Copy the text of float tokens only when a SAX handler asks for it (#5852)

After the digits are converted while scanning, copying the token text into
token_buffer (an out-of-line std::string::append) was about 10% of parsing
canada.json. The only reader is the string argument of
json_sax::number_float(), and the library's own handlers
(json_sax_dom_parser, json_sax_dom_callback_parser, json_sax_acceptor)
ignore it.

For contiguous input, the fast path now keeps a pointer to the token in the
input, and get_string() copies it on first use. The parser passes an empty
string to the handlers marked by the new internal trait
sax_ignores_number_text instead of calling get_string(), so json::parse()
and json::accept() never copy float text. User handlers, and
JSON_DIAGNOSTIC_POSITIONS (which reads get_string().size()), still see the
exact text. Keys and strings, which never have pending text, use the
unchecked get_string_token(). Input read through a window into a stream
(std::istream, FILE*) copies the text right away, because the next read may
refill the window.

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

.github/workflows/check_amalgamation.yml

@@ -37,7 +37,7 @@ jobs:
           path: pr/
 
   check:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-26.04
     env:
       MAIN_DIR: ${{ github.workspace }}/main
       INCLUDE_DIR: ${{ github.workspace }}/main/single_include/nlohmann
@@ -77,13 +77,12 @@ jobs:
           cd $MAIN_DIR
 
           # Built from this PR's own tools/macro_builder/main.cpp, not a
-          # develop checkout: unlike amalgamate.py and generate_natvis.py,
-          # this tool has no other source of truth to check against (its
-          # own README documents that it must reproduce macro_scope.hpp
-          # byte for byte), so there is nothing to gain from checking out a
-          # separate copy, and doing so would make this step fail on a PR
-          # that adds support for a new dispatch table until that PR itself
-          # merges to develop.
+          # develop checkout: unlike amalgamate.py, this tool has no other
+          # source of truth to check against (its own README documents that
+          # it must reproduce macro_scope.hpp byte for byte), so there is
+          # nothing to gain from checking out a separate copy, and doing so
+          # would make this step fail on a PR that adds support for a new
+          # dispatch table until that PR itself merges to develop.
           TMPDIR=$(mktemp -d ./macro_builder_check.XXXXXX)
           c++ -std=c++11 tools/macro_builder/main.cpp -o "$TMPDIR/macro_builder"
           "$TMPDIR/macro_builder" > "$TMPDIR/paste.hpp"
@@ -146,14 +145,14 @@ jobs:
       - name: Regenerate nlohmann_json.natvis
         run: |
           cd $MAIN_DIR
-          # The template is what this check verifies the committed natvis
-          # file against, so take it from this PR rather than from the
-          # develop checkout: generate_natvis.py loads the template from its
-          # own directory, and with develop's copy a PR that changes the
-          # template could never pass. Like the macro_builder step above,
-          # this only renders PR content in a job without credentials.
-          cp tools/generate_natvis/nlohmann_json.natvis.j2 $NATVIS_TOOL_DIR/
-          python3 $NATVIS_TOOL_DIR/generate_natvis.py $MAIN_DIR
+          # The generator and its template are what this check verifies the
+          # committed natvis file against (which namespaces it lists, and
+          # what each entry looks like), so run this PR's copy rather than
+          # the develop checkout's: with develop's, a PR that changes either
+          # could never pass. generate_natvis.py loads the template from its
+          # own directory. Like the macro_builder step above, this only runs
+          # PR content in a job without credentials.
+          python3 tools/generate_natvis/generate_natvis.py $MAIN_DIR
 
       - name: Build patch and check for differences
         id: diff

.github/workflows/codeql-analysis.yml

@@ -21,7 +21,7 @@ permissions:
 jobs:
   CodeQL-Build:
 
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-26.04
     permissions:
       security-events: write
 

.github/workflows/scorecards.yml

@@ -50,7 +50,7 @@ jobs:
           persist-credentials: false
 
       - name: "Run analysis"
-        uses: ossf/scorecard-action@2d1146689b8cda280b9bc96326124645441f03bc # v2.4.4
+        uses: ossf/scorecard-action@fa7117db83c52d787ae5a89f350999701505d78e # v2.5.0
         with:
           results_file: results.sarif
           results_format: sarif

.github/workflows/semgrep.yml

@@ -33,7 +33,7 @@ jobs:
       security-events: write # for github/codeql-action/upload-sarif to upload SARIF results
       actions: read # only required for a private repository by github/codeql-action/upload-sarif to get the Action run status
     name: Scan
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-26.04
     steps:
       - name: Harden Runner
         uses: step-security/harden-runner@ccd8616d44fd3846e67624a50d5aad6d37bf2d25 # v2.22.1

.github/workflows/ubuntu.yml

@@ -32,7 +32,7 @@ jobs:
         run: cmake --build build --target ci_test_gcc
 
   ci_meson_install:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-26.04
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
         with:
@@ -77,7 +77,7 @@ jobs:
           cmake --build build-import-split
 
   ci_infer:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-26.04
     steps:
       - name: Harden Runner
         uses: step-security/harden-runner@ccd8616d44fd3846e67624a50d5aad6d37bf2d25 # v2.22.1
@@ -106,7 +106,7 @@ jobs:
           if-no-files-found: ignore
 
   ci_static_analysis_ubuntu:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-26.04
     strategy:
       matrix:
         target: [ci_test_single_header, ci_cppcheck, ci_cpplint, ci_reproducible_tests, ci_non_git_tests, ci_offline_testdata, ci_reuse_compliance, ci_test_valgrind]
@@ -175,7 +175,7 @@ jobs:
         run: cmake --build build --target ${{ matrix.target }}
 
   ci_test_coverage:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-26.04
     steps:
       - name: Harden Runner
         uses: step-security/harden-runner@ccd8616d44fd3846e67624a50d5aad6d37bf2d25 # v2.22.1
@@ -430,7 +430,7 @@ jobs:
         run: cmake --build build --target ci_nvhpc
 
   ci_emscripten:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-26.04
     steps:
       - name: Harden Runner
         uses: step-security/harden-runner@ccd8616d44fd3846e67624a50d5aad6d37bf2d25 # v2.22.1
@@ -450,7 +450,7 @@ jobs:
         run: cmake --build build
 
   ci_test_documentation:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-26.04
     strategy:
       matrix:
         target: [ci_test_examples, ci_test_build_documentation, ci_test_documentation_mermaid]

.inferconfig

@@ -1,6 +1,7 @@
 {
   "_comment": "Used by the ci_infer CMake target (#5715 item 4b). fail-on-issue makes CI fail on Infer findings; disable-issue-type is a type-level baseline for the ~174 pre-existing findings (all PULSE_UNNECESSARY_COPY*/PULSE_RESOURCE_LEAK/PULSE_CONST_REFABLE, mostly in test code) triaged in run https://github.com/nlohmann/json/actions/runs/35829411620 on commit 1054b2097, so CI fails only on a NEW issue type. Remove an entry here once its findings have been fixed or explicitly accepted.",
   "_comment_pulse": "Pulse stops exploring paths after pulse-max-disjuncts (default 20). With the default, basic_json::replace_value() (destroy + assert_invariant) exceeds the limit, Pulse loses the stored type, and reports false NULLPTR_DEREFERENCE findings for get_ptr() results in tests/src/unit-pointer_access.cpp.",
+  "_comment_view": "Infer's model of std::unique_ptr mixes up values that are returned by value and own one: basic_json_document (parse() results) and the `built` values of the build() helper in tests/src/unit-json_view_builder.cpp. It reports USE_AFTER_DELETE for one of them after the destructor of an unrelated object (e.g. ~built() on line 345 invalidating `shallow` on line 350 of that file, or two documents parsed in one lambda in tests/src/unit-json_view_dump.cpp). AddressSanitizer and Valgrind are clean on the same tests, so reports for the json_view test files are excluded; findings reported in the library headers are still reported.",
   "pulse-max-disjuncts": 40,
   "fail-on-issue": true,
   "disable-issue-type": [
@@ -10,5 +11,8 @@
     "PULSE_UNNECESSARY_COPY_OPTIONAL",
     "PULSE_RESOURCE_LEAK",
     "PULSE_CONST_REFABLE"
+  ],
+  "report-block-list-path-regex": [
+    "tests/src/unit-json_view.*\\.cpp"
   ]
 }

cmake/ci.cmake

@@ -361,10 +361,10 @@ add_custom_target(ci_test_coverage
     COMMAND ${CMAKE_COMMAND} --build ${PROJECT_BINARY_DIR}/build_coverage32
     COMMAND cd ${PROJECT_BINARY_DIR}/build_coverage32 && ${CMAKE_CTEST_COMMAND} --parallel ${N} --output-on-failure
 
-    COMMAND ${LCOV_TOOL} --directory . --capture --output-file json.info --rc branch_coverage=1 --rc geninfo_unexecuted_blocks=1 --ignore-errors mismatch --ignore-errors unused
-    COMMAND ${LCOV_TOOL} -e json.info ${SRC_FILES} --output-file json.info.filtered --rc branch_coverage=1 --ignore-errors unused
+    COMMAND ${LCOV_TOOL} --directory . --capture --output-file json.info --rc branch_coverage=1 --rc geninfo_unexecuted_blocks=1 --ignore-errors mismatch --ignore-errors unused --ignore-errors inconsistent
+    COMMAND ${LCOV_TOOL} -e json.info ${SRC_FILES} --output-file json.info.filtered --rc branch_coverage=1 --ignore-errors unused --ignore-errors inconsistent
     COMMAND ${CMAKE_SOURCE_DIR}/tests/thirdparty/imapdl/filterbr.py json.info.filtered > json.info.filtered.noexcept
-    COMMAND ${GENHTML_TOOL} --title "JSON for Modern C++" --legend --demangle-cpp --output-directory html --show-details --branch-coverage json.info.filtered.noexcept
+    COMMAND ${GENHTML_TOOL} --title "JSON for Modern C++" --legend --demangle-cpp --output-directory html --show-details --branch-coverage --ignore-errors inconsistent json.info.filtered.noexcept
 
     COMMENT "Compile and test with coverage"
 )

cmake/requirements/requirements-cppcheck.txt

@@ -1 +1 @@
-cppcheck==1.5.2
+cppcheck==1.5.3

docs/mkdocs/docs/api/basic_json/accept.md

@@ -132,6 +132,10 @@ outside of a string, invalid) byte; see the [FAQ entry](../../home/faq.md#nul-by
 - [operator>>](../operator_gtgt.md) - deserialize from stream
 - [`JSON_STRICT_NUL_HANDLING`](../macros/json_strict_nul_handling.md) - opt in to rejecting a NUL byte in the input
   instead of treating it as end of input
+- [`JSON_DISABLE_STREAM_BULK_SCAN`](../macros/json_disable_stream_bulk_scan.md) - read a `std::istream` one character
+  at a time
+- [`JSON_DISABLE_UNLOCKED_FILE_IO`](../macros/json_disable_unlocked_file_io.md) - read a `FILE*` with the locking stdio
+  functions
 
 ## Version history
 
@@ -143,6 +147,10 @@ outside of a string, invalid) byte; see the [FAQ entry](../../home/faq.md#nul-by
 - Extended overload (2) to accept heterogeneous iterator+sentinel pairs (C++20 ranges support) in version 3.13.0.
 - `JSON_STRICT_NUL_HANDLING` added in version 3.13.0 to optionally reject a NUL byte in the input instead of treating
   it as end of input; planned to become the default in version 4.0.0.
+- Parsing from an `std::istream` scans the characters in its stream buffer in bulk in version 3.13.0
+  ([`JSON_DISABLE_STREAM_BULK_SCAN`](../macros/json_disable_stream_bulk_scan.md)).
+- Parsing from a `FILE*` locks the file once instead of once per character in version 3.13.0
+  ([`JSON_DISABLE_UNLOCKED_FILE_IO`](../macros/json_disable_unlocked_file_io.md)).
 
 !!! warning "Deprecation"
 

docs/mkdocs/docs/api/basic_json/find.md

@@ -93,7 +93,8 @@ Logarithmic in the size of the JSON object.
 
 - [count](count.md) returns the number of occurrences of a key
 - [contains](contains.md) checks whether a key exists
-- [basic_json_view::find](../basic_json_view/find.md) - the same lookup on a zero-copy view
+- [basic_json_view::find](../basic_json_view/find.md) - the same lookup on a zero-copy view (a duplicate key resolves
+  to the first member, whereas `parse()` keeps the last)
 
 ## Version history
 

docs/mkdocs/docs/api/basic_json/front.md

@@ -51,7 +51,8 @@ Constant.
 ## See also
 
 - [back](back.md) to access the last element
-- [basic_json_view::front](../basic_json_view/front.md) - the same access on a zero-copy view
+- [basic_json_view::front](../basic_json_view/front.md) - the same access on a zero-copy view (document order, not
+  sorted; duplicate keys are visited)
 
 ## Version history
 

docs/mkdocs/docs/api/basic_json/items.md

@@ -100,7 +100,7 @@ When iterating over an array, `key()` will return the index of the element as st
 - [begin](begin.md) returns an iterator to the first element
 - [end](end.md) returns an iterator to one past the last element
 - [basic_json_view::items](../basic_json_view/items.md) - the same range on a zero-copy view (`#!cpp const auto`,
-  not `#!cpp const auto&`: items are produced on the fly)
+  not `#!cpp const auto&`: items are produced on the fly; document order, not sorted; duplicate keys are visited)
 
 ## Version history
 

docs/mkdocs/docs/api/basic_json/operator[].md

@@ -275,8 +275,9 @@ Strong exception safety: if an exception occurs, the original value stays intact
 - documentation on [runtime assertions](../../features/assertions.md)
 - see [`at`](at.md) for access by reference with range checking
 - see [`value`](value.md) for access with default value
-- [basic_json_view::operator[]](../basic_json_view/operator%5B%5D.md) - the same access on a zero-copy view (always
-  returns a discarded view instead of assuming undefined behavior)
+- [basic_json_view::operator[]](../basic_json_view/operator%5B%5D.md) - the same access on a zero-copy view (a duplicate
+  key resolves to the first member, whereas `parse()` keeps the last; a missing key gives a discarded view instead of
+  undefined behavior)
 
 ## Version history
 

docs/mkdocs/docs/api/basic_json/parse.md

@@ -112,6 +112,16 @@ By default, a `'\0'` (NUL) byte anywhere in the input is treated as end of input
 outside of a string, invalid) byte; see the [FAQ entry](../../home/faq.md#nul-bytes-in-the-input) for details and the
 [`JSON_STRICT_NUL_HANDLING`](../macros/json_strict_nul_handling.md) macro to opt into rejecting it instead.
 
+!!! tip "Performance"
+
+    Contiguous inputs (strings, string views, vectors, arrays, pointer ranges) are the fastest to parse. A
+    `#!cpp std::istream` comes close: the parser scans the characters its stream buffer holds in bulk, without reading
+    ahead, so the stream is left exactly where it would be after reading one character at a time. A `#!cpp FILE*` is
+    read one character at a time, but the file is locked only once, so other threads that use the same `FILE` object
+    wait until parsing is done. See [input sources and performance](../../features/parsing/index.md#input-sources-and-performance),
+    [`JSON_DISABLE_STREAM_BULK_SCAN`](../macros/json_disable_stream_bulk_scan.md), and
+    [`JSON_DISABLE_UNLOCKED_FILE_IO`](../macros/json_disable_unlocked_file_io.md).
+
 ## Examples
 
 ??? example "Example: (1) parse from a character array"
@@ -247,6 +257,10 @@ outside of a string, invalid) byte; see the [FAQ entry](../../home/faq.md#nul-by
 - [operator>>](../operator_gtgt.md) - deserialize from stream
 - [`JSON_STRICT_NUL_HANDLING`](../macros/json_strict_nul_handling.md) - opt in to rejecting a NUL byte in the input
   instead of treating it as end of input
+- [`JSON_DISABLE_STREAM_BULK_SCAN`](../macros/json_disable_stream_bulk_scan.md) - read a `std::istream` one character
+  at a time
+- [`JSON_DISABLE_UNLOCKED_FILE_IO`](../macros/json_disable_unlocked_file_io.md) - read a `FILE*` with the locking stdio
+  functions
 
 ## Version history
 
@@ -261,6 +275,11 @@ outside of a string, invalid) byte; see the [FAQ entry](../../home/faq.md#nul-by
   it as end of input; planned to become the default in version 4.0.0.
 - Extended empty-input detection to also cover an `std::istream` without a stream buffer, and fixed a crash
   (`std::terminate`) when parsing from an `std::istream` with `eofbit` in its exception mask, in version 3.13.0.
+- Parsing from an `std::istream` scans the characters in its stream buffer in bulk in version 3.13.0
+  ([`JSON_DISABLE_STREAM_BULK_SCAN`](../macros/json_disable_stream_bulk_scan.md) restores reading one character at a
+  time).
+- Parsing from a `FILE*` locks the file once instead of once per character in version 3.13.0
+  ([`JSON_DISABLE_UNLOCKED_FILE_IO`](../macros/json_disable_unlocked_file_io.md) restores `std::fgetc`).
 
 !!! warning "Deprecation"
 

docs/mkdocs/docs/api/basic_json/sax_parse.md

@@ -116,6 +116,10 @@ consumer `sax` has a super-linear complexity.
 
 A UTF-8 byte order mark is silently ignored.
 
+When reading from a `#!c FILE*`, the file is locked while it is parsed. The lock is recursive, so the SAX handler may
+use the same file with stdio functions, but it must not wait for another thread that uses the same `FILE` object; see
+[`JSON_DISABLE_UNLOCKED_FILE_IO`](../macros/json_disable_unlocked_file_io.md).
+
 ## Examples
 
 ??? example
@@ -137,6 +141,10 @@ A UTF-8 byte order mark is silently ignored.
 
 - [parse](parse.md) - deserialize from a compatible input
 - [accept](accept.md) - check if the input is valid JSON
+- [`JSON_DISABLE_STREAM_BULK_SCAN`](../macros/json_disable_stream_bulk_scan.md) - read a `std::istream` one character
+  at a time
+- [`JSON_DISABLE_UNLOCKED_FILE_IO`](../macros/json_disable_unlocked_file_io.md) - read a `FILE*` with the locking stdio
+  functions
 
 ## Version history
 
@@ -148,6 +156,10 @@ A UTF-8 byte order mark is silently ignored.
 - Extended overload (2) to accept heterogeneous iterator+sentinel pairs (C++20 ranges support) in version 3.13.0.
 - `JSON_PRECISE_STREAM_POSITION` added in version 3.13.0 to optionally leave a `#!cpp std::istream` positioned right
   after the parsed value when `strict` is `#!cpp false`.
+- Parsing from an `std::istream` scans the characters in its stream buffer in bulk in version 3.13.0
+  ([`JSON_DISABLE_STREAM_BULK_SCAN`](../macros/json_disable_stream_bulk_scan.md)).
+- Parsing from a `FILE*` locks the file once instead of once per character in version 3.13.0
+  ([`JSON_DISABLE_UNLOCKED_FILE_IO`](../macros/json_disable_unlocked_file_io.md)).
 
 !!! warning "Deprecation"
 

docs/mkdocs/docs/api/basic_json/to_bjdata.md

@@ -123,4 +123,6 @@ Linear in the size of the JSON value `j`.
   that is not valid UTF-8 unchanged, as before; `strict` (the default if
   [`JSON_STRICT_BINARY_UTF8`](../macros/json_strict_binary_utf8.md) is enabled) throws `type_error.316`.
 - Throws `type_error.321` for a discarded value since version 3.13.0; previously, a discarded value nested in an
-  array or object was silently skipped, producing invalid BJData.
+  array or object was silently skipped, producing invalid BJData.
+- Writes unsigned integers wider than 64 bits as high-precision numbers since version 3.13.0; previously, they were
+  silently truncated to 64 bits.

docs/mkdocs/docs/api/basic_json/to_bon8.md

@@ -37,8 +37,9 @@ With (2), the bytes written before the exception remain in the output adapter.
 
 ## Exceptions
 
-- Throws [out_of_range.407](../../home/exceptions.md#jsonexceptionout_of_range407) if `j` contains an unsigned integer
-  above 9223372036854775807, which BON8 cannot represent
+- Throws [out_of_range.407](../../home/exceptions.md#jsonexceptionout_of_range407) if `j` contains an integer outside
+  the range of int64 (an unsigned integer above 9223372036854775807, or, with a number type wider than 64 bits, any
+  integer beyond int64), which BON8 cannot represent
 - Throws [type_error.316](../../home/exceptions.md#jsonexceptiontype_error316) if `j` contains a string that is not
   valid UTF-8
 - Throws [type_error.321](../../home/exceptions.md#jsonexceptiontype_error321) if `j` or a value nested in it is

docs/mkdocs/docs/api/basic_json/to_bson.md

@@ -47,6 +47,10 @@ Strong guarantee: if an exception is thrown, there are no changes in the JSON va
 
 - Throws [`type_error.317`](../../home/exceptions.md#jsonexceptiontype_error317) if the top-level type of the JSON value
   is not an object; example: `"to serialize to BSON, top-level type must be object, but is string"`
+- Throws [`out_of_range.407`](../../home/exceptions.md#jsonexceptionout_of_range407) if `j` contains a signed integer
+  outside the range of int64 or an unsigned integer outside the range of uint64, which is only possible with a number
+  type wider than 64 bits; example:
+  `"integer number 9223372036854775808 cannot be represented by BSON as it does not fit int64"`
 - Throws [`out_of_range.409`](../../home/exceptions.md#jsonexceptionout_of_range409) if a key in the JSON object contains
   a null byte (code point U+0000); example: `"BSON key cannot contain code point U+0000 (at byte 2)"`
 - Throws [`out_of_range.412`](../../home/exceptions.md#jsonexceptionout_of_range412) if the length of a document, array,
@@ -119,3 +123,5 @@ pass before anything is written.
   that is not valid UTF-8 unchanged, as before; `strict` (the default if
   [`JSON_STRICT_BINARY_UTF8`](../macros/json_strict_binary_utf8.md) is enabled) throws `type_error.316` before anything
   is written.
+- Throws `out_of_range.407` for integers that do not fit 64 bits since version 3.13.0; previously, integers of a
+  number type wider than 64 bits were silently truncated.

docs/mkdocs/docs/api/basic_json/to_cbor.md

@@ -46,6 +46,9 @@ Strong guarantee: if an exception is thrown, there are no changes in the JSON va
 
 ## Exceptions
 
+- Throws [`out_of_range.407`](../../home/exceptions.md#jsonexceptionout_of_range407) if `j` contains an integer
+  outside [-2^64, 2^64-1], which is only possible with a number type wider than 64 bits; example:
+  `"integer number 18446744073709551616 cannot be represented by CBOR as it does not fit [-2^64, 2^64-1]"`
 - Throws [type_error.316](../../home/exceptions.md#jsonexceptiontype_error316) if a string or object key in `j` is
   not valid UTF-8 and `error_handler` is `strict` (the default only if
   [`JSON_STRICT_BINARY_UTF8`](../macros/json_strict_binary_utf8.md) is enabled)
@@ -90,3 +93,5 @@ Linear in the size of the JSON value `j`.
   [`JSON_STRICT_BINARY_UTF8`](../macros/json_strict_binary_utf8.md) is enabled) throws `type_error.316`.
 - Throws `type_error.321` for a discarded value since version 3.13.0; previously, a discarded value nested in an
   array or object was silently skipped, producing invalid CBOR.
+- Throws `out_of_range.407` for integers that do not fit 64 bits since version 3.13.0; previously, integers of a
+  number type wider than 64 bits were silently truncated.

docs/mkdocs/docs/api/basic_json/to_msgpack.md

@@ -46,6 +46,9 @@ Strong guarantee: if an exception is thrown, there are no changes in the JSON va
 
 ## Exceptions
 
+- Throws [`out_of_range.407`](../../home/exceptions.md#jsonexceptionout_of_range407) if `j` contains an integer
+  outside [-2^63, 2^64-1], which is only possible with a number type wider than 64 bits; example:
+  `"integer number 18446744073709551616 cannot be represented by MessagePack as it does not fit [-2^63, 2^64-1]"`
 - Throws [`out_of_range.412`](../../home/exceptions.md#jsonexceptionout_of_range412) if the length of a string, binary
   value, array, or object exceeds 4294967295, the maximum MessagePack can store; example:
   `"MessagePack length 4294967296 exceeds maximum of 4294967295"`
@@ -112,3 +115,5 @@ Linear in the size of the JSON value `j`.
   `number_unsigned_t`.
 - Throws `type_error.321` for a discarded value since version 3.13.0; previously, a discarded value nested in an
   array or object was silently skipped, producing invalid MessagePack.
+- Throws `out_of_range.407` for integers that do not fit 64 bits since version 3.13.0; previously, integers of a
+  number type wider than 64 bits were silently truncated.

docs/mkdocs/docs/api/basic_json_document/erase.md

@@ -18,10 +18,10 @@ has `erase`; calling it on a read-only `basic_json_document` fails to compile (`
 1. Removes every member of `object` whose key is `key` (see [Notes](#notes) on duplicate keys) and returns how many
    were removed; `#!cpp 0` if `object` has no member with this key.
 2. Removes the element at index `idx` of `array`, which must already exist (`#!cpp idx < array.size()`).
-3. Removes the value the JSON pointer `ptr` refers to, relative to [`root()`](root.md), and returns how many values
-   were removed: the *parent* of the target must already exist, and the target itself is removed as in 1. (an object
-   member; `#!cpp 0` or more) or 2. (an array element; always `#!cpp 1`). `ptr` must not be empty -- [`root()`](root.md)
-   itself cannot be erased.
+3. Removes the value the JSON pointer `ptr` refers to, relative to [`root()`](root.md). `ptr` must not be empty
+   ([`root()`](root.md) itself cannot be erased), and the *parent* of the target must already exist. The target is
+   removed as in 1. if the parent is an object (returns the number of removed members, `#!cpp 0` or more) and as in 2.
+   if it is an array (always returns `#!cpp 1`).
 
 ## Template parameters
 
@@ -69,10 +69,10 @@ has `erase`; calling it on a read-only `basic_json_document` fails to compile (`
    [`out_of_range.404`](../../home/exceptions.md#jsonexceptionout_of_range404) (an empty token); otherwise (an
    object, or a primitive value the pointer's parent resolves to) throws what 1. throws.
 
-Every overload also throws [`invalid_iterator.202`](../../home/exceptions.md#jsonexceptioninvalid_iterator202) ("view
-does not belong to this document") if `object`/`array` is a [discarded](../basic_json_view/is_discarded.md) view or a
-view of a *different* document (overloads 1-2 only; overload 3 always starts from this document's own
-[`root()`](root.md)).
+Overloads 1. and 2. also throw [`invalid_iterator.202`](../../home/exceptions.md#jsonexceptioninvalid_iterator202)
+("view does not belong to this document") if `object`/`array` is a [discarded](../basic_json_view/is_discarded.md)
+view or a view of a *different* document. Overload 3. never throws `invalid_iterator.202`: it takes no view, and it
+always starts from this document's own [`root()`](root.md).
 
 ## Complexity
 
@@ -92,9 +92,10 @@ view of a *different* document (overloads 1-2 only; overload 3 always starts fro
 
 Like [`set`](set.md) and [`push_back`](push_back.md), `erase` never moves an element's *value*: a view still
 referring to a removed member or element keeps showing what it last held (see [Edits](index.md#edits)) -- it just no
-longer appears when `array`/`object` is read, dumped, or iterated. Removing an element of `array` (2.) does shift the
-*links* to the elements after it, the same way `insert`, `set`, or `push_back` on the same array would; any iterator
-already taken over `array`/`object` is invalidated by an erase, since it was walking the old layout.
+longer appears when `array`/`object` is read, dumped, or iterated. What does change is the container's sequence of
+*links* to its elements: removing an element of `array` (2.) shifts the links to the elements after it by one, as
+[`insert`](insert.md) does in the other direction. *Every* iterator already taken over `array`/`object` is invalidated
+by an erase, not only those at or after the removed element, since the container may now use a different sequence.
 
 ## Examples
 

docs/mkdocs/docs/api/basic_json_document/load.md

@@ -110,17 +110,30 @@ enum class image_check
 
 How thoroughly `load()` validates `image` before trusting it.
 
-| value    | checks                                                                                                    | guarantees |
-|----------|--------------------------------------------------------------------------------------------------------------|------------|
-| `full`   | everything the parser itself guarantees: structure and bounds; that every string is valid UTF-8 (and, for a string still in the source text, that it contains no quote, backslash, or control character); and that every number token is well-formed and matches the value stored for it; strings and float tokens may share a range only if the ranges are identical (as nodes that share a value do), and never overlap otherwise | reading and serializing a checked image is safe and always produces valid JSON, exactly as for a parsed document |
-| `bounds` | structure and bounds only -- that every offset and count in the node index stays inside the image            | reading and serializing stay memory-safe, but a crafted image can hold strings that are not valid UTF-8 or that serialize to invalid JSON ([`dump()`](../basic_json_view/dump.md) writes them unchanged or throws [`type_error.316`](../../home/exceptions.md#jsonexceptiontype_error316)), and numbers whose values differ from their text |
-| `none`   | nothing                                                                                                       | images from a trusted source only -- reading a damaged image is undefined behavior |
+`full` (the default)
+:   Checks everything the parser itself guarantees: structure and bounds; that every string is valid UTF-8 (and, for a
+    string still in the source text, that it contains no quote, backslash, or control character); and that every number
+    token is well-formed and matches the value stored for it. Strings and float tokens may share a range only if the
+    ranges are identical (as nodes that share a value do), and never overlap otherwise. Reading and serializing a
+    checked image is safe and always produces valid JSON, exactly as for a parsed document. This is the right choice for
+    an image from anything you do not fully control -- a file, a cache shared with other processes, a peer on the
+    network.
 
-`full` is the default and the right choice for an image from anything you do not fully control -- a file, a cache
-shared with other processes, a peer on the network. `bounds` skips scanning the text and the decoded strings, so it
-fits a cache your own process just wrote and reads straight back, where damage would mean a bug or a hardware fault
-rather than adversarial input; it still cannot crash or read out of bounds. `none` skips validation entirely and
-should only be used for an image you trust as much as your own memory.
+`bounds`
+:   Checks structure and bounds only: that every offset and count in the node index stays inside the image. So
+    `load()`, [`root()`](root.md), element access, and [`materialize()`](../basic_json_view/materialize.md) can never
+    read outside the image, even on a damaged one. It does *not* guarantee that the image describes valid JSON: a
+    string that a `full` check would have rejected can make [`dump()`](../basic_json_view/dump.md) write invalid UTF-8
+    or invalid JSON, or throw [`type_error.316`](../../home/exceptions.md#jsonexceptiontype_error316), and a number can
+    read back with a value that does not match how it is spelled. An [editable document](../json_editable_document.md)
+    does not take such a string over: [`set`](set.md), [`push_back`](push_back.md) and [`insert`](insert.md) throw
+    `type_error.316` when they copy it from a view of the loaded document, and leave the editable document unchanged.
+    `bounds` skips scanning the text and the decoded strings, so it fits a cache your own process just wrote and reads
+    straight back, where damage would mean a bug or a hardware fault rather than adversarial input.
+
+`none`
+:   Checks nothing. Reading a damaged image is undefined behavior, so use it only for an image you trust as much as
+    your own memory.
 
 **Lifetime.** Overloads 1 and 2 borrow `image`: it must stay alive and byte-for-byte unchanged for as long as the
 returned document, and any [view](../basic_json_view/index.md) taken from it, is used -- exactly like a document
@@ -134,16 +147,6 @@ returned document, and any [view](../basic_json_view/index.md) taken from it, is
     ("unknown format"). Use images to cache a document within one build of the library, or to hand one to another
     process running the *same* build on the *same* (little-endian) machine -- not as a long-term storage format.
 
-**What `image_check::bounds` does not guarantee.** A bounds-checked image can never make `load()`,
-[`root()`](root.md), element access, or [`materialize()`](../basic_json_view/materialize.md) read outside the image,
-so those stay safe on a damaged one. It does *not* guarantee that the image describes valid JSON: a string
-that a `full` check would have rejected can make [`dump()`](../basic_json_view/dump.md) write invalid UTF-8 or invalid
-JSON, or throw `type_error.316`, and a number can read back with a value that does not match how it is spelled.
-Reserve `bounds` for images you already trust to be well-formed, and use it only to skip the extra scan.
-An [editable document](../json_editable_document.md) does not take such a string over either:
-[`set`](set.md), [`push_back`](push_back.md) and [`insert`](insert.md) throw `type_error.316` when they copy it from
-a view of the loaded document, and leave the editable document unchanged.
-
 ## Examples
 
 ??? example "Caching a document, ownership, and a rejected image"

docs/mkdocs/docs/api/basic_json_view/at.md

@@ -41,7 +41,7 @@ basic_json_view at(const json_pointer& ptr) const;
 
 ## Exception safety
 
-Strong exception safety: if an exception is thrown, there are no changes to the view or the document it refers to.
+Strong guarantee: there are no changes to the view or the document it refers to.
 
 ## Exceptions
 
@@ -77,8 +77,8 @@ None of these exceptions carry a [`JSON_DIAGNOSTICS`](../macros/json_diagnostics
 1. Linear in the number of members: as for [`ordered_json`](../ordered_json.md), members are compared one after
    another, in document order, stopping at the first match. Each comparison first checks the
    key's length -- already known from the index, without reading the key bytes -- before comparing its content.
-   Objects with 128 or more members get a hash index while parsing, so that a lookup in them takes constant time
-   on average.
+   Objects with 128 or more members get a hash index of their keys while parsing, so that a lookup in them takes
+   constant time on average.
 2. Linear in `idx`: elements are skipped one at a time from the first one, since they are not a fixed size in the
    index (unlike `BasicJsonType`'s array, which is random-access).
 3. Linear in the number of reference tokens of `ptr` and, for each token, in the number of members of the object at

docs/mkdocs/docs/api/basic_json_view/back.md

@@ -13,7 +13,7 @@ The last element or member value. For a primitive value (number, string, boolean
 
 ## Exception safety
 
-Strong exception safety: if an exception is thrown, there are no changes to the view or the document it refers to.
+Strong guarantee: there are no changes to the view or the document it refers to.
 
 ## Exceptions
 
@@ -33,8 +33,7 @@ index.
 ## Notes
 
 Unlike [`BasicJsonType::back()`](../basic_json/back.md), which has undefined behavior for an empty array or object,
-`back()` throws `invalid_iterator.214` in that case -- the same way it already does for `#!json null` and for a
-discarded view, where `BasicJsonType::back()` also throws.
+`back()` throws `invalid_iterator.214` for an empty array, an empty object, a null value, or a discarded view.
 
 ## Examples
 

docs/mkdocs/docs/api/basic_json_view/begin.md

@@ -6,7 +6,7 @@ iterator begin() const noexcept;
 
 Returns an iterator to the first element of an array, or the first member value of an object, in **document order**
 -- the order the values appear in the source text, not sorted by key. A primitive value iterates as a range of one
-element (itself); `#!json null` and a [discarded](is_discarded.md) view iterate as an empty range.
+element (itself), and a `#!json null` value or a [discarded](is_discarded.md) view iterates as an empty range.
 
 ## Return value
 
@@ -23,7 +23,7 @@ Constant.
 ## Notes
 
 For an object, iteration visits **every** member, including all occurrences of a duplicate key -- unlike
-[`operator[]`](operator[].md), [`at`](at.md), [`find`](find.md), [`contains`](contains.md), and [`count`](count.md),
+[`operator[]`](operator[].md), [`at`](at.md), [`find`](find.md), and [`contains`](contains.md),
 which all resolve to the *first* member with a given key. See the
 [Notes on duplicate keys](operator[].md#notes) of `operator[]`.
 
@@ -31,6 +31,10 @@ Because objects are iterated in document order rather than sorted by key, the or
 iterating the [`materialize()`](materialize.md)d `BasicJsonType` value would produce: a `#!cpp basic_json` object
 (`std::map`-backed by default) sorts its keys, while a view does not.
 
+An `ordered_json` value also keeps document order, but it keeps only one entry per key: as `parse()` does, it keeps the
+*last* value, at the position of the key's *first* occurrence. The view, in contrast, visits every member, including
+the duplicates.
+
 ## Examples
 
 ??? example

docs/mkdocs/docs/api/basic_json_view/contains.md

@@ -35,8 +35,8 @@ No-throw guarantee: this function never throws exceptions.
 1. Linear in the number of members: as for [`ordered_json`](../ordered_json.md), members are compared one after
    another, in document order, stopping at the first match. Each comparison first checks the
    key's length -- already known from the index, without reading the key bytes -- before comparing its content.
-   Objects with 128 or more members get a hash index while parsing, so that a lookup in them takes constant time
-   on average.
+   Objects with 128 or more members get a hash index of their keys while parsing, so that a lookup in them takes
+   constant time on average.
 2. Linear in the number of reference tokens of `ptr` and, for each token, in the number of members of the object at
    that level or the index into the array -- as for [`operator[]`](operator[].md#complexity) and
    [`at`](at.md#complexity) with a JSON pointer.
@@ -55,7 +55,8 @@ view.
 
 !!! info "Postconditions"
 
-    If `#!cpp v.contains(key)` returns `#!cpp true`, then `#!cpp v[key]` is not [discarded](is_discarded.md). If
+    If `#!cpp v.contains(key)` returns `#!cpp true`, then `#!cpp v[key]` is not [discarded](is_discarded.md) and
+    `#!cpp v.at(key)` does not throw. If
     `#!cpp v.contains(ptr)` returns `#!cpp true`, then `#!cpp v[ptr]` is not discarded and `#!cpp v.at(ptr)` does not
     throw.
 

docs/mkdocs/docs/api/basic_json_view/count.md

@@ -6,7 +6,8 @@ size_type count(const char* key) const;
 size_type count(const string_t& key) const;
 ```
 
-Returns `#!cpp 1` if the value is an object with a member with key `key`, `#!cpp 0` otherwise.
+Returns the number of members with key `key` if the value is an object, `#!cpp 0` otherwise. Members with a repeated
+key are all counted.
 
 ## Parameters
 
@@ -15,7 +16,8 @@ Returns `#!cpp 1` if the value is an object with a member with key `key`, `#!cpp
 
 ## Return value
 
-`#!cpp 1` if the value is an object and has a member with key `key`, `#!cpp 0` otherwise.
+the number of members of the object with key `key` (duplicates included), or `#!cpp 0` if the value is not an object
+or has no such member.
 
 ## Exception safety
 
@@ -23,36 +25,30 @@ No-throw guarantee: this function never throws exceptions.
 
 ## Complexity
 
-Linear in the number of members: as for [`ordered_json`](../ordered_json.md), members are compared one after another,
-in document order, stopping at the first match. Each comparison first checks the key's length
--- already known from the index, without reading the key bytes -- before comparing its content.
-Objects with 128 or more members get a hash index while parsing, so that a lookup in them takes constant time on
-average.
+Linear in the number of members: unlike the lookup functions, `count()` does not stop at the first match, because it
+has to count every member with the key. Members are compared one after another, in document order. Each comparison
+first checks the key's length -- already known from the index, without reading the key bytes -- before comparing its
+content. The hash index of the keys of objects with 128 or more members does not help here, since every member
+is visited.
 
 ## Notes
 
-!!! warning "Duplicate keys: the first member wins"
-
-    If the source text repeats a key, this resolves to the *first* member with it, not to the last one that
-    [`materialize()`](materialize.md) and `parse()` keep. See the [Notes on duplicate keys](operator[].md#notes) of
-    `operator[]` and [Duplicate keys](../../features/json_view.md#duplicate-keys) for the reasons and for how to get
-    the last value.
-
 This method always returns `#!cpp 0` when the value is not an object -- including a [discarded](is_discarded.md)
 view.
 
-Unlike [`BasicJsonType::count()`](../basic_json/count.md), whose return value can in principle exceed `#!cpp 1` for
-an `ObjectType` that allows multiple entries per key, `count()` here never does: it is exactly
-[`contains()`](contains.md) as `#!cpp 0`/`#!cpp 1`. This holds even if the source text has a duplicate key -- see the
-[Notes on duplicate keys](operator[].md#notes) of `operator[]` -- because a `#!cpp count() > 1` result would require
-counting every member with a matching key (the lookup functions resolve to the *first* one).
+If the source text repeats a key, `count()` returns more than `#!cpp 1` for it, even though the
+[materialized](materialize.md) value (like the result of `parse()`) would have only one member with that key. Hence
+`#!cpp count(key) > 1` detects duplicate keys. This is unlike the lookup functions
+([`operator[]`](operator[].md), [`at`](at.md), [`find`](find.md), [`contains`](contains.md)), which resolve to the
+*first* member; see the [Notes on duplicate keys](operator[].md#notes) of `operator[]`.
 
 ## Examples
 
 ??? example
 
-    The example below validates that every transaction of a batch carries a mandatory `amount` field, using
-    `count()` before deciding whether to materialize a transaction at all.
+    The example below validates that every transaction of a batch carries exactly one `amount` field, using
+    `count()` before deciding whether to materialize a transaction at all. It reports a transaction with no `amount`
+    field and one with a duplicated `amount` field.
 
     ```cpp
     --8<-- "examples/basic_json_view__count.cpp"

docs/mkdocs/docs/api/basic_json_view/dump.md

@@ -60,11 +60,13 @@ Linear in the size of the output text.
 The walk over the subtree is iterative, so the nesting depth it can write is limited by available memory only, not by
 the call stack -- as for [`materialize()`](materialize.md).
 
-Strings are escaped by the same rules as [`BasicJsonType::dump`](../basic_json/dump.md). With
-`#!cpp numbers == number_format::shortest`, floats are written with the library's shortest round-trip conversion,
+Strings are escaped by the same rules as [`BasicJsonType::dump`](../basic_json/dump.md).
+
+With `#!cpp numbers == number_format::shortest`, floats are written with the library's shortest round-trip conversion,
 exactly as [`BasicJsonType::dump`](../basic_json/dump.md) would (e.g. `#!cpp 1.5`, `#!cpp 100.0`, `#!cpp 1e+100`), and
 integers are copied from the source text -- already canonical in JSON, so this matches their shortest form too --
 except that `#!cpp -0` is written as `#!cpp 0`, the way [`BasicJsonType::parse()`](../basic_json/parse.md) reads it.
+
 `#!cpp number_format::source` copies every number exactly as written in the source text instead, with no exception
 for `#!cpp -0` -- `#!cpp 1.50`, `#!cpp 1E2`, `#!cpp -0.0`, `#!cpp -0`, or all digits of an integer literal with more
 digits than any number type holds (such a literal is itself classified as a float, see

docs/mkdocs/docs/api/basic_json_view/end.md

@@ -5,9 +5,8 @@ iterator end() const noexcept;
 ```
 
 Returns an iterator to one past the last element of an array, one past the last member value of an object, in
-**document order** -- see [`begin()`](begin.md). A primitive value iterates as a range of one element (itself);
-`#!json null` and a [discarded](is_discarded.md) view iterate as an empty range, so `#!cpp begin() == end()` for
-them.
+**document order** -- see [`begin()`](begin.md). A primitive value iterates as a range of one element (itself), and a
+`#!json null` value or a [discarded](is_discarded.md) view iterates as an empty range (`#!cpp begin() == end()`).
 
 ## Return value
 

docs/mkdocs/docs/api/basic_json_view/find.md

@@ -28,8 +28,8 @@ No-throw guarantee: this function never throws exceptions.
 Linear in the number of members: as for [`ordered_json`](../ordered_json.md), members are compared one after another,
 in document order, stopping at the first match. Each comparison first checks the key's length
 -- already known from the index, without reading the key bytes -- before comparing its content.
-Objects with 128 or more members get a hash index while parsing, so that a lookup in them takes constant time on
-average.
+Objects with 128 or more members get a hash index of their keys while parsing, so that a lookup in them takes
+constant time on average.
 
 ## Notes
 

docs/mkdocs/docs/api/basic_json_view/front.md

@@ -13,7 +13,7 @@ The first element or member value. For a primitive value (number, string, boolea
 
 ## Exception safety
 
-Strong exception safety: if an exception is thrown, there are no changes to the view or the document it refers to.
+Strong guarantee: there are no changes to the view or the document it refers to.
 
 ## Exceptions
 
@@ -31,8 +31,8 @@ Constant.
 ## Notes
 
 Unlike [`BasicJsonType::front()`](../basic_json/front.md), which has undefined behavior for an empty array or
-object, `front()` throws `invalid_iterator.214` in that case -- the same way it already does for `#!json null` and
-for a discarded view, where `BasicJsonType::front()` also throws.
+object, `front()` throws `invalid_iterator.214` for an empty array, an empty object, a null value, or a discarded
+view.
 
 ## Examples
 

docs/mkdocs/docs/api/basic_json_view/get.md

@@ -10,8 +10,8 @@ Converts the value to `T`.
 For the types below, the conversion works directly on the flat index -- no `BasicJsonType` value is built for it:
 
 - `#!cpp bool`
-- arithmetic types other than `#!cpp bool` (from a number; from a boolean, as `#!cpp 0`/`#!cpp 1`, exactly as
-  [`BasicJsonType::get<T>()`](../basic_json/get.md) converts a boolean)
+- arithmetic types other than `#!cpp bool` (for a boolean, `#!cpp 0`/`#!cpp 1`, exactly as for
+  [`BasicJsonType::get<T>()`](../basic_json/get.md))
 - `#!cpp std::nullptr_t`
 - `#!cpp std::basic_string<char, Traits, Alloc>` (including `string_t`) -- a copy of the string
 - [`string_view_t`](index.md#member-types) -- **no copy**: the returned view points into the document's
@@ -19,10 +19,10 @@ For the types below, the conversion works directly on the flat index -- no `Basi
   document's own buffer of decoded strings (see [`get_string()`](get_string.md))
 - `BasicJsonType` -- equivalent to [`materialize()`](materialize.md)
 - `basic_json_view` -- returns `#!cpp *this`
-- `#!cpp std::vector<U, A>` -- element by element, each converted with `#!cpp get<U>()`; `#!cpp
+- `#!cpp std::vector<U, A>` -- from an array, element by element, each converted with `#!cpp get<U>()`; `#!cpp
   std::vector<basic_json_view>` keeps a view of every element instead of a value
 - `#!cpp std::map<K, V, C, A>` and `#!cpp std::unordered_map<K, V, H, E, A>`, if `K` is constructible from a `#!cpp
-  (const char*, std::size_t)` pair -- member by member, each value converted with `#!cpp get<V>()`; with a repeated
+  (const char*, std::size_t)` pair -- from an object, member by member, each value converted with `#!cpp get<V>()`; with a repeated
   key, the *last* value is kept, as [`BasicJsonType::parse()`](../basic_json/parse.md) (and
   [`materialize()`](materialize.md)) does; `#!cpp std::map<std::string, basic_json_view>` keeps views of the members
   instead of values
@@ -43,7 +43,7 @@ the value, converted to `T`
 
 ## Exception safety
 
-Strong exception safety: if an exception is thrown, there are no changes to the view or the document it refers to.
+Strong guarantee: there are no changes to the view or the document it refers to.
 
 ## Exceptions
 
@@ -62,6 +62,9 @@ None of the exceptions thrown directly by this function (the first three bullets
 [`JSON_DIAGNOSTICS`](../macros/json_diagnostics.md) path: the view has no `BasicJsonType` value to point at. An
 exception thrown while converting through `materialize()` (the last bullet) is different: it is thrown by a real
 `BasicJsonType` value, so it **does** carry a `JSON_DIAGNOSTICS` path if `BasicJsonType` was built with it enabled.
+That path is relative to the root of the view's subtree, because `materialize()` builds only the subtree (with parent
+pointers): it is empty if the failing value is the subtree's root itself, and it never is the path within the whole
+document.
 
 ## Complexity
 
@@ -73,8 +76,8 @@ exception thrown while converting through `materialize()` (the last bullet) is d
 - `#!cpp std::map`/`#!cpp std::unordered_map`: linear in the number of members for walking them, plus the container's
   own insertion cost per member (logarithmic for `#!cpp std::map`, amortized constant for `#!cpp
   std::unordered_map`), times the complexity of converting one member to the mapped type.
-- every other `T`: linear in the size of the subtree (building the `BasicJsonType` value), plus the complexity of
-  [`BasicJsonType::get<T>()`](../basic_json/get.md) on it.
+- every other `T`: the cost of [`materialize()`](materialize.md) and of destroying the `BasicJsonType` value
+  afterwards, plus that of [`BasicJsonType::get<T>()`](../basic_json/get.md).
 
 ## Notes
 

docs/mkdocs/docs/api/basic_json_view/get_string.md

@@ -15,7 +15,7 @@ had to be decoded once when the document was parsed).
 
 ## Exception safety
 
-Strong exception safety: if an exception is thrown, there are no changes to the view or the document it refers to.
+Strong guarantee: there are no changes to the view or the document it refers to.
 
 ## Exceptions
 
@@ -37,9 +37,14 @@ hand out a reference to a stored `string_t`. `get_string()` (equivalently, [`get
 zero-copy alternative: [`BasicJsonType::get_ref<const string_t&>()`](../basic_json/get_ref.md) is its closest
 counterpart, except that it returns a view instead of a reference to a value that must already exist.
 
+`get_string()` borrows the string: nothing is copied, whereas [`get<std::string>()`](get.md) copies it. For a borrowed
+document, the returned [`string_view_t`](index.md#member-types) is valid as long as the caller's buffer; for an owned
+one, as long as the document.
+
 The returned [`string_view_t`](index.md#member-types) is valid exactly as long as the view that produced it -- see the
 [validity rules](index.md) of `basic_json_view` -- and, for a string with no escapes, for as long as the document's
-source text.
+source text, which for a borrowed document must remain valid **and unmodified** for as long as the document exists,
+since floating-point values and unescaped strings are read from the source lazily.
 
 ## Examples
 

docs/mkdocs/docs/api/basic_json_view/get_to.md

@@ -5,12 +5,15 @@ template<typename T>
 T& get_to(T& v) const;
 ```
 
-Converts the value to `T` and assigns it to `v`. Equivalent to
+Converts the value to `T` and stores it in `v`, in one of two ways:
 
-```cpp
-v = get<T>();
-return v;
-```
+- For the types that [`get<T>()`](get.md) converts directly -- arithmetic types, `#!cpp bool`,
+  `#!cpp std::nullptr_t`, [`string_view_t`](index.md#member-types), `#!cpp std::basic_string`, `BasicJsonType`,
+  `basic_json_view`, `#!cpp std::vector`, and `#!cpp std::map`/`#!cpp std::unordered_map` with string keys -- `v` is
+  assigned: `#!cpp v = get<T>()`.
+- For every other type, the subtree is built into a `BasicJsonType` value and `#!cpp materialize().get_to(v)` is
+  called, i.e. `from_json()` is invoked with the existing `v`, exactly as
+  [`BasicJsonType::get_to()`](../basic_json/get_to.md) does.
 
 ## Template parameters
 
@@ -28,15 +31,27 @@ return v;
 
 ## Exception safety
 
-Strong exception safety: if an exception is thrown, `v` is not modified.
+- For the directly converted types: strong exception safety. If an exception is thrown, `v` is not modified.
+- For every other type: as for [`BasicJsonType::get_to()`](../basic_json/get_to.md). `materialize()` throws before `v`
+  is touched, but a `from_json()` that throws may leave `v` partly modified.
 
 ## Exceptions
 
-Whatever [`get<T>()`](get.md) throws for the same value and `T`.
+- For the directly converted types: whatever [`get<T>()`](get.md) throws for the same value and `T`.
+- For every other type: whatever [`materialize()`](materialize.md) and
+  [`BasicJsonType::get_to()`](../basic_json/get_to.md) throw.
 
 ## Complexity
 
-Whatever [`get<T>()`](get.md) has for the same `T`.
+- For the directly converted types: whatever [`get<T>()`](get.md) has for the same `T`.
+- For every other type: that of [`materialize()`](materialize.md), plus that of
+  [`BasicJsonType::get_to()`](../basic_json/get_to.md).
+
+## Notes
+
+The difference between the two ways matters for a `from_json()` that sets only the fields present in the JSON: since
+it runs on the existing `v`, the other fields of `v` keep their values, whereas assigning `get<T>()` would replace
+them with those of a default-constructed `T`.
 
 ## Examples
 

docs/mkdocs/docs/api/basic_json_view/index.md

@@ -14,18 +14,27 @@ to the document and a pointer into its index), trivially copyable. A view is val
 - the document has not been re-parsed with [`read()`](../basic_json_document/read.md) (or
   [`parse()`](../basic_json_document/parse.md) into it) or shrunk with
   [`shrink_to_fit()`](../basic_json_document/shrink_to_fit.md) since the view was taken, and
-- if the document borrows its source text, that text is still alive.
+- if the document borrows its source text, that text is still alive and unmodified (floating-point values and
+  unescaped strings are read from it lazily).
 
 Moving the document itself does not invalidate its views: the index is heap-allocated independently of the
 `basic_json_document` object.
 
-`basic_json_view` provides the read-only part of the `BasicJsonType` interface: the type-inspection functions, element
-access, lookup, iteration, conversion, and comparison -- [`get<T>()`](get.md), [`get_string()`](get_string.md),
-[`number_token()`](number_token.md), and [`materialize()`](materialize.md) to build the `BasicJsonType` value of a
-subtree on demand. [`operator[]`](operator%5B%5D.md), [`at`](at.md), [`contains`](contains.md), and
-[`value`](value.md) also accept a [`json_pointer`](../json_pointer/index.md). [`operator==`](operator_eq.md) and
-[`operator!=`](operator_ne.md) compare two views, or a view and a `BasicJsonType` value, without ever building a
-`BasicJsonType` value for a view; no ordering comparison (`#!cpp operator<`) is provided.
+`basic_json_view` provides the read-only part of the `BasicJsonType` interface:
+
+- the type-inspection functions
+- element access
+- lookup
+- iteration
+- conversion: [`get<T>()`](get.md), [`get_string()`](get_string.md), [`number_token()`](number_token.md), and
+  [`materialize()`](materialize.md) to build the `BasicJsonType` value of a subtree on demand
+- comparison: [`operator==`](operator_eq.md) and [`operator!=`](operator_ne.md) compare two views, or a view and a
+  `BasicJsonType` value, without ever building a `BasicJsonType` value for a view; no ordering comparison
+  (`#!cpp operator<`) is provided
+
+[`operator[]`](operator%5B%5D.md), [`at`](at.md), [`contains`](contains.md), and [`value`](value.md) also accept a
+[`json_pointer`](../json_pointer/index.md). Unlike `basic_json`, there are no implicit conversion operators (no
+`#!cpp operator ValueType()`), and no `#!cpp get_ptr()` or `#!cpp get_ref()`; see [`get`](get.md).
 
 `basic_json_view` itself is always read-only -- it never has a `set` or `push_back` of its own. A view of an
 **editable** document (`#!cpp Editable == true`) sees every edit made through
@@ -61,7 +70,6 @@ still refers to it -- including ones taken before the change -- reads the new va
 - **string_view_t** - `#!cpp std::string_view` on C++17 and newer, a minimal internal substitute otherwise
 - **iterator**, **const_iterator** - a forward iterator over the elements of an array or the member values of an
   object, in document order; both names refer to the same type, since a view is always read-only
-- **item** - a (key, value) pair produced by [`items()`](items.md)
 - [**number_format**](number_format.md) - how [`dump()`](dump.md) writes numbers
 
 ## Member functions

docs/mkdocs/docs/api/basic_json_view/items.md

@@ -4,9 +4,9 @@
 /* unspecified */ items() const noexcept;
 ```
 
-Returns a range of [`item`](index.md#member-types) values -- (key, value) pairs -- for use in range-based for loops.
-The key of an array element is its index, converted to a string, as for
-[`BasicJsonType::items()`](../basic_json/items.md).
+Returns a range of items -- (key, value) pairs -- for use in range-based for loops. An item's `#!cpp key()` is a
+`#!cpp string_view_t` (for an array element, its index converted to a string, as for
+[`BasicJsonType::items()`](../basic_json/items.md)), and its `#!cpp value()` is a `basic_json_view` of the element.
 
 The returned type is not part of the public API and may change between versions; use a range-based for loop (see the
 example), or `#!cpp decltype(v.items())` if you need to name it.
@@ -33,8 +33,8 @@ is nothing for a reference to bind to.
 
 ## Return value
 
-A range whose iterators dereference to [`item`](index.md#member-types) and whose `#!cpp begin()`/`#!cpp end()` are
-equivalent to [`basic_json_view::begin()`](begin.md)/[`end()`](end.md), in document order.
+A range whose iterators dereference to an item (with `#!cpp key()` and `#!cpp value()`, see above) and whose
+`#!cpp begin()`/`#!cpp end()` are equivalent to [`basic_json_view::begin()`](begin.md)/[`end()`](end.md), in document order.
 
 ## Exception safety
 
@@ -48,14 +48,14 @@ Constant.
 
 As for [`begin()`](begin.md)/[`end()`](end.md), `items()` visits **every** member of an object, including all
 occurrences of a duplicate key -- unlike [`operator[]`](operator[].md), [`at`](at.md), [`find`](find.md),
-[`contains`](contains.md), and [`count`](count.md), which resolve to the *first* member with a given key. See the
+and [`contains`](contains.md), which resolve to the *first* member with a given key. See the
 [Notes on duplicate keys](operator[].md#notes) of `operator[]`.
 
 !!! danger "Lifetime issues"
 
-    As for `BasicJsonType::items()`, calling `items()` on a temporary view (or a temporary document) is dangerous:
-    the range refers back to the document, so the document must outlive the loop. See
-    [#2040](https://github.com/nlohmann/json/issues/2040) for the `BasicJsonType` background.
+    Unlike `BasicJsonType::items()` (see [#2040](https://github.com/nlohmann/json/issues/2040)), calling `items()` on
+    a temporary view is safe, for instance `#!cpp for (auto item : v["a"].items())`: the returned range holds a copy
+    of the view. What must outlive the loop is the document and, for a borrowed document, the source buffer.
 
 ## Examples
 

docs/mkdocs/docs/api/basic_json_view/number_token.md

@@ -15,7 +15,7 @@ and [`materialize()`](materialize.md) can only represent approximately, as a `nu
 
 ## Exception safety
 
-Strong exception safety: if an exception is thrown, there are no changes to the view or the document it refers to.
+Strong guarantee: there are no changes to the view or the document it refers to.
 
 ## Exceptions
 

docs/mkdocs/docs/api/basic_json_view/operator[].md

@@ -50,7 +50,7 @@ basic_json_view operator[](const json_pointer& ptr) const;
 
 ## Exception safety
 
-Strong exception safety: if an exception is thrown, there are no changes to the view or the document it refers to.
+Strong guarantee: there are no changes to the view or the document it refers to.
 
 ## Exceptions
 
@@ -83,8 +83,8 @@ None of these exceptions carry a [`JSON_DIAGNOSTICS`](../macros/json_diagnostics
    another, in document order, stopping at the first match. Each comparison first checks the
    key's length -- already known from the index, without reading the key bytes -- before comparing its content, so a
    key of a different length than `key` is rejected without touching the source text.
-   Objects with 128 or more members get a hash index while parsing, so that a lookup in them takes constant time
-   on average.
+   Objects with 128 or more members get a hash index of their keys while parsing, so that a lookup in them takes
+   constant time on average.
 2. Linear in `idx`: elements are skipped one at a time from the first one, since they are not a fixed size in the
    index (unlike `BasicJsonType`'s array, which is random-access).
 3. Linear in the number of reference tokens of `ptr` and, for each token, in the number of members of the object at
@@ -110,13 +110,13 @@ document.
 !!! warning "Duplicate keys: the first member wins"
 
     If the source text has an object with a duplicate key, `#!cpp operator[]` (and [`at`](at.md), [`find`](find.md),
-    [`contains`](contains.md), [`count`](count.md), [`value`](value.md), and JSON pointer resolution) all resolve to
+    [`contains`](contains.md), [`value`](value.md), and JSON pointer resolution) all resolve to
     the *first* member with that key, because a lookup can stop as soon as it finds a match. This is different from
     [`materialize()`](materialize.md) (and [`BasicJsonType::parse()`](../basic_json/parse.md)), which replay every
     member in order and so keep the *last* value for a repeated key, so `#!cpp v["a"]` and
     `#!cpp v.materialize()["a"]` can differ. To get the value `parse()` would give, use
     [`materialize()`](materialize.md) or iterate the members with [`items()`](items.md) and keep the last match.
-    The hash index of a larger object (128 members or more) leads to the first member of a key as well.
+    The key hash index of a larger object (128 members or more) leads to the first member of a key as well.
     [`begin()`](begin.md)/[`end()`](end.md) and [`items()`](items.md) iterate over *all* members, including
     duplicates, in document order. See [Duplicate keys](../../features/json_view.md#duplicate-keys) and
     [`size()`](size.md#notes).

docs/mkdocs/docs/api/basic_json_view/type_name.md

@@ -14,14 +14,14 @@ same string for a discarded `BasicJsonType` value).
 a string representation of the type ([`value_t`](../basic_json/value_t.md)):
 
 | Value type                                         | return value  |
-|-----------------------------------------------------|---------------|
-| `#!json null`                                        | `"null"`      |
-| boolean                                              | `"boolean"`   |
-| string                                               | `"string"`    |
-| number (integer, unsigned integer, floating-point)   | `"number"`    |
-| object                                               | `"object"`    |
-| array                                                | `"array"`     |
-| discarded                                            | `"discarded"` |
+|----------------------------------------------------|---------------|
+| `#!json null`                                      | `"null"`      |
+| boolean                                            | `"boolean"`   |
+| string                                             | `"string"`    |
+| number (integer, unsigned integer, floating-point) | `"number"`    |
+| object                                             | `"object"`    |
+| array                                              | `"array"`     |
+| discarded                                          | `"discarded"` |
 
 `type_name()` never returns `#!cpp "binary"`, since a JSON text has no binary values (see
 [`is_binary()`](is_binary.md)); it also never returns `#!cpp "invalid"`, since a view's `#!cpp kind` always comes

docs/mkdocs/docs/api/basic_json_view/value.md

@@ -44,7 +44,7 @@ equivalent) deduce `string_t`, not `const char*`, for their return type and for
 
 ## Exception safety
 
-Strong exception safety: if an exception is thrown, there are no changes to the view or the document it refers to.
+Strong guarantee: there are no changes to the view or the document it refers to.
 
 ## Exceptions
 
@@ -70,8 +70,8 @@ None of these exceptions carry a [`JSON_DIAGNOSTICS`](../macros/json_diagnostics
 1. Linear in the number of members: as for [`operator[]`](operator[].md#complexity), members are compared one after
    another, in document order, stopping at the first match. Plus the complexity of converting
    the found member to `T` (see [`get`](get.md)).
-   Objects with 128 or more members get a hash index while parsing, so that a lookup in them takes constant time
-   on average.
+   Objects with 128 or more members get a hash index of their keys while parsing, so that a lookup in them takes
+   constant time on average.
 2. Linear in the number of reference tokens of `ptr` and, for each token, in the number of members of the object at
    that level or the index into the array -- as for the [`operator[]`](operator[].md#complexity) and
    [`at`](at.md#complexity) overloads that take a JSON pointer. Plus the complexity of converting the resolved value

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

@@ -16,6 +16,10 @@ header. See also the [macro overview page](../../features/macros.md).
 
 ## Parsing
 
+- [**JSON_DISABLE_STREAM_BULK_SCAN**](json_disable_stream_bulk_scan.md) - read a `std::istream` one character at a time
+  instead of scanning its stream buffer in bulk
+- [**JSON_DISABLE_UNLOCKED_FILE_IO**](json_disable_unlocked_file_io.md) - read a `FILE*` with `std::fgetc` instead of
+  locking it once and using the unlocked stdio functions
 - [**JSON_PRECISE_STREAM_POSITION**](json_precise_stream_position.md) - opt in to leaving an input stream positioned
   right after a parsed number
 - [**JSON_STRICT_BINARY_UTF8**](json_strict_binary_utf8.md) - opt in to checking strings for valid UTF-8 in the CBOR,

docs/mkdocs/docs/api/macros/json_disable_stream_bulk_scan.md

@@ -0,0 +1,144 @@
+# JSON_DISABLE_STREAM_BULK_SCAN
+
+```cpp
+#define JSON_DISABLE_STREAM_BULK_SCAN /* value */
+```
+
+When defined to `1`, parsing from a `#!cpp std::istream` reads the input one character at a time through the stream
+buffer, as in version 3.12.0 and earlier, instead of scanning the characters that the stream buffer already holds in
+bulk.
+
+The macro is an escape hatch for stream buffers that do not follow the usual contract of `std::basic_streambuf` (see
+the requirements on the stream buffer in the notes below). It does not change any result: values, exceptions, and the
+position of the stream after parsing are the same either way. Only the speed differs.
+
+## Default definition
+
+The default value is `0`: streams are scanned in bulk.
+
+```cpp
+#define JSON_DISABLE_STREAM_BULK_SCAN 0
+```
+
+## Notes
+
+!!! note "How streams are scanned"
+
+    A `std::streambuf` keeps the part of the input it has read but not yet handed out in its *get area*, the range
+    from `gptr()` to `egptr()`. A `std::filebuf`, for instance, reads the file in blocks of a few kilobytes into its get
+    area, and a `std::stringbuf` exposes its whole string as one get area.
+
+    When parsing from a `std::istream`, the parser scans this get area directly, with the same fast paths that it uses
+    for a `std::string`: runs of ordinary string characters, numbers, and indentation are recognized several bytes at a
+    time instead of being requested from the stream buffer one character at a time. When the parser reaches the end of
+    the get area, it asks the stream buffer for the next character as before, which makes the stream buffer refill its
+    get area.
+
+    The parser only reads characters that it would also have read one at a time, and it advances the stream buffer
+    exactly over the characters it consumes. Nothing is read ahead, buffered elsewhere, or put back. Therefore, the
+    stream is left exactly where it was left before:
+
+    - [`operator>>`](../operator_gtgt.md) and [`sax_parse`](../basic_json/sax_parse.md) with `strict = false` can still
+      read several values from one stream, and the stream can still be used with other `std::istream` functions after a
+      value has been read.
+    - With [`JSON_PRECISE_STREAM_POSITION`](json_precise_stream_position.md), the stream is still positioned right after
+      a parsed number.
+    - Exception messages, including the line, the column, and the last read token, are unchanged.
+
+!!! note "Performance"
+
+    Parsing from a `std::ifstream` or `std::istringstream` is typically 1.5 to 2 times as fast as before, and reading
+    many small values from one stream with `operator>>` 1.4 to 4 times as fast. Parsing from a stream is then within
+    about 10 to 20 % of first reading the whole file into a `std::string` and parsing that.
+
+    A stream buffer without a get area (one that only overrides `underflow()` and `uflow()` and hands out every
+    character from there) offers nothing to scan in bulk, so it is read one character at a time as before.
+
+!!! note "Requirements on the stream buffer"
+
+    The bulk scan relies on the following behavior, which every stream buffer of the standard library and every stream
+    buffer that keeps its input in its get area has:
+
+    - When `sgetc()` returns a character while `gptr() < egptr()`, that character is `*gptr()`; when `sbumpc()` returns
+      one, it was at `gptr()` before the call.
+    - Characters in the get area are not changed while they are in the get area.
+    - `underflow()` or `uflow()` either sets up a get area that holds the returned character at `gptr()` (or at
+      `gptr() - 1` after `uflow()`), or leaves the get area empty.
+
+    A stream buffer that returns characters from `uflow()` while leaving unrelated characters in its get area, or that
+    changes the characters in its get area in place, does not meet these requirements. Define the macro to `1` for
+    such stream buffers.
+
+!!! note "ABI compatibility"
+
+    The value of this macro is encoded in the [namespace](../../features/namespace.md) (tag `_dsbs`), resulting in
+    distinct symbol names. Translation units compiled with and without it can therefore be linked into the same program
+    without One Definition Rule (ODR) violations, but they cannot exchange instances of library types.
+
+!!! note "Other inputs"
+
+    Strings, string views, vectors, arrays, and pointer or iterator ranges over contiguous memory are scanned in bulk
+    anyway. Reading from a `#!cpp FILE*` is not affected by this macro; see
+    [`JSON_DISABLE_UNLOCKED_FILE_IO`](json_disable_unlocked_file_io.md) for that input.
+
+## Examples
+
+??? example "Example: reading values from a stream (default)"
+
+    The stream is scanned in bulk, and the values are read one after the other:
+
+    ```cpp
+    #include <iostream>
+    #include <sstream>
+    #include <nlohmann/json.hpp>
+
+    using json = nlohmann::json;
+
+    int main()
+    {
+        std::istringstream input(R"({"id": 1} {"id": 2} 42 rest)");
+        json j1, j2, j3;
+        input >> j1 >> j2 >> j3;  // {"id":1}, {"id":2}, 42
+
+        std::string word;
+        input >> word;            // "rest"
+    }
+    ```
+
+??? example "Example: reading one character at a time (macro defined to 1)"
+
+    With the macro, the same program reads the stream one character at a time, with the same results:
+
+    ```cpp
+    #define JSON_DISABLE_STREAM_BULK_SCAN 1
+    #include <iostream>
+    #include <sstream>
+    #include <nlohmann/json.hpp>
+
+    using json = nlohmann::json;
+
+    int main()
+    {
+        std::istringstream input(R"({"id": 1} {"id": 2} 42 rest)");
+        json j1, j2, j3;
+        input >> j1 >> j2 >> j3;  // {"id":1}, {"id":2}, 42
+
+        std::string word;
+        input >> word;            // "rest"
+    }
+    ```
+
+## See also
+
+- [**parse**](../basic_json/parse.md) - deserialize from a compatible input
+- [**operator>>**](../operator_gtgt.md) - deserialize from stream
+- [**sax_parse**](../basic_json/sax_parse.md) - generate SAX events
+- [**JSON_PRECISE_STREAM_POSITION**](json_precise_stream_position.md) - leave an input stream positioned right after a
+  parsed number
+- [**JSON_DISABLE_UNLOCKED_FILE_IO**](json_disable_unlocked_file_io.md) - read a `FILE*` with the locking stdio
+  functions
+- [Parsing: input sources and performance](../../features/parsing/index.md#input-sources-and-performance)
+
+## Version history
+
+- Added in version 3.13.0.

docs/mkdocs/docs/api/macros/json_disable_unlocked_file_io.md

@@ -0,0 +1,115 @@
+# JSON_DISABLE_UNLOCKED_FILE_IO
+
+```cpp
+#define JSON_DISABLE_UNLOCKED_FILE_IO /* value */
+```
+
+When defined to `1`, parsing from a `#!cpp FILE*` reads every character with `std::fgetc`, as in version 3.12.0 and
+earlier, instead of locking the file once and reading it with the unlocked stdio functions.
+
+It does not change any result: values, exceptions, and the position of the file after parsing are the same either way.
+Only the speed and the locking of the file differ (see the notes on locking below).
+
+Builds with MemorySanitizer use `std::fgetc` automatically, because the inline `getc_unlocked` of glibc reads the stdio
+buffer directly, which glibc (not instrumented) filled, and MemorySanitizer would report it as uninitialized.
+
+## Default definition
+
+The default value is `0`: the file is locked once and read with the unlocked stdio functions where they are available.
+
+```cpp
+#define JSON_DISABLE_UNLOCKED_FILE_IO 0
+```
+
+## Notes
+
+!!! note "Background"
+
+    The C standard library makes every `std::fgetc` call thread-safe by locking and unlocking the `FILE` object. When
+    parsing from a `FILE*`, the parser reads one character at a time, so this happens for every character of the input.
+
+    Instead, the parser locks the file once when parsing starts and unlocks it when parsing ends - also when an
+    exception is thrown - and reads the characters in between without locking:
+
+    | C library                                                  | functions                                          |
+    |------------------------------------------------------------|----------------------------------------------------|
+    | glibc (Linux), Apple (macOS, iOS), Android, FreeBSD, NetBSD, OpenBSD, DragonFly BSD | `flockfile`, `getc_unlocked`, `funlockfile` (POSIX) |
+    | Microsoft (MSVC, clang-cl)                                 | `_lock_file`, `_fgetc_nolock`, `_unlock_file`       |
+    | others (e.g., musl, MinGW, newlib)                         | `std::fgetc`, as before                             |
+
+    This makes parsing from a `FILE*` 3 to 4 times as fast with the Apple C library and 1.1 to 1.3 times as fast with
+    glibc, whose locking is cheaper. The binary formats ([`from_cbor`](../basic_json/from_cbor.md),
+    [`from_msgpack`](../basic_json/from_msgpack.md), ...) read from a `FILE*` the same way.
+
+!!! note "Locking"
+
+    While a value is parsed from a `FILE*`, the parsing thread holds the file's lock. The lock is recursive, so the
+    parsing thread itself - for instance, a [SAX](../json_sax/index.md) handler or a
+    [parser callback](../basic_json/parser_callback_t.md) - can keep using the file with stdio functions such as
+    `std::ftell`. Other threads that use the same `FILE` object wait until the parser is done. Before, they could
+    interleave their reads with the parser's, which made the parsed input unpredictable anyway.
+
+    A program in which a SAX handler or parser callback waits for **another** thread that uses the **same** `FILE`
+    object would deadlock. Define the macro to `1` for such programs.
+
+!!! note "ABI compatibility"
+
+    The value of this macro is encoded in the [namespace](../../features/namespace.md) (tag `_dufio`), resulting in
+    distinct symbol names. Translation units compiled with and without it can therefore be linked into the same program
+    without One Definition Rule (ODR) violations, but they cannot exchange instances of library types.
+
+!!! note "Other inputs"
+
+    Reading from a `#!cpp std::istream` is not affected by this macro; see
+    [`JSON_DISABLE_STREAM_BULK_SCAN`](json_disable_stream_bulk_scan.md) for that input.
+
+## Examples
+
+??? example "Example: parsing from a `FILE*` (default)"
+
+    The file is locked once while it is parsed:
+
+    ```cpp
+    #include <cstdio>
+    #include <nlohmann/json.hpp>
+
+    using json = nlohmann::json;
+
+    int main()
+    {
+        std::FILE* f = std::fopen("example.json", "rb");
+        json j = json::parse(f);  // f is locked until parse() returns
+        std::fclose(f);
+    }
+    ```
+
+??? example "Example: locking for every character (macro defined to 1)"
+
+    With the macro, every character is read with `std::fgetc`, with the same result:
+
+    ```cpp
+    #define JSON_DISABLE_UNLOCKED_FILE_IO 1
+    #include <cstdio>
+    #include <nlohmann/json.hpp>
+
+    using json = nlohmann::json;
+
+    int main()
+    {
+        std::FILE* f = std::fopen("example.json", "rb");
+        json j = json::parse(f);  // std::fgetc for every character
+        std::fclose(f);
+    }
+    ```
+
+## See also
+
+- [**parse**](../basic_json/parse.md) - deserialize from a compatible input
+- [**sax_parse**](../basic_json/sax_parse.md) - generate SAX events
+- [**JSON_DISABLE_STREAM_BULK_SCAN**](json_disable_stream_bulk_scan.md) - read a `std::istream` one character at a
+  time
+- [Parsing: input sources and performance](../../features/parsing/index.md#input-sources-and-performance)
+
+## Version history
+
+- Added in version 3.13.0.

docs/mkdocs/docs/api/macros/json_precise_stream_position.md

@@ -72,6 +72,12 @@ The default value is `0` (disabled — existing behavior is preserved).
     distinct symbol names. Translation units compiled with and without it can therefore be linked into the same program
     without One Definition Rule (ODR) violations, but they cannot exchange instances of library types.
 
+!!! note "Bulk scanning"
+
+    The parser scans the characters in the stream buffer in bulk (see
+    [`JSON_DISABLE_STREAM_BULK_SCAN`](json_disable_stream_bulk_scan.md)), but it never reads ahead. The stream is left
+    at the same position with and without bulk scanning, in both modes of this macro.
+
 !!! tip "Workaround without the macro"
 
     Separate the values in the stream with whitespace. The character consumed after a number is then the separator,

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

@@ -12,8 +12,9 @@ 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.
+diagnostic is preserved). A `std::istream` uses simdutf for the characters its stream buffer holds (see
+[`JSON_DISABLE_STREAM_BULK_SCAN`](json_disable_stream_bulk_scan.md)); other streaming inputs (`FILE*`, 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.

docs/mkdocs/docs/api/macros/json_view_no_simd.md

@@ -14,10 +14,12 @@ The same input is accepted or rejected either way, with the same values, and err
 the speed differs. The macro exists for platforms whose compilers lack the intrinsics headers, and to test the portable
 code.
 
-!!! warning "Define consistently"
+!!! note "ABI compatibility"
 
-    The macro selects between two definitions of the same inline functions. It must therefore be defined identically for
-    **every** translation unit that includes `<nlohmann/json_view.hpp>`; prefer a compile definition on the target.
+    The macro selects between two definitions of the same inline functions. Whether it is defined is encoded in the
+    [namespace](../../features/namespace.md) (tag `_vns`), resulting in distinct symbol names. Translation units
+    compiled with and without it can therefore be linked into the same program without One Definition Rule (ODR)
+    violations, but they cannot exchange instances of library types.
 
 ## Default definition
 

docs/mkdocs/docs/api/macros/json_view_use_ssse3.md

@@ -17,12 +17,12 @@ Define the macro only together with a compiler option that enables SSSE3 (e.g. `
 has it), and only for programs that run on such CPUs. It saves the check of the CPU, which costs little. The same
 input is accepted or rejected either way; only the speed of non-ASCII text differs.
 
-!!! warning "Define consistently"
+!!! note "ABI compatibility"
 
-    The macro selects between two definitions of the same inline functions. It must therefore be defined identically,
-    with the same compiler options, for **every** translation unit that includes `<nlohmann/json_view.hpp>`; mixing
-    translation units that define it with ones that do not is an ODR violation. Prefer a compile definition on the
-    target.
+    The macro selects between two definitions of the same inline functions. Whether it is defined is encoded in the
+    [namespace](../../features/namespace.md) (tag `_vssse3`, on every architecture), resulting in distinct symbol
+    names. Translation units compiled with and without it can therefore be linked into the same program without One
+    Definition Rule (ODR) violations, but they cannot exchange instances of library types.
 
 ## Default definition
 

docs/mkdocs/docs/api/macros/nlohmann_json_namespace.md

@@ -39,5 +39,6 @@ the library.
 ## Version history
 
 - Added in version 3.11.0. Changed inline namespace name in version 3.11.2. Added the ABI tag `_dp` in
-  version 3.12.0, and the ABI tags `_bics`, `_psp`, `_snul`, `_sbu8`, and `_ekmo` in version 3.13.0; see
+  version 3.12.0, and the ABI tags `_bics`, `_psp`, `_snul`, `_sbu8`, `_ekmo`, `_dsbs`, `_dufio`, `_vns`, and
+  `_vssse3` in version 3.13.0; see
   [`nlohmann` Namespace](../../features/namespace.md#structure).

docs/mkdocs/docs/api/operator_gtgt.md

@@ -121,6 +121,7 @@ being read.
 - [`JSON_STRICT_NUL_HANDLING`](macros/json_strict_nul_handling.md) - opt in to rejecting a NUL byte in the input
   instead of treating it as end of input
 - [`JSON_PRECISE_STREAM_POSITION`](macros/json_precise_stream_position.md) - opt in to leaving the stream positioned right after a number
+- [`JSON_DISABLE_STREAM_BULK_SCAN`](macros/json_disable_stream_bulk_scan.md) - read the stream one character at a time
 
 ## Version history
 
@@ -133,3 +134,6 @@ being read.
   crash (`std::terminate`) when `i` has `eofbit` in its exception mask, in version 3.13.0.
 - Changed to the strong exception safety guarantee in version 3.13.0: `j` is no longer left with a partially parsed
   value if parsing throws.
+- Scans the characters in the stream buffer of `i` in bulk in version 3.13.0; the position of `i` after the call is
+  unchanged ([`JSON_DISABLE_STREAM_BULK_SCAN`](macros/json_disable_stream_bulk_scan.md) restores reading one character
+  at a time).

docs/mkdocs/docs/community/quality_assurance.md

@@ -135,6 +135,9 @@ Note: Some modern features (like C++20 ranges or filesystem support) may be disa
   headers are included.
 - [x] On Windows, the library is compiled with `<Windows.h>` being included to detect and avoid common bugs (see
   [`unit-windows_h.cpp`](https://github.com/nlohmann/json/blob/develop/tests/src/unit-windows_h.cpp)).
+- [x] The library is compiled with Qt's keyword macros (`slots`, `signals`, `emit`, `foreach`, `forever`) being defined
+  to make sure it does not use them as identifiers (see
+  [`unit-qt_keywords.cpp`](https://github.com/nlohmann/json/blob/develop/tests/src/unit-qt_keywords.cpp)).
 - [x] The library is compiled with exceptions disabled to support alternative means of error handling.
 
 ## Stable public API

docs/mkdocs/docs/examples/basic_json_view__count.cpp

@@ -5,13 +5,14 @@ using json_document = nlohmann::json_document;
 
 int main()
 {
-    // validate that every transaction of a batch carries a mandatory
+    // validate that every transaction of a batch carries exactly one
     // "amount" field before materializing any of them into a nlohmann::json
-    // value -- count() returns 0 or 1 for an object
+    // value -- count() counts every member with the key
     json_document batch = json_document::parse(R"(
       [
         {"id": 1, "amount": 9.99},
-        {"id": 2}
+        {"id": 2},
+        {"id": 3, "amount": 5.00, "amount": 500.00}
       ]
     )");
 
@@ -19,9 +20,10 @@ int main()
     for (std::size_t i = 0; i < transactions.size(); ++i)
     {
         const auto transaction = transactions[i];
-        if (transaction.count("amount") == 0)
+        const auto amounts = transaction.count("amount");
+        if (amounts != 1)
         {
-            std::cout << "transaction " << i << " is missing \"amount\"\n";
+            std::cout << "transaction " << i << " has " << amounts << " \"amount\" fields\n";
             continue;
         }
         std::cout << "transaction " << i << ": " << transaction["amount"].materialize().dump() << '\n';

docs/mkdocs/docs/examples/basic_json_view__count.output

@@ -1,2 +1,3 @@
 transaction 0: 9.99
-transaction 1 is missing "amount"
+transaction 1 has 0 "amount" fields
+transaction 2 has 2 "amount" fields

docs/mkdocs/docs/features/index.md

@@ -14,8 +14,8 @@ C++ types, and finally serialize it again.
   [parsing untrusted input](parsing/untrusted_input.md).
 - [Zero-copy JSON views](json_view.md) — read a JSON text through a flat index instead of building a `json` tree;
   strings and numbers stay in the input and are only decoded when needed.
-  [Editable documents](json_view.md#editing-a-document) can also be modified, and [images](json_view.md#images) load a
-  parsed document again without parsing it.
+- [Editable documents](json_view.md#editing-a-document) — modify a parsed document without building a `json` tree.
+- [Images](json_view.md#images) — save a parsed document and load it again without parsing it.
 - [Comments](comments.md) and [trailing commas](trailing_commas.md) — opt-in relaxations of the JSON grammar.
 
 ## Accessing and modifying values

docs/mkdocs/docs/features/json_view.md

@@ -67,9 +67,10 @@ table). [`parse_copy()`](../api/basic_json_document/parse_copy.md) always owns a
 
 !!! warning "A borrowed document depends on your buffer"
 
-    If a document borrows its text, that text **must outlive the document** (and every view taken from it). Reading
-    or writing through a view after the underlying buffer is gone is undefined behavior, exactly as it would be for
-    a dangling `#!cpp std::string_view`.
+    If a document borrows its text, that text **must outlive the document** (and every view taken from it) and must
+    stay **unmodified** for as long as the document exists, since floating-point values and unescaped strings are
+    read from it lazily. Reading or writing through a view after the underlying buffer is gone is undefined behavior,
+    exactly as it would be for a dangling `#!cpp std::string_view`.
 
 A view is valid only while all of the following hold:
 
@@ -77,7 +78,7 @@ A view is valid only while all of the following hold:
 - the document has not been re-parsed since the view was taken (with [`read()`](../api/basic_json_document/read.md)
   or [`parse()`](../api/basic_json_document/parse.md) into it),
 - the document has not had [`shrink_to_fit()`](../api/basic_json_document/shrink_to_fit.md) called on it since, and
-- if the document borrows its source text, that text is still alive.
+- if the document borrows its source text, that text is still alive and unmodified.
 
 Moving the `basic_json_document<>` itself is fine and does **not** invalidate its views: the index is a separate heap
 allocation that keeps its address across the move. Take a fresh view from [`root()`](../api/basic_json_document/root.md)
@@ -167,10 +168,10 @@ document: `#!cpp auto v = json_document::parse(text).root();` does not compile.
 
     If an object in the source text repeats a key, [`operator[]`](../api/basic_json_view/operator%5B%5D.md),
     [`at`](../api/basic_json_view/at.md), [`find`](../api/basic_json_view/find.md),
-    [`value`](../api/basic_json_view/value.md), [`contains`](../api/basic_json_view/contains.md),
-    [`count`](../api/basic_json_view/count.md), and JSON pointer resolution all return the **first** member with that
-    key. `basic_json::parse()` instead keeps the **last** value of a repeated key, so for `#!cpp {"a":1,"a":2}`,
-    `#!cpp view["a"]` is `#!cpp 1` while `#!cpp parse(text)["a"]` is `#!cpp 2`.
+    [`value`](../api/basic_json_view/value.md), [`contains`](../api/basic_json_view/contains.md), and JSON pointer
+    resolution all return the **first** member with that key. `basic_json::parse()` instead keeps the **last** value
+    of a repeated key, so for `#!cpp {"a":1,"a":2}`, `#!cpp view["a"]` is `#!cpp 1` while `#!cpp parse(text)["a"]` is
+    `#!cpp 2`. [`count`](../api/basic_json_view/count.md) counts every member with the key, so it is `#!cpp 2` here.
 
 [`begin()`](../api/basic_json_view/begin.md)/[`end()`](../api/basic_json_view/end.md) and
 [`items()`](../api/basic_json_view/items.md) visit *every* occurrence, in document order, and
@@ -179,7 +180,7 @@ document: `#!cpp auto v = json_document::parse(text).root();` does not compile.
 
 **Why the first?** A lookup can stop as soon as it finds a match. Returning the last member would force every lookup
 to scan all members of the object, even when the key is found at the very first one: this made lookups in small
-objects 1.6 to 3.4 times slower. The hash index of objects with 128 members or more leads to the first member of a
+objects 1.6 to 3.4 times slower. The hash index of the keys of objects with 128 members or more leads to the first member of a
 key as well. Other zero-copy parsers that index the source text, such as yyjson and simdjson, also return the first
 member. RFC 8259 only says that names within an object SHOULD be unique and that the behavior of a receiver that sees
 duplicates is unpredictable, so neither choice is wrong.

docs/mkdocs/docs/features/macros.md

@@ -133,6 +133,14 @@ When defined, default parse and serialize functions for enums are excluded and h
 
 See [full documentation of `JSON_DISABLE_ENUM_SERIALIZATION`](../api/macros/json_disable_enum_serialization.md).
 
+## `JSON_DISABLE_STREAM_BULK_SCAN`
+
+When defined to `1`, parsing from a `#!cpp std::istream` reads one character at a time instead of scanning the
+characters that the stream buffer holds in bulk. Results are the same either way; the macro is an escape hatch for
+stream buffers that do not follow the usual `std::streambuf` contract. The default value is `0`.
+
+See [full documentation of `JSON_DISABLE_STREAM_BULK_SCAN`](../api/macros/json_disable_stream_bulk_scan.md).
+
 ## `JSON_DISABLE_TUPLE_REFERENCE_CONVERSION`
 
 When defined to `1`, a JSON value can no longer be created from a one-element `std::tuple` holding a reference to a JSON
@@ -141,6 +149,14 @@ is planned to become the default in version 4.0.0.
 
 See [full documentation of `JSON_DISABLE_TUPLE_REFERENCE_CONVERSION`](../api/macros/json_disable_tuple_reference_conversion.md).
 
+## `JSON_DISABLE_UNLOCKED_FILE_IO`
+
+When defined to `1`, parsing from a `#!cpp FILE*` reads every character with `std::fgetc`, which locks the file each
+time, instead of locking the file once and reading it with the unlocked stdio functions. Results are the same either
+way. The default value is `0`.
+
+See [full documentation of `JSON_DISABLE_UNLOCKED_FILE_IO`](../api/macros/json_disable_unlocked_file_io.md).
+
 ## `JSON_NO_AUTOMATIC_UDLS`
 
 When defined, `<nlohmann/json.hpp>` does not include `<nlohmann/json_literals.hpp>` with the user-defined string literals

docs/mkdocs/docs/features/namespace.md

@@ -23,6 +23,12 @@ The complete default namespace name is derived as follows:
     - [`JSON_STRICT_BINARY_UTF8`](../api/macros/json_strict_binary_utf8.md) defined non-zero appends `_sbu8`.
     - [`JSON_USE_OBJECTS_FOR_ENUM_KEYED_MAPS`](../api/macros/json_use_objects_for_enum_keyed_maps.md) defined non-zero
       appends `_ekmo`.
+    - [`JSON_DISABLE_STREAM_BULK_SCAN`](../api/macros/json_disable_stream_bulk_scan.md) defined non-zero appends
+      `_dsbs`.
+    - [`JSON_DISABLE_UNLOCKED_FILE_IO`](../api/macros/json_disable_unlocked_file_io.md) defined non-zero appends
+      `_dufio`.
+    - [`JSON_VIEW_NO_SIMD`](../api/macros/json_view_no_simd.md) defined appends `_vns`.
+    - [`JSON_VIEW_USE_SSSE3`](../api/macros/json_view_use_ssse3.md) defined appends `_vssse3`.
 - The inline namespace ends with the suffix `_v` followed by the 3 components of the version number separated by
   underscores. To omit the version component, see [Disabling the version component](#disabling-the-version-component)
   below.
@@ -105,5 +111,9 @@ follows:
 - Added ABI tags `_bics` ([`JSON_BRACE_INIT_COPY_SEMANTICS`](../api/macros/json_brace_init_copy_semantics.md)), `_psp`
   ([`JSON_PRECISE_STREAM_POSITION`](../api/macros/json_precise_stream_position.md)), `_snul`
   ([`JSON_STRICT_NUL_HANDLING`](../api/macros/json_strict_nul_handling.md)), `_sbu8`
-  ([`JSON_STRICT_BINARY_UTF8`](../api/macros/json_strict_binary_utf8.md)), and `_ekmo`
-  ([`JSON_USE_OBJECTS_FOR_ENUM_KEYED_MAPS`](../api/macros/json_use_objects_for_enum_keyed_maps.md)) in version 3.13.0.
+  ([`JSON_STRICT_BINARY_UTF8`](../api/macros/json_strict_binary_utf8.md)), `_ekmo`
+  ([`JSON_USE_OBJECTS_FOR_ENUM_KEYED_MAPS`](../api/macros/json_use_objects_for_enum_keyed_maps.md)), `_dsbs`
+  ([`JSON_DISABLE_STREAM_BULK_SCAN`](../api/macros/json_disable_stream_bulk_scan.md)), `_dufio`
+  ([`JSON_DISABLE_UNLOCKED_FILE_IO`](../api/macros/json_disable_unlocked_file_io.md)), `_vns`
+  ([`JSON_VIEW_NO_SIMD`](../api/macros/json_view_no_simd.md)), and `_vssse3`
+  ([`JSON_VIEW_USE_SSSE3`](../api/macros/json_view_use_ssse3.md)) in version 3.13.0.

docs/mkdocs/docs/features/parsing/index.md

@@ -55,6 +55,30 @@ whitespace, because `operator>>` consumes the character that terminates a number
 [`JSON_PRECISE_STREAM_POSITION`](../../api/macros/json_precise_stream_position.md) is defined to `1` — see the
 [`operator>>` notes](../../api/operator_gtgt.md#notes) for details and examples.
 
+## Input sources and performance
+
+All inputs produce the same values and the same exceptions, but they are read in different ways, which makes some of
+them faster than others:
+
+| input                                                                                      | how it is read                                                                     |
+|--------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------|
+| strings, string views, string literals, contiguous containers, pointer or iterator ranges over contiguous memory | the whole input is scanned in bulk: runs of string characters, numbers, and indentation are recognized several bytes at a time |
+| `#!cpp std::istream` (`#!cpp std::ifstream`, `#!cpp std::istringstream`, ...)            | the characters the stream buffer holds are scanned in bulk like a string; the stream buffer is asked for more when they are used up |
+| `#!cpp FILE*`                                                                              | one character at a time, with the file locked once for the whole parse              |
+| other iterator ranges (e.g., `#!cpp std::list<char>`, `#!cpp std::istreambuf_iterator<char>`), wide strings | one character at a time                                                             |
+
+Parsing a file from a `#!cpp std::ifstream` is therefore only about 10 to 20 % slower than reading the whole file into a
+`#!cpp std::string` first and parsing that, and it needs no memory for a copy of the file. The stream is never read
+ahead: after parsing, it is positioned exactly as if every character had been read one at a time, so
+[`operator>>`](../../api/operator_gtgt.md) can still read several values from one stream (see also
+[JSON Lines](json_lines.md)).
+
+Two macros switch these optimizations off without changing any result:
+[`JSON_DISABLE_STREAM_BULK_SCAN`](../../api/macros/json_disable_stream_bulk_scan.md) for stream buffers that do not
+follow the usual `std::streambuf` contract, and
+[`JSON_DISABLE_UNLOCKED_FILE_IO`](../../api/macros/json_disable_unlocked_file_io.md) for programs in which another
+thread must use the same `#!cpp FILE` object while it is parsed.
+
 ## SAX vs. DOM parsing
 
 The library offers two parsing models:

docs/mkdocs/docs/features/performance.md

@@ -20,32 +20,39 @@ below apply on every supported standard. This adapter exposes two optimizations
 
 - it can reconstruct already-consumed input on demand for error messages, instead of copying every character as it is
   read, and
-- the lexer can scan ordinary string characters directly out of the buffer, several bytes at a time, rather than one
-  character (and one function call) at a time.
+- the lexer can scan ordinary string characters, numbers, and indentation directly out of the buffer, several bytes at
+  a time, rather than one character (and one function call) at a time.
 
-A `#!cpp std::istream` (including `#!cpp std::ifstream`) or `#!cpp FILE*`, by contrast, is read through
-`input_stream_adapter` or `file_input_adapter`, which read one character (or one block, for binary formats) at a time
-and expose neither optimization -- the lexer falls back to the same byte-at-a-time path it uses for any
-non-contiguous, general-purpose iterator range. An iterator pair over non-contiguous but random-access storage (e.g.
-`#!cpp std::deque<char>::iterator`) gets the first optimization but not the second, since the byte-scanning fast path
-additionally requires contiguous storage.
+A `#!cpp std::istream` (including `#!cpp std::ifstream`) gets the same bulk scanning: `input_stream_adapter` scans the
+characters its stream buffer already holds (the *get area*) and asks the stream buffer for more only when they are used
+up. Nothing is read ahead, so the stream is left exactly where it would be after reading one character at a time, and
+[`operator>>`](../api/operator_gtgt.md) can still read several values from one stream. A stream buffer without a get
+area is read one character at a time. A `#!cpp FILE*` is still read one character at a time, but `file_input_adapter`
+locks the file once per parse instead of once per character. See
+[input sources and performance](parsing/index.md#input-sources-and-performance) for the full table. According to the
+description of [#5828](https://github.com/nlohmann/json/pull/5828), parsing from a stream is within about 10 to 20 % of
+parsing an equivalent `#!cpp std::string`.
 
-Practically: if the JSON text is already in memory, or small enough to read into memory, prefer passing a
-`#!cpp std::string`, a `#!cpp std::vector<char>`, or a pointer range to `parse` over a `#!cpp std::istream`. For a
-file, that means reading it into a string first and then parsing the string, rather than passing a
-`#!cpp std::ifstream` directly to `parse` -- the latter never benefits from either optimization:
+Inputs that are still read one character at a time: iterator ranges that are not contiguous (e.g.
+`#!cpp std::list<char>` or `#!cpp std::istreambuf_iterator<char>`), wide strings, and user-defined adapters. An
+iterator pair over non-contiguous but random-access storage (e.g. `#!cpp std::deque<char>::iterator`) gets the first
+optimization but not the second, since the byte-scanning fast path requires contiguous storage.
+
+Practically: a file can be passed directly to `parse`, which also avoids the memory for a copy of the file. Reading it
+into a string first is no longer necessary for speed:
 
 ```cpp
-// gets the contiguous fast paths
+// fast: the stream buffer's get area is scanned in bulk
 std::ifstream f("example.json");
-std::string contents((std::istreambuf_iterator<char>(f)), std::istreambuf_iterator<char>());
-json j = json::parse(contents);
-
-// does not: input_stream_adapter has no fast path
-std::ifstream f2("example.json");
-json j2 = json::parse(f2);
+json j = json::parse(f);
 ```
 
+Two macros switch the stream optimizations off without changing any result:
+[`JSON_DISABLE_STREAM_BULK_SCAN`](../api/macros/json_disable_stream_bulk_scan.md) for stream buffers that do not follow
+the usual `#!cpp std::streambuf` contract, and
+[`JSON_DISABLE_UNLOCKED_FILE_IO`](../api/macros/json_disable_unlocked_file_io.md) for programs in which another thread
+must use the same `#!cpp FILE` object while it is parsed.
+
 For contiguous input with many non-ASCII characters, [`JSON_USE_SIMDUTF`](../api/macros/json_use_simdutf.md) can
 additionally speed up UTF-8 validation by using the [simdutf](https://github.com/simdutf/simdutf) library instead of
 the built-in scalar validator; streaming inputs (files, `#!cpp std::istream`, wide strings, user-defined adapters)

docs/mkdocs/docs/home/architecture.md

@@ -194,9 +194,9 @@ packet-beta
 
 | Bytes | Field   | Type       | Contents                                                                                                                      |
 |-------|---------|------------|-------------------------------------------------------------------------------------------------------------------------------|
-| 0     | `kind`  | `uint8_t`  | the type, numbered as [`value_t`](../api/basic_json/value_t.md): 0 null, 1 object, 2 array, 3 string, 4 boolean, 5 signed integer, 6 unsigned integer, 7 float; 10 for a link (see below) |
-| 1     | `flags` | `uint8_t`  | bits 0-1: where a string's bytes (or a number's token) are (0: the source text, 1: the buffer of decoded strings, for strings with escapes, 2: the edit buffer); bit 2: the value of a boolean; bits 3 and 4: moved and new (see below) |
-| 2-3   | `extra` | `uint16_t` | numbers: the number of integer digits (low byte) and fraction digits (high byte), 255 for more; objects: the number of their hash index (1-based), or 0; otherwise 0 |
+| 0     | `kind`  | `uint8_t`  | the type, numbered as [`value_t`](../api/basic_json/value_t.md): 0 null, 1 object, 2 array, 3 string, 4 boolean, 5 signed integer, 6 unsigned integer, 7 float; 10 for a link (see below). There is no node for a discarded value: a [discarded](../api/basic_json_view/is_discarded.md) view has none |
+| 1     | `flags` | `uint8_t`  | bits 0-1: where a string's bytes (or a number's token) are (0: the source text, 1: the buffer of decoded strings, for strings with escapes, 2: the edit buffer); bit 2: the value of a boolean; bit 3: moved, bit 4: new, bit 5: linked (see below) |
+| 2-3   | `extra` | `uint16_t` | numbers: the number of integer digits (low byte) and fraction digits (high byte), 255 for more; objects: the number of their key hash index (1-based), or 0; otherwise 0 |
 | 4-7   | `off`   | `uint32_t` | where the value starts: the first byte after a string's opening quote (or its position in the buffer of decoded strings), the first byte of a number or literal, the bracket of an array or object |
 | 8-11  | `len`   | `uint32_t` | strings: the length after decoding; floats and literals: the length of the token; arrays and objects: the number of elements |
 | 12-15 | `next`  | `uint32_t` | arrays and objects: the number of nodes of the subtree, including the node itself                                              |
@@ -212,7 +212,7 @@ packet-beta
   step from element to element this way and skip whole subtrees in constant time.
 - **Offsets** are 32 bits wide, so a document is limited to 4294967279 bytes, 4 GiB minus 16 bytes (a margin below
   2^32 for positions one scanner step past the end of the text; `out_of_range.416`).
-- **Large objects** (128 members or more) get a hash index after parsing
+- **Large objects** (128 members or more) get a hash index for their keys after parsing
   ([`detail/view/object_index.hpp`](https://github.com/nlohmann/json/blob/develop/include/nlohmann/detail/view/object_index.hpp)):
   an open-addressing table whose slots hold the distance from the object's node to a key's node, so that a lookup does
   not compare every key. Of duplicate keys the table leads to the first, as a linear search does. The object's `extra`
@@ -257,6 +257,8 @@ edited:
   sequence (a header node, then the entries), whose number is in `off`. The entries are links (`kind` 10), whose bytes
   8-15 hold the address of the value's node, so values never move.
 - A node written by an edit gets the flag *new* (bit 4): it has no position in the source text.
+- A value that an entry of such a sequence links to gets the flag *linked* (bit 5): its extent in the parsed layout
+  (`next`) no longer matters, so a scalar can replace an array or object there in place.
 - Views of read-only documents compile without any of this: how views walk the index is a template parameter
   (`navigation<Editable>`).
 

docs/mkdocs/docs/home/exceptions.md

@@ -972,9 +972,19 @@ double-precision number when `number_float_t` is `#!cpp float`.
 
 ### json.exception.out_of_range.407
 
-An unsigned integer number greater than 9223372036854775807 cannot be represented in the binary format. This exception
-is only thrown by [`to_bon8`](../api/basic_json/to_bon8.md), as [BON8](../features/binary_formats/bon8.md) has no
-unsigned 64-bit integer type.
+An integer number cannot be represented by the binary format it is serialized to:
+
+- [BON8](../features/binary_formats/bon8.md) only stores integers that fit into int64, so
+  [`to_bon8`](../api/basic_json/to_bon8.md) throws for unsigned integers greater than 9223372036854775807.
+- [CBOR](../features/binary_formats/cbor.md), [MessagePack](../features/binary_formats/messagepack.md), and
+  [BSON](../features/binary_formats/bson.md) store integers in at most 64 bits. With the default number types, every
+  integer fits, but a [`number_integer_t`](../api/basic_json/number_integer_t.md) or
+  [`number_unsigned_t`](../api/basic_json/number_unsigned_t.md) wider than 64 bits (e.g., `__int128`) can hold values
+  outside the range of the format: [-2^64, 2^64-1] for CBOR, [-2^63, 2^64-1] for MessagePack, and the range of int64
+  (signed integers) or uint64 (unsigned integers) for BSON.
+
+[UBJSON](../features/binary_formats/ubjson.md) and [BJData](../features/binary_formats/bjdata.md) never throw this
+exception, because they serialize integers beyond 64 bits as high-precision numbers.
 
 !!! note "Former occurrences"
 
@@ -983,11 +993,17 @@ unsigned 64-bit integer type.
     - Since version 3.9.0, integer numbers beyond int64 are serialized as high-precision UBJSON numbers.
     - Since version 3.12.0, integer numbers beyond int64 are serialized as uint64 BSON numbers.
 
-!!! failure "Example message"
+    Before version 3.13.0, CBOR, MessagePack, and BSON silently truncated integers wider than 64 bits, and BJData
+    truncated unsigned integers wider than 64 bits.
+
+!!! failure "Example messages"
 
     ```
     integer number 18446744073709551615 cannot be represented by BON8 as it does not fit int64
     ```
+    ```
+    integer number 1267650600228229401496703205376 cannot be represented by CBOR as it does not fit [-2^64, 2^64-1]
+    ```
 
 ### json.exception.out_of_range.408
 

docs/mkdocs/mkdocs.yml

@@ -370,7 +370,9 @@ nav:
           - 'JSON_DELETE_DEPRECATED_FUNCTIONS': api/macros/json_delete_deprecated_functions.md
           - 'JSON_DIAGNOSTIC_POSITIONS': api/macros/json_diagnostic_positions.md
           - 'JSON_DISABLE_ENUM_SERIALIZATION': api/macros/json_disable_enum_serialization.md
+          - 'JSON_DISABLE_STREAM_BULK_SCAN': api/macros/json_disable_stream_bulk_scan.md
           - 'JSON_DISABLE_TUPLE_REFERENCE_CONVERSION': api/macros/json_disable_tuple_reference_conversion.md
+          - 'JSON_DISABLE_UNLOCKED_FILE_IO': api/macros/json_disable_unlocked_file_io.md
           - 'JSON_HAS_CPP_11, JSON_HAS_CPP_14, JSON_HAS_CPP_17, JSON_HAS_CPP_20, JSON_HAS_CPP_23, JSON_HAS_CPP_26': api/macros/json_has_cpp_11.md
           - 'JSON_HAS_EXPERIMENTAL_FILESYSTEM, JSON_HAS_FILESYSTEM': api/macros/json_has_filesystem.md
           - 'JSON_HAS_RANGE_VIEW_CONVERSION': api/macros/json_has_range_view_conversion.md

include/nlohmann/detail/abi_macros.hpp

@@ -54,6 +54,14 @@
     #define JSON_USE_OBJECTS_FOR_ENUM_KEYED_MAPS 0
 #endif
 
+#ifndef JSON_DISABLE_STREAM_BULK_SCAN
+    #define JSON_DISABLE_STREAM_BULK_SCAN 0
+#endif
+
+#ifndef JSON_DISABLE_UNLOCKED_FILE_IO
+    #define JSON_DISABLE_UNLOCKED_FILE_IO 0
+#endif
+
 #if JSON_DIAGNOSTICS
     #define NLOHMANN_JSON_ABI_TAG_DIAGNOSTICS _diag
 #else
@@ -102,14 +110,38 @@
     #define NLOHMANN_JSON_ABI_TAG_OBJECTS_FOR_ENUM_KEYED_MAPS
 #endif
 
+#if JSON_DISABLE_STREAM_BULK_SCAN
+    #define NLOHMANN_JSON_ABI_TAG_DISABLE_STREAM_BULK_SCAN _dsbs
+#else
+    #define NLOHMANN_JSON_ABI_TAG_DISABLE_STREAM_BULK_SCAN
+#endif
+
+#if JSON_DISABLE_UNLOCKED_FILE_IO
+    #define NLOHMANN_JSON_ABI_TAG_DISABLE_UNLOCKED_FILE_IO _dufio
+#else
+    #define NLOHMANN_JSON_ABI_TAG_DISABLE_UNLOCKED_FILE_IO
+#endif
+
+#ifdef JSON_VIEW_NO_SIMD
+    #define NLOHMANN_JSON_ABI_TAG_VIEW_NO_SIMD _vns
+#else
+    #define NLOHMANN_JSON_ABI_TAG_VIEW_NO_SIMD
+#endif
+
+#ifdef JSON_VIEW_USE_SSSE3
+    #define NLOHMANN_JSON_ABI_TAG_VIEW_USE_SSSE3 _vssse3
+#else
+    #define NLOHMANN_JSON_ABI_TAG_VIEW_USE_SSSE3
+#endif
+
 #ifndef NLOHMANN_JSON_NAMESPACE_NO_VERSION
     #define NLOHMANN_JSON_NAMESPACE_NO_VERSION 0
 #endif
 
 // Construct the namespace ABI tags component
-#define NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c, d, e, f, g, h) json_abi ## a ## b ## c ## d ## e ## f ## g ## h
-#define NLOHMANN_JSON_ABI_TAGS_CONCAT(a, b, c, d, e, f, g, h) \
-    NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c, d, e, f, g, h)
+#define NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c, d, e, f, g, h, i, j, k, l) json_abi ## a ## b ## c ## d ## e ## f ## g ## h ## i ## j ## k ## l
+#define NLOHMANN_JSON_ABI_TAGS_CONCAT(a, b, c, d, e, f, g, h, i, j, k, l) \
+    NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c, d, e, f, g, h, i, j, k, l)
 
 #define NLOHMANN_JSON_ABI_TAGS                                       \
     NLOHMANN_JSON_ABI_TAGS_CONCAT(                                   \
@@ -120,7 +152,11 @@
             NLOHMANN_JSON_ABI_TAG_PRECISE_STREAM_POSITION,           \
             NLOHMANN_JSON_ABI_TAG_STRICT_NUL_HANDLING,               \
             NLOHMANN_JSON_ABI_TAG_STRICT_BINARY_UTF8,                \
-            NLOHMANN_JSON_ABI_TAG_OBJECTS_FOR_ENUM_KEYED_MAPS)
+            NLOHMANN_JSON_ABI_TAG_OBJECTS_FOR_ENUM_KEYED_MAPS,       \
+            NLOHMANN_JSON_ABI_TAG_DISABLE_STREAM_BULK_SCAN,          \
+            NLOHMANN_JSON_ABI_TAG_DISABLE_UNLOCKED_FILE_IO,          \
+            NLOHMANN_JSON_ABI_TAG_VIEW_NO_SIMD,                      \
+            NLOHMANN_JSON_ABI_TAG_VIEW_USE_SSSE3)
 
 // Construct the namespace version component
 #define NLOHMANN_JSON_NAMESPACE_VERSION_CONCAT_EX(major, minor, patch) \

include/nlohmann/detail/conversions/to_chars.hpp

@@ -1342,14 +1342,32 @@ inline char* write_shortest(char* first, const zmij::shortest_decimal d) noexcep
     return end + (three ? 5 : 4);
 }
 
-/// the powers of ten up to 10^16
-inline const std::array<std::uint64_t, 17>& powers_of_ten_16() noexcept
+/// the powers of ten up to 10^19 (all that fit in std::uint64_t)
+inline const std::array<std::uint64_t, 20>& powers_of_ten() noexcept
 {
-    static const std::array<std::uint64_t, 17> powers =
+    static const std::array<std::uint64_t, 20> powers =
     {
         {
-            1u, 10u, 100u, 1000u, 10000u, 100000u, 1000000u, 10000000u, 100000000u, 1000000000u, 10000000000u,
-            100000000000u, 1000000000000u, 10000000000000u, 100000000000000u, 1000000000000000u, 10000000000000000u
+            1u,
+            10u,
+            100u,
+            1000u,
+            10000u,
+            100000u,
+            1000000u,
+            10000000u,
+            100000000u,
+            1000000000u,
+            10000000000u,
+            100000000000u,
+            1000000000000u,
+            10000000000000u,
+            100000000000000u,
+            1000000000000000u,
+            10000000000000000u,
+            100000000000000000u,
+            1000000000000000000u,
+            10000000000000000000u
         }
     };
     return powers;
@@ -1367,9 +1385,9 @@ JSON_HEDLEY_NON_NULL(1)
 JSON_HEDLEY_RETURNS_NON_NULL
 inline char* write_short_decimal(char* first, std::uint64_t digits, int count, int exp) noexcept
 {
-    JSON_ASSERT(digits >= powers_of_ten_16()[static_cast<std::size_t>(count - 1)] && count <= 15);
+    JSON_ASSERT(digits >= powers_of_ten()[static_cast<std::size_t>(count - 1)] && count <= 15);
     const int scale = 16 - count;
-    return write_shortest(first, zmij::shortest_decimal{digits * powers_of_ten_16()[static_cast<std::size_t>(scale)], exp - scale - 1, 0, false});
+    return write_shortest(first, zmij::shortest_decimal{digits * powers_of_ten()[static_cast<std::size_t>(scale)], exp - scale - 1, 0, false});
 }
 
 /// as write_short_decimal(), counting the digits (not 0, less than 10^15)
@@ -1380,7 +1398,7 @@ inline char* write_short_decimal(char* first, std::uint64_t digits, int exp) noe
     JSON_ASSERT(digits != 0 && digits < 1000000000000000u);
     // floor(log10(2^bits)) + 1 digits, or one less
     const int log2_bound = ((64 - count_leading_zeros(digits)) * 1233) >> 12;
-    const int count = log2_bound + (digits >= powers_of_ten_16()[static_cast<std::size_t>(log2_bound)] ? 1 : 0);
+    const int count = log2_bound + (digits >= powers_of_ten()[static_cast<std::size_t>(log2_bound)] ? 1 : 0);
     return write_short_decimal(first, digits, count, exp);
 }
 

include/nlohmann/detail/input/input_adapters.hpp

@@ -14,10 +14,12 @@
 #include <cstdint> // uint32_t
 #include <cstring> // strlen
 #include <iterator> // begin, end, iterator_traits, random_access_iterator_tag, distance, next, reverse_iterator
+#include <limits> // numeric_limits
 #include <streambuf> // streambuf
 #include <string> // string, char_traits
 #include <type_traits> // enable_if, is_base_of, is_pointer, is_integral, remove_pointer
 #include <utility> // pair, declval
+#include <vector> // vector
 
 #ifndef JSON_NO_IO
     #include <cstdio>   // FILE *
@@ -30,6 +32,33 @@
 #include <nlohmann/detail/meta/type_traits.hpp>
 #include <nlohmann/detail/string_utils.hpp>
 
+#ifndef JSON_NO_IO
+    // Which unlocked stdio functions file_input_adapter uses (see
+    // JSON_DISABLE_UNLOCKED_FILE_IO): 1 = POSIX flockfile/getc_unlocked,
+    // 2 = Microsoft CRT _lock_file/_fgetc_nolock, 0 = none (std::fgetc, which
+    // locks the file for every character). Only C libraries known to declare
+    // the functions are listed; the macros are tested after <cstdio>, which
+    // defines __GLIBC__. MemorySanitizer builds always use std::fgetc: glibc's
+    // inline getc_unlocked reads the stdio buffer directly, which glibc (not
+    // instrumented) filled, so MemorySanitizer would report a false positive.
+    #if defined(__has_feature)
+        #if __has_feature(memory_sanitizer)
+            #define JSON_UNLOCKED_FILE_IO_MSAN 1
+        #endif
+    #endif
+
+    #if JSON_DISABLE_UNLOCKED_FILE_IO || defined(JSON_UNLOCKED_FILE_IO_MSAN)
+        #define JSON_UNLOCKED_FILE_IO 0
+    #elif defined(_MSC_VER)
+        #define JSON_UNLOCKED_FILE_IO 2
+    #elif defined(__GLIBC__) || defined(__APPLE__) || defined(__ANDROID__) || defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__) || defined(__DragonFly__)
+        #define JSON_UNLOCKED_FILE_IO 1
+    #else
+        #define JSON_UNLOCKED_FILE_IO 0
+    #endif
+    #undef JSON_UNLOCKED_FILE_IO_MSAN
+#endif  // JSON_NO_IO
+
 NLOHMANN_JSON_NAMESPACE_BEGIN
 namespace detail
 {
@@ -56,18 +85,35 @@ class file_input_adapter
         : m_file(f)
     {
         JSON_ASSERT(m_file != nullptr);
+        lock();
     }
 
     // make class move-only
     file_input_adapter(const file_input_adapter&) = delete;
-    file_input_adapter(file_input_adapter&&) noexcept = default;
+    file_input_adapter(file_input_adapter&& rhs) noexcept
+        : m_file(rhs.m_file)
+    {
+        // the lock is owned by the thread, not the adapter: hand it over
+        rhs.m_file = nullptr;
+    }
     file_input_adapter& operator=(const file_input_adapter&) = delete;
     file_input_adapter& operator=(file_input_adapter&&) = delete;
-    ~file_input_adapter() = default;
+    ~file_input_adapter()
+    {
+        unlock();
+    }
 
     std::char_traits<char>::int_type get_character() noexcept
     {
+        // the file is locked for the lifetime of the adapter, so the
+        // per-character lock of std::fgetc can be skipped
+#if JSON_UNLOCKED_FILE_IO == 1
+        return getc_unlocked(m_file);
+#elif JSON_UNLOCKED_FILE_IO == 2
+        return _fgetc_nolock(m_file);
+#else
         return std::fgetc(m_file);
+#endif
     }
 
     // returns the number of characters successfully read
@@ -78,6 +124,32 @@ class file_input_adapter
     }
 
   private:
+    // Lock the file once for the whole parse instead of once per character.
+    // The lock is recursive, so fread() and the user's own stdio calls on the
+    // same thread keep working.
+    void lock() noexcept
+    {
+#if JSON_UNLOCKED_FILE_IO == 1
+        flockfile(m_file);
+#elif JSON_UNLOCKED_FILE_IO == 2
+        _lock_file(m_file);
+#endif
+    }
+
+    void unlock() noexcept
+    {
+        // a moved-from adapter has no file and holds no lock
+        if (m_file == nullptr)
+        {
+            return;
+        }
+#if JSON_UNLOCKED_FILE_IO == 1
+        funlockfile(m_file);
+#elif JSON_UNLOCKED_FILE_IO == 2
+        _unlock_file(m_file);
+#endif
+    }
+
     /// the file pointer to read from
     std::FILE* m_file;
 };
@@ -127,15 +199,26 @@ class input_stream_adapter
     input_stream_adapter& operator=(input_stream_adapter&) = delete;
     input_stream_adapter& operator=(input_stream_adapter&&) = delete;
 
-#if JSON_PRECISE_STREAM_POSITION
     input_stream_adapter(input_stream_adapter&& rhs) noexcept
-        : is(rhs.is), sb(rhs.sb), lookahead(rhs.lookahead)
+        : is(rhs.is), sb(rhs.sb)
+#if JSON_PRECISE_STREAM_POSITION
+        , lookahead(rhs.lookahead)
+#endif
+#if !JSON_DISABLE_STREAM_BULK_SCAN
+        , token_carry(std::move(rhs.token_carry))
+        , token_segment(rhs.token_segment)
+        , current_in_get_area(rhs.current_in_get_area)
+        , tracking_token(rhs.tracking_token)
+#endif
     {
         rhs.is = nullptr;
         rhs.sb = nullptr;
+#if JSON_PRECISE_STREAM_POSITION
         rhs.lookahead = false;
+#endif
     }
 
+#if JSON_PRECISE_STREAM_POSITION
     // Whether the character last returned by get_character() can be given back
     // to the input with release_lookahead().
     static constexpr bool supports_lookahead = true;
@@ -155,17 +238,41 @@ class input_stream_adapter
             sb->sbumpc();
         }
 
+#if !JSON_DISABLE_STREAM_BULK_SCAN
+        // sgetc() refills the get area if it is exhausted
+        const bool refill = streambuf_access::get_ptr(sb) == streambuf_access::get_end(sb);
+        if (JSON_HEDLEY_UNLIKELY(refill))
+        {
+            save_token_segment();
+        }
+#endif
+
         auto res = sb->sgetc();
         // set eof manually, as we don't use the istream interface.
         if (JSON_HEDLEY_UNLIKELY(res == std::char_traits<char>::eof()))
         {
             // there is nothing to step over next time
             lookahead = false;
+#if !JSON_DISABLE_STREAM_BULK_SCAN
+            current_in_get_area = false;
+#endif
             is->clear(is->rdstate() | std::ios::eofbit);
         }
         else
         {
             lookahead = true;
+#if !JSON_DISABLE_STREAM_BULK_SCAN
+            if (JSON_HEDLEY_UNLIKELY(refill))
+            {
+                // the peeked character is at gptr(), unless the streambuf has
+                // no get area
+                after_refill(res, streambuf_access::get_ptr(sb) != streambuf_access::get_end(sb));
+            }
+            else
+            {
+                current_in_get_area = true;
+            }
+#endif
         }
         return res;
     }
@@ -179,13 +286,6 @@ class input_stream_adapter
         lookahead = false;
     }
 #else
-    input_stream_adapter(input_stream_adapter&& rhs) noexcept
-        : is(rhs.is), sb(rhs.sb)
-    {
-        rhs.is = nullptr;
-        rhs.sb = nullptr;
-    }
-
     // std::istream/std::streambuf use std::char_traits<char>::to_int_type, to
     // ensure that std::char_traits<char>::eof() and the character 0xFF do not
     // end up as the same value, e.g., 0xFFFFFFFF.
@@ -194,12 +294,37 @@ class input_stream_adapter
     // stays consumed after parsing; see JSON_PRECISE_STREAM_POSITION.
     std::char_traits<char>::int_type get_character()
     {
+#if !JSON_DISABLE_STREAM_BULK_SCAN
+        // sbumpc() refills the get area if it is exhausted
+        const bool refill = streambuf_access::get_ptr(sb) == streambuf_access::get_end(sb);
+        if (JSON_HEDLEY_UNLIKELY(refill))
+        {
+            save_token_segment();
+        }
+#endif
+
         auto res = sb->sbumpc();
         // set eof manually, as we don't use the istream interface.
         if (JSON_HEDLEY_UNLIKELY(res == std::char_traits<char>::eof()))
         {
+#if !JSON_DISABLE_STREAM_BULK_SCAN
+            current_in_get_area = false;
+#endif
             is->clear(is->rdstate() | std::ios::eofbit);
         }
+#if !JSON_DISABLE_STREAM_BULK_SCAN
+        else if (JSON_HEDLEY_UNLIKELY(refill))
+        {
+            // the consumed character is at gptr() - 1, unless the streambuf
+            // has no get area and returned it from uflow()
+            const char* const next = streambuf_access::get_ptr(sb);
+            after_refill(res, next != nullptr && next != streambuf_access::get_begin(sb));
+        }
+        else
+        {
+            current_in_get_area = true;
+        }
+#endif
         return res;
     }
 #endif
@@ -218,7 +343,165 @@ class input_stream_adapter
         return res;
     }
 
+#if !JSON_DISABLE_STREAM_BULK_SCAN
+    // The lexer may scan the characters in the get area of the streambuf
+    // directly (SWAR fast paths). The get area is only a window into the
+    // input: it ends wherever the streambuf's buffer ends, and is refilled by
+    // the next get_character(). The lexer therefore keeps counting the
+    // position itself and treats the end of the window as "unknown", not as
+    // the end of the input.
+    static constexpr bool supports_bulk_scan = true;
+    static constexpr bool bulk_window = true;
+
+    // Pointer to the next unread character; only valid when
+    // bulk_remaining() > 0. The character before it is the one last returned
+    // by get_character().
+    const char* bulk_data() const
+    {
+        return current_end();
+    }
+
+    // Number of unread characters in the get area.
+    std::size_t bulk_remaining() const
+    {
+        if (!current_in_get_area)
+        {
+            return 0;
+        }
+        return static_cast<std::size_t>(streambuf_access::get_end(sb) - current_end());
+    }
+
+    // Consume @a n characters previously inspected via bulk_data(); the last
+    // of them becomes the character last returned by get_character().
+    void bulk_skip(std::size_t n)
+    {
+        JSON_ASSERT(n <= bulk_remaining());
+        // gbump() takes an int, but a get area (e.g., of a std::stringbuf) can
+        // be larger
+        constexpr auto max_bump = static_cast<std::size_t>((std::numeric_limits<int>::max)());
+        for (; n > max_bump; n -= max_bump)
+        {
+            streambuf_access::bump(sb, static_cast<int>(max_bump));
+        }
+        streambuf_access::bump(sb, static_cast<int>(n));
+    }
+
+    // The adapter keeps the current token for the lexer's error messages: the
+    // part in the get area is only remembered as a pointer, and copied out
+    // when the get area is about to be refilled.
+    static constexpr bool supports_token_tracking = true;
+
+    // Start a token with the characters read from now on. Until this is
+    // called, nothing is tracked (the binary readers never call it).
+    void begin_token() noexcept
+    {
+        tracking_token = true;
+        token_carry.clear();
+        token_segment = current_end();
+    }
+
+    // Start a new token with the character @a c last returned by
+    // get_character().
+    void mark_token_start(std::char_traits<char>::int_type c)
+    {
+        tracking_token = true;
+        token_carry.clear();
+        if (current_in_get_area)
+        {
+            token_segment = current_end() - 1;
+        }
+        else
+        {
+            token_segment = nullptr;
+            token_carry.push_back(std::char_traits<char>::to_char_type(c));
+        }
+    }
+
+    // Copy the characters read since mark_token_start() to @a out, without the
+    // last one if @a drop_last.
+    void copy_token(std::vector<char>& out, bool drop_last) const
+    {
+        out = token_carry;
+        if (token_segment != nullptr)
+        {
+            out.insert(out.end(), token_segment, current_end());
+        }
+        if (drop_last && !out.empty())
+        {
+            out.pop_back();
+        }
+    }
+#endif
+
   private:
+#if !JSON_DISABLE_STREAM_BULK_SCAN
+    // Reads the protected get-area pointers of any std::streambuf. Naming the
+    // members through a derived class makes forming the member pointers
+    // accessible; calling them on another streambuf object is then ordinary
+    // member access through a pointer to member.
+    struct streambuf_access : std::streambuf
+    {
+        static const char* get_ptr(std::streambuf* s)
+        {
+            return (s->*(&streambuf_access::gptr))();
+        }
+        static const char* get_end(std::streambuf* s)
+        {
+            return (s->*(&streambuf_access::egptr))();
+        }
+        static const char* get_begin(std::streambuf* s)
+        {
+            return (s->*(&streambuf_access::eback))();
+        }
+        static void bump(std::streambuf* s, int n)
+        {
+            (s->*(&streambuf_access::gbump))(n);
+        }
+    };
+
+    // one past the character last returned by get_character(), if that is in
+    // the get area
+    const char* current_end() const
+    {
+#if JSON_PRECISE_STREAM_POSITION
+        // the character is peeked at gptr() and not stepped over yet, unless
+        // it was given back with release_lookahead()
+        return streambuf_access::get_ptr(sb) + (lookahead ? 1 : 0);
+#else
+        return streambuf_access::get_ptr(sb);
+#endif
+    }
+
+    // the get area is about to be replaced: copy the token's part in it
+    void save_token_segment()
+    {
+        if (token_segment != nullptr)
+        {
+            token_carry.insert(token_carry.end(), token_segment, streambuf_access::get_end(sb));
+            token_segment = nullptr;
+        }
+    }
+
+    // character @a c was just read into a new get area (@a in_get_area) or
+    // returned without one
+    void after_refill(std::char_traits<char>::int_type c, bool in_get_area)
+    {
+        current_in_get_area = in_get_area;
+        if (!tracking_token)
+        {
+            return;
+        }
+        if (in_get_area)
+        {
+            token_segment = current_end() - 1;
+        }
+        else
+        {
+            token_carry.push_back(std::char_traits<char>::to_char_type(c));
+        }
+    }
+#endif
+
 #if JSON_PRECISE_STREAM_POSITION
     // Step over the character last returned by get_character(). The character
     // has already been peeked successfully, so for every streambuf with a get
@@ -240,6 +523,16 @@ class input_stream_adapter
     /// whether get_character() peeked a character that is not consumed yet
     bool lookahead = false;
 #endif
+#if !JSON_DISABLE_STREAM_BULK_SCAN
+    /// the part of the current token that was in earlier get areas
+    std::vector<char> token_carry{}; // NOLINT(readability-redundant-member-init)
+    /// start of the current token's part in the get area, or nullptr
+    const char* token_segment = nullptr;
+    /// whether the character last returned by get_character() is in the get area
+    bool current_in_get_area = false;
+    /// whether a token is tracked at all; the binary readers never start one
+    bool tracking_token = false;
+#endif
 };
 #endif  // JSON_NO_IO
 

include/nlohmann/detail/input/lexer.hpp

@@ -166,6 +166,43 @@ constexpr bool input_adapter_supports_bulk_scan(std::false_type /*detected*/)
     return false;
 }
 
+// Detect whether an input adapter keeps the current token itself and can
+// hand it out on demand (see input_stream_adapter::supports_token_tracking),
+// so the lexer need not copy every scanned character into token_string.
+template<typename InputAdapterType>
+using detect_supports_token_tracking = decltype(InputAdapterType::supports_token_tracking);
+
+template<typename InputAdapterType>
+constexpr bool input_adapter_supports_token_tracking(std::true_type /*detected*/)
+{
+    return InputAdapterType::supports_token_tracking;
+}
+
+template<typename InputAdapterType>
+constexpr bool input_adapter_supports_token_tracking(std::false_type /*detected*/)
+{
+    return false;
+}
+
+// Detect whether the contiguous block of a bulk-scan adapter is only a window
+// into the input (see input_stream_adapter::bulk_window): its end is not the
+// end of the input, so a token that touches it must be finished character by
+// character.
+template<typename InputAdapterType>
+using detect_bulk_window = decltype(InputAdapterType::bulk_window);
+
+template<typename InputAdapterType>
+constexpr bool input_adapter_bulk_window(std::true_type /*detected*/)
+{
+    return InputAdapterType::bulk_window;
+}
+
+template<typename InputAdapterType>
+constexpr bool input_adapter_bulk_window(std::false_type /*detected*/)
+{
+    return false;
+}
+
 /*!
 @brief lexical analysis
 
@@ -184,9 +221,20 @@ class lexer : public lexer_base<BasicJsonType>
     /// whether the last read token can be reconstructed from the input adapter
     /// on demand (in error paths) instead of being copied on every scanned
     /// character; see input_adapter_supports_seek
-    static constexpr bool lazy_token_string =
+    static constexpr bool seekable =
         input_adapter_supports_seek<InputAdapterType>(is_detected<detect_supports_seek, InputAdapterType> {});
 
+    /// whether the adapter keeps the token itself (mark_token_start() /
+    /// copy_token()); see input_adapter_supports_token_tracking
+    static constexpr bool adapter_token =
+        !seekable && input_adapter_supports_token_tracking<InputAdapterType>(is_detected<detect_supports_token_tracking, InputAdapterType> {});
+
+    /// whether the last read token is obtained from the input adapter on
+    /// demand (in error paths) instead of being copied on every scanned
+    /// character: seekable adapters reconstruct it from the consumed input,
+    /// token-tracking adapters keep it themselves
+    static constexpr bool lazy_token_string = seekable || adapter_token;
+
     /// whether the read position is derived from the input adapter on demand
     /// instead of being counted on every scanned character. Seekable adapters
     /// know how much input they consumed and can revisit it, so
@@ -194,7 +242,7 @@ class lexer : public lexer_base<BasicJsonType>
     /// and line and column are recomputed from the consumed input only when
     /// get_position() is called (errors). Counting per character kept two
     /// increments and a newline check in the hot path of every get().
-    static constexpr bool lazy_position = lazy_token_string;
+    static constexpr bool lazy_position = seekable;
 
     /// whether a simulated unget can be passed on to the input adapter, which
     /// then leaves the character in the input; see
@@ -203,14 +251,17 @@ class lexer : public lexer_base<BasicJsonType>
         input_adapter_supports_lookahead<InputAdapterType>(is_detected<detect_supports_lookahead, 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. It
-    /// also implies lazy_position, so the bulk paths need not advance the
-    /// position counters.
+    /// directly from a contiguous input buffer (SWAR fast path). Without
+    /// lazy_token_string / lazy_position, the bulk paths copy the consumed
+    /// characters into token_string and advance the position counters
+    /// themselves (see bulk_consume).
     static constexpr bool bulk_scan =
-        lazy_token_string
-        && input_adapter_supports_bulk_scan<InputAdapterType>(is_detected<detect_supports_bulk_scan, InputAdapterType> {});
+        input_adapter_supports_bulk_scan<InputAdapterType>(is_detected<detect_supports_bulk_scan, InputAdapterType> {});
+
+    /// whether the contiguous block is only a window into the input, so that
+    /// its end does not end a token; see input_adapter_bulk_window
+    static constexpr bool bulk_window =
+        input_adapter_bulk_window<InputAdapterType>(is_detected<detect_bulk_window, InputAdapterType> {});
 
   public:
     using token_type = typename lexer_base<BasicJsonType>::token_type;
@@ -219,7 +270,11 @@ class lexer : public lexer_base<BasicJsonType>
         : ia(std::move(adapter))
         , ignore_comments(ignore_comments_)
         , discard_number_values(discard_number_values_)
-    {}
+    {
+        // the token string of an error before the first token is everything
+        // read so far
+        begin_token(std::integral_constant<bool, adapter_token> {});
+    }
 
     // deleted because of pointer members
     lexer(const lexer&) = delete;
@@ -236,8 +291,8 @@ class lexer : public lexer_base<BasicJsonType>
     /// contiguous input: try to decode the 4 hex digits following `\\u`
     /// directly from the input buffer via hex_codepoint(), instead of 4 calls
     /// to get(). On success, advances the adapter exactly as those 4 get()
-    /// calls would (bulk_scan implies lazy_position, so there are no position
-    /// counters to advance) and leaves @a current holding the last of
+    /// calls would (including the position and token bookkeeping of a
+    /// streaming adapter, see bulk_consume) and leaves @a current holding the last of
     /// the 4 digits, just as the last such get() would; the codepoint is
     /// written to @a out. Makes no state change and returns false for a
     /// pending unget, fewer than 4 remaining bytes, or any of the 4 bytes not
@@ -256,7 +311,7 @@ class lexer : public lexer_base<BasicJsonType>
         {
             return false;
         }
-        ia.bulk_skip(4);
+        bulk_consume(reinterpret_cast<const char*>(raw), 4);
         current = char_traits<char_type>::to_int_type(raw[3]);
         out = codepoint;
         return true;
@@ -385,7 +440,36 @@ class lexer : public lexer_base<BasicJsonType>
             return;
         }
         token_buffer.append(reinterpret_cast<const typename string_t::value_type*>(data), pos);
-        ia.bulk_skip(pos);
+        bulk_consume(reinterpret_cast<const char*>(data), pos);
+    }
+
+    /// consume @a n characters at @a data (the adapter's bulk_data()) and do
+    /// the bookkeeping get() would have done for them; the characters are
+    /// never newlines
+    void bulk_consume(const char* data, std::size_t n)
+    {
+        bulk_track_position(n, std::integral_constant<bool, lazy_position> {});
+        bulk_track_token(data, n, std::integral_constant<bool, lazy_token_string> {});
+        ia.bulk_skip(n);
+    }
+
+    /// seekable adapter: the position is derived from the adapter
+    void bulk_track_position(std::size_t /*n*/, std::true_type /*lazy*/) const noexcept {}
+
+    /// streaming adapter: count the characters
+    void bulk_track_position(std::size_t n, std::false_type /*lazy*/) noexcept
+    {
+        position.chars_read_total += n;
+        position.chars_read_current_line += n;
+    }
+
+    /// the token is obtained from the adapter
+    void bulk_track_token(const char* /*data*/, std::size_t /*n*/, std::true_type /*lazy*/) const noexcept {}
+
+    /// streaming adapter: capture the characters
+    void bulk_track_token(const char* data, std::size_t n, std::false_type /*lazy*/)
+    {
+        token_string.insert(token_string.end(), data, data + n);
     }
 
     /// streaming input: no bulk fast path
@@ -1541,9 +1625,12 @@ scan_number_done:
         const std::size_t avail = rem + 1;
 
         // validate + classify the number extent (mirrors scan_number()'s grammar)
+        // and accumulate the mantissa digits on the way, so that tokens with up
+        // to 19 digits need no second pass for their value
         std::size_t i = 0;
         std::size_t dot_index = std::string::npos;
         token_type number_type = token_type::value_unsigned;
+        std::uint64_t w = 0;
         if (data[0] == '-')
         {
             number_type = token_type::value_integer;
@@ -1553,22 +1640,21 @@ scan_number_done:
                 return token_type::uninitialized;
             }
         }
+        const std::size_t integer_begin = i;
         if (data[i] == '0')
         {
             ++i;
         }
         else if (data[i] >= '1' && data[i] <= '9')
         {
-            ++i;
-            while (i < avail && data[i] >= '0' && data[i] <= '9')
-            {
-                ++i;
-            }
+            i = accumulate_digits(data, i, avail, w);
         }
         else
         {
             return token_type::uninitialized;
         }
+        std::size_t digit_count = i - integer_begin;
+        std::size_t fraction_digits = 0;
         if (i < avail && data[i] == '.')
         {
             number_type = token_type::value_float;
@@ -1578,31 +1664,45 @@ scan_number_done:
             {
                 return token_type::uninitialized;
             }
-            while (i < avail && data[i] >= '0' && data[i] <= '9')
-            {
-                ++i;
-            }
+            const std::size_t fraction_begin = i;
+            i = accumulate_digits(data, i, avail, w);
+            fraction_digits = i - fraction_begin;
+            digit_count += fraction_digits;
         }
         // the mantissa ends here, whether or not an exponent part follows
         const std::size_t mantissa_end = i;
+        std::int64_t exponent = 0;
         if (i < avail && (data[i] == 'e' || data[i] == 'E'))
         {
             number_type = token_type::value_float;
             ++i;
+            bool negative_exponent = false;
             if (i < avail && (data[i] == '+' || data[i] == '-'))
             {
+                negative_exponent = data[i] == '-';
                 ++i;
             }
             if (i >= avail || !(data[i] >= '0' && data[i] <= '9'))
             {
                 return token_type::uninitialized;
             }
+            // saturated far beyond every range, see parse_float_exponent()
+            constexpr std::int64_t saturation = 100000000000000000; // 10^17
             while (i < avail && data[i] >= '0' && data[i] <= '9')
             {
+                if (exponent < saturation)
+                {
+                    exponent = (exponent * 10) + (data[i] - '0');
+                }
                 ++i;
             }
+            exponent = negative_exponent ? -exponent : exponent;
         }
         const std::size_t len = i;
+        if (number_may_continue(len == avail, std::integral_constant<bool, bulk_window> {}))
+        {
+            return token_type::uninitialized;
+        }
 
         // reset() records where this token starts (for diagnostics), so it has
         // to run before the input position advances below
@@ -1617,10 +1717,14 @@ scan_number_done:
 #if !JSON_DIAGNOSTIC_POSITIONS
         if (number_type != token_type::value_float)
         {
-            const token_type integer_result = convert_integer(number_type, data, data + len);
+            // up to 18 digits always fit into std::uint64_t and std::int64_t
+            const token_type integer_result = (digit_count <= 18)
+                                              ? store_integer(number_type, w)
+                                              : convert_integer(number_type, data, data + len);
             if (JSON_HEDLEY_LIKELY(integer_result != token_type::uninitialized))
             {
-                ia.bulk_skip(len - 1);
+                bulk_consume(data + 1, len - 1);
+                read_number_terminator(std::integral_constant<bool, bulk_window> {});
                 return integer_result;
             }
             // The value does not fit an integer, so this token converts as a
@@ -1630,17 +1734,99 @@ scan_number_done:
         }
 #endif
 
+        decimal_point_position = dot_index;
+
+        bulk_consume(data + 1, len - 1);
+        read_number_terminator(std::integral_constant<bool, bulk_window> {});
+
+        // a float whose digits all fit into w converts from the accumulated
+        // value; longer ones (and other float formats) take the general path
+        if (number_type == token_type::value_float && digit_count <= 19)
+        {
+            float_significand s;
+            s.w = w;
+            s.exponent = exponent - static_cast<std::int64_t>(fraction_digits);
+            s.negative = data[0] == '-';
+            if (convert_accumulated_float(s, std::integral_constant<bool, has_native_float_format<number_float_t>::value> {}))
+            {
+                // the token text is only copied into token_buffer when
+                // get_string() asks for it; the library's own SAX handlers
+                // ignore it. A window into a stream may be refilled by the
+                // next read, so there the text is copied right away.
+                if (bulk_window)
+                {
+                    token_buffer.append(reinterpret_cast<const typename string_t::value_type*>(data), len);
+                }
+                else
+                {
+                    pending_number_text = data;
+                    pending_number_length = len;
+                }
+                return token_type::value_float;
+            }
+        }
+
         // materialize the token exactly as scan_number() would. 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);
-        decimal_point_position = dot_index;
-
-        ia.bulk_skip(len - 1);
-
         return convert_number(number_type, mantissa_end);
     }
 
+    /// window adapter: a number that reaches the end of the window may continue
+    /// after it, so scan_number() has to read it character by character
+    static constexpr bool number_may_continue(bool at_end, std::true_type /*window*/) noexcept
+    {
+        return at_end;
+    }
+
+    /// the end of the block is the end of the input, which ends the number
+    static constexpr bool number_may_continue(bool /*at_end*/, std::false_type /*window*/) noexcept
+    {
+        return false;
+    }
+
+    /// window adapter: read the character after a bulk-scanned number and unget
+    /// it, as scan_number() does. The adapter is a stream that is handed back to
+    /// the user, who must find the same input consumed either way.
+    void read_number_terminator(std::true_type /*window*/)
+    {
+        get();
+        unget();
+    }
+
+    /// the input is not handed back, so the terminator can stay unread
+    void read_number_terminator(std::false_type /*window*/) const noexcept {}
+
+    /// binary32/binary64: the value of a float token from its accumulated
+    /// significand (see scan_number_bulk_contiguous())
+    bool convert_accumulated_float(const float_significand& s, std::true_type /*native*/) noexcept
+    {
+        value_float = decimal_to_float<number_float_t>(s);
+        return true;
+    }
+
+    /// other formats (e.g. x87 long double): not instantiated for them, as
+    /// decimal_to_float() only supports binary32/binary64
+    static bool convert_accumulated_float(const float_significand& /*s*/, std::false_type /*native*/) noexcept
+    {
+        return false;
+    }
+
+    /// the value of an integer token with at most 18 digits, accumulated in @a w
+    token_type store_integer(token_type number_type, std::uint64_t w) noexcept
+    {
+        if (number_type == token_type::value_unsigned)
+        {
+            value_unsigned = static_cast<number_unsigned_t>(w);
+            // reject values that do not round-trip into a narrower type
+            return (static_cast<std::uint64_t>(value_unsigned) == w) ? token_type::value_unsigned : token_type::uninitialized;
+        }
+        const std::int64_t x = -static_cast<std::int64_t>(w);
+        value_integer = static_cast<number_integer_t>(x);
+        return (static_cast<std::int64_t>(value_integer) == x) ? token_type::value_integer : token_type::uninitialized;
+    }
+
     /// contiguous input: try the number fast path, else the byte-path scanner
     token_type scan_number_dispatch(std::true_type /*bulk*/)
     {
@@ -1683,6 +1869,7 @@ scan_number_done:
     void reset() noexcept
     {
         token_buffer.clear();
+        pending_number_text = nullptr;
         decimal_point_position = std::string::npos;
 
 #if JSON_DIAGNOSTIC_POSITIONS
@@ -1696,11 +1883,30 @@ scan_number_done:
     /// seekable adapter: remember where the current token starts so it can be
     /// reconstructed from the input on error; current has already been
     /// consumed, hence the -1
-    void note_token_start(std::true_type /*lazy*/) noexcept
+    void note_token_start(std::true_type /*lazy*/)
+    {
+        note_token_start_lazy(std::integral_constant<bool, adapter_token> {});
+    }
+
+    void note_token_start_lazy(std::false_type /*adapter_token*/) noexcept
     {
         token_string_start = ia.get_consumed_count() - 1;
     }
 
+    /// token-tracking adapter: start tracking the token at the first character
+    void begin_token(std::true_type /*adapter_token*/) noexcept
+    {
+        ia.begin_token();
+    }
+
+    void begin_token(std::false_type /*adapter_token*/) const noexcept {}
+
+    /// token-tracking adapter: the token starts with the current character
+    void note_token_start_lazy(std::true_type /*adapter_token*/)
+    {
+        ia.mark_token_start(current);
+    }
+
     /// streaming adapter: start copying the token eagerly, beginning with the
     /// already-read first character
     void note_token_start(std::false_type /*lazy*/) noexcept
@@ -1923,10 +2129,30 @@ scan_number_done:
     /// return current string value
     string_t& get_string()
     {
+        if (pending_number_text != nullptr)
+        {
+            token_buffer.append(reinterpret_cast<const typename string_t::value_type*>(pending_number_text), pending_number_length);
+            pending_number_text = nullptr;
+        }
         // a number token holds '.' regardless of the locale (#4084)
         return token_buffer;
     }
 
+    /// the value of a string token (key or string), which never has pending
+    /// number text, so get_string()'s check is not needed
+    string_t& get_string_token() noexcept
+    {
+        JSON_ASSERT(pending_number_text == nullptr);
+        return token_buffer;
+    }
+
+    /// the text of a float token for a SAX handler that ignores it (the
+    /// library's own handlers): empty, without copying the token
+    const string_t& get_ignored_number_text() const noexcept
+    {
+        return token_buffer;
+    }
+
     /////////////////////
     // diagnostics
     /////////////////////
@@ -2021,6 +2247,20 @@ scan_number_done:
   public:
     /// seekable adapter: rebuild the last read token from the input on demand
     const std::vector<char_type>& collect_token_chars(std::vector<char_type>& out, std::true_type /*lazy*/) const
+    {
+        return collect_token_chars_lazy(out, std::integral_constant<bool, adapter_token> {});
+    }
+
+    /// token-tracking adapter: the adapter hands out everything read since
+    /// mark_token_start(), except a pending unget (see below)
+    const std::vector<char_type>& collect_token_chars_lazy(std::vector<char_type>& out, std::true_type /*adapter_token*/) const
+    {
+        const bool pending_real_unget = next_unget && current != char_traits<char_type>::eof();
+        ia.copy_token(out, pending_real_unget);
+        return out;
+    }
+
+    const std::vector<char_type>& collect_token_chars_lazy(std::vector<char_type>& out, std::false_type /*adapter_token*/) const
     {
         // a pending unget of a real (non-EOF) character means that character
         // was consumed from the input but is not part of the token; EOF is
@@ -2142,9 +2382,8 @@ scan_number_done:
     Pretty-printed JSON indents every line with a run of spaces, which the loop
     in skip_whitespace() would otherwise read one character at a time. The
     spaces are skipped 8 bytes at a time; the last 7 bytes of the input and any
-    other whitespace are left to that loop. Skipping bypasses get(), which is
-    fine because bulk_scan implies lazy_position: the position is derived from
-    the adapter, not counted per character.
+    other whitespace are left to that loop. Skipping bypasses get();
+    bulk_consume() does its bookkeeping.
     */
     void skip_indentation(std::true_type /*bulk*/)
     {
@@ -2167,7 +2406,7 @@ scan_number_done:
                 break;
             }
         }
-        ia.bulk_skip(pos);
+        bulk_consume(data, pos);
     }
 
     /// streaming input: no bulk fast path
@@ -2309,6 +2548,11 @@ scan_number_done:
     /// whether the next get() call should just return current
     bool next_unget = false;
 
+    /// the text of the current float token in the input, not yet copied into
+    /// token_buffer (see get_string()), or nullptr
+    const char* pending_number_text = nullptr;
+    std::size_t pending_number_length = 0;
+
     /// the start position of the current token
     position_t position {};
 

include/nlohmann/detail/input/number_parse.hpp

@@ -275,13 +275,77 @@ JSON_HEDLEY_ALWAYS_INLINE std::uint32_t parse_eight_digits(std::uint64_t v) noex
     return static_cast<std::uint32_t>(((v & 0x0000FFFF0000FFFFu) * 42949672960001u) >> 32u);
 }
 
-/// whether [first, last) contains a digit other than '0'
-inline bool has_nonzero_digit(const char* first, const char* last) noexcept
+/// whether the eight bytes in @a v (see read_eight_bytes()) are all ASCII
+/// digits (after fast_float)
+JSON_HEDLEY_ALWAYS_INLINE bool is_eight_digits(std::uint64_t v) noexcept
 {
-    return std::any_of(first, last, [](char c) noexcept
+    return ((v & 0xF0F0F0F0F0F0F0F0u) | (((v + 0x0606060606060606u) & 0xF0F0F0F0F0F0F0F0u) >> 4u)) == 0x3333333333333333u;
+}
+
+/*!
+@brief skip the ASCII digits starting at data[i] and accumulate their value
+
+Eight digits are taken at a time while at least eight bytes remain. @a w wraps
+around after 19 digits; the caller only uses it for shorter digit runs.
+
+@return the index of the first byte at or after @a i that is not a digit
+*/
+JSON_HEDLEY_ALWAYS_INLINE std::size_t accumulate_digits(const char* data, std::size_t i, std::size_t avail, std::uint64_t& w) noexcept
+{
+    while (avail - i >= 8)
     {
-        return c != '0';
-    });
+        const std::uint64_t v = read_eight_bytes(data + i);
+        if (!is_eight_digits(v))
+        {
+            break;
+        }
+        w = (w * 100000000u) + parse_eight_digits(v);
+        i += 8;
+    }
+    while (i < avail && data[i] >= '0' && data[i] <= '9')
+    {
+        w = (w * 10u) + static_cast<std::uint64_t>(data[i] - '0');
+        ++i;
+    }
+    return i;
+}
+
+/// whether [first, last) contains a digit other than '0'; a plain loop, as
+/// GCC calls std::any_of out of line even for the usual empty range
+JSON_HEDLEY_ALWAYS_INLINE bool has_nonzero_digit(const char* first, const char* last) noexcept
+{
+    for (; first != last; ++first)
+    {
+        if (*first != '0')
+        {
+            return true;
+        }
+    }
+    return false;
+}
+
+/*!
+@brief accumulate the validated digits [p, last) into @a w, as long as
+       @a remaining digits still fit into it
+
+Eight digits are taken at a time while at least eight remain, and @a remaining
+is decreased by the number of digits taken.
+
+@return a pointer past the last digit taken
+*/
+JSON_HEDLEY_ALWAYS_INLINE const char* accumulate_significant_digits(const char* p, const char* last, int& remaining, std::uint64_t& w) noexcept
+{
+    while (remaining >= 8 && last - p >= 8)
+    {
+        w = (w * 100000000u) + parse_eight_digits(read_eight_bytes(p));
+        p += 8;
+        remaining -= 8;
+    }
+    for (; remaining > 0 && p != last; ++p, --remaining)
+    {
+        w = (w * 10u) + static_cast<std::uint64_t>(*p - '0');
+    }
+    return p;
 }
 
 /// the value of the validated exponent digits [+-]?[0-9]+ in [first, last),
@@ -327,7 +391,7 @@ truncated is set.
                                    std::string::npos if there is none
 @param[in] mantissa_end            index of the 'e'/'E', or the token length
 */
-inline float_significand parse_float_significand(const char* first, const char* last,
+JSON_HEDLEY_ALWAYS_INLINE float_significand parse_float_significand(const char* first, const char* last,
         std::size_t decimal_point_position, std::size_t mantissa_end) noexcept
 {
     float_significand s;
@@ -342,16 +406,7 @@ inline float_significand parse_float_significand(const char* first, const char*
     int remaining = 19; // digits that still fit into w
     if (*p != '0') // the integer part is "0" or [1-9][0-9]*
     {
-        while (remaining >= 8 && integer_last - p >= 8)
-        {
-            w = (w * 100000000u) + parse_eight_digits(read_eight_bytes(p));
-            p += 8;
-            remaining -= 8;
-        }
-        for (; remaining > 0 && p != integer_last; ++p, --remaining)
-        {
-            w = (w * 10u) + static_cast<std::uint64_t>(*p - '0');
-        }
+        p = accumulate_significant_digits(p, integer_last, remaining, w);
         s.exponent = integer_last - p;
         s.truncated = has_nonzero_digit(p, integer_last);
     }
@@ -370,16 +425,7 @@ inline float_significand parse_float_significand(const char* first, const char*
             s.exponent -= p - zeros;
         }
         const char* const digits = p;
-        while (remaining >= 8 && mantissa_last - p >= 8)
-        {
-            w = (w * 100000000u) + parse_eight_digits(read_eight_bytes(p));
-            p += 8;
-            remaining -= 8;
-        }
-        for (; remaining > 0 && p != mantissa_last; ++p, --remaining)
-        {
-            w = (w * 10u) + static_cast<std::uint64_t>(*p - '0');
-        }
+        p = accumulate_significant_digits(p, mantissa_last, remaining, w);
         s.exponent -= p - digits;
         s.truncated = s.truncated || has_nonzero_digit(p, mantissa_last);
     }

include/nlohmann/detail/input/parser.hpp

@@ -51,6 +51,20 @@ template<typename BasicJsonType>
 using parser_callback_t =
     std::function<bool(int /*depth*/, parse_event_t /*event*/, BasicJsonType& /*parsed*/)>;
 
+/// whether a SAX handler ignores the text of float tokens (the library's own
+/// handlers do), so that the parser need not materialize it
+template<typename SAX>
+struct sax_ignores_number_text : std::false_type {};
+
+template<typename BasicJsonType, typename InputAdapterType>
+struct sax_ignores_number_text<json_sax_dom_parser<BasicJsonType, InputAdapterType>> : std::true_type {};
+
+template<typename BasicJsonType, typename InputAdapterType>
+struct sax_ignores_number_text<json_sax_dom_callback_parser<BasicJsonType, InputAdapterType>> : std::true_type {};
+
+template<typename BasicJsonType>
+struct sax_ignores_number_text<json_sax_acceptor<BasicJsonType>> : std::true_type {};
+
 /*!
 @brief syntax analysis
 
@@ -171,6 +185,23 @@ class parser
     }
 
   private:
+    /// the text of the current float token for @a sax (see sax_ignores_number_text)
+    template<typename SAX>
+    const string_t& number_float_text(SAX* /*sax*/)
+    {
+        return number_float_text(std::integral_constant<bool, sax_ignores_number_text<SAX>::value> {});
+    }
+
+    const string_t& number_float_text(std::true_type /*ignored*/) const noexcept
+    {
+        return m_lexer.get_ignored_number_text();
+    }
+
+    string_t& number_float_text(std::false_type /*ignored*/)
+    {
+        return m_lexer.get_string();
+    }
+
     /*!
     @brief run a DOM SAX parser to completion and position the lexer
 
@@ -247,7 +278,7 @@ class parser
                         {
                             return syntax_error(*sax, exception_message(token_type::value_string, "object key"));
                         }
-                        if (JSON_HEDLEY_UNLIKELY(!sax->key(m_lexer.get_string())))
+                        if (JSON_HEDLEY_UNLIKELY(!sax->key(m_lexer.get_string_token())))
                         {
                             return false;
                         }
@@ -301,7 +332,7 @@ class parser
                                                     out_of_range::create(exception_id::number_overflow, concat("number overflow parsing '", m_lexer.get_token_string(), '\''), nullptr));
                         }
 
-                        if (JSON_HEDLEY_UNLIKELY(!sax->number_float(res, m_lexer.get_string())))
+                        if (JSON_HEDLEY_UNLIKELY(!sax->number_float(res, number_float_text(sax))))
                         {
                             return false;
                         }
@@ -347,7 +378,7 @@ class parser
 
                     case token_type::value_string:
                     {
-                        if (JSON_HEDLEY_UNLIKELY(!sax->string(m_lexer.get_string())))
+                        if (JSON_HEDLEY_UNLIKELY(!sax->string(m_lexer.get_string_token())))
                         {
                             return false;
                         }
@@ -456,7 +487,7 @@ class parser
                         return syntax_error(*sax, exception_message(token_type::value_string, "object key"));
                     }
 
-                    if (JSON_HEDLEY_UNLIKELY(!sax->key(m_lexer.get_string())))
+                    if (JSON_HEDLEY_UNLIKELY(!sax->key(m_lexer.get_string_token())))
                     {
                         return false;
                     }

include/nlohmann/detail/json_pointer.hpp

@@ -532,7 +532,7 @@ class json_pointer
             {}
         };
 
-        std::vector<slot> slots;
+        std::vector<slot> table;
         std::size_t used = 0;
 
         static std::size_t hash_of(std::size_t parent, const string_t& token) noexcept
@@ -547,19 +547,19 @@ class json_pointer
 
         void grow()
         {
-            std::vector<slot> old(slots.size() * 2);
-            old.swap(slots);
-            const std::size_t mask = slots.size() - 1;
+            std::vector<slot> old(table.size() * 2);
+            old.swap(table);
+            const std::size_t mask = table.size() - 1;
             for (const auto& s : old)
             {
                 if (s.token != nullptr)
                 {
                     std::size_t i = s.hash & mask;
-                    while (slots[i].token != nullptr)
+                    while (table[i].token != nullptr)
                     {
                         i = (i + 1) & mask;
                     }
-                    slots[i] = s;
+                    table[i] = s;
                 }
             }
         }
@@ -572,7 +572,7 @@ class json_pointer
 
         // start with the empty prefix only
         prefix_tree()
-            : slots(16)
+            : table(16)
             , parent(1, 0)
             , is_array(1, false)
         {}
@@ -582,20 +582,20 @@ class json_pointer
         std::size_t add_child(std::size_t id, const string_t& reference_token)
         {
             const std::size_t hash = hash_of(id, reference_token);
-            const std::size_t mask = slots.size() - 1;
+            const std::size_t mask = table.size() - 1;
             std::size_t i = hash & mask;
-            while (slots[i].token != nullptr)
+            while (table[i].token != nullptr)
             {
-                if (slots[i].hash == hash && slots[i].parent == id && *slots[i].token == reference_token)
+                if (table[i].hash == hash && table[i].parent == id && *table[i].token == reference_token)
                 {
-                    return slots[i].id;
+                    return table[i].id;
                 }
                 i = (i + 1) & mask;
             }
 
             // new prefix
             const std::size_t child = parent.size();
-            slots[i] = slot(hash, id, &reference_token, child);
+            table[i] = slot(hash, id, &reference_token, child);
             if (reference_token == "0")
             {
                 is_array[id] = true;
@@ -604,7 +604,7 @@ class json_pointer
             is_array.push_back(false);
 
             // keep the load factor below 1/2
-            if (++used * 2 > slots.size())
+            if (++used * 2 > table.size())
             {
                 grow();
             }

include/nlohmann/detail/macro_unscope.hpp

@@ -21,6 +21,7 @@
 #undef JSON_NO_UNIQUE_ADDRESS
 #undef JSON_DISABLE_ENUM_SERIALIZATION
 #undef JSON_DISABLE_TUPLE_REFERENCE_CONVERSION
+#undef JSON_UNLOCKED_FILE_IO
 #undef JSON_DTOA_SSE2
 #undef JSON_DTOA_NEON
 
@@ -47,6 +48,8 @@
     #undef JSON_STRICT_BINARY_UTF8
     #undef JSON_DELETE_DEPRECATED_FUNCTIONS
     #undef JSON_USE_OBJECTS_FOR_ENUM_KEYED_MAPS
+    #undef JSON_DISABLE_STREAM_BULK_SCAN
+    #undef JSON_DISABLE_UNLOCKED_FILE_IO
 #endif
 
 #include <nlohmann/thirdparty/hedley/hedley_undef.hpp>

include/nlohmann/detail/output/binary_writer.hpp

@@ -23,6 +23,7 @@
     #include <cstdlib> // _byteswap_ushort, _byteswap_ulong, _byteswap_uint64
 #endif
 
+#include <nlohmann/detail/bit_ops.hpp>
 #include <nlohmann/detail/input/binary_reader.hpp>
 #include <nlohmann/detail/input/string_scan.hpp>
 #include <nlohmann/detail/macro_scope.hpp>
@@ -204,24 +205,12 @@ class binary_writer
             }
 
             case value_t::number_integer:
-            {
-                if (j.m_data.m_value.number_integer >= 0)
-                {
-                    // CBOR does not differentiate between positive signed
-                    // integers and unsigned integers
-                    write_cbor_head(0x00, static_cast<std::uint64_t>(j.m_data.m_value.number_integer));
-                }
-                else
-                {
-                    // a negative integer n is encoded as -1 - n
-                    write_cbor_head(0x20, static_cast<std::uint64_t>(-1 - j.m_data.m_value.number_integer));
-                }
-                break;
-            }
-
             case value_t::number_unsigned:
             {
-                write_cbor_head(0x00, conditional_static_cast<std::uint64_t>(j.m_data.m_value.number_unsigned));
+                // one case for both, so that an array mixing negative
+                // (number_integer) and non-negative (number_unsigned) values
+                // does not mispredict the switch
+                write_cbor_integer(j, std::integral_constant<bool, integers_fit_64_bits()>());
                 break;
             }
 
@@ -256,13 +245,7 @@ class binary_writer
 
             case value_t::array:
             {
-                // step 1: write control byte and the array size
-                write_cbor_head(0x80, j.m_data.m_value.array->size());
-
-                for (const auto& el : *j.m_data.m_value.array)
-                {
-                    write_cbor(el, depth + 1);
-                }
+                write_cbor_array(j, depth);
                 break;
             }
 
@@ -420,54 +403,12 @@ class binary_writer
             }
 
             case value_t::number_integer:
-            {
-                if (j.m_data.m_value.number_integer >= 0)
-                {
-                    // MessagePack does not differentiate between positive
-                    // signed integers and unsigned integers.
-                    write_msgpack_unsigned(static_cast<std::uint64_t>(j.m_data.m_value.number_integer));
-                }
-                else
-                {
-                    if (j.m_data.m_value.number_integer >= -32)
-                    {
-                        // negative fixnum
-                        write_number(static_cast<std::int8_t>(j.m_data.m_value.number_integer));
-                    }
-                    else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int8_t>::min)() &&
-                             j.m_data.m_value.number_integer <= (std::numeric_limits<std::int8_t>::max)())
-                    {
-                        // int 8
-                        oa.write_character(to_char_type(0xD0));
-                        write_number(static_cast<std::int8_t>(j.m_data.m_value.number_integer));
-                    }
-                    else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int16_t>::min)() &&
-                             j.m_data.m_value.number_integer <= (std::numeric_limits<std::int16_t>::max)())
-                    {
-                        // int 16
-                        oa.write_character(to_char_type(0xD1));
-                        write_number(static_cast<std::int16_t>(j.m_data.m_value.number_integer));
-                    }
-                    else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int32_t>::min)() &&
-                             j.m_data.m_value.number_integer <= (std::numeric_limits<std::int32_t>::max)())
-                    {
-                        // int 32
-                        oa.write_character(to_char_type(0xD2));
-                        write_number(static_cast<std::int32_t>(j.m_data.m_value.number_integer));
-                    }
-                    else
-                    {
-                        // int 64
-                        oa.write_character(to_char_type(0xD3));
-                        write_number(static_cast<std::int64_t>(j.m_data.m_value.number_integer));
-                    }
-                }
-                break;
-            }
-
             case value_t::number_unsigned:
             {
-                write_msgpack_unsigned(static_cast<std::uint64_t>(j.m_data.m_value.number_unsigned));
+                // one case for both, so that an array mixing negative
+                // (number_integer) and non-negative (number_unsigned) values
+                // does not mispredict the switch
+                write_msgpack_integer(j, std::integral_constant<bool, integers_fit_64_bits()>());
                 break;
             }
 
@@ -485,13 +426,7 @@ class binary_writer
 
             case value_t::array:
             {
-                // step 1: write control byte and the array size
-                write_msgpack_array_prefix(j.m_data.m_value.array->size(), j);
-
-                for (const auto& el : *j.m_data.m_value.array)
-                {
-                    write_msgpack(el, depth + 1);
-                }
+                write_msgpack_array(j, depth);
                 break;
             }
 
@@ -824,6 +759,142 @@ class binary_writer
         JSON_THROW(type_error::create(exception_id::discarded_value_used, concat("cannot serialize discarded value to ", format_name), &j));
     }
 
+    /// how the range of integer type @a NumberType relates to that of @a TargetType
+    template<typename TargetType, typename NumberType>
+    struct integer_range
+    {
+        /// whether NumberType can hold values above the maximum of TargetType
+        static constexpr bool may_exceed_max = std::numeric_limits<NumberType>::digits > std::numeric_limits<TargetType>::digits;
+        /// whether NumberType can hold values below the minimum of TargetType
+        static constexpr bool may_fall_below_min = std::numeric_limits<NumberType>::is_signed
+            && (!std::numeric_limits<TargetType>::is_signed || may_exceed_max);
+    };
+
+    /*!
+    @brief whether integer @a n does not exceed the maximum of @a TargetType
+
+    The binary formats store integers in at most 64 bits, but number_integer_t
+    and number_unsigned_t may be wider (e.g., __int128). The range of the
+    integer is taken from std::numeric_limits, so a number type whose range
+    fits into @a TargetType is never checked: the function is then constant
+    true, and the default int64_t/uint64_t types pay nothing.
+    */
+    template<typename TargetType, typename NumberType>
+    static constexpr bool integer_fits_max(const NumberType n) noexcept
+    {
+        return integer_fits_max<TargetType>(n, std::integral_constant<bool, integer_range<TargetType, NumberType>::may_exceed_max>());
+    }
+
+    template<typename TargetType, typename NumberType>
+    static constexpr bool integer_fits_max(const NumberType /*unused*/, std::false_type /*may_exceed*/) noexcept
+    {
+        return true;
+    }
+
+    template<typename TargetType, typename NumberType>
+    static constexpr bool integer_fits_max(const NumberType n, std::true_type /*may_exceed*/) noexcept
+    {
+        // NumberType has more digits than TargetType, so it can hold its maximum
+        return n <= static_cast<NumberType>((std::numeric_limits<TargetType>::max)());
+    }
+
+    /*!
+    @brief whether integer @a n is not below the minimum of @a TargetType;
+           constant true if NumberType cannot hold such a value
+    */
+    template<typename TargetType, typename NumberType>
+    static constexpr bool integer_fits_min(const NumberType n) noexcept
+    {
+        return integer_fits_min<TargetType>(n, std::integral_constant<bool, integer_range<TargetType, NumberType>::may_fall_below_min>());
+    }
+
+    template<typename TargetType, typename NumberType>
+    static constexpr bool integer_fits_min(const NumberType /*unused*/, std::false_type /*may_fall_below*/) noexcept
+    {
+        return true;
+    }
+
+    template<typename TargetType, typename NumberType>
+    static constexpr bool integer_fits_min(const NumberType n, std::true_type /*may_fall_below*/) noexcept
+    {
+        // NumberType is signed and either TargetType is unsigned (minimum 0)
+        // or NumberType has more digits, so it can hold the minimum
+        return n >= static_cast<NumberType>((std::numeric_limits<TargetType>::min)());
+    }
+
+    /*!
+    @brief whether integer @a n lies in the range of @a TargetType
+    */
+    template<typename TargetType, typename NumberType>
+    static constexpr bool integer_fits(const NumberType n) noexcept
+    {
+        return integer_fits_min<TargetType>(n) && integer_fits_max<TargetType>(n);
+    }
+
+    /*!
+    @brief the decimal digits of the non-negative integer @a n
+    */
+    template<typename NumberType>
+    static std::string integer_to_string(NumberType n, std::false_type /*is_signed*/)
+    {
+        std::string result;
+        do
+        {
+            result.push_back(static_cast<char>('0' + (n % 10)));
+            n = static_cast<NumberType>(n / 10);
+        }
+        while (n != 0);
+        std::reverse(result.begin(), result.end());
+        return result;
+    }
+
+    /*!
+    @brief the decimal digits of the integer @a n, with a leading '-' if it
+           is negative
+    */
+    template<typename NumberType>
+    static std::string integer_to_string(NumberType n, std::true_type /*is_signed*/)
+    {
+        // the remainder of a negative n is not positive; each digit is
+        // negated on its own, because negating the minimum of NumberType
+        // as a whole would overflow
+        const bool negative = n < 0;
+        std::string result;
+        do
+        {
+            const auto remainder = n % 10;
+            result.push_back(static_cast<char>('0' + (remainder < 0 ? -remainder : remainder)));
+            n = static_cast<NumberType>(n / 10);
+        }
+        while (n != 0);
+        if (negative)
+        {
+            result.push_back('-');
+        }
+        std::reverse(result.begin(), result.end());
+        return result;
+    }
+
+    /*!
+    @brief throws because the integer @a j is too large for @a format_name
+    @throw out_of_range.407 always
+    */
+    JSON_HEDLEY_NO_RETURN static void throw_integer_out_of_range(const BasicJsonType& j, const char* format_name, const char* range)
+    {
+        // neither std::to_string(), which has no overload for integer types
+        // wider than 64 bits, nor dump(), which would also instantiate the
+        // serializer for objects and fail for key types it cannot write
+        using integer_limits = std::numeric_limits<typename BasicJsonType::number_integer_t>;
+        using unsigned_limits = std::numeric_limits<typename BasicJsonType::number_unsigned_t>;
+        const std::string number = j.is_number_unsigned()
+                                   ? integer_to_string(j.m_data.m_value.number_unsigned, std::integral_constant<bool, unsigned_limits::is_signed>())
+                                   : integer_to_string(j.m_data.m_value.number_integer, std::integral_constant<bool, integer_limits::is_signed>());
+        static_cast<void>(number); // unused when JSON_NOEXCEPTION is defined
+        static_cast<void>(format_name);
+        static_cast<void>(range);
+        JSON_THROW(out_of_range::create(exception_id::integer_too_large, concat("integer number ", number, " cannot be represented by ", format_name, " as it does not fit ", range), &j));
+    }
+
     void write_msgpack_array_prefix(const std::size_t N, const BasicJsonType& j)
     {
         const auto n = to_msgpack_length(N, j);
@@ -1580,6 +1651,8 @@ class binary_writer
             is neither an object nor an array
     @throw out_of_range.415 if @a j is binary with a subtype that does not fit
            into a byte, before anything is written
+    @throw out_of_range.407 if @a j is an integer that does not fit the 64 bits
+           of a BSON integer, before anything is written
     @throw type_error.316 if @a j is a string that is not valid UTF-8, before
            anything is written
     @throw type_error.321 if @a j is discarded
@@ -1598,9 +1671,17 @@ class binary_writer
                 return 8ul;
 
             case value_t::number_integer:
-                return calc_bson_integer_size(j.m_data.m_value.number_integer);
+                if (JSON_HEDLEY_UNLIKELY(!integer_fits<std::int64_t>(j.m_data.m_value.number_integer)))
+                {
+                    throw_integer_out_of_range(j, "BSON", "int64");
+                }
+                return calc_bson_integer_size(static_cast<std::int64_t>(j.m_data.m_value.number_integer));
 
             case value_t::number_unsigned:
+                if (JSON_HEDLEY_UNLIKELY(!integer_fits_max<std::uint64_t>(j.m_data.m_value.number_unsigned)))
+                {
+                    throw_integer_out_of_range(j, "BSON", "uint64");
+                }
                 return calc_bson_unsigned_size(conditional_static_cast<std::uint64_t>(j.m_data.m_value.number_unsigned));
 
             case value_t::string:
@@ -1639,8 +1720,9 @@ class binary_writer
             case value_t::number_float:
                 return write_bson_double(name, j.m_data.m_value.number_float);
 
+            // calc_bson_value_size() checked that integers fit 64 bits
             case value_t::number_integer:
-                return write_bson_integer(name, j.m_data.m_value.number_integer);
+                return write_bson_integer(name, static_cast<std::int64_t>(j.m_data.m_value.number_integer));
 
             case value_t::number_unsigned:
                 return write_bson_unsigned(name, conditional_static_cast<std::uint64_t>(j.m_data.m_value.number_unsigned));
@@ -1951,6 +2033,376 @@ class binary_writer
         }
     }
 
+    /*!
+    @brief size class of an unsigned integer from its bit length
+
+    1, 2, 3, or 4 if @a x fits 1, 2, 4, or 8 bytes. Looked up from the number
+    of leading zero bits instead of found by branches: the sizes of a stream
+    of numbers are hard to predict, and a mispredicted branch costs more than
+    writing the number. (A table, as comparisons are vectorized by clang into
+    slower code.)
+    */
+    static unsigned integer_size_class(const std::uint64_t x) noexcept
+    {
+        static constexpr std::array<std::uint8_t, 64> size_class_by_leading_zeros =
+        {
+            {
+                4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
+                3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
+                2, 2, 2, 2, 2, 2, 2, 2,
+                1, 1, 1, 1, 1, 1, 1, 1
+            }
+        };
+        // x | 1 keeps count_leading_zeros() away from 0
+        return size_class_by_leading_zeros[static_cast<std::size_t>(count_leading_zeros(x | 1u))];
+    }
+
+    /*!
+    @brief encode @a first_byte and the low-order bytes of @a value that
+           @a size_class calls for (none, or 1, 2, 4, or 8) in network order
+           to @a out
+    @return the number of characters of the encoding; up to
+            @ref integer_encoding_space characters at @a out are overwritten
+
+    The bytes are written as one byte and one 8-byte word whose bytes after
+    the value are scratch that the caller drops, so that writing them does not
+    depend on the size class. For class 0 the word is all scratch; it is
+    shifted by 60 bits rather than 64, which would be undefined.
+    */
+    std::size_t encode_first_byte_and_tail(CharType* out, const std::uint8_t first_byte, const unsigned size_class, const std::uint64_t value) const noexcept
+    {
+        out[0] = to_char_type(first_byte);
+        const unsigned unused_bits = 64u - ((8u << size_class) >> 1u);
+        const std::uint64_t tail = is_little_endian ? byte_swap(value) >> unused_bits : value << unused_bits;
+        std::memcpy(out + 1, &tail, sizeof(tail));
+        return ((std::size_t{1} << size_class) >> 1u) + 1;
+    }
+
+    /// characters encode_first_byte_and_tail() may overwrite: the first byte
+    /// and one 8-byte word
+    static constexpr std::size_t integer_encoding_space = 9;
+
+    /*!
+    @brief the 64 bits of a number_integer or number_unsigned value, and a mask
+           that is all ones if it is negative, else zero
+
+    Selected through masks rather than branches, as the type and the sign of
+    the values in an array of mixed signs alternate unpredictably (non-negative
+    integers are number_unsigned after parsing). Up to 64 bits, the sign is bit
+    63 of the result (a narrower integer is sign-extended); computing it with
+    masks keeps GCC from branching on the type. A wider integer has to be
+    compared.
+    */
+    static std::uint64_t integer_bits(const BasicJsonType& j, std::uint64_t& sign_mask) noexcept
+    {
+        const bool is_signed = j.m_data.m_type == value_t::number_integer;
+        const std::uint64_t raw = is_signed
+                                  ? static_cast<std::uint64_t>(j.m_data.m_value.number_integer)
+                                  : static_cast<std::uint64_t>(j.m_data.m_value.number_unsigned);
+        sign_mask = sizeof(typename BasicJsonType::number_integer_t) <= sizeof(std::uint64_t)
+                    ? (std::uint64_t{0} - static_cast<std::uint64_t>(is_signed)) & (std::uint64_t{0} - (raw >> 63u))
+                    : std::uint64_t{0} - static_cast<std::uint64_t>(is_signed && j.m_data.m_value.number_integer < 0);
+        return raw;
+    }
+
+    /*!
+    @brief encode the head of a CBOR integer to @a out without branching on its
+           size
+    @return the number of characters of the head
+
+    Same output as @ref write_cbor_head. The lengths of strings and containers
+    are mostly predictable, so their heads keep the branches; the sizes of a
+    run of integers are not, so an integer's head is assembled without them.
+    */
+    std::size_t encode_cbor_integer_head(CharType* out, const std::uint8_t major_type, const std::uint64_t argument) const noexcept
+    {
+        // an argument of at most 23 is in size class 1 (one byte) and is moved
+        // to class 0, the first byte, by subtracting 1
+        const auto is_immediate = static_cast<unsigned>(argument <= 0x17u);
+        const unsigned size_class = integer_size_class(argument) - is_immediate;
+
+        // the first byte: the argument itself if it is immediate, else
+        // 0x17 + size_class, selected through a mask
+        const std::uint64_t immediate_mask = std::uint64_t{0} - is_immediate;
+        const auto first_byte = static_cast<std::uint8_t>(major_type + ((argument & immediate_mask) | ((0x17u + size_class) & ~immediate_mask)));
+        return encode_first_byte_and_tail(out, first_byte, size_class, argument);
+    }
+
+    /*!
+    @brief write an array value (head and elements) as CBOR
+
+    Integer elements, the most common scalar, are written without a call to
+    @ref write_cbor. The loop is kept out of line: inlined into write_cbor, its
+    registers would be saved and restored on every call of write_cbor, which
+    is one call per value nested in an object.
+    */
+    JSON_HEDLEY_NEVER_INLINE
+    void write_cbor_array(const BasicJsonType& j, const std::size_t depth)
+    {
+        write_cbor_head(0x80, j.m_data.m_value.array->size());
+        write_cbor_array_elements(j, depth, std::integral_constant<bool, integers_fit_64_bits()>());
+    }
+
+    void write_cbor_array_elements(const BasicJsonType& j, const std::size_t depth, std::true_type /*fits_64_bits*/)
+    {
+        write_array_elements(j, [this](CharType * out, const BasicJsonType & el)
+        {
+            return encode_cbor_integer(out, el);
+        }, [this, depth](const BasicJsonType & el)
+        {
+            write_cbor(el, depth + 1);
+        });
+    }
+
+    void write_cbor_array_elements(const BasicJsonType& j, const std::size_t depth, std::false_type /*fits_64_bits*/)
+    {
+        for (const auto& el : *j.m_data.m_value.array)
+        {
+            write_cbor(el, depth + 1);
+        }
+    }
+
+    /*!
+    @brief write the elements of the array @a j, runs of integers with
+           @a encode_integer and all other values with @a write_value
+
+    Runs of integers are encoded straight into space reserved from the sink
+    for a whole run: the write pointer stays a local, where it would otherwise
+    be reloaded from the sink for every value, as a store of a character may
+    alias it.
+    */
+    template<typename EncodeInteger, typename WriteValue>
+    void write_array_elements(const BasicJsonType& j, EncodeInteger encode_integer, WriteValue write_value)
+    {
+        constexpr std::size_t max_run = max_reserve_size / integer_encoding_space;
+        auto it = j.m_data.m_value.array->begin();
+        const auto end = j.m_data.m_value.array->end();
+        while (it != end)
+        {
+            if (it->is_number_integer())
+            {
+                CharType* const start = oa.reserve(max_run * integer_encoding_space);
+                CharType* out = start;
+                for (std::size_t n = 0; n < max_run && it != end && it->is_number_integer(); ++n, ++it)
+                {
+                    out += encode_integer(out, *it);
+                }
+                oa.commit(static_cast<std::size_t>(out - start));
+            }
+            else
+            {
+                write_value(*it);
+                ++it;
+            }
+        }
+    }
+
+    /*!
+    @brief write a number_integer or number_unsigned value as a CBOR integer
+
+    A negative integer n has major type 1 and the argument -1 - n, which is ~n.
+    Kept out of line, like the other per-value paths, so that write_cbor stays
+    small (runs of integers in an array are encoded inline instead).
+    */
+    JSON_HEDLEY_NEVER_INLINE
+    void write_cbor_integer(const BasicJsonType& j, std::true_type /*fits_64_bits*/)
+    {
+        oa.commit(encode_cbor_integer(oa.reserve(integer_encoding_space), j));
+    }
+
+    /// write a number_integer or number_unsigned value of a number type wider
+    /// than 64 bits as a CBOR integer, with range checks on the whole value
+    void write_cbor_integer(const BasicJsonType& j, std::false_type /*fits_64_bits*/)
+    {
+        if (j.m_data.m_type == value_t::number_unsigned)
+        {
+            if (JSON_HEDLEY_UNLIKELY(!integer_fits_max<std::uint64_t>(j.m_data.m_value.number_unsigned)))
+            {
+                throw_integer_out_of_range(j, "CBOR", "uint64");
+            }
+            write_cbor_head(0x00, conditional_static_cast<std::uint64_t>(j.m_data.m_value.number_unsigned));
+            return;
+        }
+
+        if (j.m_data.m_value.number_integer >= 0)
+        {
+            if (JSON_HEDLEY_UNLIKELY(!integer_fits_max<std::uint64_t>(j.m_data.m_value.number_integer)))
+            {
+                throw_integer_out_of_range(j, "CBOR", "[-2^64, 2^64-1]");
+            }
+            // CBOR does not differentiate between positive signed
+            // integers and unsigned integers
+            write_cbor_head(0x00, static_cast<std::uint64_t>(j.m_data.m_value.number_integer));
+        }
+        else
+        {
+            // a negative integer n is encoded as -1 - n
+            if (JSON_HEDLEY_UNLIKELY(!integer_fits_max<std::uint64_t>(-1 - j.m_data.m_value.number_integer)))
+            {
+                throw_integer_out_of_range(j, "CBOR", "[-2^64, 2^64-1]");
+            }
+            write_cbor_head(0x20, static_cast<std::uint64_t>(-1 - j.m_data.m_value.number_integer));
+        }
+    }
+
+    /// encode a number_integer or number_unsigned value to @a out (see
+    /// @ref write_cbor_integer) and return the number of characters
+    std::size_t encode_cbor_integer(CharType* out, const BasicJsonType& j) const noexcept
+    {
+        std::uint64_t sign_mask = 0;
+        const std::uint64_t raw = integer_bits(j, sign_mask);
+        // ~raw is -1 - n in the 64 bits that are written
+        return encode_cbor_integer_head(out, static_cast<std::uint8_t>(sign_mask & 0x20u), raw ^ sign_mask);
+    }
+
+    /*!
+    @brief whether the CBOR and MessagePack integers of BasicJsonType are
+           written by @ref encode_cbor_integer and @ref encode_msgpack_integer
+
+    These pick the format from the low 64 bits; a wider number type is written
+    with range checks on the whole value instead.
+    */
+    static constexpr bool integers_fit_64_bits() noexcept
+    {
+        return sizeof(typename BasicJsonType::number_integer_t) <= sizeof(std::uint64_t)
+               && sizeof(typename BasicJsonType::number_unsigned_t) <= sizeof(std::uint64_t);
+    }
+
+    /*!
+    @brief write a number_integer or number_unsigned value as a MessagePack
+           integer without branching on its sign or size
+
+    Same output as @ref write_msgpack: a non-negative integer as positive fixint
+    (< 128) or uint 8/16/32/64, a negative one as negative fixint (>= -32) or
+    int 8/16/32/64. Kept out of line (see @ref write_cbor_integer).
+    */
+    JSON_HEDLEY_NEVER_INLINE
+    void write_msgpack_integer(const BasicJsonType& j, std::true_type /*fits_64_bits*/)
+    {
+        oa.commit(encode_msgpack_integer(oa.reserve(integer_encoding_space), j));
+    }
+
+    /// encode a number_integer or number_unsigned value to @a out (see
+    /// @ref write_msgpack_integer) and return the number of characters
+    std::size_t encode_msgpack_integer(CharType* out, const BasicJsonType& j) const noexcept
+    {
+        std::uint64_t sign_mask = 0;
+        const std::uint64_t raw = integer_bits(j, sign_mask);
+
+        // A negative n fits int 8/16/32 exactly if 2 * (-1 - n) + 1 fits
+        // uint 8/16/32, so x gives the size class of either sign.
+        const std::uint64_t is_negative = sign_mask & 1u;
+        const std::uint64_t x = ((raw ^ sign_mask) << is_negative) | is_negative;
+
+        // A fixint is at most 127, or at least -32 (x at most 63). It is in
+        // size class 1 and is moved to class 0, the first byte, by subtracting 1.
+        const auto is_fixint = static_cast<unsigned>(x <= (127u ^ (sign_mask & 64u)));
+        const unsigned size_class = integer_size_class(x) - is_fixint;
+
+        // the first byte: the low byte of the integer for a fixint, else the
+        // marker 0xCC..0xCF (uint) or 0xD0..0xD3 (int), selected through a mask
+        const std::uint64_t fixint_mask = std::uint64_t{0} - is_fixint;
+        const std::uint64_t marker = 0xCBu + size_class + (sign_mask & 4u);
+        const auto first_byte = static_cast<std::uint8_t>((raw & fixint_mask) | (marker & ~fixint_mask));
+        return encode_first_byte_and_tail(out, first_byte, size_class, raw);
+    }
+
+    /// write a number_integer or number_unsigned value of a number type wider
+    /// than 64 bits as a MessagePack integer, with range checks on the whole value
+    void write_msgpack_integer(const BasicJsonType& j, std::false_type /*fits_64_bits*/)
+    {
+        if (j.m_data.m_type == value_t::number_unsigned)
+        {
+            if (JSON_HEDLEY_UNLIKELY(!integer_fits_max<std::uint64_t>(j.m_data.m_value.number_unsigned)))
+            {
+                throw_integer_out_of_range(j, "MessagePack", "uint64");
+            }
+            write_msgpack_unsigned(static_cast<std::uint64_t>(j.m_data.m_value.number_unsigned));
+            return;
+        }
+
+        if (j.m_data.m_value.number_integer >= 0)
+        {
+            if (JSON_HEDLEY_UNLIKELY(!integer_fits_max<std::uint64_t>(j.m_data.m_value.number_integer)))
+            {
+                throw_integer_out_of_range(j, "MessagePack", "[-2^63, 2^64-1]");
+            }
+            // MessagePack does not differentiate between positive
+            // signed integers and unsigned integers.
+            write_msgpack_unsigned(static_cast<std::uint64_t>(j.m_data.m_value.number_integer));
+        }
+        else
+        {
+            if (JSON_HEDLEY_UNLIKELY(!integer_fits_min<std::int64_t>(j.m_data.m_value.number_integer)))
+            {
+                throw_integer_out_of_range(j, "MessagePack", "[-2^63, 2^64-1]");
+            }
+            if (j.m_data.m_value.number_integer >= -32)
+            {
+                // negative fixnum
+                write_number(static_cast<std::int8_t>(j.m_data.m_value.number_integer));
+            }
+            else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int8_t>::min)() &&
+                     j.m_data.m_value.number_integer <= (std::numeric_limits<std::int8_t>::max)())
+            {
+                // int 8
+                oa.write_character(to_char_type(0xD0));
+                write_number(static_cast<std::int8_t>(j.m_data.m_value.number_integer));
+            }
+            else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int16_t>::min)() &&
+                     j.m_data.m_value.number_integer <= (std::numeric_limits<std::int16_t>::max)())
+            {
+                // int 16
+                oa.write_character(to_char_type(0xD1));
+                write_number(static_cast<std::int16_t>(j.m_data.m_value.number_integer));
+            }
+            else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int32_t>::min)() &&
+                     j.m_data.m_value.number_integer <= (std::numeric_limits<std::int32_t>::max)())
+            {
+                // int 32
+                oa.write_character(to_char_type(0xD2));
+                write_number(static_cast<std::int32_t>(j.m_data.m_value.number_integer));
+            }
+            else
+            {
+                // int 64
+                oa.write_character(to_char_type(0xD3));
+                write_number(static_cast<std::int64_t>(j.m_data.m_value.number_integer));
+            }
+        }
+    }
+
+    /*!
+    @brief write an array value (head and elements) as MessagePack
+
+    See @ref write_cbor_array.
+    */
+    JSON_HEDLEY_NEVER_INLINE
+    void write_msgpack_array(const BasicJsonType& j, const std::size_t depth)
+    {
+        write_msgpack_array_prefix(j.m_data.m_value.array->size(), j);
+        write_msgpack_array_elements(j, depth, std::integral_constant<bool, integers_fit_64_bits()>());
+    }
+
+    void write_msgpack_array_elements(const BasicJsonType& j, const std::size_t depth, std::true_type /*fits_64_bits*/)
+    {
+        write_array_elements(j, [this](CharType * out, const BasicJsonType & el)
+        {
+            return encode_msgpack_integer(out, el);
+        }, [this, depth](const BasicJsonType & el)
+        {
+            write_msgpack(el, depth + 1);
+        });
+    }
+
+    void write_msgpack_array_elements(const BasicJsonType& j, const std::size_t depth, std::false_type /*fits_64_bits*/)
+    {
+        for (const auto& el : *j.m_data.m_value.array)
+        {
+            write_msgpack(el, depth + 1);
+        }
+    }
+
     /*!
     @brief write an object key as a CBOR text string
 
@@ -2147,7 +2599,7 @@ class binary_writer
         {
             return 'L';
         }
-        if (use_bjdata && std::is_unsigned<NumberType>::value)
+        if (use_bjdata && std::is_unsigned<NumberType>::value && integer_fits_max<std::uint64_t>(n))
         {
             return 'M';
         }
@@ -2272,14 +2724,16 @@ class binary_writer
     @brief checks whether a JSON number fits into @a TargetType
     @param[in] el a JSON number of either the signed or unsigned integer kind
     @return whether @a el's value can be represented by @a TargetType without
-            wrapping, regardless of which of the two kinds it is stored as
+            wrapping, regardless of which of the two kinds it is stored as;
+            false for a value that does not even fit the 64-bit type it is
+            read as (possible for number types wider than 64 bits)
     */
     template<typename TargetType>
     static bool bjdata_ndarray_value_in_range(const BasicJsonType& el)
     {
         return el.is_number_unsigned()
-               ? value_in_range_of<TargetType>(el.template get<std::uint64_t>())
-               : value_in_range_of<TargetType>(el.template get<std::int64_t>());
+               ? integer_fits_max<std::uint64_t>(el.m_data.m_value.number_unsigned) && value_in_range_of<TargetType>(el.template get<std::uint64_t>())
+               : integer_fits<std::int64_t>(el.m_data.m_value.number_integer) && value_in_range_of<TargetType>(el.template get<std::int64_t>());
     }
 
     /*!
@@ -2511,10 +2965,11 @@ class binary_writer
         for (const auto& el : dims)
         {
             // a dimension is read as an unsigned value below, so anything that
-            // is not a non-negative integer is rejected: a non-integer entry
-            // would pun unrelated bytes as the dimension, and a negative one
-            // would wrap into a nonsensical length
-            if (!el.is_number_integer() || (!el.is_number_unsigned() && el.template get<std::int64_t>() < 0))
+            // is not a non-negative integer in the range of std::uint64_t is
+            // rejected: a non-integer entry would pun unrelated bytes as the
+            // dimension, and a negative or wider one would wrap into a
+            // nonsensical length
+            if (!el.is_number_integer() || !bjdata_ndarray_value_in_range<std::uint64_t>(el))
             {
                 return true;
             }
@@ -2646,9 +3101,9 @@ class binary_writer
 
             case value_t::number_unsigned:
             {
-                if (j.m_data.m_value.number_unsigned > static_cast<typename BasicJsonType::number_unsigned_t>((std::numeric_limits<std::int64_t>::max)()))
+                if (JSON_HEDLEY_UNLIKELY(!integer_fits_max<std::int64_t>(j.m_data.m_value.number_unsigned)))
                 {
-                    JSON_THROW(out_of_range::create(exception_id::integer_too_large, concat("integer number ", std::to_string(j.m_data.m_value.number_unsigned), " cannot be represented by BON8 as it does not fit int64"), &j));
+                    throw_integer_out_of_range(j, "BON8", "int64");
                 }
                 write_bon8_integer(static_cast<std::int64_t>(j.m_data.m_value.number_unsigned));
                 string_open = false;
@@ -2657,6 +3112,10 @@ class binary_writer
 
             case value_t::number_integer:
             {
+                if (JSON_HEDLEY_UNLIKELY(!integer_fits<std::int64_t>(j.m_data.m_value.number_integer)))
+                {
+                    throw_integer_out_of_range(j, "BON8", "int64");
+                }
                 write_bon8_integer(static_cast<std::int64_t>(j.m_data.m_value.number_integer));
                 string_open = false;
                 break;

include/nlohmann/detail/output/output_adapters.hpp

@@ -8,7 +8,10 @@
 
 #pragma once
 
+#include <algorithm> // max
+#include <array> // array
 #include <cstddef> // size_t
+#include <cstring> // memcpy
 #include <memory> // shared_ptr, make_shared
 #include <string> // basic_string
 #include <type_traits> // conditional, integral_constant, is_same
@@ -48,32 +51,170 @@ template<typename CharType> struct output_adapter_protocol
 template<typename CharType>
 using output_adapter_t = std::shared_ptr<output_adapter_protocol<CharType>>;
 
+/// the largest number of characters a writer may reserve() from a sink at once;
+/// output_vector_sink also appends writes of at least this size with insert()
+constexpr std::size_t max_reserve_size = 256;
+
 /// @brief non-virtual output sink writing into a std::vector
 ///
 /// This sink is not part of the virtual output_adapter_protocol hierarchy: it is
 /// passed to binary_writer by value as a template parameter, so
 /// write_character()/write_characters() are ordinary (inlinable) calls with no
 /// vtable lookup and no shared_ptr. It is used for the common
-/// `to_cbor`/`to_msgpack`/... into a std::vector. output_vector_adapter below
-/// wraps this same sink to provide the virtual interface.
+/// `to_cbor`/`to_msgpack`/... into a std::vector.
+///
+/// Most writes are a few bytes - a type marker, or a number of 1, 2, 4, or 8
+/// bytes - and appending those with push_back() or insert() cost more than
+/// encoding them: insert() is an out-of-line call into a routine that can insert
+/// anywhere, and resize() is one as well with libstdc++. The sink therefore keeps
+/// its own write position: it grows the vector geometrically ahead of the
+/// position and copies each write to it, so a write is a capacity check and a
+/// short copy. The destructor shrinks the vector to the bytes actually written,
+/// also when a write throws; shrinking a vector never allocates. The sink owns
+/// that trim, so it can be moved but not copied, and the vector must not be used
+/// while the sink is alive.
 template<typename CharType, typename AllocatorType = std::allocator<CharType>>
 class output_vector_sink
 {
   public:
     explicit output_vector_sink(std::vector<CharType, AllocatorType>& vec) noexcept
+        : v(&vec), position(vec.size())
+    {}
+
+    output_vector_sink(const output_vector_sink&) = delete;
+    output_vector_sink& operator=(const output_vector_sink&) = delete;
+    output_vector_sink& operator=(output_vector_sink&&) = delete;
+
+    output_vector_sink(output_vector_sink&& other) noexcept
+        : v(other.v), position(other.position)
+    {
+        other.v = nullptr;
+    }
+
+    ~output_vector_sink()
+    {
+        if (v != nullptr)
+        {
+            v->resize(position);
+        }
+    }
+
+    // always inlined: GCC otherwise keeps these out of line in a large
+    // translation unit, and a call per byte or number costs more than the write
+    JSON_HEDLEY_ALWAYS_INLINE
+    void write_character(CharType c)
+    {
+        if (JSON_HEDLEY_UNLIKELY(position == v->size()))
+        {
+            grow(1);
+        }
+        (*v)[position++] = c;
+    }
+
+    // no JSON_HEDLEY_NON_NULL here: binary_writer legitimately passes a null
+    // pointer with length 0 for empty strings/binary values; memcpy must not
+    // see it.
+    JSON_HEDLEY_ALWAYS_INLINE
+    void write_characters(const CharType* s, std::size_t length)
+    {
+        if (JSON_HEDLEY_UNLIKELY(v->size() - position < length))
+        {
+            if (length >= max_reserve_size)
+            {
+                append(s, length);
+                return;
+            }
+            grow(length);
+        }
+        if (length != 0)
+        {
+            std::memcpy(v->data() + position, s, length * sizeof(CharType));
+            position += length;
+        }
+    }
+
+    /// @brief space for writing up to @a length characters directly
+    ///
+    /// Returns a pointer to at least @a length (at most @ref max_reserve_size)
+    /// writable characters at the write position; @ref commit then keeps the
+    /// first ones of them. A writer can store whole words there and keep only
+    /// the bytes it needs, and keep its write pointer in a register across many
+    /// values instead of updating the sink after each.
+    JSON_HEDLEY_ALWAYS_INLINE
+    CharType* reserve(const std::size_t length)
+    {
+        JSON_ASSERT(length <= max_reserve_size);
+        if (JSON_HEDLEY_UNLIKELY(v->size() - position < length))
+        {
+            grow(length);
+        }
+        return v->data() + position;
+    }
+
+    /// keep the first @a length characters written to the space from reserve()
+    JSON_HEDLEY_ALWAYS_INLINE
+    void commit(const std::size_t length) noexcept
+    {
+        position += length;
+    }
+
+  private:
+    /// make room for at least @a length more characters after the position;
+    /// kept out of line so that the write functions stay small enough for
+    /// GCC to inline them into the writer
+    JSON_HEDLEY_NEVER_INLINE
+    void grow(const std::size_t length)
+    {
+        // use the capacity the vector already has, else at least double it;
+        // never less than 64 characters, so that a small value does not
+        // reallocate for every few bytes
+        const std::size_t needed = position + length;
+        v->resize((std::max)(std::size_t{64}, needed <= v->capacity()
+                             ? v->capacity()
+                             : (std::max)(needed, v->capacity() * 2)));
+    }
+
+    /// append @a length (at least max_reserve_size) characters that do not fit
+    /// the space ahead of the position: insert() copies them once into the
+    /// grown vector, where grow() would first zero the new space and then copy
+    /// over it - noticeable for a large string or binary value
+    JSON_HEDLEY_NEVER_INLINE
+    void append(const CharType* s, const std::size_t length)
+    {
+        v->resize(position);
+        if (v->capacity() - position < length)
+        {
+            // grow geometrically, with room for the writes that follow, so
+            // that the next grow() need not reallocate right away
+            v->reserve((std::max)(v->capacity() * 2, position + length + max_reserve_size));
+        }
+        v->insert(v->end(), s, s + length);
+        position = v->size();
+    }
+
+    std::vector<CharType, AllocatorType>* v;
+    /// number of characters of *v written so far
+    std::size_t position;
+};
+
+/// output adapter for byte vectors
+template<typename CharType, typename AllocatorType = std::allocator<CharType>>
+class output_vector_adapter : public output_adapter_protocol<CharType>
+{
+  public:
+    explicit output_vector_adapter(std::vector<CharType, AllocatorType>& vec) noexcept
         : v(vec)
     {}
 
-    void write_character(CharType c)
+    void write_character(CharType c) override
     {
         v.push_back(c);
     }
 
     // no JSON_HEDLEY_NON_NULL here: binary_writer legitimately passes a null
-    // pointer with length 0 for empty strings/binary values. Appending an empty
-    // range is a no-op; the type-erased path tolerates this via the (unattributed)
-    // virtual base, and the concrete sink must do the same.
-    void write_characters(const CharType* s, std::size_t length)
+    // pointer with length 0 for empty strings/binary values; appending an empty
+    // range is a no-op
+    void write_characters(const CharType* s, std::size_t length) override
     {
         v.insert(v.end(), s, s + length);
     }
@@ -82,33 +223,6 @@ class output_vector_sink
     std::vector<CharType, AllocatorType>& v;
 };
 
-/// output adapter for byte vectors
-///
-/// The appending itself lives in output_vector_sink; this class only adds the
-/// virtual output_adapter_protocol interface on top of it, so both the
-/// type-erased and the templated path share one implementation.
-template<typename CharType, typename AllocatorType = std::allocator<CharType>>
-class output_vector_adapter : public output_adapter_protocol<CharType>
-{
-  public:
-    explicit output_vector_adapter(std::vector<CharType, AllocatorType>& vec) noexcept
-        : sink(vec)
-    {}
-
-    void write_character(CharType c) override
-    {
-        sink.write_character(c);
-    }
-
-    void write_characters(const CharType* s, std::size_t length) override
-    {
-        sink.write_characters(s, length);
-    }
-
-  private:
-    output_vector_sink<CharType, AllocatorType> sink;
-};
-
 #ifndef JSON_NO_IO
 /// output adapter for output streams
 template<typename CharType>
@@ -188,8 +302,24 @@ class output_adapter_sink
         oa->write_characters(s, length);
     }
 
+    /// space for writing up to @a length characters (see
+    /// output_vector_sink::reserve)
+    CharType* reserve(const std::size_t length) noexcept
+    {
+        JSON_ASSERT(length <= max_reserve_size);
+        static_cast<void>(length);
+        return reserved.data();
+    }
+
+    /// write the first @a length characters written to the space from reserve()
+    void commit(const std::size_t length)
+    {
+        oa->write_characters(reserved.data(), length);
+    }
+
   private:
     output_adapter_t<CharType> oa;
+    std::array<CharType, max_reserve_size> reserved{};
 };
 
 /// @brief whether std::basic_string<CharType> has a non-deprecated std::char_traits

include/nlohmann/detail/output/serializer.hpp

@@ -1525,7 +1525,7 @@ class serializer
         // without leading zeros
         if (abs_value < 10000000000000000u)
         {
-            const std::uint64_t value = abs_value;
+            const auto value = static_cast<std::uint64_t>(abs_value); // fits: abs_value < 10^16
             const std::uint64_t upper = value / 100000000u;
             const std::uint64_t first = dtoa_impl::eight_digit_bytes(upper != 0 ? upper : value);
             const auto leading = static_cast<unsigned>(count_leading_zeros(first) / 8); // (first is not 0)

include/nlohmann/detail/value_t.hpp

@@ -14,7 +14,6 @@
 #include <cstdint> // uint8_t
 #include <limits> // numeric_limits
 #include <string> // string
-#include <type_traits> // is_signed
 
 #include <nlohmann/detail/macro_scope.hpp>
 #if JSON_HAS_THREE_WAY_COMPARISON
@@ -208,13 +207,16 @@ FloatType compare_integer_with_float(const IntegerType i, const FloatType f) noe
 
     // values of IntegerType lie in [-bound, bound) when signed and in
     // [0, bound) when unsigned; digits excludes the sign bit, so bound is a
-    // power of two that the float represents exactly
-    const FloatType bound = std::ldexp(static_cast<FloatType>(1), std::numeric_limits<IntegerType>::digits);
+    // power of two that the float represents exactly; the signedness comes
+    // from numeric_limits as well, because std::is_signed is false for class
+    // types such as 128-bit or multiprecision integers
+    using limits = std::numeric_limits<IntegerType>;
+    const FloatType bound = std::ldexp(static_cast<FloatType>(1), limits::digits);
     if (f >= bound)
     {
         return ordered(-1);
     }
-    if (std::is_signed<IntegerType>::value ? (f < -bound) : (f < static_cast<FloatType>(0)))
+    if (limits::is_signed ? (f < -bound) : (f < static_cast<FloatType>(0)))
     {
         return ordered(1);
     }

include/nlohmann/detail/view/builder.hpp

@@ -422,8 +422,10 @@ class builder
             arena_ensure(16);
             std::memcpy(m_arena_write, r, 16);
         }
-        else
+        else if (n != 0)
         {
+            // (an empty run leaves an empty arena unallocated, and memcpy()
+            // must not be passed its null pointer even for zero bytes)
             arena_ensure(n);
             std::memcpy(m_arena_write, r, n);
         }
@@ -682,7 +684,7 @@ obj_next:
             return fail(error_code::expected_object_end);
 
 close_object:
-            // a large object gets a hash index (objects only, so that closing
+            // a large object gets a hash index for its keys (objects only, so that closing
             // an array pays nothing for this)
             if (NLOHMANN_VIEW_UNLIKELY(cur_count >= document_data::index_min_members))
             {
@@ -844,7 +846,7 @@ indent_done:
 
         /// append a node: (kind, flags, extra, off) and the second word (len, or
         /// an integer's value); two stores on little-endian targets
-        NLOHMANN_VIEW_ALWAYS_INLINE node* emit(value_t k, std::uint8_t flags, std::uint16_t extra, std::size_t off, std::uint64_t second)
+        NLOHMANN_VIEW_ALWAYS_INLINE node* push_node(value_t k, std::uint8_t flags, std::uint16_t extra, std::size_t off, std::uint64_t second)
         {
             if (NLOHMANN_VIEW_UNLIKELY(out == cap))
             {
@@ -874,9 +876,9 @@ indent_done:
 
         NLOHMANN_VIEW_ALWAYS_INLINE void open(value_t k)
         {
-            // (base is read after emit(), which moves the node array when it
+            // (base is read after push_node(), which moves the node array when it
             // grows; in one expression the order of the two is unspecified)
-            const node* const n = emit(k, 0, 0, static_cast<std::size_t>(p - b), 0);
+            const node* const n = push_node(k, 0, 0, static_cast<std::size_t>(p - b), 0);
             const auto idx = static_cast<std::uint32_t>(n - base);
             if (depth != 0)
             {
@@ -936,7 +938,7 @@ indent_done:
             {
                 return fail(error_code::invalid_literal);
             }
-            emit(k, flags, 0, static_cast<std::size_t>(p - b), n);
+            push_node(k, flags, 0, static_cast<std::size_t>(p - b), n);
             p += n;
             return true;
         }
@@ -1044,7 +1046,7 @@ indent_done:
                 // of the negative std::int64_t)
                 second = negative ? 0 - m : m;
             }
-            emit(kind, 0, layout, static_cast<std::size_t>(s - b), second);
+            push_node(kind, 0, layout, static_cast<std::size_t>(s - b), second);
             return true;
         }
 
@@ -1057,7 +1059,7 @@ indent_done:
             p = scan_string_run<Value>(p, e);
             if (NLOHMANN_VIEW_LIKELY(p != e && *p == '"'))
             {
-                emit(value_t::string, 0, 0, static_cast<std::size_t>(s - b), static_cast<std::uint64_t>(p - s));
+                push_node(value_t::string, 0, 0, static_cast<std::size_t>(s - b), static_cast<std::uint64_t>(p - s));
                 ++p;
                 return true;
             }
@@ -1067,7 +1069,7 @@ indent_done:
                 return false;
             }
             p = r.p;
-            emit(value_t::string, node_flags::escaped, 0, r.start, r.len);
+            push_node(value_t::string, node_flags::escaped, 0, r.start, r.len);
             return true;
         }
     };

include/nlohmann/detail/view/compare.hpp

@@ -201,6 +201,11 @@ class json_side
     const BasicJsonType* m_json;
 };
 
+inline bool is_number(value_t t) noexcept
+{
+    return t == value_t::number_integer || t == value_t::number_unsigned || t == value_t::number_float;
+}
+
 /// whether two sides are equal; iterative, so that the nesting depth is
 /// limited by memory only
 template<typename BasicJsonType, typename A, typename B>
@@ -225,8 +230,6 @@ bool equal(const A& a0, const B& b0)
     {
         const value_t ta = a.type();
         const value_t tb = b.type();
-        const bool numbers = (ta == value_t::number_integer || ta == value_t::number_unsigned || ta == value_t::number_float)
-                             && (tb == value_t::number_integer || tb == value_t::number_unsigned || tb == value_t::number_float);
         if (ta == value_t::discarded || tb == value_t::discarded)
         {
             // basic_json decides (JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON)
@@ -237,13 +240,13 @@ bool equal(const A& a0, const B& b0)
         }
         else
         {
-            if (!numbers && ta != tb)
+            if (ta != tb && !(is_number(ta) && is_number(tb))) // numbers of different types can be equal
             {
                 return false;
             }
             if (ta == value_t::string)
             {
-                if (!(a.string() == b.string()))
+                if (a.string() != b.string())
                 {
                     return false;
                 }
@@ -272,7 +275,7 @@ bool equal(const A& a0, const B& b0)
                 }
                 stack.push_back(std::move(f));
             }
-            else if (!(a.scalar() == b.scalar())) // numbers (also of different types), null, boolean
+            else if (a.scalar() != b.scalar()) // numbers (also of different types), null, boolean
             {
                 return false;
             }
@@ -289,7 +292,7 @@ bool equal(const A& a0, const B& b0)
             const std::size_t count = f.object ? f.members_a.size() : f.elements_a.size();
             if (f.next == count)
             {
-                stack.pop_back();
+                stack.pop_back(); // (f dangles now, and is bound again after continue)
                 continue;
             }
             if (f.object)

include/nlohmann/detail/view/document_data.hpp

@@ -11,7 +11,7 @@
 #include <array> // array
 #include <cstddef> // size_t
 #include <cstdint> // uint8_t, uint32_t
-#include <cstring> // memcpy
+#include <cstring> // memcmp, memcpy
 #include <limits> // numeric_limits
 #include <map> // map
 #include <memory> // unique_ptr
@@ -46,7 +46,7 @@ struct document_data
     std::string owned; ///< owned copy of the input, if any (a std::string, so that its NUL serves as the parser's sentinel)
     std::vector<std::uint8_t> owned_image; ///< a loaded image the document owns (the text and the decoded strings point into it)
 
-    // hash indexes of large objects (see object_index.hpp)
+    // hash indexes of the keys of large objects (see object_index.hpp)
     static constexpr std::uint32_t index_min_members = 128;
     struct object_index
     {
@@ -56,14 +56,16 @@ struct document_data
     std::vector<object_index> indexes;
     std::vector<std::uint32_t> index_slots;
     std::vector<std::uint32_t> large_objects; ///< positions of the objects to index (noted while parsing)
-    std::array<const char*, 4> base = {{nullptr, nullptr, nullptr, nullptr}}; ///< string bases: source, arena, edit arena (indexed by flags & node_flags::storage)
+    std::array<const char*, 4> base = {{nullptr, nullptr, nullptr, nullptr}}; ///< string bases: source, arena, edit arena, unused (indexed by flags & node_flags::storage)
     bool discarded = true;
 
     /// The storage of edits (editable documents only; see edit_storage.hpp).
-    /// Edits never move or resize the parsed index, so views stay valid: an
-    /// array/object whose elements change gets node_flags::moved, and its
-    /// elements then live in a separate sequence (a header node, then the
-    /// entries), whose entries link to the values.
+    /// No edit moves or resizes the parsed index, and the nodes of values
+    /// written by edits never move either, so a view stays valid whenever it
+    /// was taken, before the first edit or between later ones: an array/object
+    /// whose elements change gets node_flags::moved, and its elements then live
+    /// in a separate sequence (a header node, then the entries), whose entries
+    /// link to the values.
     struct edit_state
     {
         std::vector<node*> moved{};                    ///< element sequences of moved arrays/objects (header node first) // NOLINT(readability-redundant-member-init)
@@ -183,6 +185,12 @@ struct document_data
         return base[n.flags & node_flags::storage] + n.off;
     }
 
+    /// whether the string of node k (a key) is the n bytes at key
+    bool key_equals(const node& k, const char* key, std::size_t n) const noexcept
+    {
+        return k.len == n && (n == 0 || std::memcmp(str(k), key, n) == 0);
+    }
+
     /// the node after n's subtree (containers span `next` nodes, scalars one)
     static NLOHMANN_VIEW_ALWAYS_INLINE const node* after(const node* n) noexcept
     {

include/nlohmann/detail/view/edit.hpp

@@ -12,7 +12,7 @@
 #include <cmath> // isinf, isnan
 #include <cstddef> // size_t
 #include <cstdint> // int64_t, uint8_t, uint32_t, uint64_t
-#include <cstring> // memcmp, memmove
+#include <cstring> // memmove
 #include <limits> // numeric_limits
 #include <string> // string, to_string
 #include <type_traits> // decay, enable_if, integral_constant, is_arithmetic, is_convertible, is_floating_point, is_same, is_signed
@@ -313,7 +313,7 @@ class editor
 
     bool key_equals(const node& k, string_view_t key) const noexcept
     {
-        return k.len == key.size() && (key.size() == 0 || std::memcmp(m_doc.str(k), key.data(), key.size()) == 0);
+        return m_doc.key_equals(k, key.data(), key.size());
     }
 
     /// remove the members with this key (all, or all but the first) from an object

include/nlohmann/detail/view/edit_storage.hpp

@@ -50,7 +50,19 @@ inline node* alloc_nodes(document_data& d, std::size_t k)
     document_data::edit_state& e = edit_state_of(d);
     if (NLOHMANN_VIEW_UNLIKELY(static_cast<std::size_t>(e.chunk_end - e.chunk_cur) < k))
     {
-        const std::size_t count = (std::max)(k, e.chunk_next);
+        if (k >= e.chunk_next / 2)
+        {
+            // a large request (half of the next chunk or more) gets a chunk of
+            // its own, so that the rest of the current chunk stays available
+            // for smaller ones
+            std::unique_ptr<node[]> own(new node[k]()); // NOLINT(cppcoreguidelines-avoid-c-arrays,hicpp-avoid-c-arrays,modernize-avoid-c-arrays)
+            e.chunks.push_back(std::move(own));
+            e.bytes += k * sizeof(node);
+            return e.chunks.back().get();
+        }
+        // a smaller request leaves the rest of the current chunk unused
+        // (fewer than k nodes, and k is less than half of the new chunk)
+        const std::size_t count = e.chunk_next;
         std::unique_ptr<node[]> fresh(new node[count]()); // NOLINT(cppcoreguidelines-avoid-c-arrays,hicpp-avoid-c-arrays,modernize-avoid-c-arrays)
         e.chunks.push_back(std::move(fresh));
         e.chunk_cur = e.chunks.back().get();
@@ -81,7 +93,8 @@ inline std::size_t text_capacity(std::size_t cap, std::size_t used, std::size_t
 }
 
 /// copy n bytes into the edit arena and return their offset; a new buffer
-/// leaves the old one alive, so that string views into it remain valid
+/// leaves the old one alive, so that string views into it remain valid; n is
+/// 0 for an empty string (written by an edit, or the key of a new member)
 inline std::uint32_t append_text(document_data& d, const char* s, std::size_t n)
 {
     document_data::edit_state& e = edit_state_of(d);
@@ -89,7 +102,7 @@ inline std::uint32_t append_text(document_data& d, const char* s, std::size_t n)
     {
         const std::size_t cap = text_capacity(e.text_cap, e.text_used, n);
         std::unique_ptr<char[]> fresh(new char[cap]); // NOLINT(cppcoreguidelines-avoid-c-arrays,hicpp-avoid-c-arrays,modernize-avoid-c-arrays)
-        if (e.text_used != 0)
+        if (NLOHMANN_VIEW_LIKELY(e.text_used != 0)) // (0 only for the first buffer, and then texts is empty)
         {
             std::memcpy(fresh.get(), e.texts.back().get(), e.text_used);
         }
@@ -127,8 +140,10 @@ inline void reserve_moved(document_data& d)
     if (e.moved.size() == e.moved.capacity() || e.moved_cap.size() == e.moved_cap.capacity())
     {
         // both grow before either changes, so that the push_backs cannot throw
-        e.moved.reserve((2 * e.moved.size()) + 16);
-        e.moved_cap.reserve((2 * e.moved.size()) + 16);
+        // (both conditions are needed: reserve() may allocate more than asked,
+        // so the two capacities can differ)
+        e.moved.reserve((e.moved.size() * 2) + 16);
+        e.moved_cap.reserve((e.moved.size() * 2) + 16);
     }
 }
 

include/nlohmann/detail/view/image.hpp

@@ -8,11 +8,11 @@
 
 #pragma once
 
-#include <algorithm> // sort
+#include <algorithm> // min, sort
 #include <array> // array
-#include <cstddef> // size_t
+#include <cstddef> // offsetof, size_t
 #include <cstdint> // int64_t, uint8_t, uint16_t, uint32_t, uint64_t
-#include <cstring> // memcmp, memcpy
+#include <cstring> // memcmp, memcpy, memset
 #include <limits> // numeric_limits
 #include <string> // string
 #include <vector> // vector
@@ -105,7 +105,7 @@ inline void compact_nodes(const document_data& d, std::size_t arena_size, std::v
         }
         return r;
     };
-    const auto emit = [&](const node * v)
+    const auto copy_node = [&](const node * v)
     {
         node r = *v;
         switch (static_cast<value_t>(v->kind))
@@ -159,10 +159,10 @@ inline void compact_nodes(const document_data& d, std::size_t arena_size, std::v
         }
         out.push_back(r);
     };
-    emit(d.nodes);
+    copy_node(d.nodes);
     while (!stack.empty())
     {
-        frame& top = stack.back();
+        const frame top = stack.back();
         if (top.cur == top.end)
         {
             node& c = out[top.index];
@@ -171,20 +171,14 @@ inline void compact_nodes(const document_data& d, std::size_t arena_size, std::v
             stack.pop_back();
             continue;
         }
-        ++top.count;
-        const node* v = nullptr;
         if (top.object)
         {
             out.push_back(string_node(*top.cur));
-            v = document_data::deref(top.cur + 1);
-            top.cur = document_data::after(top.cur + 1);
         }
-        else
-        {
-            v = document_data::deref(top.cur);
-            top.cur = document_data::after(top.cur);
-        }
-        emit(v); // may grow the stack (top is not used afterwards)
+        const node* const value = top.cur + (top.object ? 1 : 0);
+        stack.back().cur = document_data::after(value);
+        ++stack.back().count;
+        copy_node(document_data::deref(value)); // may grow the stack
     }
 }
 
@@ -195,17 +189,15 @@ inline std::vector<std::uint8_t> save_image(const document_data& d)
     throw_type_error(320, "json_document images need a little-endian target"); // LCOV_EXCL_LINE
 #endif
     const std::size_t arena_size = d.arena_size;
-    const node* nodes = d.nodes;
-    std::size_t count = d.node_count;
     std::vector<node> compacted;
     std::string text_tail;
     std::string arena_tail;
     if (d.edits)
     {
         compact_nodes(d, arena_size, compacted, text_tail, arena_tail);
-        nodes = compacted.data();
-        count = compacted.size();
     }
+    const node* const nodes = d.edits ? compacted.data() : d.nodes;
+    const std::size_t count = d.edits ? compacted.size() : d.node_count;
     const std::size_t text_size = d.size + text_tail.size();
     const std::size_t total_arena = arena_size + arena_tail.size();
     if (NLOHMANN_VIEW_UNLIKELY(text_size >= image_limit || total_arena >= image_limit || count >= image_limit))
@@ -225,14 +217,13 @@ inline std::vector<std::uint8_t> save_image(const document_data& d)
     std::memcpy(o, &h, sizeof(h));
     o += sizeof(h);
     std::memcpy(o, nodes, count * sizeof(node));
-    // the hash indexes are rebuilt by load()
+    // the hash indexes are rebuilt by load(), whose check rejects any other
+    // extra of an object, and the same document saves to the same bytes
     for (std::size_t i = 0; i < count; ++i)
     {
         if (nodes[i].kind == static_cast<std::uint8_t>(value_t::object) && nodes[i].extra != 0)
         {
-            node n = nodes[i];
-            n.extra = 0;
-            std::memcpy(o + (i * sizeof(node)), &n, sizeof(node));
+            std::memset(o + (i * sizeof(node)) + offsetof(node, extra), 0, sizeof(nodes[i].extra));
         }
     }
     o += count * sizeof(node);
@@ -271,21 +262,22 @@ inline bool scan_number_token(const unsigned char* s, std::size_t len, number_to
     const unsigned char* p = s;
     t.negative = p != e && *p == '-';
     p += t.negative ? 1 : 0;
-    t.int_start = p;
     if (p == e)
     {
         return false;
     }
+    t.int_start = p;
     if (*p == '0')
     {
         ++p;
     }
     else if (*p >= '1' && *p <= '9')
     {
-        while (p != e && is_digit(*p))
+        do
         {
             ++p;
         }
+        while (p != e && is_digit(*p));
     }
     else
     {
@@ -318,11 +310,12 @@ inline bool scan_number_token(const unsigned char* s, std::size_t len, number_to
         {
             return false;
         }
-        while (p != e && is_digit(*p))
+        do
         {
             t.exponent = t.exponent < 100000 ? (t.exponent * 10) + (*p - '0') : t.exponent;
             ++p;
         }
+        while (p != e && is_digit(*p));
         t.exponent = exp_negative ? -t.exponent : t.exponent;
         t.is_float = true;
     }
@@ -333,7 +326,7 @@ inline bool scan_number_token(const unsigned char* s, std::size_t len, number_to
 /// writes "many" instead; the caller accepts both)
 inline std::uint16_t float_layout(const number_token& t)
 {
-    return static_cast<std::uint16_t>((t.int_digits < 255 ? t.int_digits : 255) | ((t.frac_digits < 255 ? t.frac_digits : 255) << 8u));
+    return static_cast<std::uint16_t>((std::min)(t.int_digits, std::size_t{255}) | ((std::min)(t.frac_digits, std::size_t{255}) << 8u));
 }
 
 /// whether a float token (well-formed, per scan_number_token) is finite as
@@ -735,14 +728,10 @@ inline void load_image(document_data& d, const std::uint8_t* image, std::size_t
     // the hash indexes of large objects, as after parsing
     for (std::size_t i = 0; i < count; ++i)
     {
-        node& n = d.nodes[i];
-        if (n.kind == static_cast<std::uint8_t>(value_t::object))
+        const node& n = d.nodes[i];
+        if (n.kind == static_cast<std::uint8_t>(value_t::object) && n.len >= document_data::index_min_members)
         {
-            n.extra = 0;
-            if (n.len >= document_data::index_min_members)
-            {
-                d.large_objects.push_back(static_cast<std::uint32_t>(i));
-            }
+            d.large_objects.push_back(static_cast<std::uint32_t>(i));
         }
     }
     build_object_indexes(d);

include/nlohmann/detail/view/lookup.hpp

@@ -112,7 +112,7 @@ const node* find_member(const document_data& d, const node* object, const char*
     }
     for (const node* m = nav::first(d, object); m != end; m = document_data::after(m + 1))
     {
-        if (m->len == n && std::memcmp(d.str(*m), key, n) == 0)
+        if (d.key_equals(*m, key, n))
         {
             return m;
         }
@@ -120,6 +120,23 @@ const node* find_member(const document_data& d, const node* object, const char*
     return nullptr;
 }
 
+/// the number of members of an object with this key (duplicates included)
+template<bool Editable>
+std::size_t count_members(const document_data& d, const node* object, const char* key, std::size_t n) noexcept
+{
+    using nav = navigation<Editable>;
+    std::size_t result = 0;
+    const node* const end = nav::end(d, object);
+    for (const node* m = nav::first(d, object); m != end; m = document_data::after(m + 1))
+    {
+        if (d.key_equals(*m, key, n))
+        {
+            ++result;
+        }
+    }
+    return result;
+}
+
 /// whether an integer type is accepted as an array index by the view's
 /// operator[] and at(): every integer type but bool and size_t, which has its
 /// own overload

include/nlohmann/detail/view/macro_scope.hpp

@@ -70,5 +70,11 @@
     #define NLOHMANN_VIEW_LITTLE_ENDIAN 0
 #endif
 
+/// eight checks at fixed offsets 0..7
+#define NLOHMANN_VIEW_REPEAT_FIRST_8(X) X(0) X(1) X(2) X(3) X(4) X(5) X(6) X(7)
+
+/// eight checks at fixed offsets 8..15
+#define NLOHMANN_VIEW_REPEAT_SECOND_8(X) X(8) X(9) X(10) X(11) X(12) X(13) X(14) X(15)
+
 /// sixteen checks at fixed offsets 0..15
-#define NLOHMANN_VIEW_REPEAT16(X) X(0) X(1) X(2) X(3) X(4) X(5) X(6) X(7) X(8) X(9) X(10) X(11) X(12) X(13) X(14) X(15)
+#define NLOHMANN_VIEW_REPEAT16(X) NLOHMANN_VIEW_REPEAT_FIRST_8(X) NLOHMANN_VIEW_REPEAT_SECOND_8(X)

include/nlohmann/detail/view/macro_unscope.hpp

@@ -19,6 +19,8 @@
 #undef NLOHMANN_VIEW_THROW
 #undef NLOHMANN_VIEW_LITTLE_ENDIAN
 #undef NLOHMANN_VIEW_REPEAT16
+#undef NLOHMANN_VIEW_REPEAT_SECOND_8
+#undef NLOHMANN_VIEW_REPEAT_FIRST_8
 #undef NLOHMANN_VIEW_NEON
 #undef NLOHMANN_VIEW_SSE2
 #undef NLOHMANN_VIEW_SSSE3

include/nlohmann/detail/view/node.hpp

@@ -34,7 +34,7 @@ struct node_flags
 {
     static constexpr std::uint8_t escaped = 1; ///< string payload lives in the decode arena, not the source
     static constexpr std::uint8_t edited = 2;  ///< string or number token lives in the edit arena (editable documents)
-    static constexpr std::uint8_t storage = 3; ///< mask: where a string or number token lives (index into document_data::base)
+    static constexpr std::uint8_t storage = escaped | edited; ///< mask: where a string or number token lives (index into document_data::base)
     static constexpr std::uint8_t is_true = 4; ///< boolean value
     static constexpr std::uint8_t moved = 8;   ///< array/object: the elements live in a separate sequence (editable documents)
     static constexpr std::uint8_t is_new = 16; ///< written by an edit: no source position
@@ -53,7 +53,7 @@ struct node
 {
     std::uint8_t kind;   ///< value_t, or kind_link
     std::uint8_t flags;  ///< node_flags
-    std::uint16_t extra; ///< numbers: integer digits (low byte) and fraction digits (high byte), 255 = "many"; objects: number of the hash index (1-based, 0 = none); otherwise 0
+    std::uint16_t extra; ///< numbers: integer digits (low byte) and fraction digits (high byte), 255 = "many"; objects: number of the object's key hash index (1-based, 0 = none); otherwise 0
     std::uint32_t off;   ///< source offset (string content, number token, literal, bracket); arena offset if escaped/edited; number of the element sequence if moved
     std::uint32_t len;   ///< string: decoded bytes; float: token bytes; array/object: element count
     std::uint32_t next;  ///< array/object: number of nodes of the subtree (its extent in the enclosing sequence)

include/nlohmann/detail/view/number.hpp

@@ -45,10 +45,11 @@ inline bool float_token_layout(const char* first, const char* last, std::size_t&
     {
         return false;
     }
-    while (digit(p))
+    do
     {
         ++p;
     }
+    while (digit(p));
     dot = std::string::npos;
     if (p != last && *p == '.')
     {
@@ -57,10 +58,11 @@ inline bool float_token_layout(const char* first, const char* last, std::size_t&
         {
             return false;
         }
-        while (digit(p))
+        do
         {
             ++p;
         }
+        while (digit(p));
     }
     mantissa_end = static_cast<std::size_t>(p - first);
     if (p != last && (*p == 'e' || *p == 'E'))
@@ -71,10 +73,11 @@ inline bool float_token_layout(const char* first, const char* last, std::size_t&
         {
             return false;
         }
-        while (digit(p))
+        do
         {
             ++p;
         }
+        while (digit(p));
     }
     return p == last;
 }

include/nlohmann/detail/view/object_index.hpp

@@ -10,14 +10,13 @@
 
 #include <cstddef> // size_t
 #include <cstdint> // uint32_t, uint64_t
-#include <cstring> // memcmp
 
 #include <nlohmann/json.hpp>
 #include <nlohmann/detail/view/document_data.hpp>
 #include <nlohmann/detail/view/macro_scope.hpp>
 #include <nlohmann/detail/view/node.hpp>
 
-// Hash indexes of large objects, so that a lookup does not compare thousands
+// Hash indexes of the keys of large objects, so that a lookup does not compare thousands
 // of keys (as Boost.JSON switches from a linear search to a hash table for
 // large objects). An object with document_data::index_min_members members or
 // more gets an open-addressing table after parsing; its node stores the
@@ -42,6 +41,12 @@ namespace view
 /// members; and every member costs at most this many steps while building)
 constexpr std::size_t index_max_displacement = 64;
 
+// a table has at least 2 * len slots (len >= index_min_members), so the probe
+// window of index_max_displacement + 1 slots stays far smaller than the table
+// and a probe never wraps around onto its own start
+static_assert(document_data::index_min_members >= 2 * index_max_displacement,
+              "a table must be much larger than the probe window of a lookup");
+
 /// hash of a key: its bytes, eight at a time, in a fixed byte order
 inline std::uint64_t key_hash(const char* s, std::size_t n) noexcept
 {
@@ -77,45 +82,36 @@ inline void build_object_index(document_data& d, node* obj)
     }
     const std::size_t start = d.index_slots.size();
     d.index_slots.resize(start + cap, 0);
-    std::uint32_t* const slots = d.index_slots.data() + start;
+    std::uint32_t* const table = d.index_slots.data() + start;
     const std::size_t mask = cap - 1;
-    bool degenerate = false;
     for (const node* k = document_data::first_child(obj), *end = document_data::child_end(obj); k != end; k = document_data::after(k + 1))
     {
         const char* const key = d.str(*k);
-        const std::uint64_t hash = key_hash(key, k->len); // (a cast of the call would be useless where std::uint64_t is std::size_t)
-        std::size_t i = static_cast<std::size_t>(hash) & mask;
+        std::size_t slot = conditional_static_cast<std::size_t>(key_hash(key, k->len)) & mask;
         bool duplicate = false;
         std::size_t distance = 0;
-        while (slots[i] != 0)
+        while (table[slot] != 0)
         {
-            const node* const other = obj + slots[i];
-            if (other->len == k->len && (k->len == 0 || std::memcmp(d.str(*other), key, k->len) == 0))
+            const node* const other = obj + table[slot];
+            if (d.key_equals(*other, key, k->len))
             {
                 duplicate = true; // keep the first
                 break;
             }
-            if (++distance > index_max_displacement)
+            if (distance == index_max_displacement)
             {
-                degenerate = true; // too many keys share a home region
-                break;
+                // too many keys share a home region
+                d.index_slots.resize(start); // no table: the object is searched linearly
+                return;
             }
-            i = (i + 1) & mask;
-        }
-        if (degenerate)
-        {
-            break;
+            ++distance;
+            slot = (slot + 1) & mask;
         }
         if (!duplicate)
         {
-            slots[i] = static_cast<std::uint32_t>(k - obj);
+            table[slot] = static_cast<std::uint32_t>(k - obj);
         }
     }
-    if (degenerate)
-    {
-        d.index_slots.resize(start); // no table: the object is searched linearly
-        return;
-    }
     d.indexes.push_back(document_data::object_index{start, static_cast<std::uint32_t>(mask)});
     obj->extra = static_cast<std::uint16_t>(d.indexes.size());
 }
@@ -134,22 +130,21 @@ inline void build_object_indexes(document_data& d)
 inline const node* find_indexed(const document_data& d, const node* obj, const char* key, std::size_t n) noexcept
 {
     const document_data::object_index& ix = d.indexes[obj->extra - 1u];
-    const std::uint32_t* const slots = d.index_slots.data() + ix.start;
-    const std::uint64_t hash = key_hash(key, n); // (a cast of the call would be useless where std::uint64_t is std::size_t)
-    std::size_t i = static_cast<std::size_t>(hash) & ix.mask;
+    const std::uint32_t* const table = d.index_slots.data() + ix.start;
+    std::size_t slot = conditional_static_cast<std::size_t>(key_hash(key, n)) & ix.mask;
     for (std::size_t distance = 0; distance <= index_max_displacement; ++distance)
     {
-        const std::uint32_t s = slots[i];
+        const std::uint32_t s = table[slot];
         if (s == 0)
         {
             return nullptr;
         }
         const node* const k = obj + s;
-        if (k->len == n && (n == 0 || std::memcmp(d.str(*k), key, n) == 0))
+        if (d.key_equals(*k, key, n))
         {
             return k;
         }
-        i = (i + 1) & ix.mask;
+        slot = (slot + 1) & ix.mask;
     }
     return nullptr; // (no key sits further from its home slot)
 }

include/nlohmann/detail/view/pointer.hpp

@@ -47,24 +47,24 @@ enum class index_status
 template<typename StringType>
 index_status array_index(const StringType& s, std::size_t& idx) noexcept
 {
+    // as basic_json: the empty token and "-" (past the last element) are
+    // unresolved rather than not a number
+    if (s.empty() || (s.size() == 1 && s[0] == '-'))
+    {
+        return index_status::unresolved;
+    }
+    if (s[0] < '0' || s[0] > '9')
+    {
+        return index_status::not_number;
+    }
     if (s.size() > 1 && s[0] == '0')
     {
         return index_status::leading_zero;
     }
-    // as basic_json: also a single character, but "-" and the empty token are
-    // left to the conversion below and reported as unresolved
-    if (!s.empty() && !(s.size() == 1 && s[0] == '-') && !(s[0] >= '0' && s[0] <= '9'))
-    {
-        return index_status::not_number;
-    }
-    if (s.empty())
-    {
-        return index_status::unresolved;
-    }
     std::uint64_t v = 0;
-    for (std::size_t i = 0; i < s.size(); ++i)
+    for (const auto c : s)
     {
-        const auto d = static_cast<unsigned>(static_cast<unsigned char>(s[i])) - '0';
+        const auto d = static_cast<unsigned>(static_cast<unsigned char>(c)) - '0';
         if (d > 9 || v > ((std::numeric_limits<std::uint64_t>::max)() - d) / 10)
         {
             return index_status::unresolved; // not a number, or beyond unsigned long long

include/nlohmann/detail/view/scan.hpp

@@ -83,14 +83,12 @@ NLOHMANN_VIEW_ALWAYS_INLINE const unsigned char* scan_string_run(const unsigned
         if (e - p >= 16)
         {
 #define NLOHMANN_VIEW_STEP(i) if (NLOHMANN_VIEW_LIKELY(plain[p[i]] != 0)) {} else { p += (i); goto stop; }
-            NLOHMANN_VIEW_STEP(0) NLOHMANN_VIEW_STEP(1) NLOHMANN_VIEW_STEP(2) NLOHMANN_VIEW_STEP(3)
-            NLOHMANN_VIEW_STEP(4) NLOHMANN_VIEW_STEP(5) NLOHMANN_VIEW_STEP(6) NLOHMANN_VIEW_STEP(7)
+            NLOHMANN_VIEW_REPEAT_FIRST_8(NLOHMANN_VIEW_STEP)
 #if NLOHMANN_VIEW_VECTOR
             if (!Value) // (a plain `!Value || !NLOHMANN_VIEW_VECTOR` is a constant condition for MSVC, C4127)
 #endif
             {
-                NLOHMANN_VIEW_STEP(8) NLOHMANN_VIEW_STEP(9) NLOHMANN_VIEW_STEP(10) NLOHMANN_VIEW_STEP(11)
-                NLOHMANN_VIEW_STEP(12) NLOHMANN_VIEW_STEP(13) NLOHMANN_VIEW_STEP(14) NLOHMANN_VIEW_STEP(15)
+                NLOHMANN_VIEW_REPEAT_SECOND_8(NLOHMANN_VIEW_STEP)
                 p += 8;
             }
 #undef NLOHMANN_VIEW_STEP
@@ -177,32 +175,7 @@ NLOHMANN_VIEW_ALWAYS_INLINE const unsigned char* skip_digits(const unsigned char
 /// powers of ten up to 10^19 as integers
 inline std::uint64_t int_pow10(unsigned k) noexcept
 {
-    static const std::array<std::uint64_t, 20> table =
-    {
-        {
-            1u,
-            10u,
-            100u,
-            1000u,
-            10000u,
-            100000u,
-            1000000u,
-            10000000u,
-            100000000u,
-            1000000000u,
-            10000000000u,
-            100000000000u,
-            1000000000000u,
-            10000000000000u,
-            100000000000000u,
-            1000000000000000u,
-            10000000000000000u,
-            100000000000000000u,
-            1000000000000000000u,
-            10000000000000000000u
-        }
-    };
-    return table[k];
+    return ::nlohmann::detail::dtoa_impl::powers_of_ten()[k];
 }
 
 /// value of 0 < k < 8 digits at p in one step if [p, p + 8) lies below

include/nlohmann/detail/view/serializer.hpp

@@ -73,6 +73,13 @@ class output_buffer
         m_pos += n;
     }
 
+    /// a string literal, without its terminating NUL
+    template<std::size_t N>
+    NLOHMANN_VIEW_ALWAYS_INLINE void put(const char (&s)[N]) // NOLINT(cppcoreguidelines-avoid-c-arrays,hicpp-avoid-c-arrays,modernize-avoid-c-arrays)
+    {
+        put(s, N - 1);
+    }
+
     void put_repeated(char c, std::size_t n)
     {
         reserve(n);
@@ -102,7 +109,7 @@ class output_buffer
     /// which a static constexpr member does not have before C++17)
     static constexpr std::size_t step() noexcept
     {
-        return 65536;
+        return 0x10000;
     }
 
     static StringType& sized(StringType& out, std::size_t estimate)
@@ -167,14 +174,17 @@ inline std::size_t plain_output_run(const unsigned char* s, std::size_t n) noexc
 /// A stack that starts in a buffer of the caller (a local array) and moves to
 /// the heap (a vector of the caller) only when that is full, so that dumps of
 /// shallow documents need no allocation. The top is a pointer, as in
-/// std::vector. The address of the stack never escapes (the growth gets the
-/// vector and returns the new storage), so its pointers stay in registers.
+/// std::vector. grow() is static and out of line: it gets the vector and the
+/// used range and returns the new storage, so the address of the stack never
+/// reaches a call. Otherwise the compiler would have to keep the three
+/// pointers in memory and load and store them on every push_back() and
+/// pop_back(), not just when growing.
 template<typename T>
 class small_stack
 {
   public:
     small_stack(T* buffer, std::size_t capacity, std::vector<T>& heap) noexcept
-        : m_begin(buffer), m_top(buffer), m_end(buffer + capacity), m_heap(&heap)
+        : m_begin(buffer), m_top(buffer), m_end(buffer + capacity), m_heap(heap)
     {}
     small_stack(const small_stack&) = delete;
     small_stack(small_stack&&) = delete;
@@ -188,7 +198,7 @@ class small_stack
         {
             const std::size_t used = size();
             const std::size_t capacity = 2 * static_cast<std::size_t>(m_end - m_begin);
-            m_begin = grow(*m_heap, m_begin, used, capacity);
+            m_begin = grow(m_heap, m_begin, used, capacity);
             m_top = m_begin + used;
             m_end = m_begin + capacity;
         }
@@ -228,9 +238,18 @@ class small_stack
     T* m_begin;
     T* m_top;
     T* m_end;
-    std::vector<T>* m_heap;
+    std::vector<T>& m_heap;
 };
 
+/// a string literal at w, without its terminating NUL (after room())
+/// @return a pointer past it
+template<std::size_t N>
+NLOHMANN_VIEW_ALWAYS_INLINE char* put_literal(char* w, const char (&s)[N]) noexcept // NOLINT(cppcoreguidelines-avoid-c-arrays,hicpp-avoid-c-arrays,modernize-avoid-c-arrays)
+{
+    std::memcpy(w, s, N - 1);
+    return w + N - 1;
+}
+
 /// how the view's dump() writes a value
 struct dump_style
 {
@@ -319,15 +338,15 @@ class view_serializer
                     m_out.finish();
                     return;
                 }
-                frame& f = stack.back();
-                if (f.pos == f.end)
+                if (stack.back().pos == stack.back().end)
                 {
-                    const bool object = f.object;
+                    const bool object = stack.back().object;
                     stack.pop_back();
                     newline(stack.size());
                     m_out.put(object ? '}' : ']');
                     continue;
                 }
+                frame& f = stack.back();
                 if (!f.first)
                 {
                     m_out.put(',');
@@ -339,7 +358,7 @@ class view_serializer
                     write_string(*f.pos);
                     if (m_style.pretty)
                     {
-                        m_out.put(": ", 2);
+                        m_out.put(": ");
                     }
                     else
                     {
@@ -454,12 +473,6 @@ class view_serializer
             }
             w += n;
         };
-        // a literal of n bytes (after room(n))
-        const auto literal = [&](const char* text, std::size_t n)
-        {
-            std::memcpy(w, text, n);
-            w += n;
-        };
         // a string that is not a plain string of the source (out of line, so
         // that the cursor stays in a register here)
         const auto escaped = [&](const node & n)
@@ -545,11 +558,11 @@ class view_serializer
                     room(8);
                     if ((n->flags & node_flags::is_true) != 0)
                     {
-                        literal("true", 4);
+                        w = put_literal(w, "true");
                     }
                     else
                     {
-                        literal("false", 5);
+                        w = put_literal(w, "false");
                     }
                     break;
                 case value_t::object:
@@ -559,7 +572,7 @@ class view_serializer
                     room(8);
                     if (n->len == 0)
                     {
-                        literal(object ? "{}" : "[]", 2);
+                        w = object ? put_literal(w, "{}") : put_literal(w, "[]");
                     }
                     else
                     {
@@ -579,7 +592,7 @@ class view_serializer
                 }
                 case value_t::null:
                     room(8);
-                    literal("null", 4);
+                    w = put_literal(w, "null");
                     break;
                 case value_t::binary:    // LCOV_EXCL_LINE (not in a document)
                 case value_t::discarded: // LCOV_EXCL_LINE
@@ -666,16 +679,16 @@ class view_serializer
         switch (static_cast<value_t>(n.kind))
         {
             case value_t::null:
-                m_out.put("null", 4);
+                m_out.put("null");
                 break;
             case value_t::boolean:
                 if ((n.flags & node_flags::is_true) != 0)
                 {
-                    m_out.put("true", 4);
+                    m_out.put("true");
                 }
                 else
                 {
-                    m_out.put("false", 5);
+                    m_out.put("false");
                 }
                 break;
             case value_t::string:
@@ -759,7 +772,7 @@ class view_serializer
                 // check also keeps an image that was only checked for bounds,
                 // whose token may not be made of digits, from the counted
                 // overload)
-                const auto& powers = ::nlohmann::detail::dtoa_impl::powers_of_ten_16();
+                const auto& powers = ::nlohmann::detail::dtoa_impl::powers_of_ten();
                 const unsigned count = int_digits + frac_digits;
                 return count - 1u < 15u && d.w >= powers[count - 1u] && d.w < powers[count]
                        ? ::nlohmann::detail::dtoa_impl::write_short_decimal(w, d.w, static_cast<int>(count), static_cast<int>(d.exponent))
@@ -802,12 +815,14 @@ class view_serializer
     {
         if (!std::isfinite(x))
         {
-            m_out.put("null", 4);
+            m_out.put("null");
             return;
         }
-        write_float(x, std::integral_constant < bool,
-                    (std::numeric_limits<number_float_t>::is_iec559 && std::numeric_limits<number_float_t>::digits == 24 && std::numeric_limits<number_float_t>::max_exponent == 128)
-                    || (std::numeric_limits<number_float_t>::is_iec559 && std::numeric_limits<number_float_t>::digits == 53 && std::numeric_limits<number_float_t>::max_exponent == 1024) > {});
+        static constexpr bool is_ieee_single_or_double
+            = std::numeric_limits<number_float_t>::is_iec559
+              && ((std::numeric_limits<number_float_t>::digits == 24 && std::numeric_limits<number_float_t>::max_exponent == 128)
+                  || (std::numeric_limits<number_float_t>::digits == 53 && std::numeric_limits<number_float_t>::max_exponent == 1024));
+        write_float(x, std::integral_constant<bool, is_ieee_single_or_double>());
     }
 
     void write_float(number_float_t x, std::true_type /*is_ieee_single_or_double*/)
@@ -902,25 +917,25 @@ class view_serializer
         switch (codepoint)
         {
             case 0x08:
-                m_out.put("\\b", 2);
+                m_out.put("\\b");
                 return;
             case 0x09:
-                m_out.put("\\t", 2);
+                m_out.put("\\t");
                 return;
             case 0x0A:
-                m_out.put("\\n", 2);
+                m_out.put("\\n");
                 return;
             case 0x0C:
-                m_out.put("\\f", 2);
+                m_out.put("\\f");
                 return;
             case 0x0D:
-                m_out.put("\\r", 2);
+                m_out.put("\\r");
                 return;
             case 0x22:
-                m_out.put("\\\"", 2);
+                m_out.put("\\\"");
                 return;
             case 0x5C:
-                m_out.put("\\\\", 2);
+                m_out.put("\\\\");
                 return;
             default:
                 break;

include/nlohmann/detail/view/value.hpp

@@ -8,10 +8,11 @@
 
 #pragma once
 
+#include <cstddef> // nullptr_t, size_t
 #include <cstdint> // int64_t
 #include <map> // map
 #include <string> // basic_string
-#include <type_traits> // enable_if, is_constructible
+#include <type_traits> // decay, enable_if, is_arithmetic, is_constructible, is_integral, is_same
 #include <unordered_map> // unordered_map
 #include <vector> // vector
 
@@ -103,6 +104,41 @@ Map map_value(const View& v)
 template<typename Key>
 struct is_string_key : std::is_constructible<Key, const char*, std::size_t> {};
 
+/// the arithmetic types arithmetic_value() converts to (bool has its own
+/// conversion)
+template<typename T>
+struct is_non_bool_arithmetic : conjunction<std::is_arithmetic<T>, negation<std::is_same<T, bool>>> {};
+
+/// the default values of value() other than C strings, which have their own
+/// overload (returning string_t)
+template<typename T>
+struct is_non_c_string_default : negation<std::is_same<typename std::decay<T>::type, const char*>> {};
+
+/// the integral types accepted as an array index by the edits of a document
+template<typename T>
+struct is_non_bool_integral : conjunction<std::is_integral<T>, negation<std::is_same<T, bool>>> {};
+
+/// whether View::get<T>() converts to T without materialize(); get_to()
+/// assigns these, and passes the others to from_json
+template<typename T, typename BasicJsonType, typename View>
+struct is_direct_value : std::integral_constant < bool,
+    std::is_arithmetic<T>::value || std::is_same<T, std::nullptr_t>::value
+    || std::is_same<T, typename View::string_view_t>::value
+    || std::is_same<T, BasicJsonType>::value || std::is_same<T, View>::value >
+{};
+
+template<typename Traits, typename Alloc, typename BasicJsonType, typename View>
+struct is_direct_value<std::basic_string<char, Traits, Alloc>, BasicJsonType, View> : std::true_type {};
+
+template<typename U, typename A, typename BasicJsonType, typename View>
+struct is_direct_value<std::vector<U, A>, BasicJsonType, View> : std::true_type {};
+
+template<typename K, typename V, typename C, typename A, typename BasicJsonType, typename View>
+struct is_direct_value<std::map<K, V, C, A>, BasicJsonType, View> : is_string_key<K> {};
+
+template<typename K, typename V, typename H, typename E, typename A, typename BasicJsonType, typename View>
+struct is_direct_value<std::unordered_map<K, V, H, E, A>, BasicJsonType, View> : is_string_key<K> {};
+
 }  // namespace view
 }  // namespace detail
 NLOHMANN_JSON_NAMESPACE_END

include/nlohmann/json_view.hpp

@@ -31,12 +31,12 @@
 #include <cstddef> // nullptr_t, size_t // IWYU pragma: keep
 #include <cstdint> // uint8_t, uint32_t
 #include <cstring> // memcpy, strlen
-#include <istream> // istream
+#include <istream> // istream // IWYU pragma: keep
 #include <iterator> // distance, input_iterator_tag, iterator_traits
 #include <map> // map
 #include <memory> // unique_ptr
 #ifndef JSON_NO_IO
-    #include <ostream> // ostream
+    #include <ostream> // ostream // IWYU pragma: keep
 #endif
 #include <string> // string
 #include <tuple> // tuple_element, tuple_size // IWYU pragma: keep
@@ -265,7 +265,7 @@ class basic_json_view
         {
             if (is_discarded())
             {
-                return basic_json_view();
+                return {};
             }
             detail::view::throw_type_error(305, "cannot use operator[] with a string argument with ", type_name());
         }
@@ -291,7 +291,7 @@ class basic_json_view
         {
             if (is_discarded())
             {
-                return basic_json_view();
+                return {};
             }
             detail::view::throw_type_error(305, "cannot use operator[] with a numeric argument with ", type_name());
         }
@@ -314,7 +314,7 @@ class basic_json_view
     {
         if (NLOHMANN_VIEW_UNLIKELY(is_discarded()))
         {
-            return basic_json_view();
+            return {};
         }
         return detail::view::resolve_pointer(*this, detail::json_pointer_access::reference_tokens(ptr), detail::view::pointer_mode::unchecked);
     }
@@ -379,7 +379,7 @@ class basic_json_view
     /// the member with this key converted to T, or the default value if there
     /// is no such member (the first one, should the key occur more than
     /// once). Throws type_error.306 if this is not an object.
-    template < typename T, typename std::enable_if < !std::is_same<typename std::decay<T>::type, const char*>::value, int >::type = 0 >
+    template<typename T, typename std::enable_if<detail::view::is_non_c_string_default<T>::value, int>::type = 0>
     T value(string_view_t key, const T& default_value) const
     {
         if (NLOHMANN_VIEW_UNLIKELY(!is_object()))
@@ -398,7 +398,7 @@ class basic_json_view
     /// the value a JSON pointer refers to converted to T, or the default
     /// value if the pointer cannot be resolved. Throws type_error.306 if this
     /// is neither an object nor an array.
-    template < typename T, typename std::enable_if < !std::is_same<typename std::decay<T>::type, const char*>::value, int >::type = 0 >
+    template<typename T, typename std::enable_if<detail::view::is_non_c_string_default<T>::value, int>::type = 0>
     T value(const json_pointer& ptr, const T& default_value) const
     {
         if (NLOHMANN_VIEW_UNLIKELY(!is_structured()))
@@ -487,11 +487,11 @@ class basic_json_view
         return !detail::view::resolve_pointer(*this, detail::json_pointer_access::reference_tokens(ptr), detail::view::pointer_mode::contains).is_discarded();
     }
 
-    /// 1 if this is an object with a member with this key, else 0 (duplicate
-    /// keys count once)
+    /// the number of members with this key (more than 1 if the key occurs
+    /// more than once); 0 if this is not an object
     size_type count(string_view_t key) const
     {
-        return contains(key) ? 1 : 0;
+        return is_object() ? detail::view::count_members<Editable>(*m_doc, m_node, key.data(), key.size()) : 0;
     }
 
     size_type count(const char* key) const
@@ -558,10 +558,12 @@ class basic_json_view
         return get_impl(detail::view::value_tag<T> {}, detail::priority_tag<2> {});
     }
 
+    /// as BasicJsonType::get_to(): the types get<T>() converts directly are
+    /// assigned, all others are passed to from_json with v as it is
     template<typename T>
     T& get_to(T& v) const
     {
-        v = get<T>();
+        get_to_impl(v, detail::view::is_direct_value<T, BasicJsonType, basic_json_view> {});
         return v;
     }
 
@@ -748,8 +750,10 @@ class basic_json_view
     {
         if (editable() && m_doc->edits != nullptr)
         {
-            // positions of moved and new values are not source offsets
-            return m_node == m_doc->nodes ? m_doc->size + m_doc->edits->text_used : 64;
+            // positions of moved and new values are not source offsets; the root
+            // (the first node) spans the source text and every edit
+            const bool is_root = m_node == m_doc->nodes;
+            return is_root ? m_doc->size + m_doc->edits->text_used : 64;
         }
         const node* const end = m_doc->nodes + m_doc->node_count;
         if ((m_node->flags & detail::view::node_flags::storage) != 0)
@@ -796,7 +800,7 @@ class basic_json_view
         return (m_node->flags & detail::view::node_flags::is_true) != 0;
     }
 
-    template < typename T, typename std::enable_if < std::is_arithmetic<T>::value && !std::is_same<T, bool>::value, int >::type = 0 >
+    template<typename T, typename std::enable_if<detail::view::is_non_bool_arithmetic<T>::value, int>::type = 0>
     NLOHMANN_VIEW_ALWAYS_INLINE T get_impl(detail::view::value_tag<T> /*unused*/, detail::priority_tag<2> /*unused*/) const
     {
         if (NLOHMANN_VIEW_UNLIKELY(m_node == nullptr))
@@ -862,6 +866,21 @@ class basic_json_view
         return materialize().template get<T>();
     }
 
+    // --- get_to() dispatch ---
+
+    template<typename T>
+    void get_to_impl(T& v, std::true_type /*direct*/) const
+    {
+        v = get<T>();
+    }
+
+    /// from_json may keep parts of v (members a JSON object lacks, say)
+    template<typename T>
+    void get_to_impl(T& v, std::false_type /*direct*/) const
+    {
+        materialize().get_to(v);
+    }
+
     const document_data* m_doc = nullptr;
     const node* m_node = nullptr;
 };
@@ -1021,7 +1040,7 @@ class basic_json_document
         return m_data ? m_data->node_count : 0;
     }
 
-    /// bytes held by the document (index, decoded strings, owned text or image)
+    /// bytes held by the document (index, decoded strings, and owned text or image)
     std::size_t memory_usage() const noexcept
     {
         if (!m_data)
@@ -1169,7 +1188,7 @@ class basic_json_document
     }
 
     /// assign an existing array element; returns a view of it
-    template < typename I, typename V, typename std::enable_if < std::is_integral<I>::value && !std::is_same<I, bool>::value, int >::type = 0 >
+    template<typename I, typename V, typename std::enable_if<detail::view::is_non_bool_integral<I>::value, int>::type = 0>
     view_type set(view_type array, I idx, V && value)
     {
         return editor().set(array, index(idx), std::forward<V>(value));
@@ -1191,14 +1210,14 @@ class basic_json_document
         const auto& token = ptr.back();
         if (parent.is_null())
         {
-            const bool digits = std::all_of(token.begin(), token.end(), [](const char c)
-            {
-                return c >= '0' && c <= '9';
-            });
             if (token == "-")
             {
                 return push_back(parent, std::forward<V>(value));
             }
+            const bool digits = std::all_of(token.begin(), token.end(), [](const char c)
+            {
+                return c >= '0' && c <= '9';
+            });
             if (digits)
             {
                 // (an invalid index is an error before the parent changes)
@@ -1236,7 +1255,7 @@ class basic_json_document
 
     /// insert into an array before position idx (idx <= size()); returns a
     /// view of the new element
-    template < typename I, typename V, typename std::enable_if < std::is_integral<I>::value && !std::is_same<I, bool>::value, int >::type = 0 >
+    template<typename I, typename V, typename std::enable_if<detail::view::is_non_bool_integral<I>::value, int>::type = 0>
     view_type insert(view_type array, I idx, V && value)
     {
         return editor().insert(array, index(idx), std::forward<V>(value));
@@ -1249,7 +1268,7 @@ class basic_json_document
     }
 
     /// remove an array element
-    template < typename I, typename std::enable_if < std::is_integral<I>::value && !std::is_same<I, bool>::value, int >::type = 0 >
+    template<typename I, typename std::enable_if<detail::view::is_non_bool_integral<I>::value, int>::type = 0>
     void erase(view_type array, I idx)
     {
         editor().erase(array, index(idx));

single_include/nlohmann/json_fwd.hpp

@@ -71,6 +71,14 @@
     #define JSON_USE_OBJECTS_FOR_ENUM_KEYED_MAPS 0
 #endif
 
+#ifndef JSON_DISABLE_STREAM_BULK_SCAN
+    #define JSON_DISABLE_STREAM_BULK_SCAN 0
+#endif
+
+#ifndef JSON_DISABLE_UNLOCKED_FILE_IO
+    #define JSON_DISABLE_UNLOCKED_FILE_IO 0
+#endif
+
 #if JSON_DIAGNOSTICS
     #define NLOHMANN_JSON_ABI_TAG_DIAGNOSTICS _diag
 #else
@@ -119,14 +127,38 @@
     #define NLOHMANN_JSON_ABI_TAG_OBJECTS_FOR_ENUM_KEYED_MAPS
 #endif
 
+#if JSON_DISABLE_STREAM_BULK_SCAN
+    #define NLOHMANN_JSON_ABI_TAG_DISABLE_STREAM_BULK_SCAN _dsbs
+#else
+    #define NLOHMANN_JSON_ABI_TAG_DISABLE_STREAM_BULK_SCAN
+#endif
+
+#if JSON_DISABLE_UNLOCKED_FILE_IO
+    #define NLOHMANN_JSON_ABI_TAG_DISABLE_UNLOCKED_FILE_IO _dufio
+#else
+    #define NLOHMANN_JSON_ABI_TAG_DISABLE_UNLOCKED_FILE_IO
+#endif
+
+#ifdef JSON_VIEW_NO_SIMD
+    #define NLOHMANN_JSON_ABI_TAG_VIEW_NO_SIMD _vns
+#else
+    #define NLOHMANN_JSON_ABI_TAG_VIEW_NO_SIMD
+#endif
+
+#ifdef JSON_VIEW_USE_SSSE3
+    #define NLOHMANN_JSON_ABI_TAG_VIEW_USE_SSSE3 _vssse3
+#else
+    #define NLOHMANN_JSON_ABI_TAG_VIEW_USE_SSSE3
+#endif
+
 #ifndef NLOHMANN_JSON_NAMESPACE_NO_VERSION
     #define NLOHMANN_JSON_NAMESPACE_NO_VERSION 0
 #endif
 
 // Construct the namespace ABI tags component
-#define NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c, d, e, f, g, h) json_abi ## a ## b ## c ## d ## e ## f ## g ## h
-#define NLOHMANN_JSON_ABI_TAGS_CONCAT(a, b, c, d, e, f, g, h) \
-    NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c, d, e, f, g, h)
+#define NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c, d, e, f, g, h, i, j, k, l) json_abi ## a ## b ## c ## d ## e ## f ## g ## h ## i ## j ## k ## l
+#define NLOHMANN_JSON_ABI_TAGS_CONCAT(a, b, c, d, e, f, g, h, i, j, k, l) \
+    NLOHMANN_JSON_ABI_TAGS_CONCAT_EX(a, b, c, d, e, f, g, h, i, j, k, l)
 
 #define NLOHMANN_JSON_ABI_TAGS                                       \
     NLOHMANN_JSON_ABI_TAGS_CONCAT(                                   \
@@ -137,7 +169,11 @@
             NLOHMANN_JSON_ABI_TAG_PRECISE_STREAM_POSITION,           \
             NLOHMANN_JSON_ABI_TAG_STRICT_NUL_HANDLING,               \
             NLOHMANN_JSON_ABI_TAG_STRICT_BINARY_UTF8,                \
-            NLOHMANN_JSON_ABI_TAG_OBJECTS_FOR_ENUM_KEYED_MAPS)
+            NLOHMANN_JSON_ABI_TAG_OBJECTS_FOR_ENUM_KEYED_MAPS,       \
+            NLOHMANN_JSON_ABI_TAG_DISABLE_STREAM_BULK_SCAN,          \
+            NLOHMANN_JSON_ABI_TAG_DISABLE_UNLOCKED_FILE_IO,          \
+            NLOHMANN_JSON_ABI_TAG_VIEW_NO_SIMD,                      \
+            NLOHMANN_JSON_ABI_TAG_VIEW_USE_SSSE3)
 
 // Construct the namespace version component
 #define NLOHMANN_JSON_NAMESPACE_VERSION_CONCAT_EX(major, minor, patch) \

tests/CMakeLists.txt

@@ -285,6 +285,10 @@ if (JSON_TestSimdutf)
     endif()
 endif()
 
+# test-file_input checks from a second thread that parsing from a FILE* releases the file's lock
+find_package(Threads REQUIRED)
+json_test_set_test_options("test-file_input;test-file_input_unlocked_io_disabled" LINK_LIBRARIES Threads::Threads)
+
 # *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}")
@@ -369,6 +373,33 @@ json_test_add_test_for(src/unit-diagnostic-positions.cpp
     MAIN test_main CXX_STANDARDS ${test_cxx_standards} ${test_force}
 )
 
+# test reading from a std::istream again without scanning the get area in bulk
+json_test_set_test_options(test-stream-bulk-scan_disabled
+    COMPILE_DEFINITIONS JSON_DISABLE_STREAM_BULK_SCAN=1
+)
+json_test_add_test_for(src/unit-stream-bulk-scan.cpp
+    NAME test-stream-bulk-scan_disabled
+    MAIN test_main CXX_STANDARDS ${test_cxx_standards} ${test_force}
+)
+
+# test reading from a std::istream again with JSON_PRECISE_STREAM_POSITION
+json_test_set_test_options(test-stream-bulk-scan_precise
+    COMPILE_DEFINITIONS JSON_PRECISE_STREAM_POSITION=1
+)
+json_test_add_test_for(src/unit-stream-bulk-scan.cpp
+    NAME test-stream-bulk-scan_precise
+    MAIN test_main CXX_STANDARDS ${test_cxx_standards} ${test_force}
+)
+
+# test reading from a FILE* again with the locking stdio functions
+json_test_set_test_options(test-file_input_unlocked_io_disabled
+    COMPILE_DEFINITIONS JSON_DISABLE_UNLOCKED_FILE_IO=1
+)
+json_test_add_test_for(src/unit-file_input.cpp
+    NAME test-file_input_unlocked_io_disabled
+    MAIN test_main CXX_STANDARDS ${test_cxx_standards} ${test_force}
+)
+
 # the json_view parser again with the portable string scanning instead of
 # NEON/SSE2, and on x86-64 with the SSSE3 UTF-8 check (JSON_VIEW_USE_SSSE3)
 json_test_set_test_options(test-json_view_builder_portable

tests/abi/config/CMakeLists.txt

@@ -21,7 +21,11 @@ foreach(test default noversion)
     target_compile_definitions(abi_config_${test}_all_tags PRIVATE
         JSON_DIAGNOSTICS=1
         JSON_DIAGNOSTIC_POSITIONS=1
-        JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON=1)
+        JSON_USE_LEGACY_DISCARDED_VALUE_COMPARISON=1
+        JSON_DISABLE_STREAM_BULK_SCAN=1
+        JSON_DISABLE_UNLOCKED_FILE_IO=1
+        JSON_VIEW_NO_SIMD
+        JSON_VIEW_USE_SSSE3)
     target_link_libraries(abi_config_${test}_all_tags PRIVATE abi_compat_main)
     add_test(
         NAME test-abi_config_${test}_all_tags

tests/abi/config/default.cpp

@@ -52,6 +52,22 @@ TEST_CASE("default namespace")
         expected += "_ekmo";
 #endif
 
+#if JSON_DISABLE_STREAM_BULK_SCAN
+        expected += "_dsbs";
+#endif
+
+#if JSON_DISABLE_UNLOCKED_FILE_IO
+        expected += "_dufio";
+#endif
+
+#ifdef JSON_VIEW_NO_SIMD
+        expected += "_vns";
+#endif
+
+#ifdef JSON_VIEW_USE_SSSE3
+        expected += "_vssse3";
+#endif
+
         expected += "_v" STRINGIZE(NLOHMANN_JSON_VERSION_MAJOR);
         expected += "_" STRINGIZE(NLOHMANN_JSON_VERSION_MINOR);
         expected += "_" STRINGIZE(NLOHMANN_JSON_VERSION_PATCH) "::basic_json";

tests/abi/config/noversion.cpp

@@ -53,6 +53,22 @@ TEST_CASE("default namespace without version component")
         expected += "_ekmo";
 #endif
 
+#if JSON_DISABLE_STREAM_BULK_SCAN
+        expected += "_dsbs";
+#endif
+
+#if JSON_DISABLE_UNLOCKED_FILE_IO
+        expected += "_dufio";
+#endif
+
+#ifdef JSON_VIEW_NO_SIMD
+        expected += "_vns";
+#endif
+
+#ifdef JSON_VIEW_USE_SSSE3
+        expected += "_vssse3";
+#endif
+
         expected += "::basic_json";
 
         // fallback for Clang

tests/benchmarks/README.md

@@ -22,6 +22,15 @@ Micro-benchmarks for parsing, serialization and the binary formats, written with
 | `ViewAccept` | validate with `json_document::accept`; compare with `Accept` |
 | `ViewMaterialize` | convert a parsed `json_document` into a `json` value |
 | `ViewDump` | serialize a parsed `json_document`; compare with `Dump` |
+| `SaxParse` | parse with a SAX handler that only counts events, so no values are built |
+| `Copy`, `CopyAssign`, `CopyRing` | copy a parsed document and destroy a copy, both timed: a temporary copy, an assignment over an earlier copy, and replacing the oldest of eight live copies |
+| `CopyOnly`, `DestroyOnly` | only the copy or only the destruction, the other half untimed; a fast destructor can leave work to the next allocations, so these tell the halves apart but do not compare versions |
+| `Equal`, `Iterate` | compare a parsed document with a copy, or visit every value with range-based `for` loops |
+| `Flatten`, `FlattenOrdered`, `Unflatten` | `flatten()` for `json` and `ordered_json`, and `unflatten()` |
+| `Diff`, `Patch` | `diff()` against the document with every eighth scalar changed, and applying that patch |
+| `JsonPointerParse`, `JsonPointerToString`, `JsonPointerAt`, `JsonPointerContains` | the JSON pointers of every scalar in a document: parse them from strings, turn them back into strings, look them up with `at()` and `contains()` |
+| `GetVectorInt`, `GetVectorDouble`, `GetMap` | `get<std::vector<std::int64_t>>()`, `get<std::vector<double>>()` and `get<std::map<std::string, json>>()` on large arrays and a large flat object |
+| `ToJsonStrings` | create a `json` array from a `std::vector` of 100,000 strings, copied or moved, and destroy it again |
 
 The input files are those of [nativejson-benchmark](https://github.com/miloyip/nativejson-benchmark) (`canada`,
 `citm_catalog`, `twitter`), a large `jeopardy` file, and number-heavy files (`floats`, `signed_ints`, ...).

tests/benchmarks/src/benchmarks.cpp

@@ -8,11 +8,16 @@
 
 #include <benchmark/benchmark.h>
 #include <nlohmann/json.hpp>
-#include <fstream>
 #include <array>
+#include <cstdint>
+#include <cstdio>
 #include <deque>
+#include <fstream>
+#include <map>
 #include <numeric>
 #include <set>
+#include <sstream>
+#include <string>
 #include <vector>
 #include <test_data.hpp>
 
@@ -51,6 +56,65 @@ BENCHMARK_CAPTURE(ParseFile, signed_ints,       TEST_DATA_DIRECTORY "/regression
 BENCHMARK_CAPTURE(ParseFile, unsigned_ints,     TEST_DATA_DIRECTORY "/regression/unsigned_ints.json");
 BENCHMARK_CAPTURE(ParseFile, small_signed_ints, TEST_DATA_DIRECTORY "/regression/small_signed_ints.json");
 
+//////////////////////////////////////////////////////////////////////////////
+// parse JSON from a FILE*
+//////////////////////////////////////////////////////////////////////////////
+
+static void ParseFILE(benchmark::State& state, const char* filename)
+{
+    while (state.KeepRunning())
+    {
+        state.PauseTiming();
+        std::FILE* f = std::fopen(filename, "rb");
+        auto* j = new json();
+        state.ResumeTiming();
+
+        *j = json::parse(f);
+
+        state.PauseTiming();
+        std::fclose(f);
+        delete j;
+        state.ResumeTiming();
+    }
+
+    std::ifstream file(filename, std::ios::binary | std::ios::ate);
+    state.SetBytesProcessed(state.iterations() * file.tellg());
+}
+BENCHMARK_CAPTURE(ParseFILE, canada,       TEST_DATA_DIRECTORY "/nativejson-benchmark/canada.json");
+BENCHMARK_CAPTURE(ParseFILE, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(ParseFILE, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+//////////////////////////////////////////////////////////////////////////////
+// read many small values from one stream (NDJSON)
+//////////////////////////////////////////////////////////////////////////////
+
+static void ParseStreamValues(benchmark::State& state, bool numbers)
+{
+    std::string input;
+    for (int i = 0; i < 10000; ++i)
+    {
+        input += numbers ? std::to_string(i * 7919) + " "
+                 : R"({"id":)" + std::to_string(i) + R"(,"name":"user)" + std::to_string(i) + R"(","active":true,"score":)" + std::to_string(i * 0.25) + "}\n";
+    }
+
+    for (auto _ : state)
+    {
+        std::istringstream is(input);
+        json j;
+        std::size_t count = 0;
+        while (!(is >> std::ws).eof())
+        {
+            is >> j;
+            ++count;
+        }
+        benchmark::DoNotOptimize(count);
+    }
+
+    state.SetBytesProcessed(static_cast<int64_t>(state.iterations() * input.size()));
+}
+BENCHMARK_CAPTURE(ParseStreamValues, ndjson,  false);
+BENCHMARK_CAPTURE(ParseStreamValues, numbers, true);
+
 //////////////////////////////////////////////////////////////////////////////
 // parse JSON from string
 //////////////////////////////////////////////////////////////////////////////
@@ -649,6 +713,584 @@ static void FromCborChunkedString(benchmark::State& state, const std::size_t chu
 
 BENCHMARK_CAPTURE(FromCborChunkedString, 10000 chunks, 10000);
 
+//////////////////////////////////////////////////////////////////////////////
+// SAX-parse JSON from string
+//
+// sax_parse() with a handler that does next to nothing runs the lexer and
+// parser without building values, so the gap to ParseString above is the cost
+// of building the DOM.
+//////////////////////////////////////////////////////////////////////////////
+
+static std::string read_file(const char* filename)
+{
+    std::ifstream f(filename);
+    return std::string((std::istreambuf_iterator<char>(f)), std::istreambuf_iterator<char>());
+}
+
+static json load_json(const char* filename)
+{
+    return json::parse(read_file(filename));
+}
+
+/// SAX handler that only counts the events, so that the handler costs next to nothing
+class counting_sax : public nlohmann::json_sax<json>
+{
+  public:
+    std::size_t events = 0;
+
+    bool null() override
+    {
+        return count();
+    }
+    bool boolean(bool /*val*/) override
+    {
+        return count();
+    }
+    bool number_integer(number_integer_t /*val*/) override
+    {
+        return count();
+    }
+    bool number_unsigned(number_unsigned_t /*val*/) override
+    {
+        return count();
+    }
+    bool number_float(number_float_t /*val*/, const string_t& /*s*/) override
+    {
+        return count();
+    }
+    bool string(string_t& /*val*/) override
+    {
+        return count();
+    }
+    bool binary(binary_t& /*val*/) override
+    {
+        return count();
+    }
+    bool start_object(std::size_t /*elements*/) override
+    {
+        return count();
+    }
+    bool key(string_t& /*val*/) override
+    {
+        return count();
+    }
+    bool end_object() override
+    {
+        return count();
+    }
+    bool start_array(std::size_t /*elements*/) override
+    {
+        return count();
+    }
+    bool end_array() override
+    {
+        return count();
+    }
+    bool parse_error(std::size_t /*position*/, const std::string& /*last_token*/, const json::exception& /*ex*/) override
+    {
+        return false;
+    }
+
+  private:
+    bool count()
+    {
+        ++events;
+        return true;
+    }
+};
+
+static void SaxParse(benchmark::State& state, const char* filename)
+{
+    const std::string str = read_file(filename);
+
+    for (auto _ : state)
+    {
+        counting_sax sax;
+        bool result = json::sax_parse(str, &sax);
+        benchmark::DoNotOptimize(result);
+        benchmark::DoNotOptimize(sax.events);
+    }
+
+    state.SetBytesProcessed(state.iterations() * str.size());
+}
+BENCHMARK_CAPTURE(SaxParse, canada,       TEST_DATA_DIRECTORY "/nativejson-benchmark/canada.json");
+BENCHMARK_CAPTURE(SaxParse, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(SaxParse, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+//////////////////////////////////////////////////////////////////////////////
+// work with values: copy, destroy, compare, iterate
+//
+// The parsed test files stand in for a document an application keeps around.
+// Each benchmark times one whole-document operation; setup and cleanup run
+// outside the timed section.
+//
+// A copy is always destroyed again, and how a document is freed shapes the
+// heap the next copy allocates from: a destructor that frees quickly can leave
+// work to the allocator's next calls. Copy, CopyAssign and CopyRing therefore
+// time the copy together with the destruction of a copy; CopyOnly and
+// DestroyOnly time one half each and are meant for telling the halves apart,
+// not for comparing versions.
+//////////////////////////////////////////////////////////////////////////////
+
+/// copy a document and destroy the copy again, as a temporary copy is used
+static void Copy(benchmark::State& state, const char* filename)
+{
+    const json j = load_json(filename);
+
+    for (auto _ : state)
+    {
+        json copy = j;
+        benchmark::DoNotOptimize(copy);
+    }
+}
+BENCHMARK_CAPTURE(Copy, canada,       TEST_DATA_DIRECTORY "/nativejson-benchmark/canada.json");
+BENCHMARK_CAPTURE(Copy, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(Copy, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+/// copy-assign a document over an earlier copy of it, which is freed in the process
+static void CopyAssign(benchmark::State& state, const char* filename)
+{
+    const json j = load_json(filename);
+    json target = j;
+
+    for (auto _ : state)
+    {
+        target = j;
+        benchmark::DoNotOptimize(target);
+    }
+}
+BENCHMARK_CAPTURE(CopyAssign, canada,       TEST_DATA_DIRECTORY "/nativejson-benchmark/canada.json");
+BENCHMARK_CAPTURE(CopyAssign, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(CopyAssign, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+/// keep eight copies alive and replace the oldest one, as a cache or an undo
+/// history does, so that the copy is not made from memory just freed by an
+/// identical document
+static void CopyRing(benchmark::State& state, const char* filename)
+{
+    const json j = load_json(filename);
+    std::vector<json> ring(8, j);
+    std::size_t oldest = 0;
+
+    for (auto _ : state)
+    {
+        ring[oldest] = j;
+        benchmark::DoNotOptimize(ring[oldest]);
+        oldest = (oldest + 1) % ring.size();
+    }
+}
+BENCHMARK_CAPTURE(CopyRing, canada,       TEST_DATA_DIRECTORY "/nativejson-benchmark/canada.json");
+BENCHMARK_CAPTURE(CopyRing, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(CopyRing, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+/// only the copy; the copy is destroyed with the timer paused (diagnostic)
+static void CopyOnly(benchmark::State& state, const char* filename)
+{
+    const json j = load_json(filename);
+
+    for (auto _ : state)
+    {
+        auto* copy = new json(j);
+
+        state.PauseTiming();
+        delete copy;
+        state.ResumeTiming();
+    }
+}
+BENCHMARK_CAPTURE(CopyOnly, canada,       TEST_DATA_DIRECTORY "/nativejson-benchmark/canada.json");
+BENCHMARK_CAPTURE(CopyOnly, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(CopyOnly, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+/// only the destruction; the copy is made with the timer paused (diagnostic)
+static void DestroyOnly(benchmark::State& state, const char* filename)
+{
+    const json j = load_json(filename);
+
+    for (auto _ : state)
+    {
+        state.PauseTiming();
+        auto* copy = new json(j);
+        state.ResumeTiming();
+
+        delete copy;
+    }
+}
+BENCHMARK_CAPTURE(DestroyOnly, canada,       TEST_DATA_DIRECTORY "/nativejson-benchmark/canada.json");
+BENCHMARK_CAPTURE(DestroyOnly, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(DestroyOnly, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+static void Equal(benchmark::State& state, const char* filename)
+{
+    // two separate values, so that the comparison cannot stop at identical addresses
+    const json j = load_json(filename);
+    const json copy = j;
+
+    for (auto _ : state)
+    {
+        bool result = (j == copy);
+        benchmark::DoNotOptimize(result);
+    }
+}
+BENCHMARK_CAPTURE(Equal, canada,       TEST_DATA_DIRECTORY "/nativejson-benchmark/canada.json");
+BENCHMARK_CAPTURE(Equal, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(Equal, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+/// visit every value with range-based for loops, as client code walks a document
+static std::size_t count_values(const json& j)
+{
+    std::size_t result = 1;
+    if (j.is_structured())
+    {
+        for (const auto& element : j)
+        {
+            result += count_values(element);
+        }
+    }
+    return result;
+}
+
+static void Iterate(benchmark::State& state, const char* filename)
+{
+    const json j = load_json(filename);
+
+    for (auto _ : state)
+    {
+        std::size_t result = count_values(j);
+        benchmark::DoNotOptimize(result);
+    }
+
+    state.SetItemsProcessed(state.iterations() * static_cast<std::int64_t>(count_values(j)));
+}
+BENCHMARK_CAPTURE(Iterate, canada,       TEST_DATA_DIRECTORY "/nativejson-benchmark/canada.json");
+BENCHMARK_CAPTURE(Iterate, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(Iterate, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+//////////////////////////////////////////////////////////////////////////////
+// flatten and unflatten
+//////////////////////////////////////////////////////////////////////////////
+
+static void Flatten(benchmark::State& state, const char* filename)
+{
+    const json j = load_json(filename);
+
+    for (auto _ : state)
+    {
+        json flat = j.flatten();
+        benchmark::DoNotOptimize(flat);
+    }
+}
+BENCHMARK_CAPTURE(Flatten, canada,       TEST_DATA_DIRECTORY "/nativejson-benchmark/canada.json");
+BENCHMARK_CAPTURE(Flatten, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(Flatten, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+// ordered_json objects look keys up linearly, which flatten() used to do for
+// every value; canada (330,000 values in one flat object) is left out, as it
+// would take minutes with such versions
+static void FlattenOrdered(benchmark::State& state, const char* filename)
+{
+    const nlohmann::ordered_json j = nlohmann::ordered_json::parse(read_file(filename));
+
+    for (auto _ : state)
+    {
+        nlohmann::ordered_json flat = j.flatten();
+        benchmark::DoNotOptimize(flat);
+    }
+}
+BENCHMARK_CAPTURE(FlattenOrdered, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(FlattenOrdered, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+static void Unflatten(benchmark::State& state, const char* filename)
+{
+    const json flat = load_json(filename).flatten();
+
+    for (auto _ : state)
+    {
+        json j = flat.unflatten();
+        benchmark::DoNotOptimize(j);
+    }
+}
+BENCHMARK_CAPTURE(Unflatten, canada,       TEST_DATA_DIRECTORY "/nativejson-benchmark/canada.json");
+BENCHMARK_CAPTURE(Unflatten, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(Unflatten, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+//////////////////////////////////////////////////////////////////////////////
+// diff and patch
+//
+// The target is the document with every eighth scalar changed, so the patch
+// has one "replace" operation per changed value and the structure stays the
+// same.
+//////////////////////////////////////////////////////////////////////////////
+
+static void change_every_eighth_scalar(json& j, std::size_t& counter)
+{
+    if (j.is_structured())
+    {
+        for (auto& element : j)
+        {
+            change_every_eighth_scalar(element, counter);
+        }
+        return;
+    }
+
+    if (++counter % 8 != 0)
+    {
+        return;
+    }
+
+    switch (j.type())
+    {
+        case json::value_t::number_integer:
+            j = j.get<json::number_integer_t>() + 1;
+            break;
+        case json::value_t::number_unsigned:
+            j = j.get<json::number_unsigned_t>() + 1;
+            break;
+        case json::value_t::number_float:
+            j = j.get<json::number_float_t>() + 0.5;
+            break;
+        case json::value_t::string:
+            j = j.get<std::string>() + "!";
+            break;
+        case json::value_t::boolean:
+            j = !j.get<bool>();
+            break;
+        case json::value_t::null:
+        case json::value_t::object:
+        case json::value_t::array:
+        case json::value_t::binary:
+        case json::value_t::discarded:
+        default:
+            j = 0;
+            break;
+    }
+}
+
+static json changed_copy(const json& j)
+{
+    json result = j;
+    std::size_t counter = 0;
+    change_every_eighth_scalar(result, counter);
+    return result;
+}
+
+static void Diff(benchmark::State& state, const char* filename)
+{
+    const json source = load_json(filename);
+    const json target = changed_copy(source);
+
+    for (auto _ : state)
+    {
+        json patch = json::diff(source, target);
+        benchmark::DoNotOptimize(patch);
+    }
+
+    state.SetItemsProcessed(state.iterations() * static_cast<std::int64_t>(json::diff(source, target).size()));
+}
+BENCHMARK_CAPTURE(Diff, canada,       TEST_DATA_DIRECTORY "/nativejson-benchmark/canada.json");
+BENCHMARK_CAPTURE(Diff, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(Diff, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+static void Patch(benchmark::State& state, const char* filename)
+{
+    const json source = load_json(filename);
+    const json patch = json::diff(source, changed_copy(source));
+
+    for (auto _ : state)
+    {
+        json result = source.patch(patch);
+        benchmark::DoNotOptimize(result);
+    }
+
+    state.SetItemsProcessed(state.iterations() * static_cast<std::int64_t>(patch.size()));
+}
+BENCHMARK_CAPTURE(Patch, canada,       TEST_DATA_DIRECTORY "/nativejson-benchmark/canada.json");
+BENCHMARK_CAPTURE(Patch, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(Patch, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+//////////////////////////////////////////////////////////////////////////////
+// JSON Pointer
+//
+// The pointers are those of every scalar in the document, i.e., the keys of
+// its flattened form.
+//////////////////////////////////////////////////////////////////////////////
+
+static std::vector<std::string> pointer_strings(const json& j)
+{
+    std::vector<std::string> result;
+    const json flat = j.flatten();
+    result.reserve(flat.size());
+    for (auto it = flat.begin(); it != flat.end(); ++it)
+    {
+        result.push_back(it.key());
+    }
+    return result;
+}
+
+static std::vector<json::json_pointer> pointers(const json& j)
+{
+    const std::vector<std::string> strings = pointer_strings(j);
+    std::vector<json::json_pointer> result;
+    result.reserve(strings.size());
+    for (const auto& s : strings)
+    {
+        result.emplace_back(s);
+    }
+    return result;
+}
+
+static void JsonPointerParse(benchmark::State& state, const char* filename)
+{
+    const std::vector<std::string> strings = pointer_strings(load_json(filename));
+
+    for (auto _ : state)
+    {
+        for (const auto& s : strings)
+        {
+            json::json_pointer ptr(s);
+            benchmark::DoNotOptimize(ptr);
+        }
+    }
+
+    state.SetItemsProcessed(state.iterations() * static_cast<std::int64_t>(strings.size()));
+}
+BENCHMARK_CAPTURE(JsonPointerParse, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(JsonPointerParse, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+static void JsonPointerToString(benchmark::State& state, const char* filename)
+{
+    const std::vector<json::json_pointer> ptrs = pointers(load_json(filename));
+
+    for (auto _ : state)
+    {
+        for (const auto& ptr : ptrs)
+        {
+            std::string s = ptr.to_string();
+            benchmark::DoNotOptimize(s);
+        }
+    }
+
+    state.SetItemsProcessed(state.iterations() * static_cast<std::int64_t>(ptrs.size()));
+}
+BENCHMARK_CAPTURE(JsonPointerToString, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(JsonPointerToString, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+static void JsonPointerAt(benchmark::State& state, const char* filename)
+{
+    const json j = load_json(filename);
+    const std::vector<json::json_pointer> ptrs = pointers(j);
+
+    for (auto _ : state)
+    {
+        for (const auto& ptr : ptrs)
+        {
+            const json& value = j.at(ptr);
+            benchmark::DoNotOptimize(&value);
+        }
+    }
+
+    state.SetItemsProcessed(state.iterations() * static_cast<std::int64_t>(ptrs.size()));
+}
+BENCHMARK_CAPTURE(JsonPointerAt, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(JsonPointerAt, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+static void JsonPointerContains(benchmark::State& state, const char* filename)
+{
+    const json j = load_json(filename);
+    const std::vector<json::json_pointer> ptrs = pointers(j);
+
+    for (auto _ : state)
+    {
+        for (const auto& ptr : ptrs)
+        {
+            bool result = j.contains(ptr);
+            benchmark::DoNotOptimize(result);
+        }
+    }
+
+    state.SetItemsProcessed(state.iterations() * static_cast<std::int64_t>(ptrs.size()));
+}
+BENCHMARK_CAPTURE(JsonPointerContains, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+BENCHMARK_CAPTURE(JsonPointerContains, twitter,      TEST_DATA_DIRECTORY "/nativejson-benchmark/twitter.json");
+
+//////////////////////////////////////////////////////////////////////////////
+// conversions from and to C++ containers
+//////////////////////////////////////////////////////////////////////////////
+
+static void GetVectorInt(benchmark::State& state, const char* filename)
+{
+    const json j = load_json(filename);
+
+    for (auto _ : state)
+    {
+        auto v = j.get<std::vector<std::int64_t>>();
+        benchmark::DoNotOptimize(v);
+    }
+
+    state.SetItemsProcessed(state.iterations() * static_cast<std::int64_t>(j.size()));
+}
+BENCHMARK_CAPTURE(GetVectorInt, signed_ints, TEST_DATA_DIRECTORY "/regression/signed_ints.json");
+
+static void GetVectorDouble(benchmark::State& state, const char* filename)
+{
+    const json j = load_json(filename);
+
+    for (auto _ : state)
+    {
+        auto v = j.get<std::vector<double>>();
+        benchmark::DoNotOptimize(v);
+    }
+
+    state.SetItemsProcessed(state.iterations() * static_cast<std::int64_t>(j.size()));
+}
+BENCHMARK_CAPTURE(GetVectorDouble, floats, TEST_DATA_DIRECTORY "/regression/floats.json");
+
+// a large flat object: the flattened document, one member per scalar
+static void GetMap(benchmark::State& state, const char* filename)
+{
+    const json j = load_json(filename).flatten();
+
+    for (auto _ : state)
+    {
+        auto m = j.get<std::map<std::string, json>>();
+        benchmark::DoNotOptimize(m);
+    }
+
+    state.SetItemsProcessed(state.iterations() * static_cast<std::int64_t>(j.size()));
+}
+BENCHMARK_CAPTURE(GetMap, citm_catalog, TEST_DATA_DIRECTORY "/nativejson-benchmark/citm_catalog.json");
+
+/// 100,000 strings, each too long for the small string optimization
+static std::vector<std::string> make_strings()
+{
+    std::vector<std::string> result;
+    result.reserve(100000);
+    for (std::size_t i = 0; i < 100000; ++i)
+    {
+        result.push_back("a string that does not fit in place: " + std::to_string(i));
+    }
+    return result;
+}
+
+static void ToJsonStrings(benchmark::State& state, const bool move)
+{
+    const std::vector<std::string> strings = make_strings();
+
+    for (auto _ : state)
+    {
+        state.PauseTiming();
+        std::vector<std::string> input = strings;
+        state.ResumeTiming();
+
+        // destroyed inside the timed section, like the copies above
+        auto* j = move ? new json(std::move(input)) : new json(input);
+        delete j;
+    }
+
+    state.SetItemsProcessed(state.iterations() * static_cast<std::int64_t>(strings.size()));
+}
+BENCHMARK_CAPTURE(ToJsonStrings, copy, false);
+BENCHMARK_CAPTURE(ToJsonStrings, move, true);
+
 // conversion of a JSON array into standard containers
 template<typename Container>
 static void GetArray(benchmark::State& state)