Binary readers and writers: faster UTF-8 checks of strings

With error_handler_t::strict (and replace/ignore), every string a binary reader reads is checked for valid UTF-8, and every string a binary writer writes. The readers used the byte-at-a-time decoder; the writers skipped ASCII eight bytes at a time. - New valid_utf8_string() decides strings of up to 16 ASCII bytes, most keys and many values, with one or two word tests, and checks longer strings 16 bytes at a time with the NEON/SSSE3 UTF-8 check the lexer uses (vector_utf8_valid_blocks: the same lookup, but quotes and control characters are ordinary characters). Other targets and JSON_NO_SIMD use valid_utf8_prefix() as before. - The readers check strings with it (for one-byte character types) in an out-of-line function, so that reading without a check does not pay for its code. - The writers' check_utf8() uses it, and the CBOR and MessagePack writers write a checked string of up to max_text_in_reservation characters in one reservation, as they write unchecked ones. Signed-off-by: Niels Lohmann <mail@nlohmann.me>
2026-10-31 13:00:39 +00:00 · 2026-10-31 09:45:14 +01:00
12 changed files with 539 additions and 532 deletions
@@ -13,15 +13,9 @@ function never throws an exception for invalid input, and the returned `#!cpp bo
 is returned.

 The result is the same as that of `BasicJsonType::accept()` by design; the difference is how the check is made. This
-function runs the document's builder, which checks the text as [`parse()`](parse.md) does, while
-`BasicJsonType::accept()` runs the SAX parser with a handler that ignores every event. Use this one if you go on to
-parse the same input with [`parse()`](parse.md) anyway, so that both use the same code; either is fine for a plain
-yes/no answer.
-
-When both `ignore_comments` and `ignore_trailing_commas` are `#!cpp false` (the default), the builder writes no index.
-With either option set, it builds the index as [`parse()`](parse.md) does and discards it: the result is the same, only
-the time differs. (Neither option needs the index; the variant that skips it exists only for the default options to
-keep the code size down.)
+function runs the document's index builder (and discards the index), while `BasicJsonType::accept()` runs the SAX
+parser with a handler that ignores every event. Use this one if you go on to parse the same input with
+[`parse()`](parse.md) anyway, so that both use the same code; either is fine for a plain yes/no answer.

 ## Template parameters

@@ -4293,6 +4293,19 @@ class binary_reader
        return get_bytes(len, "string", result);
    }

+    /// whether @a s is valid UTF-8: ASCII eight bytes at a time, the rest
+    /// sequence by sequence (as the binary writers check strings)
+    static bool is_valid_utf8_string(const string_t& s, std::true_type /*is_byte*/) noexcept
+    {
+        return valid_utf8_string(reinterpret_cast<const unsigned char*>(s.data()), s.size());
+    }
+
+    /// whether @a s, of a character type wider than a byte, is valid UTF-8
+    static bool is_valid_utf8_string(const string_t& s, std::false_type /*is_byte*/) noexcept
+    {
+        return is_valid_utf8(s);
+    }
+
    /*!
    @brief validate a decoded text string (value or object key) against @ref error_handler

@@ -4312,7 +4325,16 @@ class binary_reader
    */
    bool check_string_utf8(string_t& result, const char* context)
    {
-        if (error_handler == error_handler_t::keep || is_valid_utf8(result))
+        return error_handler == error_handler_t::keep || check_string_utf8_handler(result, context);
+    }
+
+    /// @ref check_string_utf8 for an error handler other than @ref
+    /// error_handler_t::keep; out of line, so that reading strings without a
+    /// check does not pay for the code of the check
+    JSON_HEDLEY_NEVER_INLINE
+    bool check_string_utf8_handler(string_t& result, const char* context)
+    {
+        if (is_valid_utf8_string(result, std::integral_constant<bool, sizeof(typename string_t::value_type) == 1>()))
        {
            return true;
        }
@@ -322,6 +322,78 @@ JSON_SCAN_SSSE3_TARGET JSON_HEDLEY_NEVER_INLINE inline bool vector_utf8_run(cons
    block_out = block;
    return false;
 }
+
+/*!
+@brief skip 16-byte blocks of well-formed UTF-8 at p (a character boundary)
+
+Like vector_utf8_run, but quotes, backslashes, and control characters are
+ordinary characters. Returns the start of the first block in which the check
+found an error (which may belong to a sequence begun in the block before) or
+where fewer than 16 bytes are left; a sequence may be cut at that point, so the
+caller resumes at its start, one sequence at a time.
+*/
+JSON_SCAN_SSSE3_TARGET JSON_HEDLEY_NEVER_INLINE inline const unsigned char* vector_utf8_valid_blocks(const unsigned char* block, const unsigned char* e) noexcept
+{
+    using lookup = utf8_lookup4<>;
+#if JSON_SCAN_NEON
+    const uint8x16_t t1h = vld1q_u8(lookup::byte_1_high.data());
+    const uint8x16_t t1l = vld1q_u8(lookup::byte_1_low.data());
+    const uint8x16_t t2h = vld1q_u8(lookup::byte_2_high.data());
+    uint8x16_t prev = vdupq_n_u8(0);
+    while (e - block >= 16)
+    {
+        const uint8x16_t in = vld1q_u8(block);
+        if (vmaxvq_u8(vorrq_u8(in, prev)) < 0x80)
+        {
+            // ASCII here, and no sequence from the block before to finish
+            prev = in;
+            block += 16;
+            continue;
+        }
+        const uint8x16_t prev1 = vextq_u8(prev, in, 15);
+        const uint8x16_t sc = vandq_u8(vandq_u8(vqtbl1q_u8(t1h, vshrq_n_u8(prev1, 4)), vqtbl1q_u8(t1l, vandq_u8(prev1, vdupq_n_u8(0x0F)))), vqtbl1q_u8(t2h, vshrq_n_u8(in, 4)));
+        const uint8x16_t must23 = vorrq_u8(vqsubq_u8(vextq_u8(prev, in, 14), vdupq_n_u8(0xE0 - 0x80)), vqsubq_u8(vextq_u8(prev, in, 13), vdupq_n_u8(0xF0 - 0x80)));
+        const uint8x16_t err = veorq_u8(vandq_u8(must23, vdupq_n_u8(0x80)), sc);
+        if (vmaxvq_u8(err) != 0)
+        {
+            break;
+        }
+        prev = in;
+        block += 16;
+    }
+#else
+    const __m128i t1h = _mm_loadu_si128(static_cast<const __m128i*>(static_cast<const void*>(lookup::byte_1_high.data()))); // NOLINT(bugprone-casting-through-void)
+    const __m128i t1l = _mm_loadu_si128(static_cast<const __m128i*>(static_cast<const void*>(lookup::byte_1_low.data()))); // NOLINT(bugprone-casting-through-void)
+    const __m128i t2h = _mm_loadu_si128(static_cast<const __m128i*>(static_cast<const void*>(lookup::byte_2_high.data()))); // NOLINT(bugprone-casting-through-void)
+    const __m128i nibble = _mm_set1_epi8(0x0F);
+    const __m128i zero = _mm_setzero_si128();
+    __m128i prev = zero;
+    while (e - block >= 16)
+    {
+        const __m128i in = _mm_loadu_si128(static_cast<const __m128i*>(static_cast<const void*>(block))); // NOLINT(bugprone-casting-through-void)
+        if (_mm_movemask_epi8(_mm_or_si128(in, prev)) == 0)
+        {
+            prev = in;
+            block += 16;
+            continue;
+        }
+        const __m128i prev1 = _mm_alignr_epi8(in, prev, 15);
+        const __m128i sc = _mm_and_si128(_mm_and_si128(_mm_shuffle_epi8(t1h, _mm_and_si128(_mm_srli_epi16(prev1, 4), nibble)),
+                                         _mm_shuffle_epi8(t1l, _mm_and_si128(prev1, nibble))),
+                                         _mm_shuffle_epi8(t2h, _mm_and_si128(_mm_srli_epi16(in, 4), nibble)));
+        const __m128i must23 = _mm_or_si128(_mm_subs_epu8(_mm_alignr_epi8(in, prev, 14), _mm_set1_epi8(0xE0 - 0x80)),
+                                            _mm_subs_epu8(_mm_alignr_epi8(in, prev, 13), _mm_set1_epi8(0xF0 - 0x80)));
+        const __m128i err = _mm_xor_si128(_mm_and_si128(must23, _mm_set1_epi8(static_cast<char>(-128))), sc);
+        if (_mm_movemask_epi8(_mm_cmpeq_epi8(err, zero)) != 0xFFFF)
+        {
+            break;
+        }
+        prev = in;
+        block += 16;
+    }
+#endif
+    return block;
+}
 #endif

 }  // namespace detail
@@ -236,6 +236,64 @@ inline std::size_t valid_utf8_prefix(const unsigned char* data, std::size_t n) n
    return pos;
 }

+/*!
+@brief whether [data, data+n) is valid UTF-8, for checking whole strings
+
+The same answer as valid_utf8_prefix(data, n) == n. Short ASCII strings, the
+most common keys and values, are decided in one test, and long strings are
+checked 16 bytes at a time where the vector check is available.
+*/
+inline bool valid_utf8_string(const unsigned char* data, std::size_t n) noexcept
+{
+    constexpr std::uint64_t high = 0x8080808080808080ull;
+    if (n <= 16)
+    {
+        // two overlapping words cover 8 to 16 bytes, one partial word fewer
+        std::uint64_t first = 0;
+        std::uint64_t last = 0;
+        if (n >= 8)
+        {
+            std::memcpy(&first, data, sizeof(first));
+            std::memcpy(&last, data + n - 8, sizeof(last));
+        }
+        else if (n != 0)
+        {
+            std::memcpy(&first, data, n);
+        }
+        if (((first | last) & high) == 0)
+        {
+            return true;
+        }
+    }
+#if JSON_SCAN_VECTOR_UTF8
+    else if (n >= 32
+#if JSON_SCAN_SSSE3_DISPATCH
+             && cpu_has_ssse3()
+#endif
+            )
+    {
+        const unsigned char* block = vector_utf8_valid_blocks(data, data + n);
+        // the rest from the start of a sequence that may be cut at block
+        for (int i = 1; i <= 3 && block - i >= data; ++i)
+        {
+            const unsigned char c = block[-i];
+            if (c < 0x80)
+            {
+                break;
+            }
+            if (c >= 0xC0)
+            {
+                block -= i;
+                break;
+            }
+        }
+        const auto checked = static_cast<std::size_t>(block - data);
+        return checked + valid_utf8_prefix(block, n - checked) == n;
+    }
+#endif
+    return valid_utf8_prefix(data, n) == n;
+}
+
 #if JSON_SCAN_VECTOR_UTF8
 /// the end of scan_string_vector from block, where the vector loop stopped
 /// (ill-formed UTF-8, or fewer than 16 bytes left): one byte or sequence at a
@@ -3080,6 +3080,25 @@ class binary_writer
        }
    }

+    /// write the head and the characters of @a value, at most
+    /// max_text_in_reservation of them, in one reservation
+    JSON_HEDLEY_ALWAYS_INLINE void write_cbor_string_reserved(const string_t& value)
+    {
+        CharType* const out = oa.reserve(value.size() + 2);
+        CharType* p = out;
+        if (value.size() <= 0x17)
+        {
+            *p++ = to_char_type(static_cast<std::uint8_t>(0x60 + value.size()));
+        }
+        else
+        {
+            *p++ = to_char_type(0x78);
+            *p++ = to_char_type(static_cast<std::uint8_t>(value.size()));
+        }
+        copy_characters(p, reinterpret_cast<const CharType*>(value.data()), value.size());
+        oa.commit(static_cast<std::size_t>(p - out) + value.size());
+    }
+
    /*!
    @brief write a CBOR text string

@@ -3101,19 +3120,7 @@ class binary_writer
            // the string is written as it is: head and characters in one reservation
            if (JSON_HEDLEY_LIKELY(value.size() <= max_text_in_reservation))
            {
-                CharType* const out = oa.reserve(value.size() + 2);
-                CharType* p = out;
-                if (value.size() <= 0x17)
-                {
-                    *p++ = to_char_type(static_cast<std::uint8_t>(0x60 + value.size()));
-                }
-                else
-                {
-                    *p++ = to_char_type(0x78);
-                    *p++ = to_char_type(static_cast<std::uint8_t>(value.size()));
-                }
-                copy_characters(p, reinterpret_cast<const CharType*>(value.data()), value.size());
-                oa.commit(static_cast<std::size_t>(p - out) + value.size());
+                write_cbor_string_reserved(value);
                return;
            }
        }
@@ -3127,6 +3134,13 @@ class binary_writer
    JSON_HEDLEY_NEVER_INLINE
    void write_cbor_checked_string(const string_t& value, const BasicJsonType& context)
    {
+        if (error_handler == error_handler_t::strict && value.size() <= max_text_in_reservation)
+        {
+            check_utf8(value, context);
+            write_cbor_string_reserved(value);
+            return;
+        }
+
        string_t storage;
        const string_t& sanitized = sanitize_utf8_for_write(value, context, storage);

@@ -3141,6 +3155,27 @@ class binary_writer
    // MsgPack //
    /////////////

+    /// write the head and the characters of @a value, at most
+    /// max_text_in_reservation of them, in one reservation
+    JSON_HEDLEY_ALWAYS_INLINE void write_msgpack_string_reserved(const string_t& value)
+    {
+        CharType* const out = oa.reserve(value.size() + 2);
+        CharType* p = out;
+        if (value.size() <= 31)
+        {
+            // fixstr
+            *p++ = to_char_type(static_cast<std::uint8_t>(0xA0 | value.size()));
+        }
+        else
+        {
+            // str 8
+            *p++ = to_char_type(0xD9);
+            *p++ = to_char_type(static_cast<std::uint8_t>(value.size()));
+        }
+        copy_characters(p, reinterpret_cast<const CharType*>(value.data()), value.size());
+        oa.commit(static_cast<std::size_t>(p - out) + value.size());
+    }
+
    /*!
    @brief write a MessagePack str

@@ -3158,21 +3193,7 @@ class binary_writer
            // the string is written as it is: head and characters in one reservation
            if (JSON_HEDLEY_LIKELY(value.size() <= max_text_in_reservation))
            {
-                CharType* const out = oa.reserve(value.size() + 2);
-                CharType* p = out;
-                if (value.size() <= 31)
-                {
-                    // fixstr
-                    *p++ = to_char_type(static_cast<std::uint8_t>(0xA0 | value.size()));
-                }
-                else
-                {
-                    // str 8
-                    *p++ = to_char_type(0xD9);
-                    *p++ = to_char_type(static_cast<std::uint8_t>(value.size()));
-                }
-                copy_characters(p, reinterpret_cast<const CharType*>(value.data()), value.size());
-                oa.commit(static_cast<std::size_t>(p - out) + value.size());
+                write_msgpack_string_reserved(value);
                return;
            }
        }
@@ -3185,6 +3206,13 @@ class binary_writer
    JSON_HEDLEY_NEVER_INLINE
    void write_msgpack_checked_string(const string_t& value, const BasicJsonType& context)
    {
+        if (error_handler == error_handler_t::strict && value.size() <= max_text_in_reservation)
+        {
+            check_utf8(value, context);
+            write_msgpack_string_reserved(value);
+            return;
+        }
+
        string_t storage;
        const string_t& sanitized = sanitize_utf8_for_write(value, context, storage);

@@ -4794,10 +4822,9 @@ class binary_writer
    static void check_utf8(const string_t& s, const BasicJsonType& context)
    {
        const auto* data = reinterpret_cast<const unsigned char*>(s.data());
-        const std::size_t valid = valid_utf8_prefix(data, s.size());
-        if (JSON_HEDLEY_UNLIKELY(valid != s.size()))
+        if (JSON_HEDLEY_UNLIKELY(!valid_utf8_string(data, s.size())))
        {
-            throw_invalid_utf8(data, valid, context);
+            throw_invalid_utf8(data, valid_utf8_prefix(data, s.size()), context);
        }
    }

@@ -69,8 +69,7 @@ struct parse_failure
 };

 /// FloatType: the number_float_t of the document, whose overflow parse() rejects
-/// Store: false for accept(): the grammar, number, string and UTF-8 checks run as for parse(), but no node is written
-template<typename FloatType, bool Comments, bool TrailingCommas, bool NulIsEnd, bool Sentinel, bool Store>
+template<typename FloatType, bool Comments, bool TrailingCommas, bool NulIsEnd, bool Sentinel>
 class builder
 {
  public:
@@ -246,87 +245,11 @@ class builder
        return p;
    }

-    /// the string from p, the first backslash, control character, ill-formed UTF-8 byte, or the end, checked and not
-    /// decoded (accept() stores nothing, so the arena is not written): the checks, errors and error positions of
-    /// slow_string(), for everything up to the closing quote
-    NLOHMANN_VIEW_NOINLINE decoded check_string(const unsigned char* p)
-    {
-        const unsigned char* const end = m_end;
-        for (;;)
-        {
-            if (p == end)
-            {
-                return failed(error_code::string_missing_quote, p);
-            }
-            const unsigned char c = *p;
-            if (c == '"')
-            {
-                ++p;
-                return decoded{p, 0, 0};
-            }
-            if (c != '\\')
-            {
-                return failed(c < 0x20 ? error_code::string_control_character : error_code::string_utf8, p);
-            }
-            ++p;
-            if (p == end)
-            {
-                return failed(error_code::string_missing_quote, p);
-            }
-            const unsigned char d = *p++;
-            if (d == 'u')
-            {
-                const int cp = hex4(p); // (moves p past the four digits)
-                if (NLOHMANN_VIEW_UNLIKELY(cp < 0))
-                {
-                    return failed(error_code::string_unicode_hex, p);
-                }
-                if (NLOHMANN_VIEW_UNLIKELY((cp & 0xF800) == 0xD800)) // a surrogate
-                {
-                    if (cp >= 0xDC00)
-                    {
-                        return failed(error_code::string_surrogate_low, p);
-                    }
-                    if (end - p < 2 || p[0] != '\\' || p[1] != 'u')
-                    {
-                        return failed(error_code::string_surrogate_high, p);
-                    }
-                    p += 2;
-                    const int lo = hex4(p); // (moves p past the four digits)
-                    if (lo < 0)
-                    {
-                        return failed(error_code::string_unicode_hex, p);
-                    }
-                    if (lo < 0xDC00 || lo > 0xDFFF)
-                    {
-                        return failed(error_code::string_surrogate_high, p);
-                    }
-                }
-            }
-            else if (!(d == '"' || d == '\\' || d == '/' || d == 'b' || d == 'f' || d == 'n' || d == 'r' || d == 't'))
-            {
-                --p;
-                return failed(error_code::string_escape, p);
-            }
-            if (p != end && *p == '\\')
-            {
-                continue; // (escapes in a row: no run of text between them)
-            }
-            p = scan_string_run(p, end); // the text up to the next backslash, quote, control character, or ill-formed UTF-8
-        }
-    }
-
    /// s points to the start of the string, p to a backslash, a control
    /// character, an ill-formed UTF-8 byte, or the end: decode the string into
    /// the arena
    NLOHMANN_VIEW_NOINLINE decoded slow_string(const unsigned char* s, const unsigned char* p)
    {
-        if (!enabled(Store))
-        {
-            // (accept(): the string is checked, not decoded)
-            static_cast<void>(s);
-            return check_string(p);
-        }
        // single-character escapes; 0: invalid (and 'u', handled separately)
        static const std::array<char, 128> simple_escape =
        {
@@ -687,13 +610,10 @@ run_found:

        NLOHMANN_VIEW_ALWAYS_INLINE bool run()
        {
-            if (enabled(Store))
-            {
-                cold.m_doc.reserve(estimate_nodes(reinterpret_cast<const char*>(b), static_cast<std::size_t>(e - b)));
-                base = cold.m_doc.nodes;
-                out = base;
-                cap = base + cold.m_doc.node_capacity;
-            }
+            cold.m_doc.reserve(estimate_nodes(reinterpret_cast<const char*>(b), static_cast<std::size_t>(e - b)));
+            base = cold.m_doc.nodes;
+            out = base;
+            cap = base + cold.m_doc.node_capacity;

            if (e - p >= 3 && p[0] == 0xEF && p[1] == 0xBB && p[2] == 0xBF)
            {
@@ -877,10 +797,7 @@ close_object:
            // an array pays nothing for this)
            if (NLOHMANN_VIEW_UNLIKELY(cur_count >= document_data::index_min_members))
            {
-                if (enabled(Store))
-                {
-                    cold.note_large_object(cur_idx);
-                }
+                cold.note_large_object(cur_idx);
            }

 close_container:
@@ -1040,12 +957,6 @@ indent_done:
        /// an integer's value); two stores on little-endian targets
        NLOHMANN_VIEW_ALWAYS_INLINE node* push_node(node_kind k, std::uint8_t flags, std::uint16_t extra, std::size_t off, std::uint64_t second)
        {
-            if (!enabled(Store))
-            {
-                // accept(): nothing is stored, so what computes the arguments is not run either. base and out stay
-                // null: their difference, which open() takes, is 0.
-                return out;
-            }
            if (NLOHMANN_VIEW_UNLIKELY(out == cap))
            {
                const auto n = static_cast<std::size_t>(out - base);
@@ -1104,12 +1015,9 @@ indent_done:

        NLOHMANN_VIEW_ALWAYS_INLINE void close()
        {
-            if (enabled(Store))
-            {
-                node& n = base[cur_idx];
-                n.len = cur_count;
-                n.next = static_cast<std::uint32_t>(out - base) - cur_idx;
-            }
+            node& n = base[cur_idx];
+            n.len = cur_count;
+            n.next = static_cast<std::uint32_t>(out - base) - cur_idx;
            if (--depth != 0)
            {
                if (NLOHMANN_VIEW_LIKELY(depth <= 64))
@@ -1284,44 +1192,34 @@ indent_done:
 };

 /// run the builder with compile-time options
-template<typename FloatType, bool Comments, bool TrailingCommas, bool NulIsEnd, bool Store>
+template<typename FloatType, bool Comments, bool TrailingCommas, bool NulIsEnd>
 inline bool build_with(document_data& d, const char* src, std::size_t size, bool sentinel, parse_failure& failure)
 {
    if (sentinel)
    {
-        builder<FloatType, Comments, TrailingCommas, NulIsEnd, true, Store> bld(d, src, size);
+        builder<FloatType, Comments, TrailingCommas, NulIsEnd, true> bld(d, src, size);
        const bool ok = bld.run();
        failure = bld.failure();
        return ok;
    }
-    builder<FloatType, Comments, TrailingCommas, NulIsEnd, false, Store> bld(d, src, size);
+    builder<FloatType, Comments, TrailingCommas, NulIsEnd, false> bld(d, src, size);
    const bool ok = bld.run();
    failure = bld.failure();
    return ok;
 }

 /// sentinel: src[size] is readable and 0 (e.g. std::string); FloatType: the
-/// number_float_t of the document; validate_only: accept(), only the result is
-/// used
+/// number_float_t of the document
 template<typename FloatType, bool NulIsEnd>
-inline bool build(document_data& d, const char* src, std::size_t size, bool comments, bool trailing_commas, bool sentinel, bool validate_only,
-                  parse_failure& failure)
+inline bool build(document_data& d, const char* src, std::size_t size, bool comments, bool trailing_commas, bool sentinel, parse_failure& failure)
 {
-    if (validate_only && !comments && !trailing_commas)
-    {
-        // accept() of plain JSON: nothing is stored. The other option sets
-        // could skip the stores just as well; they keep the builder that
-        // stores only so that there are two more instantiations of the parse
-        // loop, not eight.
-        return build_with<FloatType, false, false, NulIsEnd, false>(d, src, size, sentinel, failure);
-    }
    if (comments)
    {
-        return trailing_commas ? build_with<FloatType, true, true, NulIsEnd, true>(d, src, size, sentinel, failure)
-               : build_with<FloatType, true, false, NulIsEnd, true>(d, src, size, sentinel, failure);
+        return trailing_commas ? build_with<FloatType, true, true, NulIsEnd>(d, src, size, sentinel, failure)
+               : build_with<FloatType, true, false, NulIsEnd>(d, src, size, sentinel, failure);
    }
-    return trailing_commas ? build_with<FloatType, false, true, NulIsEnd, true>(d, src, size, sentinel, failure)
-           : build_with<FloatType, false, false, NulIsEnd, true>(d, src, size, sentinel, failure);
+    return trailing_commas ? build_with<FloatType, false, true, NulIsEnd>(d, src, size, sentinel, failure)
+           : build_with<FloatType, false, false, NulIsEnd>(d, src, size, sentinel, failure);
 }

 }  // namespace view
@@ -1060,7 +1060,7 @@ class basic_json_document
                                     const bool ignore_trailing_commas = false)
    {
        basic_json_document d;
-        d.read_range(first, last, error_mode(allow_exceptions), ignore_comments, ignore_trailing_commas);
+        d.read_range(first, last, allow_exceptions, ignore_comments, ignore_trailing_commas);
        return d;
    }

@@ -1075,7 +1075,7 @@ class basic_json_document
        basic_json_document d;
        std::string buf;
        append_input(buf, std::forward<InputType>(input));
-        d.build_owned(std::move(buf), error_mode(allow_exceptions), ignore_comments, ignore_trailing_commas);
+        d.build_owned(std::move(buf), allow_exceptions, ignore_comments, ignore_trailing_commas);
        return d;
    }

@@ -1088,8 +1088,7 @@ class basic_json_document
    static bool accept(InputType&& input, const bool ignore_comments = false, const bool ignore_trailing_commas = false)
    {
        basic_json_document d;
-        d.read_kind(std::forward<InputType>(input), on_error::validate, ignore_comments, ignore_trailing_commas,
-                    std::integral_constant<detail::view::input_kind, detail::view::classify_input<InputType>::value> {});
+        d.read(std::forward<InputType>(input), false, ignore_comments, ignore_trailing_commas);
        return !d.is_discarded();
    }

@@ -1105,7 +1104,7 @@ class basic_json_document
              const bool ignore_comments = false,
              const bool ignore_trailing_commas = false)
    {
-        read_kind(std::forward<InputType>(input), error_mode(allow_exceptions), ignore_comments, ignore_trailing_commas,
+        read_kind(std::forward<InputType>(input), allow_exceptions, ignore_comments, ignore_trailing_commas,
                  std::integral_constant<detail::view::input_kind, detail::view::classify_input<InputType>::value> {});
    }

@@ -1464,22 +1463,6 @@ class basic_json_document
        return idx;
    }

-    /// what a failed parse does
-    enum class on_error : std::uint8_t
-    {
-        discard, ///< the document is discarded
-        raise,   ///< parse_error is thrown
-        validate ///< accept(): the document is discarded, and no node is stored while parsing
-    };
-
-    static on_error error_mode(bool allow_exceptions) noexcept
-    {
-        return allow_exceptions ? on_error::raise : on_error::discard;
-    }
-
-    /// NulIsEnd of the builder: a null byte ends the input unless strict_nul_handling
-    static constexpr bool nul_is_end = !detail::abi_config::strict_nul_handling;
-
    /// create the storage (sized for the input) on first use
    void ensure_data(const char* src, std::size_t size)
    {
@@ -1491,7 +1474,7 @@ class basic_json_document

    /// parse a buffer the document takes ownership of (it replaces the
    /// previous owned copy)
-    void build_owned(std::string&& buf, on_error allow_exceptions, bool comments, bool trailing_commas)
+    void build_owned(std::string&& buf, bool allow_exceptions, bool comments, bool trailing_commas)
    {
        ensure_data(buf.data(), buf.size());
        m_data->owned = std::move(buf);
@@ -1501,7 +1484,7 @@ class basic_json_document
    /// parse a copy of the input; the copy is appended to the cleared owned
    /// buffer, so that its capacity is reused
    template<typename T>
-    void build_copy(T&& input, on_error allow_exceptions, bool comments, bool trailing_commas)
+    void build_copy(T&& input, bool allow_exceptions, bool comments, bool trailing_commas)
    {
        if (!m_data)
        {
@@ -1522,7 +1505,7 @@ class basic_json_document
    }

    /// sentinel: src[size] is readable and 0 (std::string, C strings)
-    void build(const char* src, std::size_t size, on_error allow_exceptions, bool comments, bool trailing_commas, bool owned, bool sentinel)
+    void build(const char* src, std::size_t size, bool allow_exceptions, bool comments, bool trailing_commas, bool owned, bool sentinel)
    {
        ensure_data(src, size);
        document_data& d = *m_data;
@@ -1549,8 +1532,7 @@ class basic_json_document
        }
        else
        {
-            ok = detail::view::build<typename BasicJsonType::number_float_t, nul_is_end>(d, src, size, comments, trailing_commas, sentinel,
-                 allow_exceptions == on_error::validate, failure);
+            ok = detail::view::build < typename BasicJsonType::number_float_t, !detail::abi_config::strict_nul_handling > (d, src, size, comments, trailing_commas, sentinel, failure);
        }
        if (NLOHMANN_VIEW_LIKELY(ok))
        {
@@ -1563,7 +1545,7 @@ class basic_json_document
            return;
        }
        std::vector<std::uint32_t>().swap(d.large_objects);
-        if (allow_exceptions == on_error::raise)
+        if (allow_exceptions)
        {
            detail::view::throw_parse_failure<BasicJsonType>(failure, src, size, comments, trailing_commas);
        }
@@ -1571,13 +1553,13 @@ class basic_json_document

    // --- input dispatch (see detail::view::input_kind) ---

-    void read_kind(std::string&& s, on_error ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::move_string> /*unused*/)
+    void read_kind(std::string&& s, bool ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::move_string> /*unused*/)
    {
        build_owned(std::move(s), ae, c, tc);
    }

    template<typename CharT>
-    void read_kind(CharT* s, on_error ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::c_string> /*unused*/)
+    void read_kind(CharT* s, bool ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::c_string> /*unused*/)
    {
        if (s == nullptr)
        {
@@ -1591,7 +1573,7 @@ class basic_json_document
    }

    template<typename Array>
-    void read_kind(Array& a, on_error ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::char_array> /*unused*/)
+    void read_kind(Array& a, bool ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::char_array> /*unused*/)
    {
        using CharT = typename std::remove_cv<typename std::remove_extent<Array>::type>::type;
        const std::size_t n = std::extent<Array>::value;
@@ -1602,7 +1584,7 @@ class basic_json_document
    }

    template<typename T>
-    void read_kind(const T& s, on_error ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::borrow_range> /*unused*/)
+    void read_kind(const T& s, bool ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::borrow_range> /*unused*/)
    {
        // std::basic_string guarantees data()[size()] == 0: use it as sentinel
        const auto size = static_cast<std::size_t>(s.size());
@@ -1611,13 +1593,13 @@ class basic_json_document
    }

    template<typename T>
-    void read_kind(const T& s, on_error ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::copy_range> /*unused*/)
+    void read_kind(const T& s, bool ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::copy_range> /*unused*/)
    {
        build_copy(s, ae, c, tc);
    }

    template<typename CharT>
-    void read_kind(CharT* s, on_error ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::wide_c_string> /*unused*/)
+    void read_kind(CharT* s, bool ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::wide_c_string> /*unused*/)
    {
        if (s == nullptr)
        {
@@ -1629,13 +1611,13 @@ class basic_json_document
    }

    template<typename T>
-    void read_kind(T&& input, on_error ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::adapter> /*unused*/)
+    void read_kind(T&& input, bool ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::adapter> /*unused*/)
    {
        build_copy(std::forward<T>(input), ae, c, tc);
    }

    template<typename IteratorType>
-    void read_range(IteratorType first, IteratorType last, on_error ae, bool c, bool tc)
+    void read_range(IteratorType first, IteratorType last, bool ae, bool c, bool tc)
    {
        using value_type = typename std::remove_cv<typename std::iterator_traits<IteratorType>::value_type>::type;
        // borrow the range where the library's input adapter scans it as one
@@ -1646,14 +1628,14 @@ class basic_json_document
    }

    template<typename IteratorType>
-    void read_range_impl(IteratorType first, IteratorType last, on_error ae, bool c, bool tc, std::true_type /*contiguous bytes*/)
+    void read_range_impl(IteratorType first, IteratorType last, bool ae, bool c, bool tc, std::true_type /*contiguous bytes*/)
    {
        const auto size = static_cast<std::size_t>(std::distance(first, last));
        build(size == 0 ? "" : reinterpret_cast<const char*>(&*first), size, ae, c, tc, false, size == 0); // NOLINT(cppcoreguidelines-pro-type-reinterpret-cast)
    }

    template<typename IteratorType>
-    void read_range_impl(IteratorType first, IteratorType last, on_error ae, bool c, bool tc, std::false_type /*other*/)
+    void read_range_impl(IteratorType first, IteratorType last, bool ae, bool c, bool tc, std::false_type /*other*/)
    {
        build_copy(detail::view::make_adapter_input(detail::input_adapter(first, last)), ae, c, tc);
    }
@@ -12377,6 +12377,78 @@ JSON_SCAN_SSSE3_TARGET JSON_HEDLEY_NEVER_INLINE inline bool vector_utf8_run(cons
    block_out = block;
    return false;
 }
+
+/*!
+@brief skip 16-byte blocks of well-formed UTF-8 at p (a character boundary)
+
+Like vector_utf8_run, but quotes, backslashes, and control characters are
+ordinary characters. Returns the start of the first block in which the check
+found an error (which may belong to a sequence begun in the block before) or
+where fewer than 16 bytes are left; a sequence may be cut at that point, so the
+caller resumes at its start, one sequence at a time.
+*/
+JSON_SCAN_SSSE3_TARGET JSON_HEDLEY_NEVER_INLINE inline const unsigned char* vector_utf8_valid_blocks(const unsigned char* block, const unsigned char* e) noexcept
+{
+    using lookup = utf8_lookup4<>;
+#if JSON_SCAN_NEON
+    const uint8x16_t t1h = vld1q_u8(lookup::byte_1_high.data());
+    const uint8x16_t t1l = vld1q_u8(lookup::byte_1_low.data());
+    const uint8x16_t t2h = vld1q_u8(lookup::byte_2_high.data());
+    uint8x16_t prev = vdupq_n_u8(0);
+    while (e - block >= 16)
+    {
+        const uint8x16_t in = vld1q_u8(block);
+        if (vmaxvq_u8(vorrq_u8(in, prev)) < 0x80)
+        {
+            // ASCII here, and no sequence from the block before to finish
+            prev = in;
+            block += 16;
+            continue;
+        }
+        const uint8x16_t prev1 = vextq_u8(prev, in, 15);
+        const uint8x16_t sc = vandq_u8(vandq_u8(vqtbl1q_u8(t1h, vshrq_n_u8(prev1, 4)), vqtbl1q_u8(t1l, vandq_u8(prev1, vdupq_n_u8(0x0F)))), vqtbl1q_u8(t2h, vshrq_n_u8(in, 4)));
+        const uint8x16_t must23 = vorrq_u8(vqsubq_u8(vextq_u8(prev, in, 14), vdupq_n_u8(0xE0 - 0x80)), vqsubq_u8(vextq_u8(prev, in, 13), vdupq_n_u8(0xF0 - 0x80)));
+        const uint8x16_t err = veorq_u8(vandq_u8(must23, vdupq_n_u8(0x80)), sc);
+        if (vmaxvq_u8(err) != 0)
+        {
+            break;
+        }
+        prev = in;
+        block += 16;
+    }
+#else
+    const __m128i t1h = _mm_loadu_si128(static_cast<const __m128i*>(static_cast<const void*>(lookup::byte_1_high.data()))); // NOLINT(bugprone-casting-through-void)
+    const __m128i t1l = _mm_loadu_si128(static_cast<const __m128i*>(static_cast<const void*>(lookup::byte_1_low.data()))); // NOLINT(bugprone-casting-through-void)
+    const __m128i t2h = _mm_loadu_si128(static_cast<const __m128i*>(static_cast<const void*>(lookup::byte_2_high.data()))); // NOLINT(bugprone-casting-through-void)
+    const __m128i nibble = _mm_set1_epi8(0x0F);
+    const __m128i zero = _mm_setzero_si128();
+    __m128i prev = zero;
+    while (e - block >= 16)
+    {
+        const __m128i in = _mm_loadu_si128(static_cast<const __m128i*>(static_cast<const void*>(block))); // NOLINT(bugprone-casting-through-void)
+        if (_mm_movemask_epi8(_mm_or_si128(in, prev)) == 0)
+        {
+            prev = in;
+            block += 16;
+            continue;
+        }
+        const __m128i prev1 = _mm_alignr_epi8(in, prev, 15);
+        const __m128i sc = _mm_and_si128(_mm_and_si128(_mm_shuffle_epi8(t1h, _mm_and_si128(_mm_srli_epi16(prev1, 4), nibble)),
+                                         _mm_shuffle_epi8(t1l, _mm_and_si128(prev1, nibble))),
+                                         _mm_shuffle_epi8(t2h, _mm_and_si128(_mm_srli_epi16(in, 4), nibble)));
+        const __m128i must23 = _mm_or_si128(_mm_subs_epu8(_mm_alignr_epi8(in, prev, 14), _mm_set1_epi8(0xE0 - 0x80)),
+                                            _mm_subs_epu8(_mm_alignr_epi8(in, prev, 13), _mm_set1_epi8(0xF0 - 0x80)));
+        const __m128i err = _mm_xor_si128(_mm_and_si128(must23, _mm_set1_epi8(static_cast<char>(-128))), sc);
+        if (_mm_movemask_epi8(_mm_cmpeq_epi8(err, zero)) != 0xFFFF)
+        {
+            break;
+        }
+        prev = in;
+        block += 16;
+    }
+#endif
+    return block;
+}
 #endif

 }  // namespace detail
@@ -12604,6 +12676,64 @@ inline std::size_t valid_utf8_prefix(const unsigned char* data, std::size_t n) n
    return pos;
 }

+/*!
+@brief whether [data, data+n) is valid UTF-8, for checking whole strings
+
+The same answer as valid_utf8_prefix(data, n) == n. Short ASCII strings, the
+most common keys and values, are decided in one test, and long strings are
+checked 16 bytes at a time where the vector check is available.
+*/
+inline bool valid_utf8_string(const unsigned char* data, std::size_t n) noexcept
+{
+    constexpr std::uint64_t high = 0x8080808080808080ull;
+    if (n <= 16)
+    {
+        // two overlapping words cover 8 to 16 bytes, one partial word fewer
+        std::uint64_t first = 0;
+        std::uint64_t last = 0;
+        if (n >= 8)
+        {
+            std::memcpy(&first, data, sizeof(first));
+            std::memcpy(&last, data + n - 8, sizeof(last));
+        }
+        else if (n != 0)
+        {
+            std::memcpy(&first, data, n);
+        }
+        if (((first | last) & high) == 0)
+        {
+            return true;
+        }
+    }
+#if JSON_SCAN_VECTOR_UTF8
+    else if (n >= 32
+#if JSON_SCAN_SSSE3_DISPATCH
+             && cpu_has_ssse3()
+#endif
+            )
+    {
+        const unsigned char* block = vector_utf8_valid_blocks(data, data + n);
+        // the rest from the start of a sequence that may be cut at block
+        for (int i = 1; i <= 3 && block - i >= data; ++i)
+        {
+            const unsigned char c = block[-i];
+            if (c < 0x80)
+            {
+                break;
+            }
+            if (c >= 0xC0)
+            {
+                block -= i;
+                break;
+            }
+        }
+        const auto checked = static_cast<std::size_t>(block - data);
+        return checked + valid_utf8_prefix(block, n - checked) == n;
+    }
+#endif
+    return valid_utf8_prefix(data, n) == n;
+}
+
 #if JSON_SCAN_VECTOR_UTF8
 /// the end of scan_string_vector from block, where the vector loop stopped
 /// (ill-formed UTF-8, or fewer than 16 bytes left): one byte or sequence at a
@@ -22069,6 +22199,19 @@ class binary_reader
        return get_bytes(len, "string", result);
    }

+    /// whether @a s is valid UTF-8: ASCII eight bytes at a time, the rest
+    /// sequence by sequence (as the binary writers check strings)
+    static bool is_valid_utf8_string(const string_t& s, std::true_type /*is_byte*/) noexcept
+    {
+        return valid_utf8_string(reinterpret_cast<const unsigned char*>(s.data()), s.size());
+    }
+
+    /// whether @a s, of a character type wider than a byte, is valid UTF-8
+    static bool is_valid_utf8_string(const string_t& s, std::false_type /*is_byte*/) noexcept
+    {
+        return is_valid_utf8(s);
+    }
+
    /*!
    @brief validate a decoded text string (value or object key) against @ref error_handler

@@ -22088,7 +22231,16 @@ class binary_reader
    */
    bool check_string_utf8(string_t& result, const char* context)
    {
-        if (error_handler == error_handler_t::keep || is_valid_utf8(result))
+        return error_handler == error_handler_t::keep || check_string_utf8_handler(result, context);
+    }
+
+    /// @ref check_string_utf8 for an error handler other than @ref
+    /// error_handler_t::keep; out of line, so that reading strings without a
+    /// check does not pay for the code of the check
+    JSON_HEDLEY_NEVER_INLINE
+    bool check_string_utf8_handler(string_t& result, const char* context)
+    {
+        if (is_valid_utf8_string(result, std::integral_constant<bool, sizeof(typename string_t::value_type) == 1>()))
        {
            return true;
        }
@@ -30306,6 +30458,25 @@ class binary_writer
        }
    }

+    /// write the head and the characters of @a value, at most
+    /// max_text_in_reservation of them, in one reservation
+    JSON_HEDLEY_ALWAYS_INLINE void write_cbor_string_reserved(const string_t& value)
+    {
+        CharType* const out = oa.reserve(value.size() + 2);
+        CharType* p = out;
+        if (value.size() <= 0x17)
+        {
+            *p++ = to_char_type(static_cast<std::uint8_t>(0x60 + value.size()));
+        }
+        else
+        {
+            *p++ = to_char_type(0x78);
+            *p++ = to_char_type(static_cast<std::uint8_t>(value.size()));
+        }
+        copy_characters(p, reinterpret_cast<const CharType*>(value.data()), value.size());
+        oa.commit(static_cast<std::size_t>(p - out) + value.size());
+    }
+
    /*!
    @brief write a CBOR text string

@@ -30327,19 +30498,7 @@ class binary_writer
            // the string is written as it is: head and characters in one reservation
            if (JSON_HEDLEY_LIKELY(value.size() <= max_text_in_reservation))
            {
-                CharType* const out = oa.reserve(value.size() + 2);
-                CharType* p = out;
-                if (value.size() <= 0x17)
-                {
-                    *p++ = to_char_type(static_cast<std::uint8_t>(0x60 + value.size()));
-                }
-                else
-                {
-                    *p++ = to_char_type(0x78);
-                    *p++ = to_char_type(static_cast<std::uint8_t>(value.size()));
-                }
-                copy_characters(p, reinterpret_cast<const CharType*>(value.data()), value.size());
-                oa.commit(static_cast<std::size_t>(p - out) + value.size());
+                write_cbor_string_reserved(value);
                return;
            }
        }
@@ -30353,6 +30512,13 @@ class binary_writer
    JSON_HEDLEY_NEVER_INLINE
    void write_cbor_checked_string(const string_t& value, const BasicJsonType& context)
    {
+        if (error_handler == error_handler_t::strict && value.size() <= max_text_in_reservation)
+        {
+            check_utf8(value, context);
+            write_cbor_string_reserved(value);
+            return;
+        }
+
        string_t storage;
        const string_t& sanitized = sanitize_utf8_for_write(value, context, storage);

@@ -30367,6 +30533,27 @@ class binary_writer
    // MsgPack //
    /////////////

+    /// write the head and the characters of @a value, at most
+    /// max_text_in_reservation of them, in one reservation
+    JSON_HEDLEY_ALWAYS_INLINE void write_msgpack_string_reserved(const string_t& value)
+    {
+        CharType* const out = oa.reserve(value.size() + 2);
+        CharType* p = out;
+        if (value.size() <= 31)
+        {
+            // fixstr
+            *p++ = to_char_type(static_cast<std::uint8_t>(0xA0 | value.size()));
+        }
+        else
+        {
+            // str 8
+            *p++ = to_char_type(0xD9);
+            *p++ = to_char_type(static_cast<std::uint8_t>(value.size()));
+        }
+        copy_characters(p, reinterpret_cast<const CharType*>(value.data()), value.size());
+        oa.commit(static_cast<std::size_t>(p - out) + value.size());
+    }
+
    /*!
    @brief write a MessagePack str

@@ -30384,21 +30571,7 @@ class binary_writer
            // the string is written as it is: head and characters in one reservation
            if (JSON_HEDLEY_LIKELY(value.size() <= max_text_in_reservation))
            {
-                CharType* const out = oa.reserve(value.size() + 2);
-                CharType* p = out;
-                if (value.size() <= 31)
-                {
-                    // fixstr
-                    *p++ = to_char_type(static_cast<std::uint8_t>(0xA0 | value.size()));
-                }
-                else
-                {
-                    // str 8
-                    *p++ = to_char_type(0xD9);
-                    *p++ = to_char_type(static_cast<std::uint8_t>(value.size()));
-                }
-                copy_characters(p, reinterpret_cast<const CharType*>(value.data()), value.size());
-                oa.commit(static_cast<std::size_t>(p - out) + value.size());
+                write_msgpack_string_reserved(value);
                return;
            }
        }
@@ -30411,6 +30584,13 @@ class binary_writer
    JSON_HEDLEY_NEVER_INLINE
    void write_msgpack_checked_string(const string_t& value, const BasicJsonType& context)
    {
+        if (error_handler == error_handler_t::strict && value.size() <= max_text_in_reservation)
+        {
+            check_utf8(value, context);
+            write_msgpack_string_reserved(value);
+            return;
+        }
+
        string_t storage;
        const string_t& sanitized = sanitize_utf8_for_write(value, context, storage);

@@ -32020,10 +32200,9 @@ class binary_writer
    static void check_utf8(const string_t& s, const BasicJsonType& context)
    {
        const auto* data = reinterpret_cast<const unsigned char*>(s.data());
-        const std::size_t valid = valid_utf8_prefix(data, s.size());
-        if (JSON_HEDLEY_UNLIKELY(valid != s.size()))
+        if (JSON_HEDLEY_UNLIKELY(!valid_utf8_string(data, s.size())))
        {
-            throw_invalid_utf8(data, valid, context);
+            throw_invalid_utf8(data, valid_utf8_prefix(data, s.size()), context);
        }
    }

@@ -1311,8 +1311,7 @@ struct parse_failure
 };

 /// FloatType: the number_float_t of the document, whose overflow parse() rejects
-/// Store: false for accept(): the grammar, number, string and UTF-8 checks run as for parse(), but no node is written
-template<typename FloatType, bool Comments, bool TrailingCommas, bool NulIsEnd, bool Sentinel, bool Store>
+template<typename FloatType, bool Comments, bool TrailingCommas, bool NulIsEnd, bool Sentinel>
 class builder
 {
  public:
@@ -1488,87 +1487,11 @@ class builder
        return p;
    }

-    /// the string from p, the first backslash, control character, ill-formed UTF-8 byte, or the end, checked and not
-    /// decoded (accept() stores nothing, so the arena is not written): the checks, errors and error positions of
-    /// slow_string(), for everything up to the closing quote
-    NLOHMANN_VIEW_NOINLINE decoded check_string(const unsigned char* p)
-    {
-        const unsigned char* const end = m_end;
-        for (;;)
-        {
-            if (p == end)
-            {
-                return failed(error_code::string_missing_quote, p);
-            }
-            const unsigned char c = *p;
-            if (c == '"')
-            {
-                ++p;
-                return decoded{p, 0, 0};
-            }
-            if (c != '\\')
-            {
-                return failed(c < 0x20 ? error_code::string_control_character : error_code::string_utf8, p);
-            }
-            ++p;
-            if (p == end)
-            {
-                return failed(error_code::string_missing_quote, p);
-            }
-            const unsigned char d = *p++;
-            if (d == 'u')
-            {
-                const int cp = hex4(p); // (moves p past the four digits)
-                if (NLOHMANN_VIEW_UNLIKELY(cp < 0))
-                {
-                    return failed(error_code::string_unicode_hex, p);
-                }
-                if (NLOHMANN_VIEW_UNLIKELY((cp & 0xF800) == 0xD800)) // a surrogate
-                {
-                    if (cp >= 0xDC00)
-                    {
-                        return failed(error_code::string_surrogate_low, p);
-                    }
-                    if (end - p < 2 || p[0] != '\\' || p[1] != 'u')
-                    {
-                        return failed(error_code::string_surrogate_high, p);
-                    }
-                    p += 2;
-                    const int lo = hex4(p); // (moves p past the four digits)
-                    if (lo < 0)
-                    {
-                        return failed(error_code::string_unicode_hex, p);
-                    }
-                    if (lo < 0xDC00 || lo > 0xDFFF)
-                    {
-                        return failed(error_code::string_surrogate_high, p);
-                    }
-                }
-            }
-            else if (!(d == '"' || d == '\\' || d == '/' || d == 'b' || d == 'f' || d == 'n' || d == 'r' || d == 't'))
-            {
-                --p;
-                return failed(error_code::string_escape, p);
-            }
-            if (p != end && *p == '\\')
-            {
-                continue; // (escapes in a row: no run of text between them)
-            }
-            p = scan_string_run(p, end); // the text up to the next backslash, quote, control character, or ill-formed UTF-8
-        }
-    }
-
    /// s points to the start of the string, p to a backslash, a control
    /// character, an ill-formed UTF-8 byte, or the end: decode the string into
    /// the arena
    NLOHMANN_VIEW_NOINLINE decoded slow_string(const unsigned char* s, const unsigned char* p)
    {
-        if (!enabled(Store))
-        {
-            // (accept(): the string is checked, not decoded)
-            static_cast<void>(s);
-            return check_string(p);
-        }
        // single-character escapes; 0: invalid (and 'u', handled separately)
        static const std::array<char, 128> simple_escape =
        {
@@ -1929,13 +1852,10 @@ run_found:

        NLOHMANN_VIEW_ALWAYS_INLINE bool run()
        {
-            if (enabled(Store))
-            {
-                cold.m_doc.reserve(estimate_nodes(reinterpret_cast<const char*>(b), static_cast<std::size_t>(e - b)));
-                base = cold.m_doc.nodes;
-                out = base;
-                cap = base + cold.m_doc.node_capacity;
-            }
+            cold.m_doc.reserve(estimate_nodes(reinterpret_cast<const char*>(b), static_cast<std::size_t>(e - b)));
+            base = cold.m_doc.nodes;
+            out = base;
+            cap = base + cold.m_doc.node_capacity;

            if (e - p >= 3 && p[0] == 0xEF && p[1] == 0xBB && p[2] == 0xBF)
            {
@@ -2119,10 +2039,7 @@ close_object:
            // an array pays nothing for this)
            if (NLOHMANN_VIEW_UNLIKELY(cur_count >= document_data::index_min_members))
            {
-                if (enabled(Store))
-                {
-                    cold.note_large_object(cur_idx);
-                }
+                cold.note_large_object(cur_idx);
            }

 close_container:
@@ -2282,12 +2199,6 @@ indent_done:
        /// an integer's value); two stores on little-endian targets
        NLOHMANN_VIEW_ALWAYS_INLINE node* push_node(node_kind k, std::uint8_t flags, std::uint16_t extra, std::size_t off, std::uint64_t second)
        {
-            if (!enabled(Store))
-            {
-                // accept(): nothing is stored, so what computes the arguments is not run either. base and out stay
-                // null: their difference, which open() takes, is 0.
-                return out;
-            }
            if (NLOHMANN_VIEW_UNLIKELY(out == cap))
            {
                const auto n = static_cast<std::size_t>(out - base);
@@ -2346,12 +2257,9 @@ indent_done:

        NLOHMANN_VIEW_ALWAYS_INLINE void close()
        {
-            if (enabled(Store))
-            {
-                node& n = base[cur_idx];
-                n.len = cur_count;
-                n.next = static_cast<std::uint32_t>(out - base) - cur_idx;
-            }
+            node& n = base[cur_idx];
+            n.len = cur_count;
+            n.next = static_cast<std::uint32_t>(out - base) - cur_idx;
            if (--depth != 0)
            {
                if (NLOHMANN_VIEW_LIKELY(depth <= 64))
@@ -2526,44 +2434,34 @@ indent_done:
 };

 /// run the builder with compile-time options
-template<typename FloatType, bool Comments, bool TrailingCommas, bool NulIsEnd, bool Store>
+template<typename FloatType, bool Comments, bool TrailingCommas, bool NulIsEnd>
 inline bool build_with(document_data& d, const char* src, std::size_t size, bool sentinel, parse_failure& failure)
 {
    if (sentinel)
    {
-        builder<FloatType, Comments, TrailingCommas, NulIsEnd, true, Store> bld(d, src, size);
+        builder<FloatType, Comments, TrailingCommas, NulIsEnd, true> bld(d, src, size);
        const bool ok = bld.run();
        failure = bld.failure();
        return ok;
    }
-    builder<FloatType, Comments, TrailingCommas, NulIsEnd, false, Store> bld(d, src, size);
+    builder<FloatType, Comments, TrailingCommas, NulIsEnd, false> bld(d, src, size);
    const bool ok = bld.run();
    failure = bld.failure();
    return ok;
 }

 /// sentinel: src[size] is readable and 0 (e.g. std::string); FloatType: the
-/// number_float_t of the document; validate_only: accept(), only the result is
-/// used
+/// number_float_t of the document
 template<typename FloatType, bool NulIsEnd>
-inline bool build(document_data& d, const char* src, std::size_t size, bool comments, bool trailing_commas, bool sentinel, bool validate_only,
-                  parse_failure& failure)
+inline bool build(document_data& d, const char* src, std::size_t size, bool comments, bool trailing_commas, bool sentinel, parse_failure& failure)
 {
-    if (validate_only && !comments && !trailing_commas)
-    {
-        // accept() of plain JSON: nothing is stored. The other option sets
-        // could skip the stores just as well; they keep the builder that
-        // stores only so that there are two more instantiations of the parse
-        // loop, not eight.
-        return build_with<FloatType, false, false, NulIsEnd, false>(d, src, size, sentinel, failure);
-    }
    if (comments)
    {
-        return trailing_commas ? build_with<FloatType, true, true, NulIsEnd, true>(d, src, size, sentinel, failure)
-               : build_with<FloatType, true, false, NulIsEnd, true>(d, src, size, sentinel, failure);
+        return trailing_commas ? build_with<FloatType, true, true, NulIsEnd>(d, src, size, sentinel, failure)
+               : build_with<FloatType, true, false, NulIsEnd>(d, src, size, sentinel, failure);
    }
-    return trailing_commas ? build_with<FloatType, false, true, NulIsEnd, true>(d, src, size, sentinel, failure)
-           : build_with<FloatType, false, false, NulIsEnd, true>(d, src, size, sentinel, failure);
+    return trailing_commas ? build_with<FloatType, false, true, NulIsEnd>(d, src, size, sentinel, failure)
+           : build_with<FloatType, false, false, NulIsEnd>(d, src, size, sentinel, failure);
 }

 }  // namespace view
@@ -10203,7 +10101,7 @@ class basic_json_document
                                     const bool ignore_trailing_commas = false)
    {
        basic_json_document d;
-        d.read_range(first, last, error_mode(allow_exceptions), ignore_comments, ignore_trailing_commas);
+        d.read_range(first, last, allow_exceptions, ignore_comments, ignore_trailing_commas);
        return d;
    }

@@ -10218,7 +10116,7 @@ class basic_json_document
        basic_json_document d;
        std::string buf;
        append_input(buf, std::forward<InputType>(input));
-        d.build_owned(std::move(buf), error_mode(allow_exceptions), ignore_comments, ignore_trailing_commas);
+        d.build_owned(std::move(buf), allow_exceptions, ignore_comments, ignore_trailing_commas);
        return d;
    }

@@ -10231,8 +10129,7 @@ class basic_json_document
    static bool accept(InputType&& input, const bool ignore_comments = false, const bool ignore_trailing_commas = false)
    {
        basic_json_document d;
-        d.read_kind(std::forward<InputType>(input), on_error::validate, ignore_comments, ignore_trailing_commas,
-                    std::integral_constant<detail::view::input_kind, detail::view::classify_input<InputType>::value> {});
+        d.read(std::forward<InputType>(input), false, ignore_comments, ignore_trailing_commas);
        return !d.is_discarded();
    }

@@ -10248,7 +10145,7 @@ class basic_json_document
              const bool ignore_comments = false,
              const bool ignore_trailing_commas = false)
    {
-        read_kind(std::forward<InputType>(input), error_mode(allow_exceptions), ignore_comments, ignore_trailing_commas,
+        read_kind(std::forward<InputType>(input), allow_exceptions, ignore_comments, ignore_trailing_commas,
                  std::integral_constant<detail::view::input_kind, detail::view::classify_input<InputType>::value> {});
    }

@@ -10607,22 +10504,6 @@ class basic_json_document
        return idx;
    }

-    /// what a failed parse does
-    enum class on_error : std::uint8_t
-    {
-        discard, ///< the document is discarded
-        raise,   ///< parse_error is thrown
-        validate ///< accept(): the document is discarded, and no node is stored while parsing
-    };
-
-    static on_error error_mode(bool allow_exceptions) noexcept
-    {
-        return allow_exceptions ? on_error::raise : on_error::discard;
-    }
-
-    /// NulIsEnd of the builder: a null byte ends the input unless strict_nul_handling
-    static constexpr bool nul_is_end = !detail::abi_config::strict_nul_handling;
-
    /// create the storage (sized for the input) on first use
    void ensure_data(const char* src, std::size_t size)
    {
@@ -10634,7 +10515,7 @@ class basic_json_document

    /// parse a buffer the document takes ownership of (it replaces the
    /// previous owned copy)
-    void build_owned(std::string&& buf, on_error allow_exceptions, bool comments, bool trailing_commas)
+    void build_owned(std::string&& buf, bool allow_exceptions, bool comments, bool trailing_commas)
    {
        ensure_data(buf.data(), buf.size());
        m_data->owned = std::move(buf);
@@ -10644,7 +10525,7 @@ class basic_json_document
    /// parse a copy of the input; the copy is appended to the cleared owned
    /// buffer, so that its capacity is reused
    template<typename T>
-    void build_copy(T&& input, on_error allow_exceptions, bool comments, bool trailing_commas)
+    void build_copy(T&& input, bool allow_exceptions, bool comments, bool trailing_commas)
    {
        if (!m_data)
        {
@@ -10665,7 +10546,7 @@ class basic_json_document
    }

    /// sentinel: src[size] is readable and 0 (std::string, C strings)
-    void build(const char* src, std::size_t size, on_error allow_exceptions, bool comments, bool trailing_commas, bool owned, bool sentinel)
+    void build(const char* src, std::size_t size, bool allow_exceptions, bool comments, bool trailing_commas, bool owned, bool sentinel)
    {
        ensure_data(src, size);
        document_data& d = *m_data;
@@ -10692,8 +10573,7 @@ class basic_json_document
        }
        else
        {
-            ok = detail::view::build<typename BasicJsonType::number_float_t, nul_is_end>(d, src, size, comments, trailing_commas, sentinel,
-                 allow_exceptions == on_error::validate, failure);
+            ok = detail::view::build < typename BasicJsonType::number_float_t, !detail::abi_config::strict_nul_handling > (d, src, size, comments, trailing_commas, sentinel, failure);
        }
        if (NLOHMANN_VIEW_LIKELY(ok))
        {
@@ -10706,7 +10586,7 @@ class basic_json_document
            return;
        }
        std::vector<std::uint32_t>().swap(d.large_objects);
-        if (allow_exceptions == on_error::raise)
+        if (allow_exceptions)
        {
            detail::view::throw_parse_failure<BasicJsonType>(failure, src, size, comments, trailing_commas);
        }
@@ -10714,13 +10594,13 @@ class basic_json_document

    // --- input dispatch (see detail::view::input_kind) ---

-    void read_kind(std::string&& s, on_error ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::move_string> /*unused*/)
+    void read_kind(std::string&& s, bool ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::move_string> /*unused*/)
    {
        build_owned(std::move(s), ae, c, tc);
    }

    template<typename CharT>
-    void read_kind(CharT* s, on_error ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::c_string> /*unused*/)
+    void read_kind(CharT* s, bool ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::c_string> /*unused*/)
    {
        if (s == nullptr)
        {
@@ -10734,7 +10614,7 @@ class basic_json_document
    }

    template<typename Array>
-    void read_kind(Array& a, on_error ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::char_array> /*unused*/)
+    void read_kind(Array& a, bool ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::char_array> /*unused*/)
    {
        using CharT = typename std::remove_cv<typename std::remove_extent<Array>::type>::type;
        const std::size_t n = std::extent<Array>::value;
@@ -10745,7 +10625,7 @@ class basic_json_document
    }

    template<typename T>
-    void read_kind(const T& s, on_error ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::borrow_range> /*unused*/)
+    void read_kind(const T& s, bool ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::borrow_range> /*unused*/)
    {
        // std::basic_string guarantees data()[size()] == 0: use it as sentinel
        const auto size = static_cast<std::size_t>(s.size());
@@ -10754,13 +10634,13 @@ class basic_json_document
    }

    template<typename T>
-    void read_kind(const T& s, on_error ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::copy_range> /*unused*/)
+    void read_kind(const T& s, bool ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::copy_range> /*unused*/)
    {
        build_copy(s, ae, c, tc);
    }

    template<typename CharT>
-    void read_kind(CharT* s, on_error ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::wide_c_string> /*unused*/)
+    void read_kind(CharT* s, bool ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::wide_c_string> /*unused*/)
    {
        if (s == nullptr)
        {
@@ -10772,13 +10652,13 @@ class basic_json_document
    }

    template<typename T>
-    void read_kind(T&& input, on_error ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::adapter> /*unused*/)
+    void read_kind(T&& input, bool ae, bool c, bool tc, std::integral_constant<input_kind, input_kind::adapter> /*unused*/)
    {
        build_copy(std::forward<T>(input), ae, c, tc);
    }

    template<typename IteratorType>
-    void read_range(IteratorType first, IteratorType last, on_error ae, bool c, bool tc)
+    void read_range(IteratorType first, IteratorType last, bool ae, bool c, bool tc)
    {
        using value_type = typename std::remove_cv<typename std::iterator_traits<IteratorType>::value_type>::type;
        // borrow the range where the library's input adapter scans it as one
@@ -10789,14 +10669,14 @@ class basic_json_document
    }

    template<typename IteratorType>
-    void read_range_impl(IteratorType first, IteratorType last, on_error ae, bool c, bool tc, std::true_type /*contiguous bytes*/)
+    void read_range_impl(IteratorType first, IteratorType last, bool ae, bool c, bool tc, std::true_type /*contiguous bytes*/)
    {
        const auto size = static_cast<std::size_t>(std::distance(first, last));
        build(size == 0 ? "" : reinterpret_cast<const char*>(&*first), size, ae, c, tc, false, size == 0); // NOLINT(cppcoreguidelines-pro-type-reinterpret-cast)
    }

    template<typename IteratorType>
-    void read_range_impl(IteratorType first, IteratorType last, on_error ae, bool c, bool tc, std::false_type /*other*/)
+    void read_range_impl(IteratorType first, IteratorType last, bool ae, bool c, bool tc, std::false_type /*other*/)
    {
        build_copy(detail::view::make_adapter_input(detail::input_adapter(first, last)), ae, c, tc);
    }
@@ -403,6 +403,7 @@ TEST_CASE("guard page")
                    CHECK(nlohmann::detail::find_string_special(p, n) <= n);
                    CHECK(nlohmann::detail::find_ascii_copyable_run(p, n) <= n);
                    CHECK(nlohmann::detail::valid_utf8_prefix(p, n) <= n);
+                    CHECK(nlohmann::detail::valid_utf8_string(p, n) == (nlohmann::detail::valid_utf8_prefix(p, n) == n));
 #if defined(JSON_USE_SIMDUTF) && defined(JSON_HAS_CPP_17)
                    CHECK(nlohmann::detail::find_string_delimiter(p, n) <= n);
 #endif
@@ -59,8 +59,7 @@ built build(const std::string& text, bool comments, bool trailing_commas, bool s
        r.copy.assign(text.begin(), text.end());
        src = r.copy.data();
    }
-    constexpr bool nul_is_end = !nlohmann::detail::abi_config::strict_nul_handling;
-    r.ok = nlohmann::detail::view::build<FloatType, nul_is_end>(*r.data, src, text.size(), comments, trailing_commas, sentinel, false, r.failure);
+    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();
@@ -173,22 +172,6 @@ void check_string(const std::string& content)
    }
 }

-// accept() stores no nodes: it must accept and reject what json::accept does, for all options and for inputs with
-// and without a terminating NUL
-void check_accept(const std::string& text)
-{
-    CAPTURE(text)
-    const std::vector<char> bytes(text.begin(), text.end());
-    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);
-        CHECK(nlohmann::json_document::accept(text, comments, trailing_commas) == accepted);
-        CHECK(nlohmann::json_document::accept(bytes, comments, trailing_commas) == accepted);
-    }
-}
-
 // a small deterministic generator of documents
 struct generator
 {
@@ -440,133 +423,6 @@ TEST_CASE("json_view builder")
        }
    }

-    SECTION("accept() without an index")
-    {
-        for (const char* text :
-                {
-                    "", " ", "\n", "null", "nul", "true", "[", "]", "[]", "{}", "[1,2,3]", "[1,2,3", "[1,2,3,]", "{\"a\":1,}", "{\"a\"}", "{\"a\":}",
-                    "[1] x", "[1][2]", "\xEF\xBB\xBF[1]", "[\"a\\", "[\"\\u12\"]", "[\"\\ud800\"]", "[\"\x01\"]", "[\"\xff\"]", "[01]", "[1.]",
-                    "[.5]", "[-]", "[1e]", "// c\n[1]", "/* c */ {\"a\" /* x */ : 1}", "/* c", "[1] // c", "1e400", "-1e400", "123456789012345678901234567890",
-                    "{\"a\":1,\"a\":2}", "[[[]]]", "\"\\u00e4\\n\\ud83d\\ude00\""
-                })
-        {
-            check_accept(text);
-        }
-        // a null byte inside the text, and after it
-        check_accept(std::string("[1,\0]", 5));
-        check_accept(std::string("[1]\0", 4));
-        check_accept(std::string("\0", 1));
-
-        // generated documents and damaged copies, as for the builder above
-        generator g;
-        for (int i = 0; i < 2000; ++i)
-        {
-            std::string text;
-            g.value(text, 0);
-            check_accept(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_accept(damaged);
-            check_accept(text.substr(0, at));
-        }
-
-        // nesting beyond the 64 levels that are kept on the builder's own stack, closed and not closed
-        for (const std::size_t depth :
-                {
-                    63, 64, 65, 66, 200, 1000
-                })
-        {
-            check_accept(std::string(depth, '[') + std::string(depth, ']'));
-            check_accept(std::string(depth, '[') + std::string(depth - 1, ']'));
-            std::string objects;
-            for (std::size_t i = 0; i < depth; ++i)
-            {
-                objects += "{\"k\":";
-            }
-            objects += "1" + std::string(depth, '}');
-            check_accept(objects);
-            check_accept(objects.substr(0, objects.size() - 1));
-        }
-
-        // objects with enough members to get a hash index when they are indexed (and duplicate keys)
-        std::string large = "{";
-        for (int i = 0; i < 300; ++i)
-        {
-            large += (i == 0 ? "\"k" : ",\"k") + std::to_string(i % 200) + "\":" + std::to_string(i);
-        }
-        check_accept(large + "}");
-        check_accept(large);
-        check_accept("[" + large + "},{" + large.substr(1) + "}]");
-
-        // the overflow check uses the floating-point type of the document
-        using float_json = nlohmann::basic_json<std::map, std::vector, std::string, bool, std::int64_t, std::uint64_t, float>;
-        for (const char* text :
-                {
-                    "1e39", "-1e39", "3.4028235e38", "3.4028236e38", "[1.5e38, 3.5e38]", "{\"a\": 1e-50, \"b\": 1e39}", "340282356779733661637539395458142568448"
-                })
-        {
-            CAPTURE(text)
-            CHECK(nlohmann::basic_json_document<float_json>::accept(text) == float_json::accept(text));
-        }
-
-        // strings are checked and not decoded: runs of text of every length from 0 to 40 between valid and invalid
-        // escapes, UTF-8 sequences and control characters, in values and in keys, cut near the end of the text
-        const std::array<const char*, 9> valid = {{"\\n", "\\\"", "\\\\", "\\/", "\\t", "\\u00e4", "\\u20ac", "\\ud83d\\ude00", "\\u0000"}};
-        const std::array<const char*, 12> invalid = {{"\\x", "\\ud800", "\\udc00", "\\u12G4", "\\u12", "\\ud83dx", "\\ud83d\\u0041", "\\ud83d\\ud83d", "\\", "\x01", "\x1f", "\x7f"}};
-        const std::array<const char*, 6> utf8 = {{"\xc3\xa4", "\xe2\x82\xac", "\xf0\x9f\x98\x80", "\xc3", "\xff", "\x80"}};
-        std::mt19937 rng(2029); // NOLINT(cert-msc32-c,cert-msc51-cpp,bugprone-random-generator-seed): reproducible
-        for (std::size_t run = 0; run <= 40; ++run)
-        {
-            for (int i = 0; i < 30; ++i)
-            {
-                std::string content = std::string(rng() % 20, 'a');
-                const auto pieces = 1 + (rng() % 6);
-                for (unsigned k = 0; k < pieces; ++k)
-                {
-                    content += valid[rng() % valid.size()];
-                    content += std::string(k == 0 ? run : rng() % 41, static_cast<char>('b' + (rng() % 20)));
-                    switch (rng() % 8)
-                    {
-                        case 0:
-                            content += utf8[rng() % 3];
-                            break;
-                        case 1:
-                            content += utf8[rng() % utf8.size()];
-                            break;
-                        case 2:
-                            content += invalid[rng() % invalid.size()];
-                            break;
-                        default:
-                            break;
-                    }
-                }
-                for (const std::string& text :
-                        {
-                            "[\"" + content + "\"]", "{\"" + content + "\":1}", "[\"" + content, "[\"" + content.substr(0, content.size() - (rng() % 4)) + "\"]"
-                        })
-                {
-                    check_accept(text);
-                }
-            }
-        }
-    }
-
-    SECTION("escapes in a row")
-    {
-        for (const char* content :
-                {
-                    "\\n", "\\n\\t\\\\", "\\u00e9\\u00e9\\u4e2d", "\\u00e9\\u00e", "\\u00e9\\", "\\n\\q", "\\n\\ud83d\\ude00\\n",
-                    "\\ud83d\\ude00\\ud83d", "\\n\\n\\n abc \\n\\n", "\\n\x01", "\\n\xC3\x28", "\\n\\"
-                })
-        {
-            check_accept(std::string("[\"") + content + "\"]");
-            check_accept(std::string("[\"") + content);
-            check_accept(std::string("{\"") + content + "\":1}");
-        }
-    }
-
    SECTION("test files")
    {
        for (const char* name :
@@ -13,6 +13,7 @@ using nlohmann::json;

 #include <algorithm>
 #include <fstream>
+#include <random>
 #include <string>
 #include "make_test_data_available.hpp"
 #include "test_utils.hpp"
@@ -1631,3 +1632,40 @@ TEST_CASE("Unicode RFC 3629")
 }

 DOCTEST_CLANG_SUPPRESS_WARNING_POP
+
+TEST_CASE("UTF-8 check of whole strings")
+{
+    // valid_utf8_string() decides short strings with a word test and long ones
+    // 16 bytes at a time where a vector check is available; it must agree with
+    // the byte-wise check for every mix of ASCII, sequences of each length,
+    // ill-formed and truncated sequences, at every length and offset
+    const char* const pieces[] =
+    {
+        "a", "\x7F", "\xC3\xA9", "\xE2\x82\xAC", "\xE0\xA0\x80", "\xEF\xBF\xBF", "\xF0\x9F\x98\x80", "\xF4\x8F\xBF\xBF",
+        "\x80", "\xC0\xAF", "\xC3", "\xE2\x82", "\xED\xA0\x80", "\xF0\x9F", "\xF4\x90\x80\x80", "\xFF"
+    };
+    // std::mt19937's sequence is fixed by the standard; no distribution is used
+    std::mt19937 rng(5);
+    std::size_t valid = 0;
+    for (int i = 0; i < 100000; ++i)
+    {
+        std::string s;
+        const std::size_t length = rng() % 100;
+        const std::uint32_t ascii = rng() % 101;
+        while (s.size() < length)
+        {
+            // mostly well-formed pieces (the first eight), so that long valid
+            // strings are common
+            s += (rng() % 100 < ascii) ? std::string("x") : std::string(pieces[rng() % (rng() % 8 == 0 ? 16 : 8)]);
+        }
+        const auto* data = reinterpret_cast<const unsigned char*>(s.data());
+        const bool expected = nlohmann::detail::is_valid_utf8(s);
+        valid += expected ? 1 : 0;
+        CAPTURE(s)
+        CHECK(nlohmann::detail::valid_utf8_string(data, s.size()) == expected);
+        CHECK((nlohmann::detail::valid_utf8_prefix(data, s.size()) == s.size()) == expected);
+    }
+    // both outcomes are common enough to be tested
+    CHECK(valid > 20000);
+    CHECK(valid < 80000);
+}