Speed up parsing of contiguous input (numbers, strings, UTF-8) (#5283)

* Speed up number parsing in the lexer (fast paths from the fast_float/simdjson world)

The number scanner converted its already-validated digit buffer with
std::strtoull/std::strtoll/std::strtod. Those pull in locale and errno
machinery and dominate number-heavy parsing (strtod runs at ~6 M/s).

Replace them with dedicated parsers over the validated buffer:

- parse_integer_unsigned / parse_integer_signed: accumulate digits with
  overflow detection, falling back to the float path on overflow exactly
  as the strtoull/strtoll round-trip check did. Overflow behavior is
  unchanged for narrower or wider custom number types.

- parse_float_fast: Clinger's exact fast path for `double` (<=19 significant
  digits, |exp10| <= 22, significand < 2^53), where significand * 10^exp is
  exact under IEEE round-to-nearest. This is the same fast path used by
  fast_float/simdjson. It is bit-identical to strtod on this subset and
  declines (falling back to strtod) otherwise. Only `double` uses it; float
  and long double keep std::strtof/std::strtold via a templated overload.

Measured on representative data (g++ 13, -O3):
  - integers:  DOM parse +11%, SAX +25-34%
  - floats:    DOM parse +37%, SAX +70%  (clang: float DOM ~1.9x)

No dependencies added; header-only and C++11-clean. Existing parser,
lexer, conversion and deserialization unit tests pass unchanged; a
3M-value random-double fuzz matches strtod bit-for-bit.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Add SWAR bulk string scanning for contiguous input (simdjson-style)

scan_string() read the input one character at a time through the input
adapter and classified every byte with a large switch. For contiguous
byte buffers we can instead scan 8 bytes at a time with a SWAR word test
that finds the first byte needing individual handling (the closing quote,
an escape, a control character, or a non-ASCII UTF-8 byte) and bulk-append
the ordinary run in one go.

- input adapters expose supports_bulk_scan / bulk_data / bulk_remaining /
  bulk_skip for provably-contiguous, same-type, 1-byte iterator ranges
  (raw pointers in every standard; std::string/std::vector/std::array and
  friends additionally in C++20 via std::contiguous_iterator).
- the lexer gains a bulk_scan capability (gated on lazy_token_string so
  bypassing the per-character capture cannot lose error diagnostics) and a
  scan_string_bulk() fast path; streaming/wide/user adapters are unchanged
  and keep the byte-at-a-time scanner.

The run contains no newline (all bytes < 0x20 are treated as special), so
position bookkeeping stays exact, and error tokens are still reconstructed
lazily from the consumed byte range. The SWAR special-byte test is pure
uint64_t arithmetic - no intrinsics, no runtime dispatch, C++11-clean.

Measured on representative data, pointer input, g++ 13 -O3
(string values discarded by accept() see the largest gains):

  long ASCII strings:  DOM +4.5x,  SAX +14x,   accept +17x  (to ~2 GB/s)
  short strings:       DOM +15%,   SAX +62%,   accept +85%
  escape-heavy:        DOM +31%,   SAX +26%,   accept +28%

Same-input parity verified: 200k randomized documents (escapes, multibyte
UTF-8, surrogate pairs) accept/parse identically via the contiguous SWAR
path and the streaming byte path; unit lexer/parser/diagnostic-position/
deserialization/conversions suites pass unchanged.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Validate UTF-8 in the bulk string scanner (portable ~2-3x on non-ASCII text)

The SWAR bulk string path stopped at the first non-ASCII byte and handed
every multibyte character to the byte-at-a-time scanner, whose per-byte
get()/next_byte_in_range()/add() machinery runs at roughly half the speed
of validating straight from the buffer. As a result, dense non-ASCII text
(CJK, emoji, accented Latin) parsed ~10-15x slower than ASCII.

Fold well-formed UTF-8 into the bulk run: scan_string_bulk() now, on a
non-ASCII lead byte, validates one sequence with validate_one_utf8() -
which mirrors scan_string()'s per-byte switch ranges exactly (rejecting
overlong forms, surrogates, and out-of-range code points) - and appends it
in place, continuing until the closing quote, an escape, a control byte,
or an ill-formed sequence. All error handling still defers to the byte
path, so error messages and positions are byte-for-byte unchanged.

Because only well-formed content is fast-pathed and every rejection falls
through to the existing scanner, behavior is identical; the win is purely
throughput. Measured on pointer input (accept, string values discarded):

  content        g++ 13         clang 18
  dense CJK      277 -> 648     ~605  MB/s   (~2.3x)
  dense emoji    299 -> 857     ~702  MB/s   (~2.6-2.9x)
  mixed 90% ASCII 246 -> 331    ~334  MB/s   (~1.35x)
  pure ASCII     unchanged (~3.2 / 4.1 GB/s)

Verified: 2,000,000 randomized documents built from arbitrary bytes
(overlong, surrogate, truncated, out-of-range sequences) accept/reject and
parse identically via the contiguous path and the streaming byte path;
lexer/parser/diagnostic-position/deserialization/conversions suites pass
unchanged. Pure C++11, no intrinsics.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Route contiguous byte containers through the pointer adapter (fast paths in C++11)

json::parse(std::string) - the most common entry point - did not benefit
from the contiguous fast paths (bulk string scanning, UTF-8 bulk
validation, memcpy for binary formats) in C++11..17: std::string::iterator
is a library wrapper, not a raw pointer, and pre-C++20 there is no portable
way to prove it contiguous, so supports_bulk_scan was false. Only raw
pointers, string literals, and C-arrays (and, in C++20, anything modelling
std::contiguous_iterator) took the fast path.

Detect contiguous single-byte containers (std::string, std::vector<char>,
std::vector<std::uint8_t>, std::string_view, ...) via is_contiguous_byte_
container and route them through an iterator_input_adapter built from
data()/data()+size(). The generic iterator-based container overload is
constrained to exclude these, so the two overloads are disjoint and there
is no ambiguity (a plain competing overload loses to the greedy
forwarding-reference container overload on reference binding, and a factory
partial-specialization is ambiguous - both were tried and rejected).

The pointer keeps the container's own element type, so char_type - and
therefore all parsing behavior - is byte-for-byte identical to the iterator
path (const char* for std::string, const std::uint8_t* for
std::vector<std::uint8_t>); only the raw pointer additionally turns on the
fast paths. Lifetimes are unchanged: the container outlives the adapter for
the full parse expression, exactly as the iterators it replaces did.

Measured, C++11, json::parse/accept(std::string), g++ 13:

  long ASCII strings:  accept 201 -> 3200 MB/s  (~16x),  parse 174 -> 1444
  dense CJK:           accept 263 ->  697 MB/s  (~2.6x)
  short strings:       accept 163 ->  243 MB/s  (~1.5x)

Verified: char_type preserved for std::string (char) and
std::vector<std::uint8_t> (uint8_t); CBOR/MsgPack round-trips from
std::vector<std::uint8_t> unchanged; 1,000,000 randomized documents accept
and parse identically via std::string and via std::istream;
deserialization/user-defined-input/parser/lexer/conversions/diagnostic-
position suites pass (20,480 assertions); warning-clean on g++ and clang in
C++11/17/20.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Add optional simdutf backend for bulk UTF-8 validation (JSON_USE_SIMDUTF)

The bulk string scanner validates UTF-8 straight from a contiguous buffer.
The scalar validator caps at ~0.3-0.7 GB/s on non-ASCII text; a SIMD
validator reaches several GB/s. Rather than hand-rolling SIMD UTF-8
validation (easy to get subtly wrong - a from-scratch SSE attempt rejected
valid CJK), wire in the vetted simdutf library behind an opt-in switch.

simdutf is not header-only (it ships simdutf.cpp and uses runtime CPU
dispatch), so it is not vendored: defining JSON_USE_SIMDUTF includes
<simdutf.h> and routes the bulk validator through simdutf::validate_utf8;
the project supplies and links simdutf. Undefined (the default), nothing
external is included and the portable C++11 scalar path is used, so the
library stays header-only and its baseline behavior is unchanged.

Design keeps behavior identical either way:
- scan_string_bulk() now finds the run up to the next quote/escape/control
  byte (non-ASCII allowed) and validates it in one shot; on the rare
  validation failure it recomputes the exact valid prefix with the scalar
  helper, so ill-formed input still falls through to the byte path and is
  reported at the same position with the same message.
- the per-sequence scalar path is factored into scalar_string_bulk_run()
  and is the default backend; the refactor is behavior-preserving and does
  not change scalar throughput.

Verified: default and JSON_USE_SIMDUTF builds accept/reject/parse
identically across 2,000,000 arbitrary-byte documents and 1,000,000
mixed-escape/UTF-8 documents (differential fuzz vs the streaming byte
path); lexer/parser/diagnostic-position/deserialization suites pass under
both configurations (20,188 assertions with the backend enabled);
warning-clean on g++ and clang, C++11 and C++20, both configurations.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Add a contiguous fast path for scanning numbers

scan_number() reads a number one character at a time through the input
adapter (get()) and appends each byte to token_buffer (add()) before
converting. For contiguous input, the per-character get()/add() overhead
dominates: it is roughly two thirds of the time spent on number-heavy
parsing, far more than the value conversion itself.

Add scan_number_bulk_contiguous(), which parses the whole number token
straight from the input buffer: it validates and classifies the extent
with the same grammar as scan_number()'s state machine, materializes
token_buffer in one copy (substituting the locale decimal point exactly as
scan_number() does), advances the adapter, and reuses the shared
convert_number() tail. On anything it does not recognize as a well-formed
number it makes no state change and returns token_type::uninitialized, so
the caller falls back to scan_number(), which then produces the exact
diagnostic. Errors and their positions are therefore unchanged.

The conversion tail is factored out of scan_number() into convert_number()
so both scanners share it; the fast path is selected by tag dispatch on the
existing bulk_scan capability, so streaming/wide/user adapters are
unaffected.

Measured on pointer input, g++ 13 -O3:
  - integers: parse +65%, accept +98%
  - floats:   parse +39%, accept +70%

Verified: 2,000,000 randomized number documents (including overflow-range
integers, long digit strings and %.17g doubles) parse identically via the
contiguous path and the streaming byte path, matching value, type and
round-trip text; the locale suite and existing parser/lexer/conversions/
deserialization tests pass; a new "lexer number fast path" test checks
contiguous-vs-streaming parity, token classification, and that malformed
numbers are rejected identically. Pure C++11, no intrinsics.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Move byte-level scan/parse helpers out of lexer.hpp

lexer.hpp had grown by ~600 lines of byte-level helpers that have no
dependency on the lexer's template parameters and clutter the state
machine. Move them, unchanged, into two focused headers as free functions
in namespace detail:

- number_parse.hpp: parse_integer_unsigned/parse_integer_signed (now
  templated on the number type) and parse_float_fast (Clinger's exact
  double fast path, with the decimal point passed as an argument instead of
  read from a lexer member).
- string_scan.hpp: the SWAR string helpers (is_string_special,
  swar_string_special, find_string_special, validate_one_utf8,
  scalar_string_bulk_run) and the backend-dispatched string_bulk_run,
  including the optional simdutf include and find_string_delimiter.

lexer.hpp now includes these and calls the free functions; the methods that
touch lexer state (scan_string, scan_number, scan_string_bulk,
scan_number_bulk_contiguous, convert_number) stay put. This is a pure code
move with no behavior change: lexer.hpp drops from 2357 to 1934 lines, the
now-unused <cstdint>/<cstring>/<limits> includes are removed, and the
free-function form makes the SWAR helpers reusable elsewhere (e.g. the
serializer's string escaping).

Verified: default and JSON_USE_SIMDUTF builds compile; 2,000,000 number and
2,000,000 arbitrary-byte-string differential-fuzz documents parse
identically to before; lexer/parser/conversions/deserialization/locale/
diagnostic-position suites pass (20,576 assertions); warning-clean on g++
and clang in C++11/17/20; the amalgamation regenerates and passes
check-amalgamation.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Fix clang-tidy findings and document JSON_USE_SIMDUTF in the nav

- number_parse.hpp: use std::array for the powers-of-ten table
  (avoid-c-arrays) and `auto` for the cast-initialized result
  (modernize-use-auto), matching the codebase style (cf. the serializer's
  utf8d table). Indexing casts keep the -Wsign-conversion build clean.
- add JSON_USE_SIMDUTF to the mkdocs navigation so the macro page is
  reachable.

No behavior change; clang-tidy is clean on the new headers and the
amalgamation is regenerated.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Fix number fast path for custom string types without assign()

The contiguous number fast path materialized token_buffer with
token_buffer.assign(data, len), but string_t is only required to provide
the minimal interface the rest of the lexer uses (push_back, append,
clear, operator[], ...). Custom string types such as the test's alt_string
do not implement assign(), so scan_number_bulk_contiguous() failed to
compile for them (unit-alt-string), breaking the gcc/clang standards and
old-compiler CI jobs.

reset() already clears token_buffer, so fill it with append() - which
alt_string and std::string both provide and which the string fast path
already relies on - instead of assign().

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Satisfy clang-tidy: parenthesize math and drop unused forwarding reference

The CI clang-tidy (newer than the locally available version) reported two
additional checks on the new code:

- readability-math-missing-parentheses: parenthesize the (a * b) + c digit
  accumulations in number_parse.hpp.
- cppcoreguidelines-missing-std-forward: the contiguous-byte-container
  input_adapter overload took a forwarding reference but only reads
  data()/size() and never forwards it. It is already disjoint from the
  generic container overload via SFINAE, so a plain const& is correct and
  clearer (and keeps the container alive for the whole parse just as before).

No behavior change; char_type and routing are unchanged (std::string and
std::vector<std::uint8_t> still take the pointer adapter with char/uint8_t
char_type), CBOR/MsgPack round-trips and the 2M number fuzz still pass.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Use std::from_chars (Eisel-Lemire) for float conversion when available

The Clinger fast path is exact only for the "easy" subset (<=19 significant
digits, |exp10| <= 22); high-precision and scientific floats fall through to
strtod, where the failed Clinger attempt actually makes parsing a net loss.
std::from_chars implements the Eisel-Lemire algorithm in modern standard
libraries: locale-independent, correctly rounded, and fast over the whole
value range.

convert_number() now tries parse_float_from_chars() first (guarded by
__cpp_lib_to_chars, so C++11 and libc++-without-float-support keep the
Clinger + strtod path unchanged), then Clinger, then strtof. from_chars is
used only when it consumes the entire token; a partial parse means a non-'.'
locale decimal point, and an under-/overflow (result_out_of_range) also
declines - in both cases the existing strtod fallback supplies the exact
value and the well-defined +/-inf/0 the parser expects, side-stepping the
P4168 divergence between implementations. float and long double now get the
fast path too (Clinger was double-only).

Measured, C++17, g++ 13 -O3, json::parse/accept:
  - canada-style floats: ~unchanged (Clinger already covered them)
  - high-precision (17 digits):  parse 2.1x, accept 2.5x
  - scientific (17 digits + exp): parse 3.6x, accept 4.1x

Verified: C++11 (Clinger/strtod) and C++17 (from_chars) parse every value -
including subnormals, boundary values, and 1e9999/1e-9999 over-/underflow -
to bit-identical results; 2M number-fuzz clean; conversions/deserialization/
locale/number-fast-path suites pass in both C++11 and C++17; clang-tidy
clean; warning-clean on g++ and clang in C++11/17/20.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Guard from_chars use on JSON_HAS_CPP_17, not just __cpp_lib_to_chars

libstdc++ 15 defines __cpp_lib_to_chars even in C++14 mode (via
bits/version.h pulled in by other headers), but <charconv> is only included
under JSON_HAS_CPP_17. That made parse_float_from_chars() reference
std::from_chars without the header in C++14 builds, breaking gcc-latest,
icpx, and the offline-testdata jobs.

Gate the use on JSON_HAS_CPP_17 && __cpp_lib_to_chars so it matches the
include condition exactly; C++11/14 always take the scalar fallback.
Verified by forcing __cpp_lib_to_chars in a C++14 build: the guard
suppresses std::from_chars and it compiles. C++17 behavior is unchanged.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Disable Clinger float fast path under extended FP precision (x87)

The contiguous number fast path uses a Clinger-style exact algorithm
(significand * 10^scale in double arithmetic), which is only correctly
rounded when double operations are evaluated in true 53-bit precision.
On the x87 FPU used by 32-bit x86 (FLT_EVAL_METHOD == 2) the single
multiply/divide is computed in 80-bit and then double-rounded to double,
so a small fraction of values land 1 ULP off.

This surfaced as test-cbor_cpp11 and test-msgpack_cpp11 failing on the
mingw (x86) job for regression/floats.json: the C++17 builds pass because
they take the correctly-rounded std::from_chars path, while C++11 falls
back to parse_float_fast(). A 5M-sample check over shortest round-trip
decimals reproduces it: 0 divergences with 53-bit doubles, ~1 in 25 000
with 80-bit intermediates; declining to std::strtod fixes all of them.

Guard parse_float_fast() on FLT_EVAL_METHOD so it declines whenever the
platform evaluates doubles in extended precision, letting the caller use
the correctly-rounded std::from_chars / std::strtod path instead. On
mainstream x86-64/ARM64 (FLT_EVAL_METHOD == 0) the fast path is unchanged.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Guard <charconv> include with __has_include for GCC 7

GCC 7 sets __cplusplus to the C++17 value under -std=gnu++1z, so
JSON_HAS_CPP_17 is defined, but its libstdc++ ships no <charconv> header
(added in GCC 8; floating-point from_chars in GCC 11). The unconditional
"#if defined(JSON_HAS_CPP_17) #include <charconv>" therefore failed to
compile there: "fatal error: charconv: No such file or directory" in the
ci_test_compilers_gcc (7) job.

Wrap the include in __has_include(<charconv>), mirroring the library's
existing handling of <version> and <filesystem> in macro_scope.hpp. When
the header is absent, __cpp_lib_to_chars stays undefined and
parse_float_from_chars() takes its scalar fallback, so the from_chars use
site (already gated on __cpp_lib_to_chars) is never reached. GCC 8-10,
which have <charconv> but no floating-point from_chars, are unaffected:
they include the header but still take the fallback. GCC 11+ is unchanged.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Restore the column when ungetting a newline

The contiguous number fast path never reads the character that terminates
a number token, while scan_number() reads it and then ungets it. When that
character is a newline, get() has already cleared chars_read_current_line,
and unget() could only restore lines_read - leaving the column at 0. The
two paths therefore reported different columns for the same document:

    json::parse("[01\n]")      -> line 1, column 3
    json::parse(stringstream)  -> line 1, column 0

Remember the column the newline was read at so unget() can restore it.
Both paths now report the position the offending token actually starts at,
which also fixes the pre-existing column-0 artifact for streaming input.

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

* Extend the bulk scan fast paths to sized sentinels

supports_bulk_scan required IteratorType and SentinelType to be the same
type, which excluded std::counted_iterator paired with std::default_sentinel_t
- the combination #5268 had already enabled for the memcpy fast path. Such
input fell back to the byte-at-a-time scanner even though it is contiguous
and its remaining length is computable in O(1).

Factor the "distance is computable in O(1)" test into sentinel_is_sized and
use it for iterator_is_contiguous, supports_seek, and supports_bulk_scan
alike, and share the std::ranges::distance/std::distance dispatch through a
remaining_count() helper.

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

* Document JSON_USE_SIMDUTF on the macro overview page

The macro was only listed in the API macro index; add it to the supported
macros overview alongside the other JSON_USE_* macros, and note that it
selects between two definitions of the same inline function and so must be
defined identically in every translation unit.

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

* Amalgamate source code

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

* Cover the counted-iterator bulk scan paths

Sized sentinels newly reach the bulk string/number scanners and the
seek-based token reconstruction, so exercise both:

- diagnostics that quote the offending token, which are rebuilt from the
  consumed input via copy_consumed_range()
- inputs whose count ends before the underlying buffer does, including a
  closing quote that exists only behind the count, a cut inside an 8-byte
  SWAR stride, and a cut inside a UTF-8 sequence

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

* Fix the SentinelType example on iterator_input_adapter

The comment offered "a C++20 sentinel or counted_iterator" as examples of a
SentinelType, but std::counted_iterator is the IteratorType - the sentinel it
pairs with is std::default_sentinel_t. #5268 corrected the same wording in the
API documentation and left the code comment behind.

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

* Do not discard the parse result in the error position check

json::parse is declared warn_unused_result, and CHECK_THROWS_WITH_AS
evaluates its expression as a discarded statement, so the assertion broke
the -Werror builds (GCC -Werror=unused-result, MSVC C4834 under /WX).
Compare against the helper that already captures the message instead.

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

* Lock the two number grammars together with a parity test

The JSON number grammar is encoded twice: as the scan_number() state machine
and as the contiguous fast path. The fast path declining on anything it does
not recognize keeps most divergence harmless, but if it ever accepted
something the state machine rejects the result would be a silent correctness
bug, and the existing test only pinned a hand-written list of numbers.

Enumerate every string of length 1..4 over "01.eE+-" (2800 tokens) and
require both paths to agree on the parsed value and on the exact error
message. Verified to fail if the fast path's grammar is perturbed.

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

* Skip token_buffer for integers on the contiguous path

An integer token does not need token_buffer: the number_integer and
number_unsigned SAX callbacks take only the value, and the overflow
diagnostic rebuilds the text from the input via get_token_string(). Convert
straight from the input buffer and materialize the token only for the
floating-point tail, which still needs a NUL-terminated buffer for strtod.

JSON_DIAGNOSTIC_POSITIONS derives a number's start position from
get_string().size(), so the copy is kept when that is enabled.

The integer dispatch is factored into convert_integer() and shared with
convert_number(), so both scanners keep using one implementation.

Integer-heavy input, 400k values, -O3:

              parse    accept
  gcc 16      +14%     +23%
  clang       +15%     +22%

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

* Amalgamate source code

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

* Cover the bulk string and UTF-8 scanners

These paths had no dedicated tests and rested on differential fuzzing only.
Add three sections, all comparing the contiguous scanner against the
byte-at-a-time one on the parsed value and on the exact error message:

- every string of length 1..3 over an alphabet of ordinary ASCII, both
  specials, a control byte, escape characters, UTF-8 lead and continuation
  bytes, and a byte that is never valid - each at offset 0 and offset 9, so
  the bulk scanner sees them with and without a run behind them
- every kind of run-ending byte at each offset across two 8-byte SWAR words,
  so multibyte sequences also straddle the word boundary
- the boundaries of every range validate_one_utf8() recognizes: shortest and
  longest encodings, overlongs, both ends of the surrogate block, U+10FFFF
  and just past it, and truncated sequences

Verified to fail if the bulk validator accepts surrogates, and if the SWAR
word test stops detecting control characters.

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

* Format the new string fast path test with astyle

The pinned astyle expands a braced-init-list used as a range-for range onto
several lines; hoist the two offsets into a named vector instead, which reads
better and leaves nothing for astyle to reformat.

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

* Fix shadowed locals and guard the exception-dependent tests

Two problems in the tests added for the bulk scanners, both found by CI:

- the inner `const json j` in the counted-iterator diagnostics shadowed the
  one declared at test-case scope, which -Wshadow rejects on GCC and clang
  and C4456 rejects on MSVC under /WX; rename them
- the new parity checks parse deliberately invalid input, which calls
  std::abort() rather than throwing when JSON_NOEXCEPTION is defined, so
  they would have crashed the no-exception build; guard them the way the
  other tests do

json::accept() does not abort, so the UTF-8 range assertions stay compiled
without exceptions and keep covering validate_one_utf8() there.

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

* Gate the C++20 iterator classification on JSON_HAS_RANGES

Making supports_bulk_scan depend on iterator_is_contiguous meant the trait
is now instantiated for every adapter, not only when get_elements() is
called. On standard libraries with an incomplete <ranges> that is fatal:
libstdc++ 10 evaluates std::contiguous_iterator<std::counted_iterator<T*>>
by calling std::to_address, which needs an operator-> its counted_iterator
does not have, so satisfaction checking is a hard error rather than false.
Reported by clang 14 + libstdc++ 10.

JSON_HAS_RANGES already encodes exactly this ("libstdc++ < 11 has incomplete
C++20 ranges", #4440), so require it for the C++20 branch. Affected
toolchains fall back to the pointer-only test and the byte-at-a-time
scanner, which parses identically, just without the bulk fast paths.

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

* Amalgamate source code

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

* Satisfy clang-tidy in the new bulk scanner tests

- give the helper lambdas an explicit std::string return type and return
  braced initializer lists (modernize-return-braced-init-list)
- replace the C-style array of test cases with a std::vector
  (modernize-avoid-c-arrays)
- silence pro-type-member-init on the two brace-initialized aggregates;
  default member initializers would stop them being aggregates in C++11

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

* Avoid escaped literals in the counted-iterator diagnostics list

clang-tidy reads "[\"\\ud834\"]" as a literal better written raw, and the
two literals written next to each other in "[\"a\x01""b\"]" as a missing
comma. The concatenation was there to stop the hex escape swallowing the
following character; build those documents from explicit bytes instead and
use raw strings elsewhere. The byte sequences are unchanged.

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

* Reattach convert_number's documentation

@gregmarr spotted that convert_integer() was inserted between convert_number()
and its doc block, leaving convert_integer() with two stacked blocks and
convert_number() with none. Comment only; no code change.

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

* Address review comments on the simdutf backend

string_bulk_run() had the same `return scalar_string_bulk_run(...)` in both
arms of the `#if`. The simdutf arm already falls through when validation
fails, so a single return after the `#endif` says the same thing.

The JSON_USE_SIMDUTF example showed `#include <simdutf.h>`, which
string_scan.hpp already does under the same guard; users only have to put the
header on the include path and link the library, not include it themselves.

Set the version history entry to 3.13.0, matching the other macro pages
documenting unreleased features.

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Pin the number error position against a non-newline terminator

The comment claimed the reported column is the one the offending token starts
at. It is the column reached after the token's last character - which is the
actual point of the unget() change: a number terminated by a newline now
reports what the same number terminated by a space always did.

Assert that equality directly, and add a multi-character token where the start
and end columns differ, so the invariant cannot be read off a single-character
example.

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Do not repeat the integer conversion that just failed

scan_number_bulk_contiguous() converts an integer token straight from the input
buffer. When the value does not fit, it materializes token_buffer and calls
convert_number(), which tried the very same integer conversion again before
falling back to floating point.

Recording the outcome in number_type skips the second attempt. The resulting
token type and value are unchanged: convert_number() reached the float tail
either way.

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Require a container's value_type to match what data() points at

is_contiguous_byte_container accepted any type with a data() returning a
pointer to a single-byte integral plus a size(). That is duck typing: the two
members say nothing about size() counting the units data() points at.

A type where it does not - fixed-size records, say - was routed to the
pointer-based adapter and parsed as [data(), data() + size()) bytes, silently
truncating input the iterator-based adapter had read in full:

    struct record_buffer {
        using value_type = std::array<char, 4>;
        std::string bytes;
        const char* data() const;                        // raw bytes
        std::size_t size() const;                        // in records
        const char* begin() const; const char* end() const;
    };
    json::parse(record_buffer{"[1,2,3,4,5]"});           // parse error at column 3

Requiring the container's own value_type to be that same element type ties the
two together. Every contiguous standard container satisfies it, so std::string,
std::vector<char>, std::array<char, N> and std::string_view keep the fast path;
anything else falls back to the iterator-based adapter, which is always correct.

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Only compile the simdutf backend from C++17 on

simdutf.h rejects anything below C++17 with an #error, so defining
JSON_USE_SIMDUTF in a C++11 or C++14 translation unit did not fail with a
message about simdutf being unavailable - it failed to compile at all, taking
the library's C++11 support with it. Nothing caught this because no build ever
compiled that path.

Gate the include and both uses on JSON_HAS_CPP_17, the same way number_parse.hpp
gates std::from_chars. Below C++17 the macro now has no effect and the scalar
validator runs; it accepts and rejects exactly the same input, so the macro is
safe to set project-wide even when some translation units use an older standard.

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Test the simdutf backend in CI

JSON_USE_SIMDUTF was documented and shipped but never compiled by anything in
the repository, so nothing held the backend to the behavior the docs promise.

Add JSON_TestSimdutf (OFF by default), which fetches simdutf and defines
JSON_USE_SIMDUTF for every test target, and a ci_test_simdutf target that runs
the whole suite in that configuration. Because simdutf needs C++17, the suite is
built at C++11 as well, so one job covers both the scalar fallback with the
macro defined and simdutf itself.

The dependency hangs off test_main, whose usage requirements every test target
inherits. The library target and the installed CMake package are deliberately
untouched: making nlohmann_json link simdutf would put a find_dependency() in
the exported package, which is a separate decision.

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Warn when JSON_TestSimdutf cannot reach the simdutf backend

simdutf needs C++17: without it the dependency does not even compile, and with
a C++17 compiler but no C++17-or-later standard under test it builds and then
goes unused. Either way the option silently did nothing useful, or broke the
configure step outright.

Resolve the tested standards first, then check them: when none of them can
reach simdutf, skip the dependency and say so, naming which of the two reasons
applies and how to fix it. The tests then run against the scalar validator,
which is what would have happened anyway.

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Use record_buffer::data() so clang does not flag it unneeded

The record_buffer test type declares data() and size() so the
is_contiguous_byte_container trait can see both and still reject the
type on its value_type. data() was never called, so clang's
-Wunneeded-member-function (under -Weverything -Werror) failed the
C++20 build. Assert that data() points at the underlying bytes: it
ODR-uses the member and documents the property the type is meant to
demonstrate.

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XAYM1qhSA2FDaDcGfPW3fG
Signed-off-by: Niels Lohmann <mail@nlohmann.me>

* Skip the float fast path when it cannot succeed

parse_float_fast() (Clinger) needs a significand below 2^53, so it always
declines once the mantissa has 17 or more significant digits. convert_number()
called it unconditionally, so those numbers were walked an extra time before
strtod had to run anyway. On streaming input, where scanning is byte-at-a-time
and there is no compensating win, that made canada.json about 6% slower than
develop.

Derive the significant-digit count from token_buffer indices - the digits are
not scanned again - and skip the call when it is guaranteed to decline. Both
scanners pass the offset where the mantissa ends; the count only has to be
corrected for a leading "0", which the JSON grammar admits nowhere else. The
integer path returns before the check, so integer-heavy input is unaffected.

Values are unchanged: this only avoids an attempt that would have failed.
Verified bit-exact against develop over every number in canada.json,
floats.json, signed_ints.json, unsigned_ints.json, small_signed_ints.json,
citm_catalog.json and twitter.json, for both the contiguous and the streaming
scanner.

  parse, streaming     develop    before     after
  canada.json           19.4ms    20.5ms    19.3ms
  floats.json          135.9ms   131.8ms   128.0ms

  parse, contiguous    develop    before     after
  canada.json           15.5ms    12.9ms    11.7ms
  floats.json           98.6ms    69.8ms    66.7ms

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

---------

Signed-off-by: Niels Lohmann <mail@nlohmann.me>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

.github/workflows/ubuntu.yml

@@ -100,7 +100,7 @@ jobs:
     container: ubuntu:focal
     strategy:
       matrix:
-        target: [ci_cmake_flags, ci_test_diagnostics, ci_test_diagnostic_positions, ci_test_noexceptions, ci_test_noimplicitconversions, ci_test_legacycomparison, ci_test_noglobaludls]
+        target: [ci_cmake_flags, ci_test_diagnostics, ci_test_diagnostic_positions, ci_test_noexceptions, ci_test_noimplicitconversions, ci_test_legacycomparison, ci_test_noglobaludls, ci_test_simdutf]
     steps:
       - name: Install build-essential
         run: apt-get update ; apt-get install -y build-essential unzip wget git libssl-dev

cmake/ci.cmake

@@ -212,6 +212,24 @@ add_custom_target(ci_test_legacycomparison
     COMMENT "Compile and test with legacy discarded value comparison enabled"
 )
 
+###############################################################################
+# Validate UTF-8 with simdutf.
+###############################################################################
+
+add_custom_target(ci_test_simdutf
+    COMMAND ${CMAKE_COMMAND}
+    -DCMAKE_BUILD_TYPE=Debug -GNinja
+    -DJSON_BuildTests=ON -DJSON_TestSimdutf=ON
+    # simdutf needs C++17, so the library falls back to its scalar validator
+    # below that: build the suite at C++11 to cover the fallback with the macro
+    # defined, and at C++17 to run every test against simdutf itself
+    "-DJSON_TestStandards=11\;17"
+    -S${PROJECT_SOURCE_DIR} -B${PROJECT_BINARY_DIR}/build_simdutf
+    COMMAND ${CMAKE_COMMAND} --build ${PROJECT_BINARY_DIR}/build_simdutf
+    COMMAND cd ${PROJECT_BINARY_DIR}/build_simdutf && ${CMAKE_CTEST_COMMAND} --parallel ${N} --output-on-failure
+    COMMENT "Compile and test with simdutf UTF-8 validation enabled"
+)
+
 ###############################################################################
 # Enable brace-init copy semantics.
 ###############################################################################

docs/mkdocs/docs/api/macros/index.md

@@ -24,6 +24,7 @@ header. See also the [macro overview page](../../features/macros.md).
 - [**JSON_NO_IO**](json_no_io.md) - switch off functions relying on certain C++ I/O headers
 - [**JSON_SKIP_UNSUPPORTED_COMPILER_CHECK**](json_skip_unsupported_compiler_check.md) - do not warn about unsupported compilers
 - [**JSON_USE_GLOBAL_UDLS**](json_use_global_udls.md) - place user-defined string literals (UDLs) into the global namespace
+- [**JSON_USE_SIMDUTF**](json_use_simdutf.md) - use the simdutf library to accelerate UTF-8 validation
 
 ## Library version
 

docs/mkdocs/docs/api/macros/json_use_simdutf.md

@@ -0,0 +1,71 @@
+# JSON_USE_SIMDUTF
+
+```cpp
+#define JSON_USE_SIMDUTF
+```
+
+When defined, the parser validates the UTF-8 content of JSON strings that come from a **contiguous byte input**
+(`std::string`, `std::vector<char>`/`<std::uint8_t>`, string literals, `const char*` ranges, …) using the
+[simdutf](https://github.com/simdutf/simdutf) library instead of the built-in scalar validator. On text with many
+non-ASCII characters (e.g. CJK or emoji) this can validate several times faster.
+
+This is an **opt-in external dependency**. The library itself remains header-only and its behavior is unchanged: the
+same input is accepted or rejected either way, and every parse error is reported at the same position with the same
+message (simdutf is only used to fast-path *valid* runs; anything it flags falls back to the scalar path so the exact
+diagnostic is preserved). Streaming inputs (files, `std::istream`, wide strings, user-defined adapters) always use the
+scalar path.
+
+When `JSON_USE_SIMDUTF` is defined you must make the `simdutf.h` header available on the include path and link the
+simdutf library. When it is not defined, no simdutf header is included and there is no dependency.
+
+!!! note "Requires C++17"
+
+    simdutf requires C++17 and its header rejects older standards with an `#!cpp #error`. The backend is therefore only
+    compiled in from C++17 on. In C++11 and C++14 the macro has no effect and the scalar validator is used, which
+    accepts and rejects exactly the same input -- only throughput differs. Setting the macro project-wide is therefore
+    safe even when some translation units are built with an older standard.
+
+!!! warning "Define consistently"
+
+    The macro selects between two definitions of the same inline validation function. It must therefore be defined
+    identically for **every** translation unit that includes the library; mixing translation units that define it with
+    ones that do not is an ODR violation. Prefer setting it as a compile definition on the target rather than with
+    `#!cpp #define` in individual source files.
+
+## Default definition
+
+By default, `#!cpp JSON_USE_SIMDUTF` is not defined and the portable C++11 scalar validator is used.
+
+```cpp
+#undef JSON_USE_SIMDUTF
+```
+
+## Examples
+
+??? example
+
+    The code below enables the simdutf backend for UTF-8 validation.
+
+    ```cpp
+    #define JSON_USE_SIMDUTF 1
+    #include <nlohmann/json.hpp>
+
+    ...
+    ```
+
+    The project must also link against simdutf, e.g. with CMake:
+
+    ```cmake
+    target_compile_definitions(your_target PRIVATE JSON_USE_SIMDUTF)
+    target_link_libraries(your_target PRIVATE simdutf::simdutf)
+    ```
+
+!!! hint "Testing this configuration"
+
+    The unit tests can be built against the simdutf backend with the CMake option `JSON_TestSimdutf` (`OFF` by
+    default), which fetches simdutf and defines `JSON_USE_SIMDUTF` for every test target. The `ci_test_simdutf` target
+    runs the whole test suite in that configuration.
+
+## Version history
+
+- Added in version 3.13.0.

docs/mkdocs/docs/features/macros.md

@@ -137,6 +137,14 @@ behavior is deprecated and switched off (`0`) by default.
 
 See [full documentation of `JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON`](../api/macros/json_use_legacy_discarded_value_comparison.md).
 
+## `JSON_USE_SIMDUTF`
+
+When defined, UTF-8 validation of JSON strings read from contiguous byte input is delegated to the
+[simdutf](https://github.com/simdutf/simdutf) library instead of the built-in scalar validator. This is an opt-in
+external dependency and is not defined by default.
+
+See [full documentation of `JSON_USE_SIMDUTF`](../api/macros/json_use_simdutf.md).
+
 ## `NLOHMANN_DEFINE_TYPE_*(...)`, `NLOHMANN_DEFINE_DERIVED_TYPE_*(...)`
 
 The library defines 12 macros to simplify the serialization/deserialization of types. See the page on

docs/mkdocs/mkdocs.yml

@@ -296,6 +296,7 @@ nav:
           - 'JSON_USE_GLOBAL_UDLS': api/macros/json_use_global_udls.md
           - 'JSON_USE_IMPLICIT_CONVERSIONS': api/macros/json_use_implicit_conversions.md
           - 'JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON': api/macros/json_use_legacy_discarded_value_comparison.md
+          - 'JSON_USE_SIMDUTF': api/macros/json_use_simdutf.md
           - 'NLOHMANN_DEFINE_DERIVED_TYPE_INTRUSIVE, NLOHMANN_DEFINE_DERIVED_TYPE_INTRUSIVE_WITH_DEFAULT, NLOHMANN_DEFINE_DERIVED_TYPE_INTRUSIVE_ONLY_SERIALIZE, NLOHMANN_DEFINE_DERIVED_TYPE_NON_INTRUSIVE, NLOHMANN_DEFINE_DERIVED_TYPE_NON_INTRUSIVE_WITH_DEFAULT, NLOHMANN_DEFINE_DERIVED_TYPE_NON_INTRUSIVE_ONLY_SERIALIZE': api/macros/nlohmann_define_derived_type.md
           - 'NLOHMANN_DEFINE_TYPE_INTRUSIVE, NLOHMANN_DEFINE_TYPE_INTRUSIVE_WITH_DEFAULT, NLOHMANN_DEFINE_TYPE_INTRUSIVE_ONLY_SERIALIZE': api/macros/nlohmann_define_type_intrusive.md
           - 'NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE, NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE_WITH_DEFAULT, NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE_ONLY_SERIALIZE': api/macros/nlohmann_define_type_non_intrusive.md

include/nlohmann/detail/input/input_adapters.hpp

@@ -155,11 +155,31 @@ class input_stream_adapter
 
 // General-purpose iterator-based adapter. It might not be as fast as
 // theoretically possible for some containers, but it is extremely versatile.
-// SentinelType defaults to IteratorType for backward compatibility, but may
-// be a different type (e.g., a C++20 sentinel or counted_iterator).
+// SentinelType defaults to IteratorType for backward compatibility, but may be
+// a different type, e.g. a C++20 sentinel such as std::default_sentinel_t when
+// IteratorType is a std::counted_iterator.
 template<typename IteratorType, typename SentinelType = IteratorType>
 class iterator_input_adapter
 {
+    // Whether the number of elements between two positions can be computed in
+    // O(1): either the iterator and the sentinel have the same type (plain
+    // std::distance) or, in C++20, the sentinel is a sized sentinel for the
+    // iterator (std::ranges::distance), e.g. std::default_sentinel_t paired
+    // with std::counted_iterator.
+    //
+    // JSON_HAS_RANGES gates the C++20 branch: on standard libraries with an
+    // incomplete <ranges> (libstdc++ < 11, see #4440) evaluating
+    // std::contiguous_iterator on a std::counted_iterator is a hard error
+    // instead of yielding false, and these traits are instantiated for every
+    // adapter. Such toolchains fall back to the pointer-only test and simply
+    // use the byte-at-a-time scanner.
+    static constexpr bool sentinel_is_sized =
+#if JSON_HAS_RANGES && defined(__cpp_lib_concepts) && defined(JSON_HAS_CPP_20)
+        std::is_same<IteratorType, SentinelType>::value || std::sized_sentinel_for<SentinelType, IteratorType>;
+#else
+        std::is_same<IteratorType, SentinelType>::value;
+#endif
+
   public:
     using char_type = typename std::iterator_traits<IteratorType>::value_type;
 
@@ -171,7 +191,7 @@ class iterator_input_adapter
     // in wide_string_input_adapter, which does not expose this).
     static constexpr bool supports_seek =
         std::is_same<typename std::iterator_traits<IteratorType>::iterator_category, std::random_access_iterator_tag>::value
-        && std::is_same<IteratorType, SentinelType>::value
+        && sentinel_is_sized
         && sizeof(char_type) == 1;
 
     iterator_input_adapter(IteratorType first, SentinelType last)
@@ -219,30 +239,60 @@ class iterator_input_adapter
   private:
     // whether IteratorType refers to a contiguous range and therefore supports
     // a std::memcpy fast path (pointers always do; in C++20 we can also detect
-    // library iterators such as those of std::vector and std::string).
-    // Computing the available element count needs either same-type iterators
-    // (plain std::distance) or, in C++20, a sized sentinel (std::ranges::distance),
-    // e.g. std::counted_iterator paired with std::default_sentinel_t.
-    static constexpr bool iterator_is_contiguous =
-#if defined(__cpp_lib_concepts) && defined(JSON_HAS_CPP_20)
-        (std::is_same<IteratorType, SentinelType>::value || std::sized_sentinel_for<SentinelType, IteratorType>)
-        && (std::contiguous_iterator<IteratorType> || std::is_pointer<IteratorType>::value);
+    // library iterators such as those of std::vector and std::string). The
+    // available element count must also be computable in O(1), hence
+    // sentinel_is_sized.
+    static constexpr bool iterator_is_contiguous = sentinel_is_sized &&
+#if JSON_HAS_RANGES && defined(__cpp_lib_concepts) && defined(JSON_HAS_CPP_20)
+        (std::contiguous_iterator<IteratorType> || std::is_pointer<IteratorType>::value);
 #else
-        std::is_same<IteratorType, SentinelType>::value && std::is_pointer<IteratorType>::value;
+        std::is_pointer<IteratorType>::value;
 #endif
 
+    // number of unread elements in [current, end)
+    std::size_t remaining_count() const
+    {
+#if JSON_HAS_RANGES && defined(__cpp_lib_concepts) && defined(JSON_HAS_CPP_20)
+        // std::ranges::distance also supports sized sentinels of a different
+        // type (e.g. std::counted_iterator + std::default_sentinel_t)
+        return static_cast<std::size_t>(std::ranges::distance(current, end));
+#else
+        return static_cast<std::size_t>(std::distance(current, end));
+#endif
+    }
+
+  public:
+    // Whether the remaining input is a single contiguous block of 1-byte
+    // elements that the lexer can inspect directly (used for the SWAR string
+    // fast path).
+    static constexpr bool supports_bulk_scan =
+        iterator_is_contiguous && sizeof(char_type) == 1;
+
+    // Pointer to the next unread element; only valid when bulk_remaining() > 0.
+    const char_type* bulk_data() const
+    {
+        return &*current;
+    }
+
+    // Number of unread elements available as one contiguous block.
+    std::size_t bulk_remaining() const
+    {
+        return remaining_count();
+    }
+
+    // Consume @a n elements previously inspected via bulk_data().
+    void bulk_skip(std::size_t n)
+    {
+        std::advance(current, static_cast<typename std::iterator_traits<IteratorType>::difference_type>(n));
+    }
+
+  private:
     // contiguous fast path: bulk copy the remaining range with std::memcpy
     template<class T>
     std::size_t get_elements_impl(T* dest, std::size_t count, std::true_type /*contiguous*/)
     {
         const std::size_t wanted = count * sizeof(T);
-#if defined(__cpp_lib_concepts) && defined(JSON_HAS_CPP_20)
-        // std::ranges::distance also supports sized sentinels of a different
-        // type (e.g. std::counted_iterator + std::default_sentinel_t)
-        const std::size_t available = static_cast<std::size_t>(std::ranges::distance(current, end)) * sizeof(char_type);
-#else
-        const std::size_t available = static_cast<std::size_t>(std::distance(current, end)) * sizeof(char_type);
-#endif
+        const std::size_t available = remaining_count() * sizeof(char_type);
         const std::size_t copied = (std::min)(wanted, available);
         if (JSON_HEDLEY_LIKELY(copied != 0))
         {
@@ -570,6 +620,46 @@ typename iterator_input_adapter_factory<IteratorType, SentinelType>::adapter_typ
     return factory_type::create(first, last);
 }
 
+// The element type a container's data() points at, cv-qualifiers removed.
+// Ill-formed - and therefore SFINAE-friendly - for types without data().
+template<typename ContainerType>
+using container_data_t = typename std::remove_cv<typename std::remove_pointer <
+                         decltype(std::declval<const ContainerType&>().data()) >::type >::type;
+
+// The container's own element type, cv-qualifiers removed. It is looked up on
+// the bare type so it is also found when ContainerType is deduced as a
+// reference by the forwarding-reference overload below.
+template<typename ContainerType>
+using container_value_t = typename std::remove_cv <
+                          typename std::remove_cv<typename std::remove_reference<ContainerType>::type>::type::value_type >::type;
+
+// Detect a container that stores its elements contiguously as single bytes
+// (std::string, std::vector<char/unsigned char>, std::array<char, N>,
+// std::string_view, ...). Such inputs are wrapped in a pointer-based adapter so
+// they benefit from the contiguous fast paths (bulk string scanning, memcpy for
+// binary formats) in every C++ standard - not only in C++20, where the standard
+// library iterators model std::contiguous_iterator and are detected directly.
+//
+// data() and size() on their own would be duck typing: they say nothing about
+// size() counting the units data() points at, and reading [data(), data() +
+// size()) as bytes would be wrong for a type where it does not. Requiring the
+// container's own value_type to be that same single-byte element ties the two
+// together; every contiguous standard container satisfies it. Anything else
+// keeps the iterator-based adapter, which is always correct - only slower.
+template<typename ContainerType, typename = void>
+struct is_contiguous_byte_container : std::false_type {};
+
+template<typename ContainerType>
+struct is_contiguous_byte_container < ContainerType, void_t <
+    container_data_t<ContainerType>,
+    container_value_t<ContainerType>,
+decltype(std::declval<const ContainerType&>().size()) >>
+            : std::integral_constant < bool,
+        std::is_pointer<decltype(std::declval<const ContainerType&>().data())>::value&&
+        std::is_integral<container_data_t<ContainerType>>::value&&
+        sizeof(container_data_t<ContainerType>) == 1 &&
+        std::is_same<container_data_t<ContainerType>, container_value_t<ContainerType>>::value > {};
+
 // Convenience shorthand from container to iterator
 // Enables ADL on begin(container) and end(container)
 // Encloses the using declarations in namespace for not to leak them to outside scope
@@ -597,12 +687,32 @@ struct container_input_adapter_factory< ContainerType,
 
 }  // namespace container_input_adapter_factory_impl
 
-template<typename ContainerType>
-typename container_input_adapter_factory_impl::container_input_adapter_factory<ContainerType>::adapter_type input_adapter(ContainerType&& container)
+// General container path (iterator-based). Contiguous single-byte containers
+// are excluded here and routed through the pointer-based overload below.
+template < typename ContainerType,
+           enable_if_t < !is_contiguous_byte_container<ContainerType>::value, int > = 0 >
+typename container_input_adapter_factory_impl::container_input_adapter_factory<ContainerType>::adapter_type input_adapter(ContainerType && container)
 {
     return container_input_adapter_factory_impl::container_input_adapter_factory<ContainerType>::create(std::forward<ContainerType>(container));
 }
 
+// Contiguous single-byte containers (std::string, std::vector<char>, ...) are
+// wrapped in a pointer-based adapter so the contiguous fast paths apply in every
+// standard. The pointer keeps the container's own element type (const char* for
+// std::string, const std::uint8_t* for std::vector<std::uint8_t>, ...), so the
+// resulting char_type - and therefore the parsing behavior - is byte-for-byte
+// identical to the iterator-based path; only the raw pointer additionally
+// enables the bulk fast paths. The container outlives the adapter for the whole
+// parse (temporaries live until the end of the full expression), exactly as the
+// iterators it replaces did.
+template < typename ContainerType,
+           enable_if_t < is_contiguous_byte_container<ContainerType>::value, int > = 0 >
+auto input_adapter(const ContainerType& container)
+-> decltype(input_adapter(container.data(), container.data() + container.size()))
+{
+    return input_adapter(container.data(), container.data() + container.size());
+}
+
 // specialization for std::string
 using string_input_adapter_type = decltype(input_adapter(std::declval<std::string>()));
 

include/nlohmann/detail/input/lexer.hpp

@@ -19,7 +19,9 @@
 #include <vector> // vector
 
 #include <nlohmann/detail/input/input_adapters.hpp>
+#include <nlohmann/detail/input/number_parse.hpp>
 #include <nlohmann/detail/input/position_t.hpp>
+#include <nlohmann/detail/input/string_scan.hpp>
 #include <nlohmann/detail/macro_scope.hpp>
 #include <nlohmann/detail/meta/type_traits.hpp>
 
@@ -125,6 +127,25 @@ constexpr bool input_adapter_supports_seek(std::false_type /*detected*/)
     return false;
 }
 
+// Detect whether an input adapter exposes a contiguous byte block that the
+// lexer can scan directly (see iterator_input_adapter::supports_bulk_scan).
+// Adapters without the flag - file, stream, wide-string, user-defined - fall
+// back to the character-at-a-time string scanner.
+template<typename InputAdapterType>
+using detect_supports_bulk_scan = decltype(InputAdapterType::supports_bulk_scan);
+
+template<typename InputAdapterType>
+constexpr bool input_adapter_supports_bulk_scan(std::true_type /*detected*/)
+{
+    return InputAdapterType::supports_bulk_scan;
+}
+
+template<typename InputAdapterType>
+constexpr bool input_adapter_supports_bulk_scan(std::false_type /*detected*/)
+{
+    return false;
+}
+
 /*!
 @brief lexical analysis
 
@@ -146,6 +167,14 @@ class lexer : public lexer_base<BasicJsonType>
     static constexpr bool lazy_token_string =
         input_adapter_supports_seek<InputAdapterType>(is_detected<detect_supports_seek, InputAdapterType> {});
 
+    /// whether string scanning may bulk-consume runs of ordinary characters
+    /// directly from a contiguous input buffer (SWAR fast path). This requires
+    /// the token to be reconstructible lazily (lazy_token_string), so bypassing
+    /// the per-character capture in get() cannot lose error diagnostics.
+    static constexpr bool bulk_scan =
+        lazy_token_string
+        && input_adapter_supports_bulk_scan<InputAdapterType>(is_detected<detect_supports_bulk_scan, InputAdapterType> {});
+
   public:
     using token_type = typename lexer_base<BasicJsonType>::token_type;
 
@@ -266,6 +295,40 @@ class lexer : public lexer_base<BasicJsonType>
         return true;
     }
 
+    /// contiguous input: bulk-append the run of ordinary characters and complete
+    /// well-formed UTF-8 sequences starting at the current read position, leaving
+    /// the first byte that needs individual handling (the closing quote, an
+    /// escape, a control character, or an ill-formed UTF-8 byte) for get()
+    void scan_string_bulk(std::true_type /*bulk*/)
+    {
+        // a pending unget must be consumed through the normal path first
+        if (next_unget)
+        {
+            return;
+        }
+        const std::size_t remaining = ia.bulk_remaining();
+        if (remaining == 0)
+        {
+            return;
+        }
+        const auto* const data = reinterpret_cast<const unsigned char*>(ia.bulk_data());
+
+        const std::size_t pos = string_bulk_run(data, remaining);
+        if (pos == 0)
+        {
+            return;
+        }
+        token_buffer.append(reinterpret_cast<const typename string_t::value_type*>(data), pos);
+        ia.bulk_skip(pos);
+        // the run contains no newline (all bytes < 0x20 are treated as special),
+        // so only the flat character counters advance
+        position.chars_read_total += pos;
+        position.chars_read_current_line += pos;
+    }
+
+    /// streaming input: no bulk fast path
+    void scan_string_bulk(std::false_type /*bulk*/) const noexcept {}
+
     /*!
     @brief scan a string literal
 
@@ -291,6 +354,10 @@ class lexer : public lexer_base<BasicJsonType>
 
         while (true)
         {
+            // bulk-consume ordinary characters from contiguous input, then
+            // handle the next special byte through the switch below
+            scan_string_bulk(std::integral_constant<bool, bulk_scan> {});
+
             // get the next character
             switch (get())
             {
@@ -1009,6 +1076,12 @@ class lexer : public lexer_base<BasicJsonType>
         // changed if minus sign, decimal point, or exponent is read
         token_type number_type = token_type::value_unsigned;
 
+        // offset just past the last mantissa byte in token_buffer (i.e. the
+        // index of 'e'/'E', or the whole token when there is no exponent).
+        // convert_number() uses it to count significant digits; npos means
+        // "not seen an exponent yet" and is resolved at scan_number_done
+        std::size_t mantissa_end = std::string::npos;
+
         // state (init): we just found out we need to scan a number
         switch (current)
         {
@@ -1194,6 +1267,9 @@ scan_number_decimal2:
 scan_number_exponent:
         // we just parsed an exponent
         number_type = token_type::value_float;
+        // this label is reached only right after the 'e'/'E' was appended (from
+        // the zero, any1, and decimal2 states), so the mantissa ends before it
+        mantissa_end = token_buffer.size() - 1;
         switch (get())
         {
             case '+':
@@ -1280,6 +1356,116 @@ scan_number_done:
         // we are done scanning a number)
         unget();
 
+        // no exponent was scanned: the mantissa spans the whole token
+        if (mantissa_end == std::string::npos)
+        {
+            mantissa_end = token_buffer.size();
+        }
+
+        return convert_number(number_type, mantissa_end);
+    }
+
+    /*!
+    @brief convert an already-validated integer token to its value
+
+    The digit sequence in [first, last) has been validated by the caller, so a
+    dedicated parser can avoid the locale/errno overhead of std::strtoull.
+
+    @return the token type on success; token_type::uninitialized if @a
+            number_type is not an integer type or the value does not fit, in
+            which case the caller falls back to the floating-point conversion
+            (matching the previous std::strtoull/std::strtoll behavior)
+    */
+    token_type convert_integer(token_type number_type, const char* first, const char* last)
+    {
+        if (number_type == token_type::value_unsigned)
+        {
+            if (parse_integer_unsigned(first, last, value_unsigned))
+            {
+                return token_type::value_unsigned;
+            }
+        }
+        else if (number_type == token_type::value_integer)
+        {
+            if (parse_integer_signed(first, last, value_integer))
+            {
+                return token_type::value_integer;
+            }
+        }
+
+        return token_type::uninitialized;
+    }
+
+    /*!
+    @brief check whether Clinger's fast path can still succeed for this token
+
+    parse_float_fast() needs a significand below 2^53. A mantissa with 17 or
+    more significant digits is at least 10^16 and therefore always exceeds it,
+    so calling the fast path would walk the token one extra time only to
+    decline before strtod has to run anyway.
+
+    Significant digits are the mantissa's digits from the first nonzero one on;
+    the sign, the decimal point, leading zeros, and the exponent do not count.
+    The answer is derived from indices - the digits are not scanned again - so
+    this stays off the hot path of the number scanners.
+
+    @param[in] mantissa_end  offset just past the last mantissa byte in
+                             token_buffer
+    @return false if parse_float_fast() is guaranteed to decline
+    */
+    bool mantissa_fits_clinger(std::size_t mantissa_end) const
+    {
+        // 10^16 already exceeds 2^53, so 17 digits can never fit
+        constexpr std::size_t limit = 17;
+
+        const std::size_t neg = (!token_buffer.empty() && token_buffer[0] == '-') ? 1u : 0u;
+        const std::size_t has_dot = (decimal_point_position != std::string::npos) ? 1u : 0u;
+        // the JSON grammar restricts the integer part to "0" or [1-9][0-9]*, so
+        // a leading zero can only be a lone "0", which is not significant
+        const std::size_t lead_zero = (token_buffer[neg] == '0') ? 1u : 0u;
+        JSON_ASSERT(mantissa_end >= neg + has_dot + lead_zero);
+        std::size_t digits = mantissa_end - neg - has_dot - lead_zero;
+
+        if (JSON_HEDLEY_LIKELY(digits < limit))
+        {
+            return true;
+        }
+
+        // Only a number below 1 can carry further insignificant zeros, and only
+        // while the count stays at the limit does removing them change the
+        // answer - so this loop is skipped for all but a few tokens. Note
+        // token_buffer holds the locale's decimal point, so the fraction is
+        // located through decimal_point_position rather than by searching '.'.
+        if (lead_zero != 0)
+        {
+            JSON_ASSERT(has_dot != 0); // an integer "0" cannot reach the limit
+            for (std::size_t i = decimal_point_position + 1;
+                    digits >= limit && i < mantissa_end && token_buffer[i] == '0'; ++i)
+            {
+                --digits;
+            }
+        }
+
+        return digits < limit;
+    }
+
+    /*!
+    @brief convert the number text in token_buffer to its value and token type
+
+    The digit sequence in token_buffer has already been validated (by the
+    scan_number() state machine or by the contiguous fast path) and holds the
+    locale decimal point in place of '.'. Integers are parsed first and fall
+    back to floating point on overflow. This is shared so both scanners produce
+    identical results.
+
+    @param[in] mantissa_end  offset just past the last mantissa byte in
+                             token_buffer (the index of 'e'/'E', or
+                             token_buffer.size() when there is no exponent);
+                             used to skip Clinger's fast path when it cannot
+                             possibly succeed - see mantissa_fits_clinger()
+    */
+    token_type convert_number(token_type number_type, std::size_t mantissa_end)
+    {
         // If the caller does not need the converted value (only whether the
         // input is syntactically valid; see json_sax_acceptor/accept()), an
         // unsigned/integer token can be reported without calling
@@ -1332,45 +1518,37 @@ scan_number_done:
             }
         }
 
-        char* endptr = nullptr; // NOLINT(misc-const-correctness,cppcoreguidelines-pro-type-vararg,hicpp-vararg)
-        errno = 0;
+        const char* const num_begin = token_buffer.data();
+        const char* const num_end = num_begin + token_buffer.size();
 
-        // try to parse integers first and fall back to floats
-        if (number_type == token_type::value_unsigned)
+        if (number_type != token_type::value_float)
         {
-            const auto x = std::strtoull(token_buffer.data(), &endptr, 10);
-
-            // we checked the number format before
-            JSON_ASSERT(endptr == token_buffer.data() + token_buffer.size());
-
-            if (errno != ERANGE)
+            const token_type integer_result = convert_integer(number_type, num_begin, num_end);
+            if (integer_result != token_type::uninitialized)
             {
-                value_unsigned = static_cast<number_unsigned_t>(x);
-                if (value_unsigned == x)
-                {
-                    return token_type::value_unsigned;
-                }
-            }
-        }
-        else if (number_type == token_type::value_integer)
-        {
-            const auto x = std::strtoll(token_buffer.data(), &endptr, 10);
-
-            // we checked the number format before
-            JSON_ASSERT(endptr == token_buffer.data() + token_buffer.size());
-
-            if (errno != ERANGE)
-            {
-                value_integer = static_cast<number_integer_t>(x);
-                if (value_integer == x)
-                {
-                    return token_type::value_integer;
-                }
+                return integer_result;
             }
         }
 
         // this code is reached if we parse a floating-point number or if an
-        // integer conversion above failed
+        // integer conversion above overflowed. Prefer std::from_chars
+        // (Eisel-Lemire, locale-independent, correctly rounded) when available;
+        // otherwise the exact Clinger fast path (double only); otherwise the
+        // locale-aware strtof/strtod.
+        if (parse_float_from_chars(num_begin, num_end, value_float))
+        {
+            return token_type::value_float;
+        }
+        // Skipping a fast path that cannot succeed is lossless and saves a full
+        // extra pass over the token's bytes, which otherwise shows up on
+        // high-precision inputs such as canada.json
+        if (mantissa_fits_clinger(mantissa_end)
+                && parse_float_fast(num_begin, num_end, decimal_point_char, value_float))
+        {
+            return token_type::value_float;
+        }
+
+        char* endptr = nullptr; // NOLINT(misc-const-correctness,cppcoreguidelines-pro-type-vararg,hicpp-vararg)
         strtof(value_float, token_buffer.data(), &endptr);
 
         // we checked the number format before
@@ -1379,6 +1557,158 @@ scan_number_done:
         return token_type::value_float;
     }
 
+    /*!
+    @brief contiguous fast path for scanning a number
+
+    Parses the whole number token straight from the input buffer, avoiding the
+    per-character get()/add() of scan_number(). On success it fills token_buffer
+    (with the locale decimal point substituted, as scan_number() does) and
+    returns the token type. On anything it does not fully recognize as a
+    well-formed number it makes no state change and returns
+    token_type::uninitialized, so the caller falls back to scan_number(), which
+    then produces the exact diagnostic. @a current is the first digit or the
+    leading minus (already read); the remaining bytes are taken from the adapter.
+    */
+    token_type scan_number_bulk_contiguous()
+    {
+        // a pending unget offsets the buffer position from current; fall back
+        if (next_unget)
+        {
+            return token_type::uninitialized;
+        }
+        const std::size_t rem = ia.bulk_remaining();
+        if (rem == 0)
+        {
+            // the first digit is the last input byte; let scan_number() finish
+            return token_type::uninitialized;
+        }
+        // the byte before the next unread one is current (contiguous input)
+        const char* const data = reinterpret_cast<const char*>(ia.bulk_data()) - 1;
+        const std::size_t avail = rem + 1;
+
+        // validate + classify the number extent (mirrors scan_number()'s grammar)
+        std::size_t i = 0;
+        std::size_t dot_index = std::string::npos;
+        token_type number_type = token_type::value_unsigned;
+        if (data[0] == '-')
+        {
+            number_type = token_type::value_integer;
+            i = 1;
+            if (i >= avail)
+            {
+                return token_type::uninitialized;
+            }
+        }
+        if (data[i] == '0')
+        {
+            ++i;
+        }
+        else if (data[i] >= '1' && data[i] <= '9')
+        {
+            ++i;
+            while (i < avail && data[i] >= '0' && data[i] <= '9')
+            {
+                ++i;
+            }
+        }
+        else
+        {
+            return token_type::uninitialized;
+        }
+        if (i < avail && data[i] == '.')
+        {
+            number_type = token_type::value_float;
+            dot_index = i;
+            ++i;
+            if (i >= avail || !(data[i] >= '0' && data[i] <= '9'))
+            {
+                return token_type::uninitialized;
+            }
+            while (i < avail && data[i] >= '0' && data[i] <= '9')
+            {
+                ++i;
+            }
+        }
+        // the mantissa ends here, whether or not an exponent part follows
+        const std::size_t mantissa_end = i;
+        if (i < avail && (data[i] == 'e' || data[i] == 'E'))
+        {
+            number_type = token_type::value_float;
+            ++i;
+            if (i < avail && (data[i] == '+' || data[i] == '-'))
+            {
+                ++i;
+            }
+            if (i >= avail || !(data[i] >= '0' && data[i] <= '9'))
+            {
+                return token_type::uninitialized;
+            }
+            while (i < avail && data[i] >= '0' && data[i] <= '9')
+            {
+                ++i;
+            }
+        }
+        const std::size_t len = i;
+
+        // reset() records where this token starts (for diagnostics), so it has
+        // to run before the input position advances below
+        reset();
+
+        // An integer token needs no token_buffer: the SAX callbacks for
+        // number_integer/number_unsigned take only the value, and the overflow
+        // diagnostic rebuilds the text from the input. Convert straight from the
+        // input buffer and leave token_buffer empty. (JSON_DIAGNOSTIC_POSITIONS
+        // derives a number's start position from get_string().size(), so there
+        // the token still has to be materialized.)
+#if !JSON_DIAGNOSTIC_POSITIONS
+        if (number_type != token_type::value_float)
+        {
+            const token_type integer_result = convert_integer(number_type, data, data + len);
+            if (JSON_HEDLEY_LIKELY(integer_result != token_type::uninitialized))
+            {
+                ia.bulk_skip(len - 1);
+                position.chars_read_total += (len - 1);
+                position.chars_read_current_line += (len - 1);
+                return integer_result;
+            }
+            // The value does not fit an integer, so this token converts as a
+            // float. Recording that here keeps convert_number() below from
+            // repeating the integer attempt that just failed.
+            number_type = token_type::value_float;
+        }
+#endif
+
+        // materialize the token exactly as scan_number() would, substituting the
+        // locale decimal point so convert_number()'s strtof fallback stays valid.
+        // reset() already cleared token_buffer, so append() fills it (assign() is
+        // avoided because custom string_t types need not provide it)
+        token_buffer.append(reinterpret_cast<const typename string_t::value_type*>(data), len);
+        if (dot_index != std::string::npos)
+        {
+            token_buffer[dot_index] = static_cast<typename string_t::value_type>(decimal_point_char);
+            decimal_point_position = dot_index;
+        }
+
+        ia.bulk_skip(len - 1);
+        position.chars_read_total += (len - 1);
+        position.chars_read_current_line += (len - 1);
+
+        return convert_number(number_type, mantissa_end);
+    }
+
+    /// contiguous input: try the number fast path, else the byte-path scanner
+    token_type scan_number_dispatch(std::true_type /*bulk*/)
+    {
+        const token_type t = scan_number_bulk_contiguous();
+        return (t != token_type::uninitialized) ? t : scan_number();
+    }
+
+    /// streaming input: always use the byte-path scanner
+    token_type scan_number_dispatch(std::false_type /*bulk*/)
+    {
+        return scan_number();
+    }
+
     /*!
     @param[in] literal_text  the literal text to expect
     @param[in] length        the length of the passed literal text
@@ -1482,6 +1812,9 @@ scan_number_done:
         if (current == '\n')
         {
             ++position.lines_read;
+            // remember the column the newline was read at: chars_read_current_line
+            // is about to be cleared, and a matching unget() cannot reconstruct it
+            chars_read_before_newline = position.chars_read_current_line;
             position.chars_read_current_line = 0;
         }
 
@@ -1538,12 +1871,20 @@ scan_number_done:
         --position.chars_read_total;
 
         // in case we "unget" a newline, we have to also decrement the lines_read
+        // and restore the column that get() cleared when it saw the newline;
+        // chars_read_current_line == 0 can only mean the last get() read one
         if (position.chars_read_current_line == 0)
         {
             if (position.lines_read > 0)
             {
                 --position.lines_read;
             }
+
+            // chars_read_before_newline counts the newline itself, which is the
+            // character being ungotten, hence the -1
+            position.chars_read_current_line = (chars_read_before_newline > 0)
+                                               ? chars_read_before_newline - 1
+                                               : 0;
         }
         else
         {
@@ -1810,7 +2151,7 @@ scan_number_done:
             case '7':
             case '8':
             case '9':
-                return scan_number();
+                return scan_number_dispatch(std::integral_constant<bool, bulk_scan> {});
 
             // end of input (the null byte is needed when parsing from
             // string literals)
@@ -1841,6 +2182,10 @@ scan_number_done:
     /// the start position of the current token
     position_t position {};
 
+    /// the value chars_read_current_line had when the last newline was read, so
+    /// that unget() can restore the column instead of leaving it at 0
+    std::size_t chars_read_before_newline = 0;
+
     /// raw input token string for error messages; only populated for streaming
     /// adapters (seekable adapters reconstruct it lazily via token_string_start)
     std::vector<char_type> token_string {};

include/nlohmann/detail/input/number_parse.hpp

@@ -0,0 +1,302 @@
+//     __ _____ _____ _____
+//  __|  |   __|     |   | |  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 <cfloat> // FLT_EVAL_METHOD
+#include <cstddef> // size_t
+#include <cstdint> // int64_t, uint64_t
+#include <limits> // numeric_limits
+
+#include <nlohmann/detail/macro_scope.hpp>
+
+// std::from_chars lives in <charconv>, but being in C++17 mode does not
+// guarantee the header exists: GCC 7 sets __cplusplus to C++17 yet ships no
+// <charconv> (added in GCC 8; floating-point support in GCC 11). Guard the
+// include with __has_include so such toolchains fall back to the scalar path.
+#if defined(JSON_HAS_CPP_17) && defined(__has_include)
+    #if __has_include(<charconv>)
+        #include <charconv> // from_chars (only used when __cpp_lib_to_chars is defined)
+        #include <system_error> // errc
+    #endif
+#endif
+
+// This file contains the value-conversion helpers used by the lexer to turn an
+// already-validated number token into a value, without the locale/errno
+// overhead of std::strtoull/std::strtod. They are free functions so the lexer
+// stays focused on scanning; see lexer::convert_number().
+
+NLOHMANN_JSON_NAMESPACE_BEGIN
+namespace detail
+{
+
+/*!
+@brief fast integer parser for an already-validated unsigned integer
+
+The number scanner has already checked that [first, last) is a valid JSON
+integer, so this only needs to accumulate the digits and detect overflow. This
+avoids the locale/errno machinery of std::strtoull, which dominates
+integer-heavy inputs.
+
+@param[in]  first   pointer to the first character (a digit)
+@param[in]  last    pointer past the last character
+@param[out] value   the parsed value on success
+@return true if the value fit into @a NumberUnsignedType; false on overflow, in
+        which case the caller falls back to floating-point parsing (matching the
+        previous std::strtoull behavior)
+*/
+template<typename NumberUnsignedType>
+bool parse_integer_unsigned(const char* first, const char* last, NumberUnsignedType& value) noexcept
+{
+    // accumulate in the widest unsigned type used by the previous strtoull
+    // path so the overflow behavior is unchanged for custom number types
+    std::uint64_t x = 0;
+    constexpr std::uint64_t cutoff = (std::numeric_limits<std::uint64_t>::max)() / 10u;
+    constexpr std::uint64_t cutlim = (std::numeric_limits<std::uint64_t>::max)() % 10u;
+    for (const char* p = first; p != last; ++p)
+    {
+        const auto digit = static_cast<std::uint64_t>(static_cast<unsigned char>(*p) - static_cast<unsigned char>('0'));
+        if (JSON_HEDLEY_UNLIKELY(x > cutoff || (x == cutoff && digit > cutlim)))
+        {
+            return false;
+        }
+        x = (x * 10u) + digit;
+    }
+    value = static_cast<NumberUnsignedType>(x);
+    // reject values that do not round-trip into a narrower NumberUnsignedType
+    return static_cast<std::uint64_t>(value) == x;
+}
+
+/*!
+@brief fast integer parser for an already-validated negative integer
+
+@param[in]  first   pointer to the leading '-'
+@param[in]  last    pointer past the last character
+@param[out] value   the parsed (negative) value on success
+@return true on success; false on overflow (caller falls back to float)
+*/
+template<typename NumberIntegerType>
+bool parse_integer_signed(const char* first, const char* last, NumberIntegerType& value) noexcept
+{
+    // the state machine only reaches the signed path via a leading '-'
+    JSON_ASSERT(first != last && *first == '-');
+    std::uint64_t magnitude = 0;
+    // |INT64_MIN| == INT64_MAX + 1; this is the largest admissible magnitude
+    constexpr std::uint64_t limit = static_cast<std::uint64_t>((std::numeric_limits<std::int64_t>::max)()) + 1u;
+    for (const char* p = first + 1; p != last; ++p)
+    {
+        const auto digit = static_cast<std::uint64_t>(static_cast<unsigned char>(*p) - static_cast<unsigned char>('0'));
+        if (JSON_HEDLEY_UNLIKELY(magnitude > (limit - digit) / 10u))
+        {
+            return false;
+        }
+        magnitude = (magnitude * 10u) + digit;
+    }
+    const std::int64_t x = (magnitude == limit)
+                           ? (std::numeric_limits<std::int64_t>::min)()
+                           : -static_cast<std::int64_t>(magnitude);
+    value = static_cast<NumberIntegerType>(x);
+    // reject values that do not round-trip into a narrower NumberIntegerType
+    return static_cast<std::int64_t>(value) == x;
+}
+
+/*!
+@brief exact fast path for parsing a `double` (Clinger's algorithm)
+
+For the common case - at most 19 significant digits, a decimal exponent in
+[-22, 22], and a significand below 2^53 - the value equals significand *
+10^exp computed in IEEE-754 double arithmetic, which is exact under
+round-to-nearest because both operands are exactly representable. This is the
+same fast path used by fast_float/simdjson; the general cases are left to
+std::strtod. The parser only activates for number_float_t == double; float and
+long double keep the std::strtof/std::strtold paths (see the templated overload
+below).
+
+@param[in]  first          pointer to the first character of the number
+@param[in]  last           pointer past the last character
+@param[in]  decimal_point  the (locale-dependent) decimal point character
+@param[out] out            the parsed value on success
+@return true if the value was parsed exactly; false to fall back to strtod
+*/
+template<typename DecimalPointType>
+bool parse_float_fast(const char* first, const char* last, DecimalPointType decimal_point, double& out) noexcept
+{
+#if defined(FLT_EVAL_METHOD) && FLT_EVAL_METHOD != 0
+    // Clinger's fast path is only exact when double operations are evaluated in
+    // true double precision. On platforms that keep intermediates in extended
+    // precision (e.g. the x87 FPU on 32-bit x86, where FLT_EVAL_METHOD == 2) the
+    // single significand * 10^scale step is double-rounded and can be 1 ULP off,
+    // so decline and let the caller fall back to the correctly-rounded
+    // std::from_chars / std::strtod path.
+    static_cast<void>(first);
+    static_cast<void>(last);
+    static_cast<void>(decimal_point);
+    static_cast<void>(out);
+    return false;
+#else
+    static const std::array<double, 23> powers_of_ten =
+    {
+        {
+            1e0, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6, 1e7, 1e8, 1e9, 1e10, 1e11,
+            1e12, 1e13, 1e14, 1e15, 1e16, 1e17, 1e18, 1e19, 1e20, 1e21, 1e22
+        }
+    };
+
+    const char* p = first;
+    bool negative = false;
+    if (p != last && (*p == '-' || *p == '+'))
+    {
+        negative = (*p == '-');
+        ++p;
+    }
+
+    std::uint64_t significand = 0;
+    int num_digits = 0;
+    int fractional_digits = 0;
+    bool seen_dot = false;
+    bool any_digit = false;
+    for (; p != last; ++p)
+    {
+        const char c = *p;
+        if (c >= '0' && c <= '9')
+        {
+            any_digit = true;
+            if (JSON_HEDLEY_UNLIKELY(num_digits >= 19))
+            {
+                return false; // significand may not fit into uint64_t
+            }
+            significand = (significand * 10u) + static_cast<std::uint64_t>(c - '0');
+            ++num_digits;
+            fractional_digits += static_cast<int>(seen_dot);
+        }
+        else if (static_cast<DecimalPointType>(c) == decimal_point)
+        {
+            if (JSON_HEDLEY_UNLIKELY(seen_dot))
+            {
+                return false;
+            }
+            seen_dot = true;
+        }
+        else if (c == 'e' || c == 'E')
+        {
+            ++p;
+            break;
+        }
+        else
+        {
+            return false;
+        }
+    }
+    if (JSON_HEDLEY_UNLIKELY(!any_digit))
+    {
+        return false;
+    }
+
+    int exponent = 0;
+    if (p != last) // an exponent part remains
+    {
+        bool exp_negative = false;
+        if (p != last && (*p == '-' || *p == '+'))
+        {
+            exp_negative = (*p == '-');
+            ++p;
+        }
+        bool any_exp_digit = false;
+        for (; p != last; ++p)
+        {
+            if (JSON_HEDLEY_UNLIKELY(*p < '0' || *p > '9'))
+            {
+                return false;
+            }
+            exponent = (exponent * 10) + (*p - '0');
+            any_exp_digit = true;
+            if (JSON_HEDLEY_UNLIKELY(exponent > 9999))
+            {
+                return false;
+            }
+        }
+        if (JSON_HEDLEY_UNLIKELY(!any_exp_digit))
+        {
+            return false;
+        }
+        if (exp_negative)
+        {
+            exponent = -exponent;
+        }
+    }
+
+    const int scale = exponent - fractional_digits;
+    if (JSON_HEDLEY_UNLIKELY(significand >= (static_cast<std::uint64_t>(1) << 53)))
+    {
+        return false; // significand not exactly representable as double
+    }
+
+    auto result = static_cast<double>(significand);
+    if (scale >= 0)
+    {
+        if (JSON_HEDLEY_UNLIKELY(scale > 22))
+        {
+            return false;
+        }
+        result *= powers_of_ten[static_cast<std::size_t>(scale)];
+    }
+    else
+    {
+        if (JSON_HEDLEY_UNLIKELY(-scale > 22))
+        {
+            return false;
+        }
+        result /= powers_of_ten[static_cast<std::size_t>(-scale)];
+    }
+    out = negative ? -result : result;
+    return true;
+#endif
+}
+
+/// fast float path is only exact for `double`; decline for float/long double
+template<typename DecimalPointType, typename FloatType>
+bool parse_float_fast(const char* /*first*/, const char* /*last*/, DecimalPointType /*decimal_point*/, FloatType& /*out*/) noexcept
+{
+    return false;
+}
+
+/*!
+@brief parse a float with std::from_chars (Eisel-Lemire) when available
+
+std::from_chars is locale-independent, correctly rounded, and - via the
+Eisel-Lemire algorithm in modern standard libraries - much faster than strtod
+over the whole value range (not just the Clinger subset). It is used only when
+__cpp_lib_to_chars indicates full floating-point support and only when it
+consumes the entire token ([first, last)); a partial parse means the buffer
+uses a non-'.' locale decimal point, in which case the caller falls back to the
+locale-aware path. An under-/overflow (result_out_of_range) also declines, so
+the caller's strtod fallback supplies the well-defined ±inf/0 result the parser
+expects (side-stepping the P4168 divergence between implementations).
+
+@return true if the value was parsed exactly and fully; false to fall back
+*/
+template<typename FloatType>
+bool parse_float_from_chars(const char* first, const char* last, FloatType& out) noexcept
+{
+    // JSON_HAS_CPP_17 must gate the use as well as the <charconv> include above:
+    // some standard libraries (e.g. libstdc++ 15) define __cpp_lib_to_chars even
+    // in C++14 mode, where <charconv> is not included.
+#if defined(JSON_HAS_CPP_17) && defined(__cpp_lib_to_chars)
+    const auto result = std::from_chars(first, last, out);
+    return result.ec == std::errc() && result.ptr == last;
+#else
+    static_cast<void>(first);
+    static_cast<void>(last);
+    static_cast<void>(out);
+    return false;
+#endif
+}
+
+}  // namespace detail
+NLOHMANN_JSON_NAMESPACE_END

include/nlohmann/detail/input/string_scan.hpp

@@ -0,0 +1,241 @@
+//     __ _____ _____ _____
+//  __|  |   __|     |   | |  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> // uint64_t
+#include <cstring> // memcpy
+
+#include <nlohmann/detail/macro_scope.hpp>
+
+// Optional SIMD backend for bulk UTF-8 validation. This is an opt-in external
+// dependency: nlohmann/json itself stays header-only and the C++11 scalar
+// validator below is always available; defining JSON_USE_SIMDUTF additionally
+// requires the simdutf headers on the include path and linking the simdutf
+// library. See string_bulk_run().
+//
+// simdutf.h itself requires C++17 - it rejects older standards with an #error -
+// so the backend is only compiled in from C++17 on. Below that the macro has no
+// effect and the scalar validator is used; it accepts and rejects exactly the
+// same input, so only throughput differs. macro_scope.hpp is included above to
+// have JSON_HAS_CPP_17 available for this test.
+#if defined(JSON_USE_SIMDUTF) && defined(JSON_HAS_CPP_17)
+    #include <simdutf.h>
+#endif
+
+// This file contains the byte-level string-scanning helpers used by the lexer's
+// contiguous fast path. They operate purely on raw bytes (no dependency on the
+// lexer's template parameters) so they are free functions, keeping the lexer
+// itself focused on the state machine; see lexer::scan_string_bulk().
+
+NLOHMANN_JSON_NAMESPACE_BEGIN
+namespace detail
+{
+
+// classify a single byte as needing individual string handling: the closing
+// quote, an escape, a control character, or a non-ASCII (UTF-8)
+// lead/continuation byte. Ordinary bytes (0x20..0x7F except '"' and '\\') are
+// copied verbatim, which the bulk scanner does 8 bytes at a time.
+inline bool is_string_special(unsigned char c) noexcept
+{
+    return c == '\"' || c == '\\' || c < 0x20u || c >= 0x80u;
+}
+
+// SWAR helper: return a word whose high bit is set in every byte of @a v that
+// is_string_special(); zero if the 8 bytes are all ordinary.
+inline std::uint64_t swar_string_special(std::uint64_t v) noexcept
+{
+    constexpr std::uint64_t ones = 0x0101010101010101ull;
+    constexpr std::uint64_t high = 0x8080808080808080ull;
+    const std::uint64_t q = v ^ 0x2222222222222222ull; // '"'  (0x22)
+    const std::uint64_t b = v ^ 0x5C5C5C5C5C5C5C5Cull; // '\\' (0x5C)
+    const std::uint64_t has_quote     = (q - ones) & ~q & high;
+    const std::uint64_t has_backslash = (b - ones) & ~b & high;
+    const std::uint64_t has_control   = (v - 0x2020202020202020ull) & ~v & high; // < 0x20
+    const std::uint64_t has_non_ascii = v & high;                                // >= 0x80
+    return has_quote | has_backslash | has_control | has_non_ascii;
+}
+
+// return the index of the first is_string_special() byte in [data, data+n), or
+// n if every byte is ordinary; scans 8 bytes at a time
+inline std::size_t find_string_special(const unsigned char* data, std::size_t n) noexcept
+{
+    std::size_t i = 0;
+    for (; i + 8 <= n; i += 8)
+    {
+        std::uint64_t word = 0;
+        std::memcpy(&word, data + i, sizeof(word));
+        if (swar_string_special(word) != 0)
+        {
+            // a special byte is in this word; locate it (endian-agnostic)
+            for (std::size_t j = 0; j < 8; ++j)
+            {
+                if (is_string_special(data[i + j]))
+                {
+                    return i + j;
+                }
+            }
+        }
+    }
+    for (; i < n; ++i)
+    {
+        if (is_string_special(data[i]))
+        {
+            return i;
+        }
+    }
+    return n;
+}
+
+// Validate one UTF-8 sequence at the front of [data, data+avail). Returns its
+// length (2..4) only when the bytes form a *well-formed* sequence using exactly
+// the same ranges as scan_string()'s per-byte switch, so the bulk path accepts
+// precisely what the byte path accepts. Returns 0 for anything that is invalid,
+// incomplete, or that the byte path must diagnose (the caller then defers to
+// that path, keeping error messages unchanged). Lead bytes < 0x80 are handled
+// by the caller and never passed here.
+inline std::size_t validate_one_utf8(const unsigned char* data, std::size_t avail) noexcept
+{
+    const unsigned char c0 = data[0];
+    if (c0 >= 0xC2 && c0 <= 0xDF)  // U+0080..U+07FF
+    {
+        if (avail >= 2 && data[1] >= 0x80 && data[1] <= 0xBF)
+        {
+            return 2;
+        }
+    }
+    else if (c0 == 0xE0)  // U+0800..U+0FFF
+    {
+        if (avail >= 3 && data[1] >= 0xA0 && data[1] <= 0xBF && data[2] >= 0x80 && data[2] <= 0xBF)
+        {
+            return 3;
+        }
+    }
+    else if ((c0 >= 0xE1 && c0 <= 0xEC) || c0 == 0xEE || c0 == 0xEF)  // U+1000..U+CFFF, U+E000..U+FFFF
+    {
+        if (avail >= 3 && data[1] >= 0x80 && data[1] <= 0xBF && data[2] >= 0x80 && data[2] <= 0xBF)
+        {
+            return 3;
+        }
+    }
+    else if (c0 == 0xED)  // U+D000..U+D7FF (excludes surrogates)
+    {
+        if (avail >= 3 && data[1] >= 0x80 && data[1] <= 0x9F && data[2] >= 0x80 && data[2] <= 0xBF)
+        {
+            return 3;
+        }
+    }
+    else if (c0 == 0xF0)  // U+10000..U+3FFFF
+    {
+        if (avail >= 4 && data[1] >= 0x90 && data[1] <= 0xBF && data[2] >= 0x80 && data[2] <= 0xBF && data[3] >= 0x80 && data[3] <= 0xBF)
+        {
+            return 4;
+        }
+    }
+    else if (c0 >= 0xF1 && c0 <= 0xF3)  // U+40000..U+FFFFF
+    {
+        if (avail >= 4 && data[1] >= 0x80 && data[1] <= 0xBF && data[2] >= 0x80 && data[2] <= 0xBF && data[3] >= 0x80 && data[3] <= 0xBF)
+        {
+            return 4;
+        }
+    }
+    else if (c0 == 0xF4)  // U+100000..U+10FFFF
+    {
+        if (avail >= 4 && data[1] >= 0x80 && data[1] <= 0x8F && data[2] >= 0x80 && data[2] <= 0xBF && data[3] >= 0x80 && data[3] <= 0xBF)
+        {
+            return 4;
+        }
+    }
+    return 0; // invalid, incomplete, or must be diagnosed by the byte path
+}
+
+// Scalar (C++11) computation of the bulk run length: the number of leading
+// bytes in [data, data+n) that are ordinary ASCII or complete well-formed UTF-8
+// sequences, stopping before the first byte that needs individual handling (the
+// closing quote, an escape, a control character, or an ill-formed/truncated
+// sequence). ASCII is skipped 8 bytes at a time.
+inline std::size_t scalar_string_bulk_run(const unsigned char* data, std::size_t n) noexcept
+{
+    std::size_t pos = 0;
+    while (pos < n)
+    {
+        pos += find_string_special(data + pos, n - pos);
+        if (pos >= n || data[pos] < 0x80u)
+        {
+            break; // end of buffer, or a quote/escape/control byte
+        }
+        const std::size_t seq = validate_one_utf8(data + pos, n - pos);
+        if (seq == 0)
+        {
+            break; // ill-formed or truncated: let the byte path diagnose it
+        }
+        pos += seq;
+    }
+    return pos;
+}
+
+#if defined(JSON_USE_SIMDUTF) && defined(JSON_HAS_CPP_17)
+// Index of the first quote/escape/control byte in [data, data+n) (non-ASCII
+// bytes are *not* stops here - the whole run is handed to simdutf), or n.
+inline std::size_t find_string_delimiter(const unsigned char* data, std::size_t n) noexcept
+{
+    constexpr std::uint64_t ones = 0x0101010101010101ull;
+    constexpr std::uint64_t high = 0x8080808080808080ull;
+    std::size_t i = 0;
+    for (; i + 8 <= n; i += 8)
+    {
+        std::uint64_t v = 0;
+        std::memcpy(&v, data + i, sizeof(v));
+        const std::uint64_t q = v ^ 0x2222222222222222ull;
+        const std::uint64_t b = v ^ 0x5C5C5C5C5C5C5C5Cull;
+        const std::uint64_t hit = ((q - ones) & ~q & high)
+                                  | ((b - ones) & ~b & high)
+                                  | ((v - 0x2020202020202020ull) & ~v & high);
+        if (hit != 0)
+        {
+            for (std::size_t j = 0; j < 8; ++j)
+            {
+                const unsigned char c = data[i + j];
+                if (c == '\"' || c == '\\' || c < 0x20u)
+                {
+                    return i + j;
+                }
+            }
+        }
+    }
+    for (; i < n; ++i)
+    {
+        const unsigned char c = data[i];
+        if (c == '\"' || c == '\\' || c < 0x20u)
+        {
+            return i;
+        }
+    }
+    return n;
+}
+#endif
+
+// Backend-dispatched bulk run length. With JSON_USE_SIMDUTF the run up to the
+// next delimiter is validated in one shot by simdutf; on the rare failure the
+// scalar helper recomputes the exact valid prefix so the byte path still
+// produces the precise diagnostic. Without it, the pure scalar path is used.
+inline std::size_t string_bulk_run(const unsigned char* data, std::size_t n) noexcept
+{
+#if defined(JSON_USE_SIMDUTF) && defined(JSON_HAS_CPP_17)
+    const std::size_t run = find_string_delimiter(data, n);
+    if (run != 0 && simdutf::validate_utf8(reinterpret_cast<const char*>(data), run))
+    {
+        return run;
+    }
+#endif
+    return scalar_string_bulk_run(data, n);
+}
+
+}  // namespace detail
+NLOHMANN_JSON_NAMESPACE_END

tests/CMakeLists.txt

@@ -2,6 +2,9 @@ cmake_minimum_required(VERSION 3.13...4.0)
 
 option(JSON_Valgrind    "Execute test suite with Valgrind." OFF)
 option(JSON_FastTests   "Skip expensive/slow tests." OFF)
+option(JSON_TestSimdutf "Build the unit tests against the simdutf UTF-8 validation backend." OFF)
+
+set(JSON_SIMDUTF_VERSION 9.1.0 CACHE STRING "The simdutf version used by JSON_TestSimdutf.")
 
 set(JSON_32bitTest     AUTO CACHE STRING "Enable the 32bit unit test (ON/OFF/AUTO/ONLY).")
 set(JSON_TestStandards "" CACHE STRING "The list of standards to test explicitly.")
@@ -194,6 +197,71 @@ if(test_force)
 endif()
 message(STATUS "${msg}")
 
+#############################################################################
+# optionally validate UTF-8 with simdutf (JSON_USE_SIMDUTF)
+#############################################################################
+
+# The simdutf backend is opt-in and not vendored, so it is fetched here rather
+# than being a checked-in dependency. Everything below hangs off test_main,
+# whose usage requirements every test target inherits; the library target and
+# the installed CMake package are deliberately left untouched.
+if (JSON_TestSimdutf)
+    # simdutf requires C++17, both to compile itself and to be reachable from
+    # the library, which keeps its scalar validator below that. Find a tested
+    # standard that satisfies it.
+    set(simdutf_standard "")
+    foreach(cxx_standard ${test_cxx_standards})
+        if(NOT cxx_standard LESS 17 AND compiler_supports_cpp_${cxx_standard})
+            set(simdutf_standard ${cxx_standard})
+            break()
+        endif()
+    endforeach()
+
+    if("${simdutf_standard}" STREQUAL "")
+        # Building simdutf would fail outright without a C++17 compiler, and
+        # even with one it would go unused if no C++17-or-later standard is
+        # tested. Say so and fall back to the scalar validator rather than
+        # failing the build.
+        if(NOT compiler_supports_cpp_17)
+            set(simdutf_reason "the compiler does not support C++17")
+        else()
+            set(simdutf_reason "no tested standard is C++17 or later (testing ${msg_standards})")
+        endif()
+        message(WARNING
+            "JSON_TestSimdutf is enabled, but ${simdutf_reason}. simdutf requires C++17, so it "
+            "is not fetched and JSON_USE_SIMDUTF is not defined: the tests run against the "
+            "built-in scalar UTF-8 validator instead. Set JSON_TestStandards to include 17 or "
+            "later, or build with a compiler that supports C++17.")
+    else()
+        if (CMAKE_VERSION VERSION_LESS 3.18)
+            message(FATAL_ERROR "JSON_TestSimdutf requires CMake 3.18 or later (simdutf's minimum).")
+        endif()
+
+        include(FetchContent)
+
+        # simdutf builds its tests and tools by default, and its tests pull
+        # further dependencies of their own; only the library is needed here
+        set(SIMDUTF_TESTS OFF CACHE BOOL "" FORCE)
+        set(SIMDUTF_TOOLS OFF CACHE BOOL "" FORCE)
+        set(SIMDUTF_BENCHMARKS OFF CACHE BOOL "" FORCE)
+        set(SIMDUTF_ICONV OFF CACHE BOOL "" FORCE)
+
+        FetchContent_Declare(simdutf
+            URL https://github.com/simdutf/simdutf/archive/refs/tags/v${JSON_SIMDUTF_VERSION}.tar.gz
+            DOWNLOAD_EXTRACT_TIMESTAMP TRUE
+        )
+        FetchContent_MakeAvailable(simdutf)
+
+        target_compile_definitions(test_main PUBLIC JSON_USE_SIMDUTF)
+        target_link_libraries(test_main PUBLIC simdutf::simdutf)
+
+        # simdutf.h requires C++17; below that the library keeps its scalar
+        # validator, so any C++11/14 test targets exercise the fallback and the
+        # C++17-and-later ones exercise simdutf. Both must agree.
+        message(STATUS "UTF-8 validation delegated to simdutf ${JSON_SIMDUTF_VERSION} for C++17 and later (JSON_USE_SIMDUTF)")
+    endif()
+endif()
+
 # *DO* use json_test_set_test_options() above this line
 
 json_test_should_build_32bit_test(json_32bit_test json_32bit_test_only "${JSON_32bitTest}")

tests/src/unit-class_lexer.cpp

@@ -12,6 +12,11 @@
 #include <nlohmann/json.hpp>
 using nlohmann::json;
 
+#include <cstdlib> // strtod
+#include <sstream> // stringstream
+#include <string> // string
+#include <vector> // vector
+
 namespace
 {
 // shortcut to scan a string literal
@@ -224,3 +229,431 @@ TEST_CASE("lexer class")
         CHECK((scan_string("/**//**//**/", true) == json::lexer::token_type::end_of_input));
     }
 }
+
+TEST_CASE("lexer number fast path")
+{
+    // The contiguous fast path (used for pointer/string input) must agree with
+    // the streaming byte path (used for std::istream) on token type, numeric
+    // value, and round-trip text for every well-formed number, and reject the
+    // same malformed numbers with the same message.
+    SECTION("contiguous vs streaming parity")
+    {
+        const std::vector<std::string> numbers =
+        {
+            "0", "-0", "1", "-1", "42", "-42", "10", "100", "1234567890",
+            "0.0", "-0.0", "3.14", "-3.14", "0.5", "-0.001", "123.456789",
+            "1e0", "1E0", "1e10", "1e-10", "1e+10", "1.5e3", "-2.5E-4",
+            "9223372036854775807",             // INT64_MAX -> unsigned
+            "9223372036854775808",             // INT64_MAX + 1 -> unsigned
+            "18446744073709551615",            // UINT64_MAX -> unsigned
+            "18446744073709551616",            // UINT64_MAX + 1 -> float
+            "-9223372036854775808",            // INT64_MIN -> integer
+            "-9223372036854775809",            // INT64_MIN - 1 -> float
+            "123456789012345678901234567890",  // huge -> float
+            "0.30000000000000004", "2.2250738585072014e-308", "1e308",
+            // high-precision / wide-exponent values that exercise the
+            // std::from_chars (Eisel-Lemire) path beyond the Clinger subset
+            "1.7976931348623157e308", "1.2345678901234567e-250",
+            "9007199254740993", "5e-324", "1e-320"
+        };
+
+        for (const auto& n : numbers)
+        {
+            const std::string doc = "[" + n + "]";
+
+            // contiguous fast path
+            const json a = json::parse(doc);
+            // streaming byte path
+            std::stringstream ss(doc);
+            const json b = json::parse(ss);
+
+            CAPTURE(n);
+            CHECK(a == b);
+            CHECK(a.dump() == b.dump());
+            CHECK(a[0].type() == b[0].type());
+        }
+    }
+
+    SECTION("significant-digit gate for the Clinger fast path")
+    {
+        // Clinger's fast path needs a significand below 2^53, so it cannot
+        // succeed once the mantissa has 17 or more significant digits (the
+        // significand would be at least 10^16). The lexer skips the attempt
+        // there. That is only allowed to save work: every value must still come
+        // out bit-exactly, and both scanners must agree. In particular the gate
+        // must not fire for tokens whose leading zeros merely look like extra
+        // digits - "0.1234567890123456" has 16 significant digits, not 17.
+        const std::vector<std::string> numbers =
+        {
+            "1234567890123456",                  // 16 significant digits
+            "12345678901234567",                 // 17 -> attempt skipped
+            "123456789012345678",                // 18 -> attempt skipped
+            "0.1234567890123456",                // 16: the leading "0" is not significant
+            "0.12345678901234567",               // 17
+            "0.00000000000000001",               // 1, in a long token
+            "0.000000000000000012345678901234",  // 14, in a long token
+            "-0.0000000000000000000001",         // 1, negative
+            "1.0000000000000000",                // 17: trailing zeros are significant here
+            "10000000000000000",                 // 17
+            "9007199254740992",                  // 2^53
+            "9007199254740993",                  // 2^53 + 1
+            "-65.613616999999977",               // canada.json shape
+            "1.2345678901234567e-250",           // 17 with an exponent
+            "1.234567890123456e-250",            // 16 with an exponent
+            "1e10", "0.0", "-0.0", "0e0", "0.000123"
+        };
+
+        for (const auto& n : numbers)
+        {
+            CAPTURE(n);
+            const std::string doc = "[" + n + "]";
+
+            const json a = json::parse(doc);   // contiguous fast path
+            std::stringstream ss(doc);
+            const json b = json::parse(ss);    // streaming byte path
+
+            CHECK(a[0].type() == b[0].type());
+            CHECK(a == b);
+
+            if (a[0].is_number_float())
+            {
+                const double expected = std::strtod(n.c_str(), nullptr);
+                CHECK(a[0].get<double>() == expected);
+                CHECK(b[0].get<double>() == expected);
+            }
+        }
+    }
+
+    SECTION("token type classification")
+    {
+        CHECK((scan_string("0") == json::lexer::token_type::value_unsigned));
+        CHECK((scan_string("-1") == json::lexer::token_type::value_integer));
+        CHECK((scan_string("1.5") == json::lexer::token_type::value_float));
+        CHECK((scan_string("1e5") == json::lexer::token_type::value_float));
+        CHECK((scan_string("18446744073709551615") == json::lexer::token_type::value_unsigned));
+        CHECK((scan_string("18446744073709551616") == json::lexer::token_type::value_float));
+        CHECK((scan_string("-9223372036854775808") == json::lexer::token_type::value_integer));
+        CHECK((scan_string("-9223372036854775809") == json::lexer::token_type::value_float));
+    }
+
+    SECTION("malformed numbers are rejected identically")
+    {
+        for (const char* bad :
+                {"-", "1.", "1e", "1e+", "1.2e", "01", "-01", "1..2", "1.2.3"
+                })
+        {
+            CAPTURE(bad);
+            // the contiguous fast path must decline and let the byte path report
+            const std::string doc = std::string("[") + bad + "]";
+            CHECK_FALSE(json::accept(doc));
+            std::stringstream ss(doc);
+            CHECK_FALSE(json::accept(ss));
+        }
+    }
+
+#if !defined(JSON_NOEXCEPTION)
+    // these sections parse invalid input, which aborts when exceptions are off
+    SECTION("exhaustive grammar parity with the streaming path")
+    {
+        // The JSON number grammar is encoded twice: once as the scan_number()
+        // state machine and once as the contiguous fast path. Enumerate every
+        // short string over the number alphabet and require the two encodings to
+        // agree exactly - on acceptance, on the reported error, and on the parsed
+        // value - so they cannot drift apart.
+        const std::string alphabet = "01.eE+-";
+
+        // full outcome of parsing @a doc, so a mismatch in type, value, or error
+        // message is caught, not just a mismatch in acceptance
+        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 std::string(j[0].type_name()) + '|' + j.dump();
+                }
+                const json j = json::parse(doc);
+                return std::string(j[0].type_name()) + '|' + j.dump();
+            }
+            catch (const json::parse_error& e)
+            {
+                return {e.what()};
+            }
+        };
+
+        std::vector<std::string> mismatches;
+        std::vector<std::string> tokens{""};
+        for (std::size_t length = 1; length <= 4; ++length)
+        {
+            std::vector<std::string> next;
+            next.reserve(tokens.size() * alphabet.size());
+            for (const auto& prefix : tokens)
+            {
+                for (const char c : alphabet)
+                {
+                    next.push_back(prefix + c);
+                }
+            }
+            tokens = next;
+
+            for (const auto& token : tokens)
+            {
+                const std::string doc = "[" + token + "]";
+                if (outcome(doc, false) != outcome(doc, true))
+                {
+                    mismatches.push_back(doc);
+                }
+            }
+        }
+
+        // 7 + 49 + 343 + 2401 tokens
+        CHECK(tokens.size() == 2401);
+        CAPTURE(mismatches);
+        CHECK(mismatches.empty());
+    }
+
+    SECTION("error positions match the streaming path")
+    {
+        // Rejecting identically is not enough: the fast path must also report the
+        // error at the same position as the byte path. A number directly followed
+        // by a newline is the interesting case, because the byte path reaches the
+        // newline (which resets the column) and then ungets it.
+        // returns the parse_error message, or "" if the document parsed
+        const auto contiguous_error = [](const std::string & doc) -> std::string
+        {
+            try
+            {
+                const json j = json::parse(doc);
+                static_cast<void>(j);
+            }
+            catch (const json::parse_error& e)
+            {
+                return {e.what()};
+            }
+            return {};
+        };
+        const auto streaming_error = [](const std::string & doc) -> std::string
+        {
+            try
+            {
+                std::stringstream ss(doc);
+                const json j = json::parse(ss);
+                static_cast<void>(j);
+            }
+            catch (const json::parse_error& e)
+            {
+                return {e.what()};
+            }
+            return {};
+        };
+
+        for (const char* bad :
+                {"[01\n]", "[00\n]", "[-01\n]", "{1\n}", "[1\n2]", "[1.2.3\n]",
+                 "[1 \n2]", "[\n1\n2]", "1\n2", "[01\r\n]", "[1e\n]", "[-\n]"
+                })
+        {
+            CAPTURE(bad);
+            const std::string doc = bad;
+            const std::string contiguous_what = contiguous_error(doc);
+
+            CHECK_FALSE(contiguous_what.empty());
+            CHECK(contiguous_what == streaming_error(doc));
+        }
+
+        // A number terminated by a newline must report the same position as the
+        // same number terminated by anything else: scan_number() reads the
+        // terminator and ungets it, so the reported column is the one reached
+        // after the number's last character - not the 0 that an unget() across
+        // the newline used to leave behind.
+        CHECK(contiguous_error("[01\n]") == contiguous_error("[01 ]"));
+        CHECK(contiguous_error("[01\n]") ==
+              "[json.exception.parse_error.101] parse error at line 1, column 3: "
+              "syntax error while parsing array - unexpected number literal; expected ']'");
+
+        // the same for a multi-character token, where the column of the last
+        // character (the '3' of "-2.5e3") differs from the column it starts at
+        CHECK(contiguous_error("null -2.5e3\nfalse") == contiguous_error("null -2.5e3 false"));
+        CHECK(contiguous_error("null -2.5e3\nfalse") ==
+              "[json.exception.parse_error.101] parse error at line 1, column 11: "
+              "syntax error while parsing value - unexpected number literal; expected end of input");
+    }
+#endif
+}
+
+TEST_CASE("lexer string fast path")
+{
+    // Build a byte string from explicit values: a hex escape in a string
+    // literal swallows every following hex digit, which makes sequences like
+    // "\xC3\xA9b" mean something other than they look like.
+    const auto bytes = [](std::initializer_list<int> values)
+    {
+        std::string result;
+        for (const int value : values)
+        {
+            result.push_back(static_cast<char>(value));
+        }
+        return result;
+    };
+
+#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. Only usable with exceptions
+    // on: parsing invalid input aborts when they are off.
+    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();
+        }
+        // not just parse_error: if a bulk scanner ever let ill-formed UTF-8
+        // through, dump() would throw type_error.316, and that has to surface
+        // as a reported mismatch rather than as an uncaught exception
+        catch (const json::exception& e)
+        {
+            return {e.what()};
+        }
+    };
+#endif
+
+    // once at the start of the string, once past the first 8-byte SWAR word, so
+    // the bulk scanner sees each case with and without a run behind it
+    const std::vector<std::size_t> offsets{0, 9};
+
+#if !defined(JSON_NOEXCEPTION)
+    SECTION("exhaustive contiguous vs streaming parity")
+    {
+        // ordinary ASCII, both specials, a control byte, characters that make
+        // the preceding backslash a valid escape, a UTF-8 lead byte of each
+        // length, a continuation byte, and a byte that is never valid
+        const std::vector<std::string> alphabet =
+        {
+            "a", "\"", "\\", "n", "u", "0", bytes({0x01}),
+            bytes({0xC3}), bytes({0xA9}), bytes({0xE4}), bytes({0xF0}),
+            bytes({0x80}), bytes({0xFF})
+        };
+
+        std::vector<std::string> mismatches;
+        std::vector<std::string> tokens{""};
+        for (std::size_t length = 1; length <= 3; ++length)
+        {
+            std::vector<std::string> next;
+            next.reserve(tokens.size() * alphabet.size());
+            for (const auto& prefix : tokens)
+            {
+                for (const auto& symbol : alphabet)
+                {
+                    next.push_back(prefix + symbol);
+                }
+            }
+            tokens = next;
+
+            for (const auto& token : tokens)
+            {
+                for (const std::size_t offset : offsets)
+                {
+                    const std::string doc = "[\"" + std::string(offset, 'a') + token + "\"]";
+                    if (outcome(doc, false) != outcome(doc, true))
+                    {
+                        mismatches.push_back(doc);
+                    }
+                }
+            }
+        }
+
+        // 13 + 169 + 2197 tokens, each at two offsets
+        CHECK(tokens.size() == 2197);
+        CAPTURE(mismatches);
+        CHECK(mismatches.empty());
+    }
+
+    SECTION("special bytes at every offset of the SWAR stride")
+    {
+        // The bulk scanner consumes 8 bytes at a time and then a tail; place
+        // every kind of byte that ends a run at each offset across two words,
+        // so multibyte sequences also straddle the word boundary.
+        const std::vector<std::string> specials =
+        {
+            "\"", "\\", bytes({0x01}), bytes({0x1F}), bytes({0x7F}),
+            bytes({0xC3, 0xA9}), bytes({0xE4, 0xB8, 0xAD}), bytes({0xF0, 0x9F, 0x98, 0x80}),
+            bytes({0xFF}), bytes({0xC3}), bytes({0xE4, 0xB8})
+        };
+
+        std::vector<std::string> mismatches;
+        for (std::size_t offset = 0; offset <= 17; ++offset)
+        {
+            for (const auto& special : specials)
+            {
+                const std::string doc = "[\"" + std::string(offset, 'a') + special + "\"]";
+                if (outcome(doc, false) != outcome(doc, true))
+                {
+                    mismatches.push_back(doc);
+                }
+            }
+        }
+        CAPTURE(mismatches);
+        CHECK(mismatches.empty());
+    }
+#endif
+
+    // json::accept() never throws, so the ranges stay covered without exceptions
+    SECTION("UTF-8 ranges are accepted and rejected as documented")
+    {
+        // The bulk validator must accept exactly what the byte-at-a-time
+        // scanner accepts, so pin the boundaries of every range it recognizes.
+        // aggregate, only ever brace-initialized below; default member
+        // initializers would stop it being an aggregate in C++11
+        struct utf8_case // NOLINT(cppcoreguidelines-pro-type-member-init,hicpp-member-init)
+        {
+            std::string sequence;
+            bool valid;
+            const char* description;
+        };
+        const std::vector<utf8_case> cases =
+        {
+            {bytes({0xC2, 0x80}), true, "U+0080, shortest two-byte"},
+            {bytes({0xDF, 0xBF}), true, "U+07FF, longest two-byte"},
+            {bytes({0xC1, 0xBF}), false, "overlong two-byte"},
+            {bytes({0xC2, 0x7F}), false, "two-byte with bad continuation"},
+            {bytes({0xE0, 0xA0, 0x80}), true, "U+0800, shortest three-byte"},
+            {bytes({0xE0, 0x9F, 0xBF}), false, "overlong three-byte"},
+            {bytes({0xED, 0x9F, 0xBF}), true, "U+D7FF, just below the surrogates"},
+            {bytes({0xED, 0xA0, 0x80}), false, "surrogate U+D800"},
+            {bytes({0xED, 0xBF, 0xBF}), false, "surrogate U+DFFF"},
+            {bytes({0xEE, 0x80, 0x80}), true, "U+E000, just above the surrogates"},
+            {bytes({0xEF, 0xBF, 0xBF}), true, "U+FFFF"},
+            {bytes({0xF0, 0x90, 0x80, 0x80}), true, "U+10000, shortest four-byte"},
+            {bytes({0xF0, 0x8F, 0xBF, 0xBF}), false, "overlong four-byte"},
+            {bytes({0xF4, 0x8F, 0xBF, 0xBF}), true, "U+10FFFF, highest code point"},
+            {bytes({0xF4, 0x90, 0x80, 0x80}), false, "above U+10FFFF"},
+            {bytes({0xF5, 0x80, 0x80, 0x80}), false, "lead byte out of range"},
+            {bytes({0x80}), false, "bare continuation byte"},
+            {bytes({0xFF}), false, "byte that never appears in UTF-8"},
+            {bytes({0xC3}), false, "truncated two-byte"},
+            {bytes({0xE4, 0xB8}), false, "truncated three-byte"},
+            {bytes({0xF0, 0x9F, 0x98}), false, "truncated four-byte"}
+        };
+
+        for (const auto& test_case : cases)
+        {
+            CAPTURE(test_case.description);
+            for (const std::size_t offset : offsets)
+            {
+                CAPTURE(offset);
+                const std::string doc = "[\"" + std::string(offset, 'a') + test_case.sequence + "\"]";
+                CHECK(json::accept(doc) == test_case.valid);
+#if !defined(JSON_NOEXCEPTION)
+                CHECK(outcome(doc, false) == outcome(doc, true));
+#endif
+            }
+        }
+    }
+}

tests/src/unit-user_defined_input.cpp

@@ -18,7 +18,12 @@
 #include <nlohmann/json.hpp>
 using nlohmann::json;
 
+#include <array> // array
+#include <cstddef> // size_t
+#include <cstdint> // uint8_t
 #include <list>
+#include <string> // string
+#include <vector> // vector
 
 #if defined(__cpp_lib_concepts) && defined(JSON_HAS_CPP_20)
     #include <iterator>
@@ -212,6 +217,66 @@ TEST_CASE("Parse with heterogeneous iterator and sentinel types")
     CHECK(j2.at(0) == 1);
 }
 
+// A type whose data() hands out raw bytes but whose size() counts something
+// else - here fixed-size records. Reading [data(), data() + size()) as bytes
+// would silently truncate the input, so data() and size() alone must not be
+// taken as evidence of contiguous byte storage.
+struct record_buffer
+{
+    using value_type = std::array<char, 4>;
+
+    std::string bytes;
+
+    const char* data() const noexcept
+    {
+        return bytes.data();
+    }
+    std::size_t size() const noexcept
+    {
+        return bytes.size() / sizeof(value_type);
+    }
+    const char* begin() const noexcept
+    {
+        return bytes.data();
+    }
+    const char* end() const noexcept
+    {
+        return bytes.data() + bytes.size();
+    }
+};
+
+TEST_CASE("Contiguous byte containers take the pointer adapter")
+{
+    // Containers with contiguous single-byte storage are routed through the
+    // pointer-based adapter so the bulk fast paths apply in every standard, not
+    // only in C++20 where the library iterators model std::contiguous_iterator.
+    CHECK(nlohmann::detail::is_contiguous_byte_container<std::string>::value);
+    CHECK(nlohmann::detail::is_contiguous_byte_container<std::vector<char>>::value);
+    CHECK(nlohmann::detail::is_contiguous_byte_container<std::vector<std::uint8_t>>::value);
+    CHECK(nlohmann::detail::is_contiguous_byte_container<std::array<char, 4>>::value);
+
+    // input_adapter() takes its container by forwarding reference, so the trait
+    // is also asked about reference types
+    CHECK(nlohmann::detail::is_contiguous_byte_container<std::string&>::value);
+    CHECK(nlohmann::detail::is_contiguous_byte_container<const std::string&>::value);
+
+    // everything else keeps the iterator-based adapter
+    CHECK_FALSE(nlohmann::detail::is_contiguous_byte_container<std::list<char>>::value);
+    CHECK_FALSE(nlohmann::detail::is_contiguous_byte_container<std::vector<int>>::value);
+    CHECK_FALSE(nlohmann::detail::is_contiguous_byte_container<const char*>::value);
+
+    // including a type that has data() and size() but whose size() does not
+    // count the units data() points at: its value_type says so
+    CHECK_FALSE(nlohmann::detail::is_contiguous_byte_container<record_buffer>::value);
+
+    // and such a container still parses through its iterators, in full - taking
+    // it for a byte container would stop after data() + size() bytes
+    const record_buffer buffer{"[1,2,3,4,5]"};
+    CHECK(buffer.data() == buffer.bytes.data());
+    CHECK(buffer.size() * sizeof(record_buffer::value_type) < buffer.bytes.size());
+    CHECK(json::parse(buffer) == json({1, 2, 3, 4, 5}));
+}
+
 #if defined(__cpp_lib_concepts) && defined(JSON_HAS_CPP_20)
 // JSON_HAS_CPP_20 (do not remove; see note at top of file)
 TEST_CASE("Parse with std::counted_iterator and std::default_sentinel_t")
@@ -228,6 +293,180 @@ TEST_CASE("Parse with std::counted_iterator and std::default_sentinel_t")
     const std::counted_iterator<iterator_type> first2(json_str.begin(), len);
     CHECK(json::accept(first2, std::default_sentinel));
 }
+
+TEST_CASE("std::counted_iterator reaches the contiguous fast paths")
+{
+    // A sized sentinel makes the remaining element count computable in O(1), so
+    // std::counted_iterator over a contiguous iterator must reach the same bulk
+    // string/number scanners as a plain pointer - not just the byte-at-a-time
+    // fallback (see #5268 for the equivalent memcpy fast path).
+#if JSON_HAS_RANGES
+    // JSON_HAS_RANGES is 0 on standard libraries with an incomplete <ranges>
+    // (libstdc++ < 11, libc++ < 16), where the adapter deliberately falls back
+    // to the byte-at-a-time scanner; everything below still has to work there.
+    using adapter_type = nlohmann::detail::iterator_input_adapter<std::counted_iterator<const char*>, std::default_sentinel_t>;
+    CHECK(adapter_type::supports_bulk_scan);
+    CHECK(adapter_type::supports_seek);
+#endif
+
+    // exercise every fast path: long ASCII run, multibyte UTF-8, escapes, and
+    // integer/floating-point numbers
+    const std::string json_str =
+        R"({"ascii":"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",)"
+        "\"utf8\":\"\xe4\xb8\xad\xe6\x96\x87\xf0\x9f\x98\x80\xc3\xa9\","
+        R"("escaped":"aéb\n\\","ints":[0,-1,18446744073709551615,-9223372036854775808],)"
+        R"("floats":[1.5,-2.25e3,0.30000000000000004]})";
+    const auto len = static_cast<std::iter_difference_t<const char*>>(json_str.size());
+
+    const std::counted_iterator<const char*> first(json_str.data(), len);
+    const json j = json::parse(first, std::default_sentinel);
+
+    // parsing through the pointer adapter must give exactly the same result
+    CHECK(j == json::parse(json_str));
+
+#if !defined(JSON_NOEXCEPTION)
+    // Diagnostics that quote the offending token are reconstructed from the
+    // already-consumed input (supports_seek), a path a sized sentinel only
+    // reaches now; check a few that include the "last read" text. Parsing
+    // invalid input aborts when exceptions are off, hence the guard.
+    // Raw strings and explicit bytes: an escaped literal and two literals
+    // written next to each other both read as mistakes to static analysis.
+    const auto byte = [](int value)
+    {
+        return std::string(1, static_cast<char>(value));
+    };
+    const std::vector<std::string> diagnostic_docs =
+    {
+        "1\nx",
+        "truX",
+        "[tru]",
+        R"("abc)",
+        R"(["\ud834"])",
+        R"(["a)" + byte(0x01) + R"(b"])",
+        R"([")" + byte(0xC3) + byte(0x28) + R"("])",
+        "[1e]",
+        R"(["aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaX)"
+    };
+
+    for (const auto& text : diagnostic_docs)
+    {
+        CAPTURE(text);
+        const std::counted_iterator<const char*> it(text.data(), static_cast<std::iter_difference_t<const char*>>(text.size()));
+        std::string counted_message;
+        std::string string_message;
+        try
+        {
+            const json counted_result = json::parse(it, std::default_sentinel);
+            static_cast<void>(counted_result);
+        }
+        catch (const json::parse_error& e)
+        {
+            counted_message = e.what();
+        }
+        try
+        {
+            const json string_result = json::parse(text);
+            static_cast<void>(string_result);
+        }
+        catch (const json::parse_error& e)
+        {
+            string_message = e.what();
+        }
+        CHECK_FALSE(counted_message.empty());
+        CHECK(counted_message == string_message);
+    }
+
+    // and errors must still be reported identically
+    const std::string bad = "[01\n]";
+    const std::counted_iterator<const char*> bad_first(bad.data(), static_cast<std::iter_difference_t<const char*>>(bad.size()));
+    std::string counted_what;
+    std::string string_what;
+    try
+    {
+        const json counted_result = json::parse(bad_first, std::default_sentinel);
+        static_cast<void>(counted_result);
+    }
+    catch (const json::parse_error& e)
+    {
+        counted_what = e.what();
+    }
+    try
+    {
+        const json string_result = json::parse(bad);
+        static_cast<void>(string_result);
+    }
+    catch (const json::parse_error& e)
+    {
+        string_what = e.what();
+    }
+    CHECK_FALSE(counted_what.empty());
+    CHECK(counted_what == string_what);
+#endif
+}
+
+#if !defined(JSON_NOEXCEPTION)
+// several cases below are truncated on purpose, and parsing invalid input
+// aborts when exceptions are off
+TEST_CASE("std::counted_iterator bulk scanning stops at the counted end")
+{
+    // The count, not the size of the underlying buffer, is the end of the
+    // input: the bulk scanners must never look at the bytes behind it, even
+    // though they are readable. Each case is compared against parsing the
+    // equivalent prefix as a std::string.
+    const auto via_counted = [](const std::string & buf, std::size_t n) -> std::string
+    {
+        const std::counted_iterator<const char*> first(buf.data(), static_cast<std::iter_difference_t<const char*>>(n));
+        try
+        {
+            const json j = json::parse(first, std::default_sentinel);
+            return "OK|" + j.dump();
+        }
+        catch (const json::parse_error& e)
+        {
+            return {e.what()};
+        }
+    };
+    const auto via_prefix = [](const std::string & buf, std::size_t n) -> std::string
+    {
+        try
+        {
+            const json j = json::parse(buf.substr(0, n));
+            return "OK|" + j.dump();
+        }
+        catch (const json::parse_error& e)
+        {
+            return {e.what()};
+        }
+    };
+
+    struct testcase // NOLINT(cppcoreguidelines-pro-type-member-init,hicpp-member-init)
+    {
+        const char* buffer;
+        std::size_t count;
+    };
+    const std::vector<testcase> cases =
+    {
+        {"[\"abc\"]____TRAILING____", 7},                    // exact fit, tail hidden
+        {"[\"abcdefghijklmnop\"]____", 8},                   // cut inside a string
+        {"[\"abc\"]____", 6},                                // cut just before the closing quote
+        {"[12345]xxxxx", 4},                                 // cut inside a number
+        {"[123]999999", 5},                                  // number ends exactly at the count
+        {"[\"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\"]", 12},        // closing quote only behind the count
+        {"[\"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\"]", 19},      // cut inside an 8-byte SWAR stride
+        {"[\"\xe4\xb8\xad\xe6\x96\x87\"]", 5},               // cut inside a UTF-8 sequence
+        {"[\"\xe4\xb8\xad\xe6\x96\x87\"]____", 10},          // complete UTF-8, tail hidden
+        {"[1.25e3]TRAILINGDIGITS999", 7},                    // number token reaches the count
+    };
+
+    for (const auto& tc : cases)
+    {
+        CAPTURE(tc.buffer);
+        CAPTURE(tc.count);
+        const std::string buffer = tc.buffer;
+        CHECK(via_counted(buffer, tc.count) == via_prefix(buffer, tc.count));
+    }
+}
+#endif
 #endif
 
 } // namespace