier error
before, so the previous CI run did not show this one.

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

BUILD.bazel

@@ -20,6 +20,7 @@ cc_library(
     hdrs = [
         "include/nlohmann/adl_serializer.hpp",
         "include/nlohmann/byte_container_with_subtype.hpp",
+        "include/nlohmann/detail/abi_config.hpp",
         "include/nlohmann/detail/abi_macros.hpp",
         "include/nlohmann/detail/bit_ops.hpp",
         "include/nlohmann/detail/conversions/from_json.hpp",
@@ -65,6 +66,13 @@ cc_library(
         "include/nlohmann/detail/string_escape.hpp",
         "include/nlohmann/detail/string_utils.hpp",
         "include/nlohmann/detail/value_t.hpp",
+        "include/nlohmann/detail/view/builder.hpp",
+        "include/nlohmann/detail/view/document_data.hpp",
+        "include/nlohmann/detail/view/macro_scope.hpp",
+        "include/nlohmann/detail/view/macro_unscope.hpp",
+        "include/nlohmann/detail/view/node.hpp",
+        "include/nlohmann/detail/view/scan.hpp",
+        "include/nlohmann/detail/view/string_ref.hpp",
         "include/nlohmann/json.hpp",
         "include/nlohmann/json_fwd.hpp",
         "include/nlohmann/json_literals.hpp",

include/nlohmann/detail/abi_config.hpp

@@ -0,0 +1,34 @@
+//     __ _____ _____ _____
+//  __|  |   __|     |   | |  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
+
+#pragma once
+
+#include <nlohmann/detail/abi_macros.hpp>
+
+NLOHMANN_JSON_NAMESPACE_BEGIN
+namespace detail
+{
+
+/*!
+@brief the configuration macros that change the library's behavior
+
+json.hpp undefines these macros at its end (see macro_unscope.hpp), so code
+that builds on the library after it (json_view.hpp) reads them here. Like the
+macros, they are part of the ABI namespace, so they always match the
+basic_json they are used with.
+*/
+struct abi_config
+{
+    /// JSON_STRICT_NUL_HANDLING: a null byte is an error, not the end of input
+    static constexpr bool strict_nul_handling = JSON_STRICT_NUL_HANDLING != 0;
+    /// JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON
+    static constexpr bool legacy_discarded_value_comparison = JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON != 0;
+};
+
+}  // namespace detail
+NLOHMANN_JSON_NAMESPACE_END

include/nlohmann/detail/input/lexer.hpp

@@ -219,6 +219,44 @@ class lexer : public lexer_base<BasicJsonType>
     // scan functions
     /////////////////////
 
+    /// contiguous input: try to decode the 4 hex digits following `\u`
+    /// directly from the input buffer via hex_codepoint(), instead of 4 calls
+    /// to get(). On success, advances the adapter and the position counters
+    /// exactly as those 4 get() calls would (a hex digit is never '\n', so
+    /// only the flat counters move) and leaves @a current holding the last of
+    /// the 4 digits, just as the last such get() would; the codepoint is
+    /// written to @a out. Makes no state change and returns false - for a
+    /// pending unget, fewer than 4 remaining bytes, or any of the 4 bytes not
+    /// being a hex digit - so the caller falls back unchanged to the
+    /// per-character loop, which then reports the same diagnostic (stopping
+    /// at the first invalid digit) as before this optimization.
+    bool get_codepoint_bulk(std::true_type /*bulk*/, int& out)
+    {
+        if (next_unget || ia.bulk_remaining() < 4)
+        {
+            return false;
+        }
+        const char_type* const raw = ia.bulk_data();
+        const int codepoint = hex_codepoint(reinterpret_cast<const unsigned char*>(raw));
+        if (codepoint < 0)
+        {
+            return false;
+        }
+        ia.bulk_skip(4);
+        // a hex digit is never a newline, so only the flat counters advance
+        position.chars_read_total += 4;
+        position.chars_read_current_line += 4;
+        current = char_traits<char_type>::to_int_type(raw[3]);
+        out = codepoint;
+        return true;
+    }
+
+    /// streaming input: no bulk fast path
+    bool get_codepoint_bulk(std::false_type /*bulk*/, int& /*out*/) const noexcept
+    {
+        return false;
+    }
+
     /*!
     @brief get codepoint from 4 hex characters following `\u`
 
@@ -238,6 +276,14 @@ class lexer : public lexer_base<BasicJsonType>
     {
         // this function only makes sense after reading `\u`
         JSON_ASSERT(current == 'u');
+
+        // contiguous input: decode all 4 hex digits directly from the buffer
+        int fast_codepoint = 0;
+        if (get_codepoint_bulk(std::integral_constant<bool, bulk_scan> {}, fast_codepoint))
+        {
+            return fast_codepoint;
+        }
+
         int codepoint = 0;
 
         const auto factors = { 12u, 8u, 4u, 0u };

include/nlohmann/detail/input/string_scan.hpp

@@ -8,8 +8,9 @@
 
 #pragma once
 
+#include <array> // array
 #include <cstddef> // size_t
-#include <cstdint> // uint64_t
+#include <cstdint> // uint64_t, uint8_t
 #include <cstring> // memcpy
 
 #include <nlohmann/detail/bit_ops.hpp>
@@ -315,5 +316,50 @@ inline std::size_t string_bulk_run(const unsigned char* data, std::size_t n) noe
     return scalar_string_bulk_run(data, n);
 }
 
+// Decode the 4 hex digits at [data, data+4) - the digits following a `\u`
+// escape - into a codepoint 0x0000..0xFFFF via one table lookup per byte
+// (after yyjson's read_hex_u16), or return -1 if any of the 4 bytes is not a
+// hex digit ('0'..'9', 'A'..'F', 'a'..'f'). The caller must already have
+// checked that 4 bytes are available; used by lexer::get_codepoint()'s
+// contiguous fast path. On -1 it falls back to the byte-at-a-time loop, which
+// stops at the first invalid digit, so the reported error and position are
+// unaffected by this fast path.
+inline int hex_codepoint(const unsigned char* data) noexcept
+{
+    static const std::array<std::uint8_t, 256> hex_digit_table = // NOLINT(cppcoreguidelines-avoid-non-const-global-variables)
+    {
+        {
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 00..0F
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 10..1F
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 20..2F
+            0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 30..3F ('0'..'9')
+            0xFF, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 40..4F ('A'..'F')
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 50..5F
+            0xFF, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 60..6F ('a'..'f')
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 70..7F
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 80..8F
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 90..9F
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // A0..AF
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // B0..BF
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // C0..CF
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // D0..DF
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // E0..EF
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF  // F0..FF
+        }
+    };
+
+    const std::uint8_t d0 = hex_digit_table[data[0]];
+    const std::uint8_t d1 = hex_digit_table[data[1]];
+    const std::uint8_t d2 = hex_digit_table[data[2]];
+    const std::uint8_t d3 = hex_digit_table[data[3]];
+    // every valid digit is <= 0xF; the combined OR only exceeds it if at
+    // least one of the four bytes was not a hex digit (looked up as 0xFF)
+    if ((d0 | d1 | d2 | d3) > 0x0F)
+    {
+        return -1;
+    }
+    return (d0 << 12) | (d1 << 8) | (d2 << 4) | d3;
+}
+
 }  // namespace detail
 NLOHMANN_JSON_NAMESPACE_END

include/nlohmann/detail/view/document_data.hpp

@@ -0,0 +1,130 @@
+//     __ _____ _____ _____
+//  __|  |   __|     |   | |  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
+
+#pragma once
+
+#include <array> // array
+#include <cstddef> // size_t
+#include <cstring> // memcpy
+#include <new> // operator new, placement new
+#include <string> // string
+
+#include <nlohmann/json.hpp>
+#include <nlohmann/detail/view/macro_scope.hpp>
+#include <nlohmann/detail/view/node.hpp>
+
+NLOHMANN_JSON_NAMESPACE_BEGIN
+namespace detail
+{
+namespace view
+{
+
+/// storage of a parsed document; heap-allocated (header and an initial node
+/// array in one block) so that views survive moves of the owning document
+struct document_data
+{
+    const char* src = nullptr;
+    std::size_t size = 0;
+    node* tape = nullptr;
+    std::size_t tape_size = 0;
+    std::size_t tape_cap = 0;
+    node* inline_tape = nullptr; ///< node array allocated together with this header
+    std::size_t inline_cap = 0;
+    std::string arena{}; ///< decoded strings that contained escapes // NOLINT(readability-redundant-member-init)
+    std::string owned{}; ///< owned copy of the input, if any // NOLINT(readability-redundant-member-init)
+    std::array<const char*, 4> base = {{nullptr, nullptr, nullptr, nullptr}}; ///< string bases: source, arena (indexed by flags & node_flags::storage)
+    bool discarded = true;
+
+    /// one allocation for the header and room for `nodes` nodes; large
+    /// documents get a separate node array instead (so it can be trimmed)
+    static document_data* create(std::size_t nodes)
+    {
+        nodes = nodes <= 256 ? nodes : 0;
+        void* mem = ::operator new (sizeof(document_data) + (nodes * sizeof(node)));
+        auto* d = new (mem) document_data(); // NOLINT(cppcoreguidelines-owning-memory): owned by the returned pointer, freed by deleter
+        // (aligned: sizeof is a multiple of the alignment; through void*, as GCC's -Wcast-align wants)
+        d->inline_tape = static_cast<node*>(static_cast<void*>(static_cast<char*>(mem) + sizeof(document_data))); // NOLINT(bugprone-casting-through-void)
+        d->inline_cap = nodes;
+        d->tape = d->inline_tape;
+        d->tape_cap = nodes;
+        return d;
+    }
+
+    struct deleter
+    {
+        void operator()(document_data* d) const noexcept
+        {
+            d->~document_data();
+            ::operator delete (d);
+        }
+    };
+
+    document_data() noexcept = default;
+    document_data(const document_data&) = delete;
+    document_data(document_data&&) = delete;
+    document_data& operator=(const document_data&) = delete;
+    document_data& operator=(document_data&&) = delete;
+    ~document_data()
+    {
+        release();
+    }
+
+    void release() noexcept
+    {
+        if (tape != inline_tape)
+        {
+            ::operator delete (tape);
+        }
+        tape = inline_tape;
+        tape_cap = inline_cap;
+    }
+
+    /// make room for n nodes; keeps the first tape_size nodes
+    void reserve(std::size_t n)
+    {
+        if (n <= tape_cap)
+        {
+            return;
+        }
+        node* fresh = static_cast<node*>(::operator new (n * sizeof(node)));
+        if (tape_size != 0)
+        {
+            std::memcpy(fresh, tape, tape_size * sizeof(node));
+        }
+        release();
+        tape = fresh;
+        tape_cap = n;
+    }
+
+    const char* str(const node& n) const noexcept
+    {
+        return base[n.flags & node_flags::storage] + n.off;
+    }
+
+    /// the node after n's subtree (containers span `next` nodes, scalars one)
+    static NLOHMANN_VIEW_ALWAYS_INLINE const node* after(const node* n) noexcept
+    {
+        return n + (is_container(*n) ? n->next : 1u);
+    }
+
+    /// first element (array) or first key (object) of a container
+    static NLOHMANN_VIEW_ALWAYS_INLINE const node* first_child(const node* n) noexcept
+    {
+        return n + 1;
+    }
+
+    /// end of the elements of a container
+    static NLOHMANN_VIEW_ALWAYS_INLINE const node* child_end(const node* n) noexcept
+    {
+        return n + n->next;
+    }
+};
+
+}  // namespace view
+}  // namespace detail
+NLOHMANN_JSON_NAMESPACE_END

include/nlohmann/detail/view/macro_scope.hpp

@@ -0,0 +1,63 @@
+//     __ _____ _____ _____
+//  __|  |   __|     |   | |  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
+
+#pragma once
+
+// Macros of json_view.hpp and its detail headers. json.hpp undefines its own
+// macros at its end (macro_unscope.hpp), so the view defines the few it needs
+// under its own prefix; json_view.hpp undefines them all at its end
+// (detail/view/macro_unscope.hpp). Configuration that json.hpp undefines is
+// read from detail::abi_config instead.
+
+#if (defined(__cplusplus) && __cplusplus >= 201703L) || (defined(_MSVC_LANG) && _MSVC_LANG >= 201703L)
+    #define NLOHMANN_VIEW_HAS_CPP_17 1
+#else
+    #define NLOHMANN_VIEW_HAS_CPP_17 0
+#endif
+
+#if defined(__GNUC__) || defined(__clang__)
+    #define NLOHMANN_VIEW_LIKELY(x) __builtin_expect(!!(x), 1)
+    #define NLOHMANN_VIEW_UNLIKELY(x) __builtin_expect(!!(x), 0)
+    #define NLOHMANN_VIEW_ALWAYS_INLINE inline __attribute__((always_inline))
+    #define NLOHMANN_VIEW_NOINLINE __attribute__((noinline))
+#elif defined(_MSC_VER)
+    #define NLOHMANN_VIEW_LIKELY(x) (x)
+    #define NLOHMANN_VIEW_UNLIKELY(x) (x)
+    #define NLOHMANN_VIEW_ALWAYS_INLINE __forceinline
+    #define NLOHMANN_VIEW_NOINLINE __declspec(noinline)
+#else
+    #define NLOHMANN_VIEW_LIKELY(x) (x)
+    #define NLOHMANN_VIEW_UNLIKELY(x) (x)
+    #define NLOHMANN_VIEW_ALWAYS_INLINE inline
+    #define NLOHMANN_VIEW_NOINLINE
+#endif
+
+// exceptions as in json.hpp (JSON_NOEXCEPTION, JSON_THROW_USER)
+#if (defined(__cpp_exceptions) || defined(__EXCEPTIONS) || defined(_CPPUNWIND)) && !defined(JSON_NOEXCEPTION)
+    #define NLOHMANN_VIEW_THROW(exception) throw exception
+#else
+    #include <cstdlib>
+    // (the exception is built first, so that the arguments of the throwing
+    // helpers count as used; the program ends anyway)
+    #define NLOHMANN_VIEW_THROW(exception) (static_cast<void>(exception), std::abort())
+#endif
+#if defined(JSON_THROW_USER)
+    #undef NLOHMANN_VIEW_THROW
+    #define NLOHMANN_VIEW_THROW JSON_THROW_USER
+#endif
+
+// the parser stores a node's first word at once where the layout of `node` is
+// known to be little-endian (MSVC targets are); elsewhere field by field
+#if (defined(__BYTE_ORDER__) && defined(__ORDER_LITTLE_ENDIAN__) && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__) || defined(_MSC_VER)
+    #define NLOHMANN_VIEW_LITTLE_ENDIAN 1
+#else
+    #define NLOHMANN_VIEW_LITTLE_ENDIAN 0
+#endif
+
+/// sixteen checks at fixed offsets 0..15
+#define NLOHMANN_VIEW_REPEAT16(X) X(0) X(1) X(2) X(3) X(4) X(5) X(6) X(7) X(8) X(9) X(10) X(11) X(12) X(13) X(14) X(15)

include/nlohmann/detail/view/macro_unscope.hpp

@@ -0,0 +1,20 @@
+//     __ _____ _____ _____
+//  __|  |   __|     |   | |  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
+
+#pragma once
+
+// undefine the macros of detail/view/macro_scope.hpp (at the end of json_view.hpp)
+
+#undef NLOHMANN_VIEW_HAS_CPP_17
+#undef NLOHMANN_VIEW_LIKELY
+#undef NLOHMANN_VIEW_UNLIKELY
+#undef NLOHMANN_VIEW_ALWAYS_INLINE
+#undef NLOHMANN_VIEW_NOINLINE
+#undef NLOHMANN_VIEW_THROW
+#undef NLOHMANN_VIEW_LITTLE_ENDIAN
+#undef NLOHMANN_VIEW_REPEAT16

include/nlohmann/detail/view/node.hpp

@@ -0,0 +1,97 @@
+//     __ _____ _____ _____
+//  __|  |   __|     |   | |  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
+
+#pragma once
+
+#include <cstddef> // size_t
+#include <cstdint> // uint8_t, uint16_t, uint32_t, uint64_t
+#include <cstring> // memcpy
+
+#include <nlohmann/json.hpp>
+#include <nlohmann/detail/view/macro_scope.hpp>
+
+NLOHMANN_JSON_NAMESPACE_BEGIN
+namespace detail
+{
+namespace view
+{
+
+// the node kinds are value_t values; the tests of is_container() and of the
+// number kinds depend on this numbering
+static_assert(static_cast<std::uint8_t>(value_t::null) == 0 && static_cast<std::uint8_t>(value_t::object) == 1
+              && static_cast<std::uint8_t>(value_t::array) == 2 && static_cast<std::uint8_t>(value_t::string) == 3
+              && static_cast<std::uint8_t>(value_t::boolean) == 4 && static_cast<std::uint8_t>(value_t::number_integer) == 5
+              && static_cast<std::uint8_t>(value_t::number_unsigned) == 6 && static_cast<std::uint8_t>(value_t::number_float) == 7,
+              "the node format depends on the numbering of value_t");
+
+/// node flags
+struct node_flags
+{
+    static constexpr std::uint8_t escaped = 1; ///< string payload lives in the decode arena, not the source
+    static constexpr std::uint8_t storage = 3; ///< mask: where a string or number token lives (index into document_data::base)
+    static constexpr std::uint8_t is_true = 4; ///< boolean value
+};
+
+/// One entry of the flat index, in document order. An object's members are
+/// stored as key node followed by the value's subtree. Integers keep their
+/// converted 64-bit value in the len/next bytes (the node after a scalar is
+/// always the next one, and the token length follows from `extra`).
+struct node
+{
+    std::uint8_t kind;   ///< value_t
+    std::uint8_t flags;  ///< node_flags
+    std::uint16_t extra; ///< numbers: integer digits (low byte) and fraction digits (high byte), 255 = "many"; otherwise 0
+    std::uint32_t off;   ///< source offset (string content, number token, literal, bracket); arena offset if node_flags::escaped
+    std::uint32_t len;   ///< string: decoded bytes; float: token bytes; array/object: element count
+    std::uint32_t next;  ///< array/object: number of nodes of the subtree (its extent in the enclosing sequence)
+};
+static_assert(sizeof(node) == 16, "node must stay 16 bytes");
+
+NLOHMANN_VIEW_ALWAYS_INLINE bool is_container(const node& n) noexcept
+{
+    return static_cast<unsigned>(n.kind) - 1u <= 1u;
+}
+
+/// the converted value of an integer node (stored in len/next)
+NLOHMANN_VIEW_ALWAYS_INLINE std::uint64_t integer_bits(const node& n) noexcept
+{
+    std::uint64_t v = 0;
+    std::memcpy(&v, reinterpret_cast<const unsigned char*>(&n) + 8, 8); // NOLINT(cppcoreguidelines-pro-type-reinterpret-cast)
+    return v;
+}
+
+NLOHMANN_VIEW_ALWAYS_INLINE void set_integer_bits(node& n, std::uint64_t v) noexcept
+{
+    std::memcpy(reinterpret_cast<unsigned char*>(&n) + 8, &v, 8); // NOLINT(cppcoreguidelines-pro-type-reinterpret-cast)
+}
+
+/// token length of a number node
+NLOHMANN_VIEW_ALWAYS_INLINE std::uint32_t number_length(const node& n) noexcept
+{
+    return n.kind == static_cast<std::uint8_t>(value_t::number_float) ? n.len
+           : (n.extra & 0xFFu) + (n.kind == static_cast<std::uint8_t>(value_t::number_integer) ? 1u : 0u);
+}
+
+/// estimated number of nodes for an input of `size` bytes (one node per ~12
+/// bytes covers typical documents without regrowth)
+inline std::size_t estimate_nodes(std::size_t size) noexcept
+{
+    return (size / 12) + 16;
+}
+
+/// estimated number of nodes for the input [src, src + size): pretty-printed
+/// input (whitespace after the first byte) needs about a node per 12 bytes,
+/// minified input up to one per 4 (yyjson tells the two apart the same way)
+inline std::size_t estimate_nodes(const char* src, std::size_t size) noexcept
+{
+    return size >= 2 && (src[1] == ' ' || src[1] == '\n' || src[1] == '\r' || src[1] == '\t') ? estimate_nodes(size) : (size / 4) + 16;
+}
+
+}  // namespace view
+}  // namespace detail
+NLOHMANN_JSON_NAMESPACE_END

include/nlohmann/detail/view/scan.hpp

@@ -0,0 +1,189 @@
+//     __ _____ _____ _____
+//  __|  |   __|     |   | |  JSON for Modern C++
+// |  |  |__   |  |  | | | |  version 3.12.0
+// |_____|_____|_____|_|___|  https://github.com/nlohmann/json
+//
+// SPDX-FileCopyrightText: 2013-2026 Niels Lohmann <https://nlohmann.me>
+// SPDX-FileCopyrightText: 2020 YaoYuan <https://github.com/ibireme/yyjson>
+// SPDX-License-Identifier: MIT
+
+#pragma once
+
+#include <array> // array
+#include <cstddef> // size_t
+#include <cstdint> // uint8_t, uint16_t, uint64_t
+#include <cstring> // memcpy
+
+#include <nlohmann/json.hpp>
+#include <nlohmann/detail/view/macro_scope.hpp>
+
+// Scanning primitives of the view's parser. The unrolled checks at fixed
+// offsets follow yyjson (https://github.com/ibireme/yyjson, MIT license): the
+// loads do not depend on each other, so the CPU can run ahead. Words are read
+// with read_eight_bytes(), so nothing here depends on the byte order.
+
+NLOHMANN_JSON_NAMESPACE_BEGIN
+namespace detail
+{
+namespace view
+{
+
+/// 1 for bytes that may appear verbatim in a string: 0x20..0x7F except '"' and '\\'
+inline const std::uint8_t* string_plain() noexcept
+{
+    static const std::array<std::uint8_t, 256> table =
+    {
+        {
+            0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 0x00..0x1F
+            1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 0x20..0x3F ('"')
+            1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, // 0x40..0x5F ('\\')
+            1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, // 0x60..0x7F
+            0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 0x80..0x9F
+            0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 0xA0..0xBF
+            0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 0xC0..0xDF
+            0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // 0xE0..0xFF
+        }
+    };
+    return table.data();
+}
+
+NLOHMANN_VIEW_ALWAYS_INLINE bool is_digit(unsigned char c) noexcept
+{
+    return static_cast<unsigned char>(c - '0') <= 9;
+}
+
+/// two bytes as they are in memory (only compared with byte-symmetric patterns)
+NLOHMANN_VIEW_ALWAYS_INLINE std::uint16_t load16(const unsigned char* p) noexcept
+{
+    std::uint16_t w = 0;
+    std::memcpy(&w, p, 2);
+    return w;
+}
+
+/// Advance over plain string bytes and well-formed UTF-8. Stops at a quote,
+/// a backslash, a control character, ill-formed UTF-8, or the end. The first
+/// 16 bytes are checked one by one, so that the position advances by
+/// constants in predicted branches (most strings are short); longer runs
+/// continue eight bytes at a time.
+NLOHMANN_VIEW_ALWAYS_INLINE const unsigned char* scan_string_run(const unsigned char* p, const unsigned char* e) noexcept
+{
+    const std::uint8_t* plain = string_plain();
+    for (;;)
+    {
+        if (e - p >= 16)
+        {
+#define NLOHMANN_VIEW_STEP(i) if (NLOHMANN_VIEW_LIKELY(plain[p[i]] != 0)) {} else { p += (i); goto stop; }
+            NLOHMANN_VIEW_REPEAT16(NLOHMANN_VIEW_STEP)
+#undef NLOHMANN_VIEW_STEP
+            p += 16;
+            while (e - p >= 8)
+            {
+                const std::uint64_t special = swar_string_special(read_eight_bytes(p));
+                if (special != 0)
+                {
+                    p += count_trailing_zeros(special) / 8;
+                    goto stop;
+                }
+                p += 8;
+            }
+            continue;
+        }
+        while (p != e && plain[*p] != 0)
+        {
+            ++p;
+        }
+        if (p == e)
+        {
+            return p;
+        }
+stop:
+        if (*p < 0x80)
+        {
+            return p; // quote, backslash, or control character
+        }
+        // non-ASCII: a run of well-formed sequences (the library's check, so
+        // that exactly what json::parse accepts is accepted)
+        do
+        {
+            const std::size_t n = validate_one_utf8(p, static_cast<std::size_t>(e - p));
+            if (n == 0)
+            {
+                return p;
+            }
+            p += n;
+        }
+        while (p != e && *p >= 0x80);
+    }
+}
+
+/// advance over ASCII digits
+NLOHMANN_VIEW_ALWAYS_INLINE const unsigned char* skip_digits(const unsigned char* p, const unsigned char* e) noexcept
+{
+    while (e - p >= 16)
+    {
+#define NLOHMANN_VIEW_STEP(i) if (NLOHMANN_VIEW_LIKELY(is_digit(p[i]))) {} else { return p + (i); }
+        NLOHMANN_VIEW_REPEAT16(NLOHMANN_VIEW_STEP)
+#undef NLOHMANN_VIEW_STEP
+        p += 16;
+    }
+    while (p != e && is_digit(*p))
+    {
+        ++p;
+    }
+    return p;
+}
+
+/// powers of ten up to 10^19 as integers
+inline std::uint64_t int_pow10(unsigned k) noexcept
+{
+    static const std::array<std::uint64_t, 20> table =
+    {
+        {
+            1u, 10u, 100u, 1000u, 10000u, 100000u, 1000000u, 10000000u, 100000000u, 1000000000u,
+            10000000000u, 100000000000u, 1000000000000u, 10000000000000u, 100000000000000u, 1000000000000000u,
+            10000000000000000u, 100000000000000000u, 1000000000000000000u, 10000000000000000000u
+        }
+    };
+    return table[k];
+}
+
+/// value of 0 < k < 8 digits at p in one step if [p, p + 8) lies below
+/// limit, else one digit at a time (whole blocks of eight digits are read by
+/// parse_upto19() directly)
+NLOHMANN_VIEW_ALWAYS_INLINE std::uint64_t parse_upto8(const unsigned char* p, unsigned k, const unsigned char* limit) noexcept
+{
+    if (NLOHMANN_VIEW_LIKELY(limit - p >= 8))
+    {
+        // move the k digits to the top and pad the vacated low bytes with '0'
+        const unsigned shift = 8 * (8 - k);
+        return parse_eight_digits((read_eight_bytes(p) << shift) | (0x3030303030303030u >> (8 * k)));
+    }
+    std::uint64_t v = 0;
+    for (unsigned i = 0; i < k; ++i)
+    {
+        v = (v * 10) + static_cast<std::uint64_t>(p[i] - '0');
+    }
+    return v;
+}
+
+/// value of k <= 19 digits at p
+NLOHMANN_VIEW_ALWAYS_INLINE std::uint64_t parse_upto19(const unsigned char* p, unsigned k, const unsigned char* limit) noexcept
+{
+    std::uint64_t w = 0;
+    while (k >= 8)
+    {
+        // (eight digits of the token: they lie below limit)
+        w = (w * 100000000u) + parse_eight_digits(read_eight_bytes(p));
+        p += 8;
+        k -= 8;
+    }
+    if (k != 0)
+    {
+        w = (w * int_pow10(k)) + parse_upto8(p, k, limit);
+    }
+    return w;
+}
+
+}  // namespace view
+}  // namespace detail
+NLOHMANN_JSON_NAMESPACE_END

include/nlohmann/detail/view/string_ref.hpp

@@ -0,0 +1,111 @@
+//     __ _____ _____ _____
+//  __|  |   __|     |   | |  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
+
+#pragma once
+
+#include <algorithm> // min
+#include <cstddef> // size_t
+#include <cstring> // memcmp, strlen
+#include <string> // basic_string
+
+#include <nlohmann/json.hpp>
+#include <nlohmann/detail/view/macro_scope.hpp>
+
+#if NLOHMANN_VIEW_HAS_CPP_17
+    #include <string_view> // string_view
+#endif
+#ifndef JSON_NO_IO
+    #include <ostream> // ostream
+#endif
+
+NLOHMANN_JSON_NAMESPACE_BEGIN
+namespace detail
+{
+namespace view
+{
+
+#if NLOHMANN_VIEW_HAS_CPP_17
+using string_ref = std::string_view;
+#else
+/// minimal C++11 stand-in for std::string_view
+class string_ref
+{
+  public:
+    using size_type = std::size_t;
+    using const_iterator = const char*;
+
+    string_ref() noexcept = default;
+    string_ref(const char* s) : m_data(s), m_size(std::strlen(s)) {} // NOLINT(google-explicit-constructor,hicpp-explicit-conversions)
+    string_ref(const char* s, std::size_t n) noexcept : m_data(s), m_size(n) {}
+    template<typename Traits, typename Alloc>
+    string_ref(const std::basic_string<char, Traits, Alloc>& s) noexcept : m_data(s.data()), m_size(s.size()) {} // NOLINT(google-explicit-constructor,hicpp-explicit-conversions)
+
+    const char* data() const noexcept
+    {
+        return m_data;
+    }
+    std::size_t size() const noexcept
+    {
+        return m_size;
+    }
+    std::size_t length() const noexcept
+    {
+        return m_size;
+    }
+    bool empty() const noexcept
+    {
+        return m_size == 0;
+    }
+    const char* begin() const noexcept
+    {
+        return m_data;
+    }
+    const char* end() const noexcept
+    {
+        return m_data + m_size;
+    }
+    char operator[](std::size_t i) const noexcept
+    {
+        return m_data[i];
+    }
+
+    template<typename Traits, typename Alloc>
+    explicit operator std::basic_string<char, Traits, Alloc>() const
+    {
+        return std::basic_string<char, Traits, Alloc>(m_data, m_size);
+    }
+
+    friend bool operator==(string_ref a, string_ref b) noexcept
+    {
+        return a.m_size == b.m_size && (a.m_size == 0 || std::memcmp(a.m_data, b.m_data, a.m_size) == 0);
+    }
+    friend bool operator!=(string_ref a, string_ref b) noexcept
+    {
+        return !(a == b);
+    }
+    friend bool operator<(string_ref a, string_ref b) noexcept
+    {
+        const int c = std::memcmp(a.m_data, b.m_data, (std::min)(a.m_size, b.m_size));
+        return c != 0 ? c < 0 : a.m_size < b.m_size;
+    }
+#ifndef JSON_NO_IO
+    friend std::ostream& operator<<(std::ostream& o, string_ref s)
+    {
+        return o.write(s.m_data, static_cast<std::streamsize>(s.m_size));
+    }
+#endif
+
+  private:
+    const char* m_data = "";
+    std::size_t m_size = 0;
+};
+#endif
+
+}  // namespace view
+}  // namespace detail
+NLOHMANN_JSON_NAMESPACE_END

include/nlohmann/json.hpp

@@ -45,6 +45,7 @@
 
 #include <nlohmann/adl_serializer.hpp>
 #include <nlohmann/byte_container_with_subtype.hpp>
+#include <nlohmann/detail/abi_config.hpp>
 #include <nlohmann/detail/conversions/from_json.hpp>
 #include <nlohmann/detail/conversions/to_json.hpp>
 #include <nlohmann/detail/exceptions.hpp>

single_include/nlohmann/json.hpp

@@ -7236,6 +7236,43 @@ class byte_container_with_subtype : public BinaryType
 
 NLOHMANN_JSON_NAMESPACE_END
 
+// #include <nlohmann/detail/abi_config.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 <nlohmann/detail/abi_macros.hpp>
+
+
+NLOHMANN_JSON_NAMESPACE_BEGIN
+namespace detail
+{
+
+/*!
+@brief the configuration macros that change the library's behavior
+
+json.hpp undefines these macros at its end (see macro_unscope.hpp), so code
+that builds on the library after it (json_view.hpp) reads them here. Like the
+macros, they are part of the ABI namespace, so they always match the
+basic_json they are used with.
+*/
+struct abi_config
+{
+    /// JSON_STRICT_NUL_HANDLING: a null byte is an error, not the end of input
+    static constexpr bool strict_nul_handling = JSON_STRICT_NUL_HANDLING != 0;
+    /// JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON
+    static constexpr bool legacy_discarded_value_comparison = JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON != 0;
+};
+
+}  // namespace detail
+NLOHMANN_JSON_NAMESPACE_END
+
 // #include <nlohmann/detail/conversions/from_json.hpp>
 
 // #include <nlohmann/detail/conversions/to_json.hpp>
@@ -10047,8 +10084,9 @@ NLOHMANN_JSON_NAMESPACE_END
 
 
 
+#include <array> // array
 #include <cstddef> // size_t
-#include <cstdint> // uint64_t
+#include <cstdint> // uint64_t, uint8_t
 #include <cstring> // memcpy
 
 // #include <nlohmann/detail/bit_ops.hpp>
@@ -10356,6 +10394,51 @@ inline std::size_t string_bulk_run(const unsigned char* data, std::size_t n) noe
     return scalar_string_bulk_run(data, n);
 }
 
+// Decode the 4 hex digits at [data, data+4) - the digits following a `\u`
+// escape - into a codepoint 0x0000..0xFFFF via one table lookup per byte
+// (after yyjson's read_hex_u16), or return -1 if any of the 4 bytes is not a
+// hex digit ('0'..'9', 'A'..'F', 'a'..'f'). The caller must already have
+// checked that 4 bytes are available; used by lexer::get_codepoint()'s
+// contiguous fast path. On -1 it falls back to the byte-at-a-time loop, which
+// stops at the first invalid digit, so the reported error and position are
+// unaffected by this fast path.
+inline int hex_codepoint(const unsigned char* data) noexcept
+{
+    static const std::array<std::uint8_t, 256> hex_digit_table = // NOLINT(cppcoreguidelines-avoid-non-const-global-variables)
+    {
+        {
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 00..0F
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 10..1F
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 20..2F
+            0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 30..3F ('0'..'9')
+            0xFF, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 40..4F ('A'..'F')
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 50..5F
+            0xFF, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 60..6F ('a'..'f')
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 70..7F
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 80..8F
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 90..9F
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // A0..AF
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // B0..BF
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // C0..CF
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // D0..DF
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // E0..EF
+            0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF  // F0..FF
+        }
+    };
+
+    const std::uint8_t d0 = hex_digit_table[data[0]];
+    const std::uint8_t d1 = hex_digit_table[data[1]];
+    const std::uint8_t d2 = hex_digit_table[data[2]];
+    const std::uint8_t d3 = hex_digit_table[data[3]];
+    // every valid digit is <= 0xF; the combined OR only exceeds it if at
+    // least one of the four bytes was not a hex digit (looked up as 0xFF)
+    if ((d0 | d1 | d2 | d3) > 0x0F)
+    {
+        return -1;
+    }
+    return (d0 << 12) | (d1 << 8) | (d2 << 4) | d3;
+}
+
 }  // namespace detail
 NLOHMANN_JSON_NAMESPACE_END
 
@@ -10560,6 +10643,44 @@ class lexer : public lexer_base<BasicJsonType>
     // scan functions
     /////////////////////
 
+    /// contiguous input: try to decode the 4 hex digits following `\u`
+    /// directly from the input buffer via hex_codepoint(), instead of 4 calls
+    /// to get(). On success, advances the adapter and the position counters
+    /// exactly as those 4 get() calls would (a hex digit is never '\n', so
+    /// only the flat counters move) and leaves @a current holding the last of
+    /// the 4 digits, just as the last such get() would; the codepoint is
+    /// written to @a out. Makes no state change and returns false - for a
+    /// pending unget, fewer than 4 remaining bytes, or any of the 4 bytes not
+    /// being a hex digit - so the caller falls back unchanged to the
+    /// per-character loop, which then reports the same diagnostic (stopping
+    /// at the first invalid digit) as before this optimization.
+    bool get_codepoint_bulk(std::true_type /*bulk*/, int& out)
+    {
+        if (next_unget || ia.bulk_remaining() < 4)
+        {
+            return false;
+        }
+        const char_type* const raw = ia.bulk_data();
+        const int codepoint = hex_codepoint(reinterpret_cast<const unsigned char*>(raw));
+        if (codepoint < 0)
+        {
+            return false;
+        }
+        ia.bulk_skip(4);
+        // a hex digit is never a newline, so only the flat counters advance
+        position.chars_read_total += 4;
+        position.chars_read_current_line += 4;
+        current = char_traits<char_type>::to_int_type(raw[3]);
+        out = codepoint;
+        return true;
+    }
+
+    /// streaming input: no bulk fast path
+    bool get_codepoint_bulk(std::false_type /*bulk*/, int& /*out*/) const noexcept
+    {
+        return false;
+    }
+
     /*!
     @brief get codepoint from 4 hex characters following `\u`
 
@@ -10579,6 +10700,14 @@ class lexer : public lexer_base<BasicJsonType>
     {
         // this function only makes sense after reading `\u`
         JSON_ASSERT(current == 'u');
+
+        // contiguous input: decode all 4 hex digits directly from the buffer
+        int fast_codepoint = 0;
+        if (get_codepoint_bulk(std::integral_constant<bool, bulk_scan> {}, fast_codepoint))
+        {
+            return fast_codepoint;
+        }
+
         int codepoint = 0;
 
         const auto factors = { 12u, 8u, 4u, 0u };

tests/CMakeLists.txt

@@ -279,6 +279,10 @@ if(json_32bit_test_only)
 elseif(NOT json_32bit_test)
     list(FILTER files EXCLUDE REGEX src/unit-32bit.cpp)
 endif()
+if(NOT JSON_MultipleHeaders)
+    # the internal headers of json_view are not part of a single header yet
+    list(FILTER files EXCLUDE REGEX src/unit-json_view_builder.cpp)
+endif()
 
 foreach(file ${files})
     json_test_add_test_for(${file} MAIN test_main CXX_STANDARDS ${test_cxx_standards} ${test_force})

tests/src/unit-class_lexer.cpp

@@ -18,6 +18,7 @@ using nlohmann::json;
 #include <cstdlib> // strtod
 #include <cstring> // memcpy
 #include <map> // map
+#include <random> // mt19937
 #include <sstream> // stringstream
 #include <string> // string
 #include <utility> // pair
@@ -664,6 +665,184 @@ TEST_CASE("lexer string fast path")
     }
 }
 
+TEST_CASE("lexer escape fast path")
+{
+    // json::accept() never throws, so this section stays covered without
+    // exceptions; it pins which of the cases below are valid/invalid and
+    // checks the contiguous and streaming paths agree on that classification.
+    SECTION("accept() parity")
+    {
+        const std::vector<std::pair<std::string, bool>> cases =
+        {
+            {"\\u0041", true}, {"\\u00e4", true}, {"\\u00E4", true},
+            {"\\uD83D\\uDE00", true},
+            {"\\u12", false}, {"\\u12G4", false}, {"\\uXYZW", false},
+            {"\\uD800", false}, {"\\uD800A", false}, {"\\uD800\\u0041", false},
+            {"\\uDC00", false}, {"\\u", false}
+        };
+
+        for (const auto& c : cases)
+        {
+            for (const std::size_t offset :
+                    {
+                        std::size_t{0}, std::size_t{9}
+                    })
+            {
+                const std::string doc = "[\"" + std::string(offset, 'a') + c.first + "\"]";
+                CAPTURE(doc);
+                CHECK(json::accept(doc) == c.second);
+                std::stringstream ss(doc);
+                CHECK(json::accept(ss) == c.second);
+            }
+        }
+    }
+
+#if !defined(JSON_NOEXCEPTION)
+    // the full outcome of parsing @a doc: the parsed value, or the exact
+    // error message, so a mismatch in either is caught
+    const auto outcome = [](const std::string & doc, bool streaming) -> std::string
+    {
+        try
+        {
+            if (streaming)
+            {
+                std::stringstream ss(doc);
+                const json j = json::parse(ss);
+                return j.dump();
+            }
+            const json j = json::parse(doc);
+            return j.dump();
+        }
+        catch (const json::exception& e)
+        {
+            return {e.what()};
+        }
+    };
+
+    SECTION("contiguous vs streaming parity")
+    {
+        const std::vector<std::string> escapes =
+        {
+            "\\u0041",             // "A"
+            "\\u00e4",             // "ä" (lowercase hex)
+            "\\u00E4",             // "ä" (uppercase hex)
+            "\\uD83D\\uDE00",      // valid surrogate pair (an emoji)
+            "\\u12",               // truncated: only 2 hex digits before the closing quote
+            "\\u12G4",             // invalid hex digit at the 3rd position
+            "\\uXYZW",             // all 4 bytes invalid
+            "\\uD800",             // lone high surrogate, string ends right after
+            "\\uD800A",            // high surrogate not followed by another \u escape
+            "\\uD800\\u0041",      // high surrogate followed by \u, but not a low surrogate
+            "\\uDC00",             // lone low surrogate
+            "\\u",                 // '\u' with nothing after (closing quote right away)
+        };
+
+        // once at the start of the string and once past the first 8-byte SWAR
+        // word of the outer string_bulk_run, so the escape is reached both
+        // right after the opening quote and mid-run
+        for (const auto& escape : escapes)
+        {
+            for (const std::size_t offset :
+                    {
+                        std::size_t{0}, std::size_t{9}
+                    })
+            {
+                const std::string doc = "[\"" + std::string(offset, 'a') + escape + "\"]";
+                CAPTURE(doc);
+                CHECK(outcome(doc, false) == outcome(doc, true));
+            }
+
+            // the escape is the last thing before end of input: no closing
+            // quote at all
+            const std::string truncated_doc = "[\"" + escape;
+            CAPTURE(truncated_doc);
+            CHECK(outcome(truncated_doc, false) == outcome(truncated_doc, true));
+        }
+    }
+
+    SECTION("truncated \\u escape at every distance from the end of input")
+    {
+        // ia.bulk_remaining() must correctly report fewer than 4 bytes for
+        // every possible count of trailing hex-looking bytes (0, 1, 2, or 3)
+        // before end of input, so the fast path declines and the byte path
+        // alone reports the "must be followed by 4 hex digits" error, at the
+        // same position, in every case
+        for (const std::string& tail :
+                {
+                    std::string{}, std::string("1"), std::string("12"), std::string("123")
+                })
+        {
+            const std::string doc = "[\"\\u" + tail;
+            CAPTURE(doc);
+            CHECK(outcome(doc, false) == outcome(doc, true));
+            CHECK(outcome(doc, false).find("must be followed by 4 hex digits") != std::string::npos);
+        }
+    }
+
+    SECTION("invalid hex digit at every position of the 4")
+    {
+        // the fast path must decline for *any* invalid byte among the 4, not
+        // just the first, and the byte path must then stop at exactly that
+        // position - same as it always has
+        for (std::size_t bad_pos = 0; bad_pos < 4; ++bad_pos)
+        {
+            std::string digits = "1234";
+            digits[bad_pos] = 'g'; // not a hex digit
+            const std::string doc = "[\"\\u" + digits + "\"]";
+            CAPTURE(doc);
+            CHECK(outcome(doc, false) == outcome(doc, true));
+            CHECK(outcome(doc, false).find("must be followed by 4 hex digits") != std::string::npos);
+        }
+    }
+
+    SECTION("random escapes")
+    {
+        // A seeded PRNG builds the 4 bytes following `\u` from a mix of hex
+        // digits and non-hex bytes, at varying distances from the start of
+        // the string, to compare the two scanners on many more shapes than
+        // are practical to enumerate by hand.
+        std::mt19937 gen(7654321); // NOLINT(cert-msc32-c,cert-msc51-cpp)
+        const std::string hex_alphabet = "0123456789AaBbCcDdEeFf";
+        std::uniform_int_distribution<std::size_t> pick_hex(0, hex_alphabet.size() - 1);
+        std::uniform_int_distribution<int> pick_byte(1, 255); // never NUL
+        std::uniform_int_distribution<int> pick_is_hex(0, 4); // 4-in-5 chance of a hex digit
+        std::uniform_int_distribution<std::size_t> pick_offset(0, 12);
+
+        std::vector<std::string> mismatches;
+        for (int iter = 0; iter < 3000; ++iter)
+        {
+            std::string digits;
+            for (int i = 0; i < 4; ++i)
+            {
+                if (pick_is_hex(gen) != 0)
+                {
+                    digits += hex_alphabet[pick_hex(gen)];
+                }
+                else
+                {
+                    char c = static_cast<char>(pick_byte(gen));
+                    if (c == '"' || c == '\\')
+                    {
+                        // keep the string well-formed apart from the escape
+                        // itself, so any mismatch is attributable to the \u
+                        // handling and not to an unrelated quote/escape
+                        c = 'z';
+                    }
+                    digits += c;
+                }
+            }
+            const std::string doc = "[\"" + std::string(pick_offset(gen), 'a') + "\\u" + digits + "\"]";
+            if (outcome(doc, false) != outcome(doc, true))
+            {
+                mismatches.push_back(doc);
+            }
+        }
+        CAPTURE(mismatches);
+        CHECK(mismatches.empty());
+    }
+#endif
+}
+
 namespace
 {
 // the index of the decimal point (or npos) and of the end of the mantissa of a

tests/src/unit-json_view_builder.cpp

@@ -0,0 +1,399 @@
+//     __ _____ _____ _____
+//  __|  |   __|     |   | |  JSON for Modern C++ (supporting code)
+// |  |  |__   |  |  | | | |  version 3.12.0
+// |_____|_____|_____|_|___|  https://github.com/nlohmann/json
+//
+// SPDX-FileCopyrightText: 2013-2026 Niels Lohmann <https://nlohmann.me>
+// SPDX-License-Identifier: MIT
+
+#include "doctest_compatibility.h"
+
+#include <nlohmann/json.hpp>
+#include <nlohmann/detail/view/builder.hpp>
+#include <nlohmann/detail/view/string_ref.hpp>
+using nlohmann::json;
+
+#include <cstdint>
+#include <fstream>
+#include <map>
+#include <memory>
+#include <random>
+#include <sstream>
+#include <string>
+#include <utility>
+#include <vector>
+
+#include <test_data.hpp>
+
+namespace
+{
+using nlohmann::detail::view::document_data;
+using nlohmann::detail::view::node;
+
+// the node index of a text; a vector input has no terminating NUL, so that
+// AddressSanitizer catches any read past the last byte
+struct built
+{
+    std::unique_ptr<document_data, document_data::deleter> data{}; // NOLINT(readability-redundant-member-init)
+    std::vector<char> copy{}; // NOLINT(readability-redundant-member-init)
+    bool ok = false;
+    nlohmann::detail::view::parse_failure failure{};
+};
+
+template<typename FloatType = double>
+built build(const std::string& text, bool comments, bool trailing_commas, bool sentinel)
+{
+    built r;
+    r.data.reset(document_data::create(nlohmann::detail::view::estimate_nodes(text.data(), text.size())));
+    const char* src = text.c_str();
+    if (!sentinel)
+    {
+        r.copy.assign(text.begin(), text.end());
+        src = r.copy.data();
+    }
+    r.ok = nlohmann::detail::view::build < FloatType, !nlohmann::detail::abi_config::strict_nul_handling > (*r.data, src, text.size(), comments, trailing_commas, sentinel, r.failure);
+    r.data->src = src;
+    r.data->base[0] = src;
+    r.data->base[1] = r.data->arena.data();
+    return r;
+}
+
+// the value of a subtree, as json::parse would build it
+json value_of(const document_data& d, const node*& n)
+{
+    const node& x = *n;
+    ++n;
+    switch (static_cast<json::value_t>(x.kind))
+    {
+        case json::value_t::object:
+        {
+            json o = json::object();
+            const node* const end = &x + x.next;
+            while (n != end)
+            {
+                const std::string key(d.str(*n), n->len);
+                ++n;
+                o[key] = value_of(d, n);
+            }
+            return o;
+        }
+        case json::value_t::array:
+        {
+            json a = json::array();
+            const node* const end = &x + x.next;
+            while (n != end)
+            {
+                a.push_back(value_of(d, n));
+            }
+            return a;
+        }
+        case json::value_t::string:
+            return std::string(d.str(x), x.len);
+        case json::value_t::boolean:
+            return (x.flags & nlohmann::detail::view::node_flags::is_true) != 0;
+        case json::value_t::number_integer:
+            return static_cast<std::int64_t>(nlohmann::detail::view::integer_bits(x));
+        case json::value_t::number_unsigned:
+            return nlohmann::detail::view::integer_bits(x);
+        case json::value_t::number_float:
+            return json::parse(std::string(d.src + x.off, x.len)).get<double>();
+        case json::value_t::null:
+        case json::value_t::binary:
+        case json::value_t::discarded:
+        default:
+            return nullptr;
+    }
+}
+
+json value_of(const built& b)
+{
+    const node* n = b.data->tape;
+    json v = value_of(*b.data, n);
+    CHECK(n == b.data->tape + b.data->tape_size);
+    return v;
+}
+
+// accept/reject and the value must match json::parse, for all options and
+// with and without a NUL after the text
+void check_same(const std::string& text)
+{
+    CAPTURE(text);
+    for (int options = 0; options < 4; ++options)
+    {
+        const bool comments = (options & 1) != 0;
+        const bool trailing_commas = (options & 2) != 0;
+        const bool accepted = json::accept(text, comments, trailing_commas);
+        for (const bool sentinel :
+                {
+                    true, false
+                })
+        {
+            const built b = build(text, comments, trailing_commas, sentinel);
+            CHECK(b.ok == accepted);
+            if (b.ok && accepted)
+            {
+                CHECK(value_of(b) == json::parse(text, nullptr, true, comments, trailing_commas));
+            }
+        }
+    }
+}
+
+// a small deterministic generator of documents
+struct generator
+{
+    std::mt19937 rng{5295}; // NOLINT(cert-msc32-c,cert-msc51-cpp,bugprone-random-generator-seed)
+
+    int r(int n)
+    {
+        return static_cast<int>(rng() % static_cast<unsigned>(n));
+    }
+
+    void ws(std::string& o)
+    {
+        for (int n = r(4) == 0 ? r(12) : r(2); n > 0; --n)
+        {
+            o += " \n\t\r  "[r(6)];
+        }
+    }
+
+    void str(std::string& o)
+    {
+        static const char* const pieces[] = {"a", "Z", " ", "~", "\\n", "\\\"", "\\\\", "\\/", "\\u00e9", "\\ud83d\\ude00", "\xc3\xa9", "\xe3\x81\x82", "\xf0\x9f\x98\x80", "\x7f", "\\u001f", "long enough text to leave the first 16 bytes"}; // NOLINT(cppcoreguidelines-avoid-c-arrays,hicpp-avoid-c-arrays,modernize-avoid-c-arrays)
+        o += '"';
+        for (int n = r(3) == 0 ? r(20) : r(6); n > 0; --n)
+        {
+            o += pieces[r(16)];
+        }
+        o += '"';
+    }
+
+    void num(std::string& o)
+    {
+        static const char* const numbers[] = {"0", "-0", "1", "-1", "12", "123456789", "1234567890123456789", "9223372036854775807", "-9223372036854775808", // NOLINT(cppcoreguidelines-avoid-c-arrays,hicpp-avoid-c-arrays,modernize-avoid-c-arrays)
+                                              "9223372036854775808", "18446744073709551615", "18446744073709551616", "-9223372036854775809",
+                                              "1.5", "-2.25e-3", "1e10", "1E+2", "0.000001", "3.141592653589793238462643", "1e308", "-1e-400", "123.456e7"
+                                             };
+        o += numbers[r(22)];
+    }
+
+    void value(std::string& o, int depth)
+    {
+        ws(o);
+        const int k = depth > 5 ? 2 + r(6) : r(8);
+        if (k == 0 || k == 1)
+        {
+            const bool object = k == 0;
+            o += object ? '{' : '[';
+            for (int i = r(5); i > 0; --i)
+            {
+                ws(o);
+                if (object)
+                {
+                    str(o);
+                    ws(o);
+                    o += ':';
+                }
+                value(o, depth + 1);
+                o += i > 1 ? "," : "";
+            }
+            ws(o);
+            o += object ? '}' : ']';
+        }
+        else if (k < 4)
+        {
+            str(o);
+        }
+        else if (k < 6)
+        {
+            num(o);
+        }
+        else
+        {
+            static const char* const literals[] = {"true", "false", "null"}; // NOLINT(cppcoreguidelines-avoid-c-arrays,hicpp-avoid-c-arrays,modernize-avoid-c-arrays)
+            o += literals[r(3)];
+        }
+        ws(o);
+    }
+};
+} // namespace
+
+TEST_CASE("json_view string_ref")
+{
+    // std::string_view in C++17, a stand-in with the same members before
+    using nlohmann::detail::view::string_ref;
+    const std::string text = "abc";
+    const string_ref r(text);
+    CHECK(r.length() == 3);
+    CHECK(std::string(r.begin(), r.end()) == "abc");
+    CHECK(r[1] == 'b');
+    CHECK(r != string_ref("abd"));
+    CHECK_FALSE(r != string_ref("abcd", 3));
+    std::ostringstream o;
+    o << r;
+    CHECK(o.str() == "abc");
+}
+
+TEST_CASE("json_view builder")
+{
+    SECTION("scalars and containers")
+    {
+        for (const char* text :
+                {
+                    "null", "true", "false", "0", "-0", "42", "-42", "1.5", "\"\"", "\"abc\"", "[]", "{}", "[1,2,3]", "{\"a\":1,\"b\":[true,null]}", // NOLINT(modernize-raw-string-literal)
+                    " [ 1 , 2 ] ", "{\"a\" : {\"b\" : {}}}", "[[[]]]", "\"\\u00e4\\n\\ud83d\\ude00\"", "{\"a\":1,\"a\":2}", "18446744073709551616", // NOLINT(modernize-raw-string-literal)
+                    "-9223372036854775809", "123456789012345678901234567890", "1e400", "-1e400", "1.7976931348623157e308"
+                })
+        {
+            check_same(text);
+        }
+
+        // the midpoint between the largest double and 2^1024 rounds to
+        // infinity (an overflow), one less to the largest double: with more
+        // than 19 digits, Eisel-Lemire cannot decide these, and the overflow
+        // check needs the exact comparison with the midpoint
+        const std::string midpoint = "179769313486231580793728971405303415079934132710037826936173778980444968292764750946649017977587207096330286416692887910946555547851940402630657488671505820681908902000708383676273854845817711531764475730270069855571366959622842914819860834936475292719074168444365510704342711559699508093042880177904174497792";
+        const std::string below = "179769313486231580793728971405303415079934132710037826936173778980444968292764750946649017977587207096330286416692887910946555547851940402630657488671505820681908902000708383676273854845817711531764475730270069855571366959622842914819860834936475292719074168444365510704342711559699508093042880177904174497791";
+        check_same(midpoint);
+        check_same("-" + midpoint);
+        check_same(below);
+        check_same("[" + below + "," + midpoint + "]");
+
+        // the check uses the floating-point type of the document: with float,
+        // the view rejects what parse() rejects (out_of_range.406), and a
+        // double document is not affected
+        using float_json = nlohmann::basic_json<std::map, std::vector, std::string, bool, std::int64_t, std::uint64_t, float>;
+        CHECK_FALSE(float_json::accept("1e39"));
+        CHECK(float_json::accept("3.4028235e38"));
+        CHECK_FALSE(float_json::accept("3.4028236e38"));
+        for (const char* text :
+                {
+                    "1e39", "-1e39", "3.4028235e38", "-3.4028235e38", "3.4028236e38", "-3.4028236e38", "3.4028234663852886e38", "1e38",
+                    "340282356779733661637539395458142568448", "340282356779733661637539395458142568447.99", "0.00034028236e42",
+                    "[1.5e38, 3.5e38]", "{\"a\": 1e-50, \"b\": 1e39}"
+                })
+        {
+            CAPTURE(text);
+            const bool float_accepted = float_json::accept(text);
+            for (const bool sentinel :
+                    {
+                        true, false
+                    })
+            {
+                const built f = build<float>(text, false, false, sentinel);
+                CHECK(f.ok == float_accepted);
+                if (!f.ok)
+                {
+                    CHECK(f.failure.code == nlohmann::detail::view::error_code::number_overflow);
+                }
+                CHECK(build<double>(text, false, false, sentinel).ok == json::accept(text));
+            }
+        }
+    }
+
+    SECTION("malformed input")
+    {
+        for (const char* text :
+                {
+                    "", " ", "[", "]", "{", "}", "[1,]", "{\"a\":1,}", "[1 2]", "{\"a\" 1}", "{1:2}", "tru", "nul", "fals", "truex", "-", "01", "1.", ".5", "1e", "1e+",
+                    "\"", "\"abc", "\"\\x\"", "\"\\u12\"", "\"\\u12G4\"", "\"\\ud800\"", "\"\\udc00\"", "\"\\ud800\\u0041\"", "\"\x01\"", "\"\xff\"", "\"\xc3\"", // NOLINT(modernize-raw-string-literal)
+                    "\"\xe0\x80\x80\"", "\"\xed\xa0\x80\"", "[1]x", "[1] [2]", "/", "/*", "/* */ 1", "// c\n1", "1 // c", "[1,/*c*/2]", "[1,2,]"
+                })
+        {
+            check_same(text);
+        }
+    }
+
+    SECTION("NUL, BOM, and whitespace")
+    {
+        // a NUL inside a string is a control character, as for json::parse
+        // (where a NUL ends the input, it does so only between values)
+        for (const bool sentinel :
+                {
+                    true, false
+                })
+        {
+            const built b = build(std::string("[\"ab\0cd\"]", 9), false, false, sentinel);
+            CHECK(!b.ok);
+            CHECK(b.failure.code == nlohmann::detail::view::error_code::string_control_character);
+            CHECK(b.failure.offset == 4);
+        }
+        check_same(std::string("[1]\0garbage", 11));
+        check_same(std::string("[1\0]", 4));
+        check_same(std::string("[1, // c\0\n2]", 12));
+        check_same(std::string("[1, /* c\0 */ 2]", 15));
+        check_same("\xEF\xBB\xBF[1]");
+        check_same("\xEF\xBB[1]");
+        check_same(" \t\r\n 7 \n");
+        for (const char* text :
+                {"[1]\r", "[1]\n", "[1]\r\n", "[1,\r2]", "[1,\r\n2]", "7\r", "\"x\"\r", "{\"a\":\r\n1}\r", "[\n  1,\n  2\n]", "{\n    \"a\": [\n        1\n    ]\n}"
+                })
+        {
+            check_same(text);
+        }
+    }
+
+    SECTION("deep nesting")
+    {
+        // the open containers beyond 64 levels live on the heap
+        for (const std::size_t depth :
+                {
+                    63u, 64u, 65u, 1000u, 100000u
+                })
+        {
+            const std::string arrays = std::string(depth, '[') + std::string(depth, ']');
+            const built b = build(arrays, false, false, false);
+            REQUIRE(b.ok);
+            CHECK(b.data->tape_size == depth);
+            CHECK(b.data->tape[0].next == depth);
+            std::string objects;
+            for (std::size_t i = 0; i < depth; ++i)
+            {
+                objects += "{\"a\":";
+            }
+            objects += '1' + std::string(depth, '}');
+            const built o = build(objects, false, false, false);
+            REQUIRE(o.ok);
+            CHECK(o.data->tape_size == (2 * depth) + 1);
+            CHECK(!build(std::string(depth, '[') + std::string(depth - 1, ']'), false, false, false).ok);
+        }
+    }
+
+    SECTION("generated documents and damaged copies")
+    {
+        generator g;
+        for (int i = 0; i < 3000; ++i)
+        {
+            std::string text;
+            g.value(text, 0);
+            check_same(text);
+            // damage: flip one byte, or cut the text
+            std::string damaged = text;
+            const auto at = static_cast<std::size_t>(g.r(static_cast<int>(damaged.size())));
+            static const char replacements[] = {'x', '"', '\\', ',', ':', ']', '}', '[', '{', '1', '-', '.', 'e', '\0', '\n', '/'}; // NOLINT(cppcoreguidelines-avoid-c-arrays,hicpp-avoid-c-arrays,modernize-avoid-c-arrays)
+            damaged[at] = replacements[g.r(16)];
+            check_same(damaged);
+            check_same(text.substr(0, at));
+        }
+    }
+
+    SECTION("test files")
+    {
+        for (const char* name :
+                {
+                    "/json.org/1.json", "/json.org/2.json", "/json.org/3.json", "/json.org/4.json", "/json.org/5.json",
+                    "/json_testsuite/sample.json", "/nativejson-benchmark/canada.json", "/nativejson-benchmark/citm_catalog.json",
+                    "/nativejson-benchmark/twitter.json", "/json_tests/pass1.json", "/json_tests/pass2.json", "/json_tests/pass3.json"
+                })
+        {
+            CAPTURE(name);
+            std::ifstream f(std::string(TEST_DATA_DIRECTORY) + name, std::ios::binary);
+            std::stringstream ss;
+            ss << f.rdbuf();
+            const std::string text = ss.str();
+            REQUIRE(!text.empty());
+            const built b = build(text, false, false, true);
+            REQUIRE(b.ok);
+            CHECK(value_of(b) == json::parse(text));
+        }
+    }
+}