🚶 fix format

Signed-off-by: Niels Lohmann <mail@nlohmann.me>
📡 Fix documentation gaps found in a full GitHub Discussions review
2026-07-10 12:35:09 +00:00 · 2026-07-10 13:38:52 +02:00 · 2026-07-10 12:01:11 +02:00 · 2026-07-09 23:05:56 +02:00 · 2026-07-09 20:57:41 +02:00 · 2026-07-09 20:39:59 +02:00
75 changed files with 759 additions and 324 deletions
@@ -234,11 +234,22 @@ jobs:

  ci_cuda_example:
    runs-on: ubuntu-latest
-    container: ghcr.io/nlohmann/json-ci:v2.4.0
+    strategy:
+      fail-fast: false
+      matrix:
+        # 11.8.0: newest pre-C++20 CUDA release, exercises the C++17 fallback
+        # path (tests/cuda_example/CMakeLists.txt picks the standard per nvcc
+        # version); 12.1.1: permanent regression guard for #3907 (nvcc 12.0/12.1
+        # choke on enable_borrowed_range at C++20, fixed in 12.2); 12.6.3: recent
+        # CUDA/C++20 coverage.
+        cuda: ['11.8.0', '12.1.1', '12.6.3']
+    container: nvidia/cuda:${{ matrix.cuda }}-devel-ubuntu22.04
    steps:
      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
        with:
          persist-credentials: false
+      - name: Get latest CMake and ninja
+        uses: lukka/get-cmake@f5b8fbb4d77cec1acc5a5f9f0df4beffaf5d98d9 # v4.3.4
      - name: Run CMake
        run: cmake -S . -B build -DJSON_CI=On
      - name: Build
@@ -669,7 +669,6 @@ add_custom_target(ci_test_compiler_default
 add_custom_target(ci_cuda_example
    COMMAND ${CMAKE_COMMAND}
        -DCMAKE_BUILD_TYPE=Debug -GNinja
-        -DCMAKE_CUDA_HOST_COMPILER=g++-8
        -S${PROJECT_SOURCE_DIR}/tests/cuda_example -B${PROJECT_BINARY_DIR}/build_cuda_example
    COMMAND ${CMAKE_COMMAND} --build ${PROJECT_BINARY_DIR}/build_cuda_example
 )
@@ -720,6 +719,11 @@ add_custom_target(ci_icpx
 # to zero and does not honor NaN ordering; -Kieee restores strict IEEE 754 behavior
 # (needed for the dtoa/grisu and NaN-comparison code paths).
 #
+# -tp=px pins the target processor to the generic x86-64 baseline (SSE2-only) to avoid
+# a nvc++ 25.5 / LLVM issue: when nvc++ auto-detects -tp from the runner's CPU (e.g. -tp znver4),
+# certain attribute combinations trigger an llc instruction-selection crash on std::ldexp<unsigned>.
+# Pinning to px removes this variability and is robust to future llc/nvc++ updates.
+#
 # The following tests are excluded as they trigger known nvc++ 25.5 defects (not
 # library bugs); see https://github.com/nlohmann/json for tracking. Only the
 # affected language-standard variants are excluded so coverage is otherwise kept:
@@ -733,7 +737,7 @@ add_custom_target(ci_nvhpc
    COMMAND ${CMAKE_COMMAND}
        -DCMAKE_BUILD_TYPE=Debug -GNinja
        -DCMAKE_C_COMPILER=nvc -DCMAKE_CXX_COMPILER=nvc++
-        -DCMAKE_CXX_FLAGS=-Kieee
+        -DCMAKE_CXX_FLAGS="-Kieee;-tp=px"
        -DJSON_BuildTests=ON -DJSON_FastTests=ON
        -S${PROJECT_SOURCE_DIR} -B${PROJECT_BINARY_DIR}/build_nvhpc
    COMMAND ${CMAKE_COMMAND} --build ${PROJECT_BINARY_DIR}/build_nvhpc
@@ -5,8 +5,11 @@
 # -Wno-extra-semi-stmt            The library uses assert which triggers this warning.
 # -Wno-padded                     We do not care about padding warnings.
 # -Wno-covered-switch-default     All switches list all cases and a default case.
-# -Wno-unsafe-buffer-usage        Otherwise Doctest would not compile.
-# -Wno-missing-noreturn           We found no way to silence this warning otherwise, see PR #4871
+# -Wno-unsafe-buffer-usage        Pervasive: the library's own low-level numeric/buffer code
+#                                 (to_chars, serializer, lexer, binary reader/writer, input
+#                                 adapters, json_pointer) plus vendored Doctest itself (~208
+#                                 distinct sites measured 2026-07-08 on clang trunk) all use
+#                                 raw pointer arithmetic / libc string calls by necessity.

 set(CLANG_CXXFLAGS
    -Werror
@@ -18,5 +21,4 @@ set(CLANG_CXXFLAGS
    -Wno-padded
    -Wno-covered-switch-default
    -Wno-unsafe-buffer-usage
-    -Wno-missing-noreturn
 )
@@ -4,7 +4,7 @@
  "archive": "JSON_for_Modern_C++.tgz",
  "author": {
    "name": "Niels Lohmann",
-    "link": "https://twitter.com/nlohmann"
+    "link": "https://nlohmann.me"
  },
  "aliases": ["nlohmann/json"]
 }
@@ -109,7 +109,7 @@ A UTF-8 byte order mark is silently ignored.
 - Added in version 3.0.0.
 - Ignoring comments via `ignore_comments` added in version 3.9.0.
 - Changed [runtime assertion](../../features/assertions.md) in case of `FILE*` null pointers to exception in version 3.12.0.
- Added `ignore_trailing_commas` in version 3.12.x.
+- Added `ignore_trailing_commas` in version 3.13.0.

 !!! warning "Deprecation"

@@ -115,7 +115,22 @@ basic_json(basic_json&& other) noexcept;
   
    Function [`array()`](array.md) and [`object()`](object.md) force array and object creation from initializer lists,
    respectively.
-        
+
+    !!! warning "Brace initialization yields arrays"
+
+        Because this constructor takes an `initializer_list_t`, brace-initializing a `json`/`ordered_json` from
+        another `json` value wraps it in a single-element array rather than copying it:
+
+        ```cpp
+        json j1 = "hello";
+        json j2{j1};   // [!] j2 is ["hello"], NOT a copy of j1
+        json j3(j1);   // j3 is "hello" -- parentheses copy as expected
+        ```
+
+        See the FAQ entry on [brace initialization](../../home/faq.md#brace-initialization-yields-arrays) for the
+        full explanation, an opt-in macro to change this behavior, and how to explicitly create a single-element
+        array (`json::array({value})`) if that is what you want.
+
 6. Constructs a JSON array value by creating `cnt` copies of a passed value. In case `cnt` is `0`, an empty array is
   created.

@@ -37,6 +37,14 @@ represent a byte array in modern C++.
 `BinaryType`
 :   container type to store arrays

+    Although not formally expressed as a C++ concept, `BinaryType` must be default-constructible,
+    copy/move-constructible, and support `push_back()`, `.data()`, and `.size()`, because
+    [`byte_container_with_subtype`](../byte_container_with_subtype/index.md) derives directly from it. Its
+    `value_type` must additionally be exactly one byte wide (e.g., `std::uint8_t`/`char`/`std::byte`): the binary
+    serializers (CBOR, MessagePack, BSON, UBJSON) read and write the container's raw bytes via
+    `reinterpret_cast`, which is only correct for byte-sized elements -- a container like
+    `#!cpp std::vector<std::intptr_t>` will not work as `BinaryType`.
+
 ## Notes

 #### Default type
@@ -92,4 +92,4 @@ std::string format_as(const BasicJsonType& j)

 ## Version history

- Added in version 3.12.x.
+- Added in version 3.13.0.
@@ -114,6 +114,13 @@ overload (3).
    See [Number conversion](../../features/types/number_handling.md#number-conversion)
    for more information.

+!!! note "`std::optional` conversions"
+
+    Prior to version 3.13.0, `#!cpp get<std::optional<T>>()` (and other conversions to `std::optional<T>`) failed to
+    compile in every configuration, due to an internal implementation bug that made the `from_json` overload for
+    `std::optional` unreachable regardless of the [`JSON_USE_IMPLICIT_CONVERSIONS`](../macros/json_use_implicit_conversions.md)
+    setting. This has been fixed.
+
 ## Examples

 ??? example
@@ -46,6 +46,17 @@ for (auto& [key, val] : j_object.items())
 }
 ```

+If you need to name the type of the dereferenced element explicitly (e.g., to write a standalone function that
+takes it as a parameter, or to use `items()` with `std::for_each`), use `decltype`:
+
+```cpp
+using element_type = decltype(*j_object.items().begin());
+```
+
+The per-element type (`iteration_proxy_value`) lives in the library's internal `detail` namespace and is
+intentionally unspecified as a stable, named type -- `decltype` is the supported way to obtain it, but its exact
+name/definition may change between versions.
+
 ## Return value

 iteration proxy object wrapping the current value with an interface to use in range-based for loops
@@ -63,7 +63,8 @@ behavior:
  object will agree on the name-value mappings.
 - When the names within an object are not unique, it is unspecified which one of the values for a given key will be
  chosen. For instance, `#!json {"key": 2, "key": 1}` could be equal to either `#!json {"key": 1}` or
-  `#!json {"key": 2}`.
+  `#!json {"key": 2}`. To reject duplicate keys instead of silently resolving them one way or another, see
+  [this parsing recipe](../../features/parsing/parser_callbacks.md#recipe-rejecting-duplicate-object-keys).
 - Internally, name/value pairs are stored in lexicographical order of the names. Objects will also be serialized (see
  [`dump`](dump.md)) in this order. For instance, `#!json {"b": 1, "a": 2}` and `#!json {"a": 2, "b": 1}` will be stored
  and serialized as `#!json {"a": 2, "b": 1}`.
@@ -124,6 +124,15 @@ Strong exception safety: if an exception occurs, the original value stays intact
      filled with `#!json null`.
    - The special value `-` is treated as a synonym for the index past the end.

+    !!! note "Creating intermediate levels that don't exist yet"
+
+        When the JSON pointer traverses intermediate levels that don't exist at all yet (not just a missing
+        leaf), each missing level is created as an array or an object depending on whether the corresponding
+        pointer token parses as a non-negative integer: a numeric token creates an array, a non-numeric token
+        creates an object. For example, on an initially `#!json null` value, `/foo/0/0/0` creates nested arrays,
+        while `/foo/one/one/one` creates nested objects. This is not specified by the JSON Pointer RFC; it is
+        this library's own, intentional disambiguation rule. See also [JSON Pointer](../../features/json_pointer.md).
+
 ## Examples

 ??? example "Example: (1) access specified array element"
@@ -251,5 +260,6 @@ Strong exception safety: if an exception occurs, the original value stays intact
 1. Added in version 1.0.0.
 2. Added in version 1.0.0. Added overloads for `T* key` in version 1.1.0. Removed overloads for `T* key` (replaced by 3)
   in version 3.11.0.
-3. Added in version 3.11.0.
+3. Added in version 3.11.0. Fixed in version 3.13.0 to consistently accept `std::string_view`-convertible keys, as
+   already supported by [`at`](at.md), [`value`](value.md), [`find`](find.md), and other lookup functions.
 4. Added in version 2.0.0.
@@ -19,10 +19,8 @@ class basic_json {
 };
 ```

-1. Compares two JSON values for inequality according to the following rules:
-    - The comparison always yields `#!cpp false` if (1) either operand is discarded, or (2) either operand is `NaN` and
-      the other operand is either `NaN` or any other number.
-    - Otherwise, returns the result of `#!cpp !(lhs == rhs)` (until C++20) or `#!cpp !(*this == rhs)` (since C++20).
+1. Compares two JSON values for inequality. Returns `#!cpp !(lhs == rhs)` (until C++20) or `#!cpp !(*this == rhs)` (since C++20).
+    - This means the comparison is simply the logical negation of `operator==`, including for special values like `NaN` and `discarded`.

 2. Compares a JSON value and a scalar or a scalar and a JSON value for inequality by converting the scalar to a JSON
   value and comparing both JSON values according to 1.
@@ -54,13 +52,12 @@ Linear.

 ## Notes

-!!! note "Comparing `NaN`"
+!!! note "Comparing `NaN` and `discarded`"

-    `NaN` values are unordered within the domain of numbers.
-    The following comparisons all yield `#!cpp false`:
-      1. Comparing a `NaN` with itself.
-      2. Comparing a `NaN` with another `NaN`.
-      3. Comparing a `NaN` and any other number.
+    Since `operator!=` is defined as `!(a == b)`, the behavior for special values follows that of `operator==`:
+    
+    - For `NaN` values: `NaN == NaN` yields `#!cpp false`, so `NaN != NaN` yields `#!cpp true`.
+    - For `discarded` values: `discarded == x` yields `#!cpp false` for any `x`, so `discarded != x` yields `#!cpp true`.

 ## Examples

@@ -94,5 +91,7 @@ Linear.

 ## Version history

-1. Added in version 1.0.0. Added C++20 member functions in version 3.11.0.
-2. Added in version 1.0.0. Added C++20 member functions in version 3.11.0.
+1. Added in version 1.0.0. Added C++20 member functions in version 3.11.0. Changed in version 3.13.0 to remove 
+   special-casing for `NaN` and `discarded` values; `operator!=` now consistently means `!(a == b)`.
+2. Added in version 1.0.0. Added C++20 member functions in version 3.11.0. Changed in version 3.13.0 to remove
+   special-casing for `NaN` and `discarded` values; `operator!=` now consistently means `!(a == b)`.
@@ -235,7 +235,7 @@ Invalid Unicode escapes and unpaired surrogates in the input are reported as
 - Overload for contiguous containers (1) added in version 2.0.3.
 - Ignoring comments via `ignore_comments` added in version 3.9.0.
 - Changed [runtime assertion](../../features/assertions.md) in case of `FILE*` null pointers to exception in version 3.12.0.
- Added `ignore_trailing_commas` in version 3.12.x.
+- Added `ignore_trailing_commas` in version 3.13.0.

 !!! warning "Deprecation"

@@ -74,4 +74,4 @@ is thrown. In any case, the original value is not changed: the patch is applied

 - Added in version 2.0.0.
 - Added [`out_of_range.411`](../../home/exceptions.md#jsonexceptionout_of_range411) and stopped relying on an internal assertion when an "add" operation's
-  target location has a non-object/non-array parent in version 3.12.x.
+  target location has a non-object/non-array parent in version 3.13.0.
@@ -71,4 +71,4 @@ function throws an exception.

 - Added in version 3.11.0.
 - Added [`out_of_range.411`](../../home/exceptions.md#jsonexceptionout_of_range411) and stopped relying on an internal assertion when an "add" operation's
-  target location has a non-object/non-array parent in version 3.12.x.
+  target location has a non-object/non-array parent in version 3.13.0.
@@ -126,7 +126,7 @@ A UTF-8 byte order mark is silently ignored.

 - Added in version 3.2.0.
 - Ignoring comments via `ignore_comments` added in version 3.9.0.
- Added `ignore_trailing_commas` in version 3.12.x.
+- Added `ignore_trailing_commas` in version 3.13.0.

 !!! warning "Deprecation"

@@ -54,4 +54,4 @@ provides `<format>`, controlled by the [`JSON_HAS_STD_FORMAT`](../macros/json_ha

 ## Version history

- Added in version 3.12.x.
+- Added in version 3.13.0.
@@ -6,18 +6,9 @@ namespace std {
 }
 ```

-Return a hash value for a JSON object. The hash function tries to rely on `std::hash` where possible. To satisfy the
-`std::hash` contract, numeric JSON values that compare equal must hash to the same value. This means:
-
- `json(42)`, `json(42u)`, and `json(42.0)` all hash to the same value
- `json(0)`, `json(0u)`, and `json(0.0)` all hash to the same value
-
-Different types hash differently for non-numeric types (e.g., `#!json null`, `#!cpp false`, and strings all have distinct hashes).
-
-**Edge case:** For very large integers outside the exact representable range of the floating-point type (beyond ~2^53 for
-typical `double`), the hash values for integer and floating-point values may differ, even if the floating-point value
-was obtained by casting the integer (due to precision loss). This is a documented limitation arising from how the
-comparison operator normalizes numeric types.
+Return a hash value for a JSON object. The hash function tries to rely on `std::hash` where possible. Furthermore, the
+type of the JSON value is taken into account to have different hash values for `#!json null`, `#!cpp 0`, `#!cpp 0U`, and
+`#!cpp false`, etc.

 ## Examples

@@ -18,6 +18,11 @@ JSON class into byte-sized characters during deserialization.
 :   the container to store strings (e.g., `std::string`). Note this container is used for keys/names in objects, see
    [object_t](object_t.md).

+    `StringType` must have a `char`-compatible `value_type`: the library relies on UTF-8/`char`-based storage and
+    processing internally, so `std::wstring`, `std::u16string`, and `std::u32string` are **not** valid choices for
+    `StringType`. To work with wide-character data, convert it to/from UTF-8 at the boundary instead -- see the
+    FAQ's [wide string handling](../../home/faq.md#wide-string-handling) section for a conversion recipe.
+
 ## Notes

 #### Default type
@@ -21,6 +21,12 @@ a string representation of the type ([`value_t`](value_t.md)):
 | array                                              | `"array"`     |
 | binary                                             | `"binary"`    |
 | discarded                                          | `"discarded"` |
+| invalid (corrupted value)                          | `"invalid"`   |
+
+!!! note "The \"invalid\" type"
+
+    The `"invalid"` return value indicates a corrupted JSON value — this can occur if an enum value falls outside the
+    range of valid `value_t` values. This is useful for diagnosing data corruption or internal errors.

 ## Exception safety

@@ -52,3 +58,4 @@ Constant.
 - Part of the public API version since 2.1.0.
 - Changed return value to `const char*` and added `noexcept` in version 3.0.0.
 - Added support for binary type in version 3.8.0.
+- Added `"invalid"` return value for corrupted JSON values in version 3.13.0.
@@ -17,6 +17,8 @@ ValueType value(const json_pointer& ptr,
                const ValueType& default_value) const;
 ```

+This is equivalent to Python's `dict.get(key, default)`.
+
 1. Returns either a copy of an object's element at the specified key `key` or a given default value if no element with
   key `key` exists.
   
@@ -184,4 +186,6 @@ changes to any JSON value.

 1. Added in version 1.0.0. Changed parameter `default_value` type from `const ValueType&` to `ValueType&&` in version 3.11.0.
 2. Added in version 3.11.0. Made `ValueType` the first template parameter in version 3.11.2.
-3. Added in version 2.0.2. Extended to work with arrays in version 3.12.x.
+3. Added in version 2.0.2. Extended to work with arrays in version 3.13.0, including fixing an issue where resolving
+   `ptr` through an array unexpectedly threw `out_of_range` instead of returning the resolved element (or
+   `default_value`, as documented).
@@ -36,4 +36,4 @@ Constant.

 ## Version history

- Added in version 3.12.x.
+- Added in version 3.13.0.
@@ -32,4 +32,4 @@ Linear in the number of reference tokens in the `json_pointer`.

 ## Version history

- Added in version 3.12.x.
+- Added in version 3.13.0.
@@ -35,4 +35,4 @@ Linear in the number of reference tokens in the `json_pointer`.

 ## Version history

- Added in version 3.12.x.
+- Added in version 3.13.0.
@@ -92,4 +92,4 @@ The default value is `0` (disabled — existing behavior is preserved).

 ## Version history

- Added in version 3.12.x.
+- Added in version 3.13.0.
@@ -44,4 +44,4 @@ The default value is detected based on preprocessor macros such as `#!cpp __cplu

 - Added in version 3.10.5.
 - Added `JSON_HAS_CPP_23` in version 3.12.0.
- Added `JSON_HAS_CPP_26` in version 3.12.x.
+- Added `JSON_HAS_CPP_26` in version 3.13.0.
@@ -19,6 +19,20 @@ The default value is detected based on the preprocessor macros `#!cpp __cpp_lib_
 `#!cpp __cpp_lib_experimental_filesystem`, `#!cpp __has_include(<filesystem>)`, or
 `#!cpp __has_include(<experimental/filesystem>)`.

+!!! info "Known compiler/stdlib exclusions"
+
+    Even when the feature-test macro indicates filesystem support is available, the library disables it on the following broken toolchains:
+    
+    - **MinGW + GCC 8** — disabled entirely (broken `std::filesystem` implementation; [MinGW-w64 bug 737](https://sourceforge.net/p/mingw-w64/bugs/737/))
+    - **GCC (non-Clang) < 8** — disabled (no filesystem support)
+    - **Clang < 7** — disabled (no filesystem support)
+    - **MSVC < 19.14** — disabled (no filesystem support)
+    - **iOS < 13** — disabled (no filesystem support)
+    - **macOS < Catalina (10.15)** — disabled (no filesystem support)
+    
+    If `JSON_HAS_FILESYSTEM` or `JSON_HAS_EXPERIMENTAL_FILESYSTEM` is `0` despite `__cpp_lib_filesystem` being defined, one
+    of the exclusions above likely applies to your toolchain.
+
 ## Notes

 - Note that older compilers or older versions of libstdc++ also require the library `stdc++fs` to be linked to for
@@ -13,6 +13,20 @@ The default value is detected based on the preprocessor macro `#!cpp __cpp_lib_r

 When the macro is not defined, the library will define it to its default value.

+!!! info "Known compiler/stdlib exclusions"
+
+    Even when the feature-test macro `__cpp_lib_ranges` indicates ranges support is available, the library disables it on
+    the following incomplete or broken toolchains:
+    
+    - **GCC 11.1.0** — disabled (the shipped `<ranges>` header has a syntax error; [issue #4440](https://github.com/nlohmann/json/issues/4440))
+    - **libstdc++ < 11** — disabled (incomplete C++20 ranges support; [issue #4440](https://github.com/nlohmann/json/issues/4440))
+    - **Clang < 16 with libstdc++** — disabled (incomplete ranges support; [issue #4440](https://github.com/nlohmann/json/issues/4440))
+    - **libc++ < 160000** — disabled (incomplete C++20 ranges support; [issue #4440](https://github.com/nlohmann/json/issues/4440))
+    - **nvcc (CUDA) 12.0.x and 12.1.x** — disabled (the `enable_borrowed_range` variable-template syntax triggers a parse error
+      under these two toolkit versions; fixed in CUDA 12.2; [issue #3907](https://github.com/nlohmann/json/issues/3907))
+    
+    If `JSON_HAS_RANGES` is `0` despite `__cpp_lib_ranges` being defined, one of the exclusions above likely applies to your toolchain.
+
 ## Examples

 ??? example
@@ -38,4 +38,4 @@ When the macro is not defined, the library will define it to its default value.

 ## Version history

- Added in version 3.12.x.
+- Added in version 3.13.0.
@@ -62,6 +62,9 @@ See the examples below for the concrete generated code.

    - The current implementation is limited to at most 63 member variables. If you want to serialize/deserialize types
      with more than 63 member variables, you need to define the `to_json`/`from_json` functions manually.
+    - These macros always produce object-style (named-key) JSON, one key per member. There is no macro variant
+      that serializes a struct's members positionally into a JSON array; for that, write `to_json`/`from_json` by
+      hand, building/reading a `json::array()` of the members in order.

 ## Examples

@@ -63,6 +63,9 @@ See the examples below for the concrete generated code.

    - The current implementation is limited to at most 63 member variables. If you want to serialize/deserialize types
      with more than 63 member variables, you need to define the `to_json`/`from_json` functions manually.
+    - These macros always produce object-style (named-key) JSON, one key per member. There is no macro variant
+      that serializes a struct's members positionally into a JSON array; for that, write `to_json`/`from_json` by
+      hand, building/reading a `json::array()` of the members in order.

 ## Examples

@@ -75,4 +75,4 @@ For further information please refer to the corresponding macros without `WITH_N

 ## Version history

-1. Added in version 3.12.x.
+1. Added in version 3.13.0.
@@ -102,4 +102,4 @@ inline void from_json(const BasicJsonType& j, type& e);

 ## Version history

-Added in version 3.12.x.
+Added in version 3.13.0.
@@ -33,6 +33,18 @@ A UTF-8 byte order mark is silently ignored.
 Invalid Unicode escapes and unpaired surrogates in the input are reported as
 [`parse_error.101`](../home/exceptions.md#jsonexceptionparse_error101) with a detailed message.

+`operator>>` parses exactly one JSON value and leaves the stream positioned right after it, so it can be called
+repeatedly to read a sequence of concatenated JSON values from the same stream:
+
+```cpp
+json j1, j2;
+input >> j1;  // parses the first value, stream now positioned right after it
+input >> j2;  // parses the next value
+```
+
+Note this does **not** work for [JSON Lines](../features/parsing/json_lines.md) (newline-delimited JSON) input --
+see that page for why and for the recommended alternative.
+
 !!! warning "Deprecation"

    This function replaces function `#!cpp std::istream& operator<<(basic_json& j, std::istream& i)` which has
@@ -64,4 +64,4 @@ Linear.

 - Added in version 1.0.0.
 - Moved to namespace `nlohmann::literals::json_literals` in 3.11.0.
- Added `char8_t*` overload in 3.12.x.
+- Added `char8_t*` overload in 3.13.0.
@@ -63,4 +63,4 @@ Linear.

 - Added in version 2.0.0.
 - Moved to namespace `nlohmann::literals::json_literals` in 3.11.0.
- Added `char8_t*` overload in 3.12.x.
+- Added `char8_t*` overload in 3.13.0.
@@ -10,6 +10,10 @@ violations will result in a failed build.

    Any compiler with complete C++11 support can compile the library without warnings.

+Note: C++20 modules support may hit compiler-specific issues not covered by the general compiler matrix below. See [Modules](../features/modules.md#known-issues) for known issues and workarounds.
+
+Note: Some modern features (like C++20 ranges or filesystem support) may be disabled on specific broken or incomplete toolchains even when standard feature-test macros indicate support. See [`JSON_HAS_RANGES`](../api/macros/json_has_ranges.md) and [`JSON_HAS_FILESYSTEM`](../api/macros/json_has_filesystem.md) for details on known exclusions.
+
 - [x] The library is compiled with 50+ different C++ compilers with different operating systems and platforms,
  including the oldest versions known to compile the library.

@@ -62,7 +66,9 @@ violations will result in a failed build.
        | Clang 20.1.1                                 | x86_64       | Ubuntu 22.04.1 LTS                | GitHub    |
        | Clang 20.1.8 with GNU-like command-line      | x86_64       | Windows Server 2022 (Build 20348) | GitHub    |
        | Clang 21.1.8                                 | x86_64       | Ubuntu 22.04.1 LTS                | GitHub    |
-        | CUDA 11.0.221 (nvcc)                         | x86_64       | Ubuntu 20.04 LTS                  | GitHub    |
+        | CUDA 11.8.0 (nvcc)                           | x86_64       | Ubuntu 22.04 LTS                  | GitHub    |
+        | CUDA 12.1.1 (nvcc)                           | x86_64       | Ubuntu 22.04 LTS                  | GitHub    |
+        | CUDA 12.6.3 (nvcc)                           | x86_64       | Ubuntu 22.04 LTS                  | GitHub    |
        | Emscripten 4.0.6                             | x86_64       | Ubuntu 22.04.1 LTS                | GitHub    |
        | GNU 4.8.5                                    | x86_64       | Ubuntu 22.04.1 LTS                | GitHub    |
        | GNU 4.9.3                                    | x86_64       | Ubuntu 22.04.1 LTS                | GitHub    |
@@ -0,0 +1,61 @@
+#include <iostream>
+#include <nlohmann/json.hpp>
+#include <stdexcept>
+#include <string>
+#include <unordered_set>
+#include <vector>
+
+using json = nlohmann::json;
+
+json parse_strict(const std::string& input)
+{
+    // one key set per nesting depth, reused across sibling objects
+    std::vector<std::unordered_set<std::string>> keys;
+
+    auto reject_duplicate_keys = [&](int depth, json::parse_event_t event, json & parsed)
+    {
+        if (event == json::parse_event_t::object_start)
+        {
+            // keys of this object are reported at depth+1 (see the event table above)
+            const auto child_depth = static_cast<std::size_t>(depth) + 1;
+            if (keys.size() <= child_depth)
+            {
+                keys.resize(child_depth + 1);
+            }
+            keys[child_depth].clear();
+            return true;
+        }
+
+        if (event == json::parse_event_t::key)
+        {
+            auto& seen = keys[static_cast<std::size_t>(depth)];
+            const auto& key = parsed.get_ref<const std::string&>();
+            if (!seen.insert(key).second)
+            {
+                throw std::runtime_error("duplicate JSON object key: " + key);
+            }
+            return true;
+        }
+
+        return true;
+    };
+
+    return json::parse(input, reject_duplicate_keys);
+}
+
+int main()
+{
+    // parsing succeeds when all keys are unique
+    json j = parse_strict(R"({"one": 1, "two": 2})");
+    std::cout << j << '\n';
+
+    // parsing throws when a key is repeated
+    try
+    {
+        parse_strict(R"({"one": 1, "one": 2})");
+    }
+    catch (const std::exception& e)
+    {
+        std::cout << e.what() << '\n';
+    }
+}
@@ -0,0 +1,2 @@
+{"one":1,"two":2}
+duplicate JSON object key: one
@@ -11,7 +11,6 @@ int main()
              << "hash(false) = " << std::hash<json> {}(json(false)) << '\n'
              << "hash(0) = " << std::hash<json> {}(json(0)) << '\n'
              << "hash(0U) = " << std::hash<json> {}(json(0U)) << '\n'
-              << "hash(0.0) = " << std::hash<json> {}(json(0.0)) << '\n'
              << "hash(\"\") = " << std::hash<json> {}(json("")) << '\n'
              << "hash({}) = " << std::hash<json> {}(json::object()) << '\n'
              << "hash([]) = " << std::hash<json> {}(json::array()) << '\n'
@@ -1,9 +1,8 @@
 hash(null) = 2654435769
 hash(false) = 2654436030
-hash(0) = 2654436221
-hash(0U) = 2654436221
-hash(0.0) = 2654436221
-hash("") = 11160318156688833227
+hash(0) = 2654436095
+hash(0U) = 2654436156
+hash("") = 6142509191626859748
 hash({}) = 2654435832
 hash([]) = 2654435899
-hash({"hello": "world"}) = 3701319991624763853
+hash({"hello": "world"}) = 4469488738203676328
@@ -180,6 +180,49 @@ For _derived_ classes and structs, use the following macros
    }
    ```

+!!! warning "Overriding conversions for natively-supported types"
+
+    The library already provides built-in `to_json`/`from_json` conversions for STL containers such as
+    `std::vector`, `std::array`, and `std::map`. Defining your own free-function `to_json`/`from_json` overload
+    for one of these container types directly (instead of for your own type) can conflict with the built-in
+    overload during overload resolution, producing compiler errors ("no matching overloaded function",
+    "call is ambiguous") that vary by compiler and library version. If you need different conversion behavior
+    for a container type the library already handles, wrap it in your own type (or use `adl_serializer`
+    specialization, as shown [above](#how-do-i-convert-third-party-types) for `boost::optional`) instead of
+    trying to re-specialize `to_json`/`from_json` for the container type itself.
+
+!!! warning "Raw C-style arrays"
+
+    Members declared as raw C-style arrays (e.g., `char buf[1024]`) do not round-trip safely through
+    `NLOHMANN_DEFINE_TYPE_*` macros or the default (de)serializers: `to_json` serializes any `char` array as a
+    JSON *string* (matching the `std::string`-constructible overload), but the `from_json` overload for
+    fixed-size arrays expects a JSON *array* and iterates it element-wise, which fails with a `type_error` when
+    given a string. Use `std::string`, `std::array<char, N>`, or a manually written `to_json`/`from_json` pair
+    for such members instead.
+
+!!! note "Macros and `nlohmann::ordered_json`"
+
+    The `NLOHMANN_DEFINE_TYPE_*`/`NLOHMANN_DEFINE_DERIVED_TYPE_*` macros are generic over any `basic_json`
+    specialization, including `nlohmann::ordered_json`. Simply use `ordered_json` as the target type and members
+    are serialized in declaration order -- no separate macro or extra code is needed.
+
+    ```cpp
+    namespace ns {
+        NLOHMANN_DEFINE_TYPE_NON_INTRUSIVE(person, name, address, age)
+    }
+
+    ns::person p{"Ned Flanders", "744 Evergreen Terrace", 60};
+    nlohmann::ordered_json j = p; // keys appear in declaration order: name, address, age
+    ```
+
+!!! note "No macro for non-default-constructible types"
+
+    There is currently no `NLOHMANN_DEFINE_TYPE_*`-style macro for types that are not
+    [DefaultConstructible](https://en.cppreference.com/w/cpp/named_req/DefaultConstructible). This is not an
+    intentional omission of documentation -- no such macro exists yet; see
+    [How can I use `get()` for non-default constructible/non-copyable types?](#how-can-i-use-get-for-non-default-constructiblenon-copyable-types)
+    for the manual pattern to use instead.
+
 ## How do I convert third-party types?

 This requires a bit more advanced technique. But first, let us see how this conversion mechanism works:
@@ -270,6 +313,49 @@ namespace nlohmann {
 }
 ```

+## Why can't I convert to/from `std::any`?
+
+`std::any` is intentionally excluded from `get<T>()`/generic conversion support, so `get<std::any>()` and
+containers like `std::map<std::string, std::any>` fail to compile by design -- there is no way to know, from a
+`json` value alone, which concrete type to store inside the `std::any`. To work with heterogeneous JSON values,
+dispatch on the value's type manually and construct the `std::any` (or extract from it) yourself:
+
+```cpp
+std::any value_to_any(const json& j) {
+    if (j.is_boolean())      { return j.get<bool>(); }
+    if (j.is_number_integer()) { return j.get<int>(); }
+    if (j.is_number_float()) { return j.get<double>(); }
+    if (j.is_string())       { return j.get<std::string>(); }
+    // ... handle other types (arrays, objects) as needed for your use case
+    return {};
+}
+
+json any_to_json(const std::any& a) {
+    if (a.type() == typeid(bool))        { return std::any_cast<bool>(a); }
+    if (a.type() == typeid(int))         { return std::any_cast<int>(a); }
+    if (a.type() == typeid(double))      { return std::any_cast<double>(a); }
+    if (a.type() == typeid(std::string)) { return std::any_cast<std::string>(a); }
+    return nullptr;
+}
+```
+
+## Why does serializing a `std::map`/`std::unordered_map` with non-string keys produce an array?
+
+A `std::map`/`std::unordered_map` whose key type is not string-like (e.g., `std::map<int, std::string>`) is
+serialized as a JSON *array* of 2-element `[key, value]` arrays, not as a JSON object -- JSON object keys must be
+strings, so the library cannot represent an integer-keyed map as an object.
+
+```cpp
+std::map<int, std::string> m{{1, "one"}, {2, "two"}};
+json j = m;
+// j is [[1,"one"],[2,"two"]], not {"1":"one","2":"two"}
+```
+
+## Why does `std::wstring` convert or dump incorrectly?
+
+The library assumes UTF-8 encoding internally, so `std::wstring` is not supported out of the box -- see the FAQ
+entry on [wide string handling](../home/faq.md#wide-string-handling) for why, and for a UTF-8 conversion recipe.
+
 ## Can I write my own serializer? (Advanced use)

 Yes. You might want to take a look at [`unit-udt.cpp`](https://github.com/nlohmann/json/blob/develop/tests/src/unit-udt.cpp) in the test suite, to see a few examples.
@@ -66,7 +66,15 @@ see "binary" cells in the table above.

 !!! info "NaN/infinity handling"

-    If NaN or Infinity are stored inside a JSON number, they are serialized properly. This behavior differs from the normal JSON serialization which serializes NaN or Infinity to `null`.
+    `NaN`, `Infinity`, and `-Infinity` are serialized as a CBOR half-precision float (type 0xF9, 3 bytes total):
+    `NaN` as `0xF9 0x7E 0x00`, `Infinity` as `0xF9 0x7C 0x00`, and `-Infinity` as `0xF9 0xFC 0x00`. This behavior
+    differs from the normal JSON serialization which serializes NaN or Infinity to `null`.
+
+!!! note
+
+    Prior to version 3.13.0, NaN and Infinity were instead serialized as a CBOR double-precision float (type 0xFB,
+    9 bytes total), because the check used to select a smaller encoding compared magnitudes with NaN, which is
+    always `false` and caused the intended half-precision path to be skipped.

 !!! info "Unused CBOR types"

@@ -160,6 +168,13 @@ The library maps CBOR types to JSON value types as follows:
     - simple values (0xE0..0xF3, 0xF8)
     - undefined (0xF7)

+!!! warning "Negative integer overflow"
+
+    CBOR negative integers (major type 1) are decoded as `-1 - n`. If the encoded magnitude `n` is too large for the
+    result to fit into `number_integer_t` (`std::int64_t` by default), parsing fails with a
+    [`parse_error.112`](../../home/exceptions.md#jsonexceptionparse_error112) exception rather than overflowing
+    silently.
+
 !!! warning "Object keys"

    CBOR allows map keys of any type, whereas JSON only allows strings as keys in object values. Therefore, CBOR maps with keys other than UTF-8 strings are rejected.
@@ -67,8 +67,15 @@ specification:

 !!! info "NaN/infinity handling"

-    If NaN or Infinity are stored inside a JSON number, they are serialized properly in contrast to the
-    [dump](../../api/basic_json/dump.md) function which serializes NaN or Infinity to `null`.
+    `NaN`, `Infinity`, and `-Infinity` are serialized as a MessagePack float 32 (type 0xCA, 5 bytes total),
+    regardless of magnitude, in contrast to the [dump](../../api/basic_json/dump.md) function which serializes NaN
+    or Infinity to `null`.
+
+!!! note
+
+    Prior to version 3.13.0, NaN and Infinity were instead serialized as a MessagePack float 64 (type 0xCB, 9 bytes
+    total), because the check used to select the smaller float 32 encoding compared magnitudes with NaN, which is
+    always `false` and caused the float 32 path to be skipped.

 ??? example

@@ -11,7 +11,7 @@ This library does not support comments *by default*. It does so for three reason

 3. It is dangerous for interoperability if some libraries add comment support while others do not. Please check [The Harmful Consequences of the Robustness Principle](https://tools.ietf.org/html/draft-iab-protocol-maintenance-01) on this.

-However, you can set parameter `ignore_comments` to `#!cpp true` in the [`parse`](../api/basic_json/parse.md) function to ignore `//` or `/* */` comments. Comments will then be treated as whitespace.
+However, you can set parameter `ignore_comments` to `#!cpp true` in the [`parse`](../api/basic_json/parse.md) function to ignore `//` or `/* */` comments. Comments will then be treated as whitespace. Combined with `ignore_trailing_commas` (also a `parse` parameter), this covers what is commonly referred to as **JSONC** (JSON with Comments, as used e.g. by Visual Studio Code's `.jsonc` files) -- comments and trailing commas, nothing more. This is a different, smaller extension than [JSON5](https://json5.org), which additionally allows unquoted keys, single-quoted strings, and other syntax changes that this library does not support.

 For more information, see [JSON With Commas and Comments (JWCC)](https://nigeltao.github.io/blog/2021/json-with-commas-comments.html).

@@ -66,6 +66,67 @@ which forces the explicit `get` form and can catch unintended conversions at com
    floating-point value as an integer truncates it, and narrowing conversions may overflow. See
    [number conversion](types/number_handling.md#number-conversion) for details and how to guard against it.

+!!! warning "std::optional direct construction from JSON null throws"
+
+    Constructing or assigning `std::optional<T>` directly from a JSON value does not correctly produce
+    `std::nullopt` for a JSON `null`:
+
+    ```cpp
+    json j_null;
+    std::optional<std::string> opt = j_null;  // ❌ throws type_error 302
+    ```
+
+    This is due to C++ language rules: `std::optional<T>` has its own converting constructor that is chosen over
+    `basic_json::operator T()` when both are viable. Use `get<std::optional<T>>()` or `get_to()` instead:
+
+    ```cpp
+    auto opt = j_null.get<std::optional<std::string>>();  // ✅ std::nullopt
+    j_null.get_to(opt);                                     // ✅ std::nullopt
+    ```
+
+!!! warning "`static_cast` and `get<std::optional<T>>()` are not guaranteed equivalent"
+
+    `operator ValueType()` (used by `static_cast` and implicit conversions) intentionally excludes
+    `std::optional<T>` from delegating to `get<T>()`, to avoid a constructor ambiguity with
+    `std::optional<T>`'s own converting constructor from `basic_json`. As a result,
+    `static_cast<std::optional<T>>(json_value)` goes through `std::optional<T>`'s own converting
+    constructor rather than through `get<std::optional<T>>()`, which can behave differently -- for example,
+    with a custom `adl_serializer<std::optional<T>>` specialization. Prefer `get<std::optional<T>>()`/`get_to()`
+    over `static_cast` for optional types.
+
+!!! warning "Converting to a fixed-size `std::array` does not check length"
+
+    Converting a JSON array to `#!cpp std::array<T, N>` does not check that the JSON array's size matches `N`:
+    if the JSON array is longer, the extra elements are silently dropped; if it is shorter, the remaining
+    `std::array` elements are left default-constructed. No exception is thrown in either case.
+
+    ```cpp
+    json j = {1, 2, 3, 4, 5};
+    auto a = j.get<std::array<int, 3>>();  // {1, 2, 3} -- elements 4 and 5 silently dropped
+    ```
+
+## Omitting a field when serializing `std::optional`
+
+By default, `to_json` for `std::optional<T>` writes either the value or `#!json null` -- there is no built-in way
+to make a field disappear from the serialized object entirely when the `std::optional` is `std::nullopt`. Because
+a specialization of `adl_serializer<std::optional<T>>` only controls how the *value* is converted (it cannot
+prevent the containing object's `to_json` from inserting the key in the first place), omission has to be
+implemented in the *containing* type's `to_json`:
+
+```cpp
+struct person {
+    std::string name;
+    std::optional<int> age;
+};
+
+void to_json(json& j, const person& p) {
+    j = json{{"name", p.name}};
+    if (p.age) {
+        j["age"] = *p.age;  // key is only inserted when the optional has a value
+    }
+}
+```
+
 ## Putting values in

 The reverse direction works the same way: assigning or constructing a `json` from a C++ value converts it to JSON.
@@ -4,7 +4,8 @@

 In many situations, such as configuration files, missing values are not exceptional, but may be treated as if a default
 value was present. For this case, use [`value(key, default_value)`](../../api/basic_json/value.md) which takes the key
-you want to access and a default value in case there is no value stored with that key.
+you want to access and a default value in case there is no value stored with that key. This is equivalent to Python's
+`dict.get(key, default)`.

 ## Example

@@ -102,6 +102,20 @@ that the passed index is the new maximal index. Intermediate values are filled w
    `operator[]` can only be used with objects (with a string argument) or with arrays (with a numeric argument). For
    other types, a [`basic_json::type_error`](../../home/exceptions.md#jsonexceptiontype_error305) is thrown.

+## Performance: reserving array capacity
+
+There is no public `reserve(count)` member on `basic_json` for pre-allocating array capacity. If you are building
+a large array incrementally (e.g., via repeated `push_back()`) and know its final size ahead of time, you can
+reserve capacity via `get_ref()` to access the underlying `array_t` directly:
+
+```cpp
+json j = json::array();
+j.get_ref<json::array_t&>().reserve(1000);
+for (int i = 0; i < 1000; ++i) {
+    j.push_back(i);
+}
+```
+
 ## Summary

 | scenario                          | non-const value                                                                                                                                      | const value                                                                 |
@@ -77,6 +77,11 @@ auto val2 = j.at(json::json_pointer("/nested/three/1"));    // false
 auto val3 = j.value(json::json_pointer("/nested/four"), 0); // 0
 ```

+!!! note "Creating intermediate levels that don't exist"
+
+    See the [`operator[]` notes](../api/basic_json/operator%5B%5D.md#return-value) for how array vs. object is
+    decided when a pointer creates intermediate levels that don't exist yet.
+
 ## Flatten / unflatten

 The library implements a function [`flatten`](../api/basic_json/flatten.md) to convert any JSON document into a JSON
@@ -27,6 +27,7 @@ json data = json::parse(f);
 It should be noted that as modules do not export macros, the `nlohmann.json` module will not export any macros.

 ## Exported symbols
+
 Only the following symbols are exported from `nlohmann.json`:

 - `nlohmann::adl_serializer`
@@ -38,3 +39,21 @@ Only the following symbols are exported from `nlohmann.json`:
 - `nlohmann::to_string`
 - `nlohmann::literals::json_literals::operator""_json`
 - `nlohmann::literals::json_literals::operator""_json_pointer`
+
+Additionally, the following `nlohmann::detail` symbols are exported, solely to work around an MSVC compilation issue
+([#3970](https://github.com/nlohmann/json/issues/3970)). They are implementation details, not part of the public API,
+and should not be used directly:
+
+- `nlohmann::detail::json_sax_dom_callback_parser`
+- `nlohmann::detail::unknown_size`
+
+## Known issues
+
+C++20 modules support is exercised in CI against current GCC and Clang on Ubuntu, and the default MSVC toolset on Windows Server 2022 — there is no documented minimum compiler version, unlike feature-test-macro-gated features such as [`JSON_HAS_RANGES`](../api/macros/json_has_ranges.md).
+
+!!! info "Known compiler issues"
+
+    - **GCC** may emit "redefinition" errors when `#include <nlohmann/json.hpp>` appears in a module preamble together with other imports. This is an upstream GCC bug, not yet resolved as of GCC 16. Workarounds: include `nlohmann/json.hpp` before other `#include`s, use `import nlohmann.json;` instead, or upgrade GCC. ([issue #5103](https://github.com/nlohmann/json/issues/5103))
+    - **MSVC** could fail with `C2039: 'json_sax_dom_callback_parser' is not a member of ... detail`; fixed by exporting the required internal symbols from `json.cppm` (see [Exported symbols](#exported-symbols) above). ([issue #3970](https://github.com/nlohmann/json/issues/3970))
+
+    If you hit a different module-related build failure, search [existing issues](https://github.com/nlohmann/json/issues?q=is%3Aissue+modules) before filing a new one.
@@ -47,3 +47,6 @@ JSON Lines input with more than one value is treated as invalid JSON by the [`pa
    ```
    
    with a JSON Lines input does not work, because the parser will try to parse one value after the last one.
+
+    This is different from parsing a stream of *concatenated* (non-newline-delimited) JSON values, for which
+    `operator>>` does work -- see its [notes](../../api/operator_gtgt.md#notes) for details.
@@ -58,6 +58,14 @@ table describes the values of the parameters `depth`, `event`, and `parsed`.
    | `array_end`    | 1     | `#!json [52.519444,13.406667]` |
    | `object_end`   | 0     | `#!json {"location":[52.519444,13.406667],"name":"Berlin"}` |

+!!! note "No built-in nesting depth limit"
+
+    The library has no built-in limit on recursion/nesting depth while parsing. A parser callback can only
+    *discard* content it has already parsed (by returning `#!c false`); it cannot make parsing fail once a
+    nesting limit is exceeded partway through reading a deeply nested value. If you need to reject over-deep
+    untrusted input outright, track `depth` in a callback and `throw` from it once your limit is exceeded (a
+    thrown exception propagates out of `parse()` as usual).
+
 ## Return value

 Discarding a value (i.e., returning `#!c false`) has different effects depending on the context in which the function
@@ -81,3 +89,82 @@ was called:
    ```json
    --8<-- "examples/parse__string__parser_callback_t.output"
    ```
+
+## Recipe: rejecting duplicate object keys
+
+The JSON specification leaves the handling of objects with repeated keys up to the implementation. As described in
+[`object_t`](../../api/basic_json/object_t.md#behavior), it is unspecified which value for a repeated key ends up in
+the resulting `#!c json` value -- once parsing has produced that value, the duplicate is already gone, because object
+storage maps each key to a single value. If duplicate keys should instead be treated as an error, a parser callback
+can detect them while the object is still being read, before that ambiguity ever applies.
+
+??? example
+
+    ```cpp
+    --8<-- "examples/reject_duplicate_keys.cpp"
+    ```
+
+    Output:
+
+    ```json
+    --8<-- "examples/reject_duplicate_keys.output"
+    ```
+
+This approach has two limitations:
+
+- The depth-indexed bookkeeping must account for the fact that `object_start` reports the depth of the *parent* of
+  the object, while the `key` events inside that object are reported one depth deeper (see the event table above);
+  it is easy to get this off by one for nested objects.
+- The thrown exception cannot carry a `parse_error`-style byte offset, because position tracking only exists inside
+  the parser and lexer, not at the callback layer.
+
+For strict validation with precise error positions, implementing a [SAX interface](sax_interface.md) instead gives
+access to the parser's position information directly.
+
+## Recipe: streaming a large homogeneous array
+
+A common use case is a huge top-level array of many similarly-shaped objects, too large to hold entirely in
+memory as a `#!c json` value. A parser callback can hand off each completed element to a user function and then
+discard it, so memory usage stays bounded by a single element (plus the not-yet-parsed tail of the input) rather
+than the whole document. Since the top-level array's `array_start`/`array_end` are reported at `depth == 0` (its
+parent is the document root), the object elements it contains are reported at `depth == 1`:
+
+??? example
+
+    ```cpp
+    std::ifstream input("large_array.json");
+
+    auto callback = [](int depth, json::parse_event_t event, json& parsed) -> bool {
+        if (depth == 1 && event == json::parse_event_t::object_end) {
+            handle_element(parsed); // process the element, e.g. write it elsewhere
+            return false;           // discard it -- frees its memory before the next one is parsed
+        }
+        return true; // keep everything else, including the (by then empty) top-level array
+    };
+
+    json::parse(input, callback);
+    ```
+
+If the array's elements are scalars or nested arrays instead of objects, check for `parse_event_t::value` or
+`parse_event_t::array_end` at `depth == 1` instead. The same approach works for a top-level *object* of many
+homogeneous values by checking `object_end`/`value` events at `depth == 1` there too.
+
+## Recipe: max nesting depth via a callback
+
+Since there is no built-in nesting-depth limit (see the note above), a callback can enforce one manually by
+tracking the maximum `depth` seen and throwing once it is exceeded:
+
+??? example
+
+    ```cpp
+    constexpr int max_depth = 32;
+
+    auto callback = [](int depth, json::parse_event_t /*event*/, json& /*parsed*/) -> bool {
+        if (depth > max_depth) {
+            throw std::runtime_error("maximum nesting depth exceeded");
+        }
+        return true;
+    };
+
+    json::parse(input, callback);
+    ```
@@ -131,7 +131,7 @@ std::map<
 The choice of `object_t` influences the behavior of the JSON class. With the default type, objects have the following behavior:

 - When all names are unique, objects will be interoperable in the sense that all software implementations receiving that object will agree on the name-value mappings.
- When the names within an object are not unique, it is unspecified which one of the values for a given key will be chosen. For instance, `#!json {"key": 2, "key": 1}` could be equal to either `#!json {"key": 1}` or `#!json {"key": 2}`.
+- When the names within an object are not unique, it is unspecified which one of the values for a given key will be chosen. For instance, `#!json {"key": 2, "key": 1}` could be equal to either `#!json {"key": 1}` or `#!json {"key": 2}`. To reject duplicate keys instead of silently resolving them one way or another, see [this parsing recipe](../parsing/parser_callbacks.md#recipe-rejecting-duplicate-object-keys).
 - Internally, name/value pairs are stored in lexicographical order of the names. Objects will also be serialized (see `dump`) in this order. For instance, both `#!json {"b": 1, "a": 2}` and `#!json {"a": 2, "b": 1}` will be stored and serialized as `#!json {"a": 2, "b": 1}`.
 - When comparing objects, the order of the name/value pairs is irrelevant. This makes objects interoperable in the sense that they will not be affected by these differences. For instance, `#!json {"b": 1, "a": 2}` and `#!json {"a": 2, "b": 1}` will be treated as equal.

@@ -151,6 +151,18 @@ In this class, the object's limit of nesting is not explicitly constrained. Howe

 Objects are stored as pointers in a `basic_json` type. That is, for any access to object values, a pointer of type `object_t*` must be dereferenced.

+### Converting maps with non-string keys
+
+A `std::map`/`std::unordered_map` whose key type is not string-like (e.g., `std::map<int, std::string>`) is
+converted to a JSON *array* of 2-element `[key, value]` arrays rather than a JSON object, because JSON object
+keys must be strings:
+
+```cpp
+std::map<int, std::string> m{{1, "one"}, {2, "two"}};
+json j = m;
+// j is [[1,"one"],[2,"two"]], not {"1":"one","2":"two"}
+```
+

 ## Arrays

@@ -63,6 +63,10 @@ In the default [`json`](../../api/json.md) type, numbers are stored as `#!c std:
 number without loss of precision. If this is impossible (e.g., if the number is too large), the number is stored as
 `#!c double`.

+Positive integers are stored as `#!c std::uint64_t`, while negative integers are stored as `#!c std::int64_t`. This
+distinction is determined at parse time: if the JSON number has a leading minus sign, it uses signed integer storage;
+otherwise, it uses unsigned integer storage.
+
 !!! info "Notes"

    - Numbers with a decimal digit or scientific notation are always stored as `#!c double`.
@@ -0,0 +1,31 @@
+# Debugging
+
+This page collects the library's built-in debugger integrations and other debugging-related features. They are
+not linked from a single place elsewhere in the docs, so are collected here.
+
+## Visual Studio (natvis)
+
+The repository ships [`nlohmann_json.natvis`](https://github.com/nlohmann/json/blob/develop/nlohmann_json.natvis)
+at its root, a [Natvis](https://learn.microsoft.com/en-us/visualstudio/debugger/create-custom-views-of-native-objects)
+file that gives `json`/`ordered_json` values a friendly, key/value debugger view instead of showing raw internal
+fields, when debugging with the MSVC debug engine (`cppvsdbg`) in Visual Studio or VS Code.
+
+Debug engines that wrap LLDB instead of the MSVC debug engine (for example, `codelldb` in VS Code) only have
+partial/experimental Natvis support, and commonly fall back to showing raw internal fields even with the
+`.natvis` file present. Switching to `cppvsdbg` where available, or checking your debug extension's own Natvis
+support/version, are the next things to try if this happens. There is currently no bundled LLDB-native
+pretty-printer script in this repository.
+
+## GDB
+
+The repository ships a [GDB Python pretty printer](https://github.com/nlohmann/json/tree/develop/tools/gdb_pretty_printer)
+under `tools/gdb_pretty_printer`, with its own usage instructions in that directory's `README.md`.
+
+## Extended exception diagnostics
+
+Defining [`JSON_DIAGNOSTICS`](../api/macros/json_diagnostics.md) before including the library augments
+`type_error`/`out_of_range`-style exceptions with a JSON Pointer to the offending value, which can help pinpoint
+where in a large document a runtime error occurred. This only applies to exceptions thrown *after* a value
+exists (e.g. during element access); parse errors, which happen before any value exists to point at, are not
+covered by this mechanism -- see [Parsing and exceptions](../features/parsing/parse_exceptions.md) for how parse
+errors report their own location instead.
@@ -326,6 +326,9 @@ An unexpected byte was read in a [binary format](../features/binary_formats/inde
    ```
    [json.exception.parse_error.112] parse error at byte 15: syntax error while parsing BSON binary: byte array length cannot be negative, is -1
    ```
+    ```
+    [json.exception.parse_error.112] parse error at byte 9: syntax error while parsing CBOR value: negative integer overflow
+    ```

 ### json.exception.parse_error.113

@@ -893,7 +896,7 @@ A JSON Patch `add` operation cannot be applied because the target location's par

 !!! note

-    This exception was added in version 3.12.x. Before that, this situation hit an internal assertion (aborting the program in debug builds) or was silently ignored when assertions were disabled.
+    This exception was added in version 3.13.0. Before that, this situation hit an internal assertion (aborting the program in debug builds) or was silently ignored when assertions were disabled.

 ## Further exceptions

@@ -129,6 +129,29 @@ As described [above](#parse-errors-reading-non-ascii-characters), the library as
    }
    ```

+## Usage
+
+### Thread safety
+
+!!! question
+
+    Is `basic_json` thread-safe?
+
+No. `basic_json` provides no built-in synchronization, the same as `std::map` or `std::vector`. Concurrent reads of
+the same value from multiple threads are safe, as are concurrent (non-overlapping) accesses to independent `json`
+objects. However, any concurrent write to a `json` object -- or a concurrent read while another thread writes to the
+same object -- is a data race and requires external synchronization (e.g., a `std::mutex`) by the caller.
+
+### Schema validation
+
+!!! question
+
+    Does this library support JSON Schema validation?
+
+Not directly, but the companion project [json-schema-validator](https://github.com/pboettch/json-schema-validator)
+builds JSON Schema (draft 4, 6, 7, and 2019-09) validation on top of this library and is a common recommendation
+for this use case.
+
 ## Exceptions

 ### Parsing without exceptions
@@ -178,7 +201,7 @@ See [this section](../features/types/number_handling.md#number-serialization) on
    - Can I use `std::format("{}", j)` on a JSON value?
    - Can I use `fmt::format("{}", j)` or `fmt::print("{}", j)` (the [{fmt}](https://github.com/fmtlib/fmt) library) on a JSON value?

-`std::format` works out of the box since version 3.12.x, as long as the standard library provides
+`std::format` works out of the box since version 3.13.0, as long as the standard library provides
 `<format>` (see [`JSON_HAS_STD_FORMAT`](../api/macros/json_has_std_format.md)); see
 [`std::formatter<basic_json>`](../api/basic_json/std_formatter.md) for details, including the `#!cpp "{:#}"`
 pretty-print spec, indent widths (`#!cpp "{:2}"`), and custom indent characters (`#!cpp "{:.>#}"`).
@@ -181,3 +181,15 @@ Execute the test suite with [Valgrind](https://valgrind.org). This option is `OF

 Build the experimental [C++ module](../features/modules.md) `nlohmann.json` (requires CMake 3.28 or later and C++20).
 This option is `OFF` by default.
+
+A consuming project must link the dedicated `nlohmann_json_modules` CMake target (not just
+`nlohmann_json::nlohmann_json`) for `import nlohmann.json;` to resolve:
+
+```cmake
+set(NLOHMANN_JSON_BUILD_MODULES ON)
+add_subdirectory(path/to/json)
+
+add_executable(myproject main.cpp)
+target_link_libraries(myproject PRIVATE nlohmann_json_modules)
+target_compile_definitions(myproject PRIVATE NLOHMANN_JSON_BUILD_MODULES)
+```
@@ -930,6 +930,15 @@ If you are using [CocoaPods](https://cocoapods.org), you can use the library by
 to your podfile (see [an example](https://bitbucket.org/benman/nlohmann_json-cocoapod/src/master/)). Please file issues
 [here](https://bitbucket.org/benman/nlohmann_json-cocoapod/issues?status=new&status=open).

+## ESP-IDF and PlatformIO
+
+There is no official package published to the [ESP-IDF Component Registry](https://components.espressif.com) or the
+[PlatformIO Registry](https://registry.platformio.org). A community-maintained fork,
+[Johboh/nlohmann-json](https://github.com/Johboh/nlohmann-json), publishes this library to both registries on each
+new release and can be used as an unofficial component/package for ESP-IDF and PlatformIO projects. As the library
+is header-only, it can otherwise be used directly by adding its `include/` directory to your component's/project's
+include paths, like any other integration method described on this page.
+
 ![](https://img.shields.io/cocoapods/v/nlohmann_json)

 !!! warning
@@ -56,6 +56,7 @@ nav:
      - home/architecture.md
      - home/customers.md
      - home/sponsors.md
+      - home/debugging.md
  - Features:
      - features/index.md
      - features/arbitrary_types.md
@@ -11,8 +11,6 @@
 #include <cstdint> // uint8_t
 #include <cstddef> // size_t
 #include <functional> // hash
-#include <limits> // numeric_limits
-#include <cmath> // isfinite

 #include <nlohmann/detail/abi_macros.hpp>
 #include <nlohmann/detail/value_t.hpp>
@@ -28,73 +26,12 @@ inline std::size_t combine(std::size_t seed, std::size_t h) noexcept
    return seed;
 }

-// Check if a number_integer_t value is exactly representable as number_float_t
-// Returns true if static_cast<number_integer_t>(static_cast<number_float_t>(val)) == val
-template<typename BasicJsonType>
-inline bool is_exactly_representable_as_float(typename BasicJsonType::number_integer_t val) noexcept
-{
-    using number_integer_t = typename BasicJsonType::number_integer_t;
-    using number_float_t = typename BasicJsonType::number_float_t;
-
-    // If the float type's mantissa covers the integer type's entire range, all values round-trip
-    constexpr int float_digits = std::numeric_limits<number_float_t>::digits;
-    constexpr int int_digits = std::numeric_limits<number_integer_t>::digits;
-
-#ifdef JSON_HEDLEY_MSVC_VERSION
-#pragma warning(push )
-#pragma warning(disable : 4127) // ignore warning to replace if with if constexpr
-#endif
-    if (float_digits >= int_digits)
-    {
-        return true;
-    }
-#ifdef JSON_HEDLEY_MSVC_VERSION
-#pragma warning( pop )
-#endif
-
-    // For values outside float's exact range, they don't round-trip
-    // The safe way to check: compute the max magnitude that round-trips
-    // Using unsigned arithmetic to avoid UB with negating INT_MIN
-
-    // Max magnitude representable exactly: 2^(digits-1) - 1 for signed, 2^digits - 1 for unsigned range
-    // But we're checking a signed value, so use 2^digits as the threshold
-    constexpr auto max_exact = static_cast<number_integer_t>(1) << (float_digits - 1);
-
-    // Check absolute value against this threshold
-    if (val >= 0)
-    {
-        if (val >= max_exact)
-        {
-            return false;
-        }
-    }
-    else
-    {
-        // For negative values, check via unsigned wrapping arithmetic
-        // -val in unsigned domain; if it wraps, the value is too negative
-        auto unsigned_abs = static_cast<typename BasicJsonType::number_unsigned_t>(-val);
-        if (unsigned_abs >= static_cast<typename BasicJsonType::number_unsigned_t>(max_exact))
-        {
-            return false;
-        }
-    }
-
-    // For values within the exact range, verify the round-trip
-    const auto f = static_cast<number_float_t>(val);
-    return std::isfinite(f) && static_cast<number_integer_t>(f) == val;
-}
-
 /*!
@brief hash a JSON value

 The hash function tries to rely on std::hash where possible. Furthermore, the
 type of the JSON value is taken into account to have different hash values for
-most types. However, numeric types (number_integer, number_unsigned, number_float)
-are hashed to satisfy the std::hash contract: if two json values compare equal,
-they must have equal hash values. This means json(42), json(42u), and json(42.0)
-all hash to the same value (since they compare equal). For large integer values
-outside the exact representable range of the float type, integer values are hashed
-in their own domain to avoid precision loss.
+null, 0, 0U, and false, etc.

@tparam BasicJsonType basic_json specialization
@param j JSON value to hash
@@ -153,36 +90,14 @@ std::size_t hash(const BasicJsonType& j)

        case BasicJsonType::value_t::number_integer:
        {
-            const auto v = j.template get<number_integer_t>();
-            // Use a shared numeric type tag so all numeric types that are equal hash the same
-            const auto numeric_type = static_cast<std::size_t>(BasicJsonType::value_t::number_float);
-
-            if (is_exactly_representable_as_float<BasicJsonType>(v))
-            {
-                const auto h = std::hash<number_float_t> {}(static_cast<number_float_t>(v));
-                return combine(numeric_type, h);
-            }
-
-            const auto h = std::hash<number_integer_t> {}(v);
-            return combine(numeric_type, h);
+            const auto h = std::hash<number_integer_t> {}(j.template get<number_integer_t>());
+            return combine(type, h);
        }

        case BasicJsonType::value_t::number_unsigned:
        {
-            const auto v = j.template get<number_unsigned_t>();
-            // Normalize to signed (matching operator== behavior for U-vs-I comparison)
-            const auto v_as_signed = static_cast<number_integer_t>(v);
-            // Use a shared numeric type tag so all numeric types that are equal hash the same
-            const auto numeric_type = static_cast<std::size_t>(BasicJsonType::value_t::number_float);
-
-            if (is_exactly_representable_as_float<BasicJsonType>(v_as_signed))
-            {
-                const auto h = std::hash<number_float_t> {}(static_cast<number_float_t>(v_as_signed));
-                return combine(numeric_type, h);
-            }
-
-            const auto h = std::hash<number_integer_t> {}(v_as_signed);
-            return combine(numeric_type, h);
+            const auto h = std::hash<number_unsigned_t> {}(j.template get<number_unsigned_t>());
+            return combine(type, h);
        }

        case BasicJsonType::value_t::number_float:
@@ -430,7 +430,7 @@ class wide_string_input_adapter

    // parsing binary with wchar doesn't make sense, but since the parsing mode can be runtime, we need something here
    template<class T>
-    std::size_t get_elements(T* /*dest*/, std::size_t /*count*/ = 1)
+    JSON_HEDLEY_NO_RETURN std::size_t get_elements(T* /*dest*/, std::size_t /*count*/ = 1)
    {
        JSON_THROW(parse_error::create(112, 1, "wide string type cannot be interpreted as binary data", nullptr));
    }
@@ -146,6 +146,11 @@
        #define JSON_HAS_RANGES 0
    #elif defined(_LIBCPP_VERSION) && _LIBCPP_VERSION < 160000
        #define JSON_HAS_RANGES 0
+        // nvcc CUDA 12.0/12.1 chokes on the enable_borrowed_range variable-template
+        // syntax when compiling as CUDA source; fixed in CUDA 12.2 (issue #3907)
+    #elif defined(__CUDACC__) && defined(__CUDACC_VER_MAJOR__) && __CUDACC_VER_MAJOR__ == 12 \
+        && defined(__CUDACC_VER_MINOR__) && (__CUDACC_VER_MINOR__ == 0 || __CUDACC_VER_MINOR__ == 1)
+        #define JSON_HAS_RANGES 0
    #elif defined(__cpp_lib_ranges)
        #define JSON_HAS_RANGES 1
    #else
@@ -3776,17 +3776,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        return *this == basic_json(rhs);
    }

-    /// @brief comparison: not equal
-    /// @sa https://json.nlohmann.me/api/basic_json/operator_ne/
-    bool operator!=(const_reference rhs) const noexcept
-    {
-        if (compares_unordered(rhs, true))
-        {
-            return false;
-        }
-        return !operator==(rhs);
-    }
-
    /// @brief comparison: 3-way
    /// @sa https://json.nlohmann.me/api/basic_json/operator_spaceship/
    std::partial_ordering operator<=>(const_reference rhs) const noexcept // *NOPAD*
@@ -3892,10 +3881,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
    /// @sa https://json.nlohmann.me/api/basic_json/operator_ne/
    friend bool operator!=(const_reference lhs, const_reference rhs) noexcept
    {
-        if (compares_unordered(lhs, rhs, true))
-        {
-            return false;
-        }
        return !(lhs == rhs);
    }

@@ -2520,6 +2520,11 @@ JSON_HEDLEY_DIAGNOSTIC_POP
        #define JSON_HAS_RANGES 0
    #elif defined(_LIBCPP_VERSION) && _LIBCPP_VERSION < 160000
        #define JSON_HAS_RANGES 0
+        // nvcc CUDA 12.0/12.1 chokes on the enable_borrowed_range variable-template
+        // syntax when compiling as CUDA source; fixed in CUDA 12.2 (issue #3907)
+    #elif defined(__CUDACC__) && defined(__CUDACC_VER_MAJOR__) && __CUDACC_VER_MAJOR__ == 12 \
+        && defined(__CUDACC_VER_MINOR__) && (__CUDACC_VER_MINOR__ == 0 || __CUDACC_VER_MINOR__ == 1)
+        #define JSON_HAS_RANGES 0
    #elif defined(__cpp_lib_ranges)
        #define JSON_HAS_RANGES 1
    #else
@@ -6677,8 +6682,6 @@ NLOHMANN_JSON_NAMESPACE_END
 #include <cstdint> // uint8_t
 #include <cstddef> // size_t
 #include <functional> // hash
-#include <limits> // numeric_limits
-#include <cmath> // isfinite

 // #include <nlohmann/detail/abi_macros.hpp>

@@ -6696,73 +6699,12 @@ inline std::size_t combine(std::size_t seed, std::size_t h) noexcept
    return seed;
 }

-// Check if a number_integer_t value is exactly representable as number_float_t
-// Returns true if static_cast<number_integer_t>(static_cast<number_float_t>(val)) == val
-template<typename BasicJsonType>
-inline bool is_exactly_representable_as_float(typename BasicJsonType::number_integer_t val) noexcept
-{
-    using number_integer_t = typename BasicJsonType::number_integer_t;
-    using number_float_t = typename BasicJsonType::number_float_t;
-
-    // If the float type's mantissa covers the integer type's entire range, all values round-trip
-    constexpr int float_digits = std::numeric_limits<number_float_t>::digits;
-    constexpr int int_digits = std::numeric_limits<number_integer_t>::digits;
-
-#ifdef JSON_HEDLEY_MSVC_VERSION
-#pragma warning(push )
-#pragma warning(disable : 4127) // ignore warning to replace if with if constexpr
-#endif
-    if (float_digits >= int_digits)
-    {
-        return true;
-    }
-#ifdef JSON_HEDLEY_MSVC_VERSION
-#pragma warning( pop )
-#endif
-
-    // For values outside float's exact range, they don't round-trip
-    // The safe way to check: compute the max magnitude that round-trips
-    // Using unsigned arithmetic to avoid UB with negating INT_MIN
-
-    // Max magnitude representable exactly: 2^(digits-1) - 1 for signed, 2^digits - 1 for unsigned range
-    // But we're checking a signed value, so use 2^digits as the threshold
-    constexpr auto max_exact = static_cast<number_integer_t>(1) << (float_digits - 1);
-
-    // Check absolute value against this threshold
-    if (val >= 0)
-    {
-        if (val >= max_exact)
-        {
-            return false;
-        }
-    }
-    else
-    {
-        // For negative values, check via unsigned wrapping arithmetic
-        // -val in unsigned domain; if it wraps, the value is too negative
-        auto unsigned_abs = static_cast<typename BasicJsonType::number_unsigned_t>(-val);
-        if (unsigned_abs >= static_cast<typename BasicJsonType::number_unsigned_t>(max_exact))
-        {
-            return false;
-        }
-    }
-
-    // For values within the exact range, verify the round-trip
-    const auto f = static_cast<number_float_t>(val);
-    return std::isfinite(f) && static_cast<number_integer_t>(f) == val;
-}
-
 /*!
@brief hash a JSON value

 The hash function tries to rely on std::hash where possible. Furthermore, the
 type of the JSON value is taken into account to have different hash values for
-most types. However, numeric types (number_integer, number_unsigned, number_float)
-are hashed to satisfy the std::hash contract: if two json values compare equal,
-they must have equal hash values. This means json(42), json(42u), and json(42.0)
-all hash to the same value (since they compare equal). For large integer values
-outside the exact representable range of the float type, integer values are hashed
-in their own domain to avoid precision loss.
+null, 0, 0U, and false, etc.

@tparam BasicJsonType basic_json specialization
@param j JSON value to hash
@@ -6821,36 +6763,14 @@ std::size_t hash(const BasicJsonType& j)

        case BasicJsonType::value_t::number_integer:
        {
-            const auto v = j.template get<number_integer_t>();
-            // Use a shared numeric type tag so all numeric types that are equal hash the same
-            const auto numeric_type = static_cast<std::size_t>(BasicJsonType::value_t::number_float);
-
-            if (is_exactly_representable_as_float<BasicJsonType>(v))
-            {
-                const auto h = std::hash<number_float_t> {}(static_cast<number_float_t>(v));
-                return combine(numeric_type, h);
-            }
-
-            const auto h = std::hash<number_integer_t> {}(v);
-            return combine(numeric_type, h);
+            const auto h = std::hash<number_integer_t> {}(j.template get<number_integer_t>());
+            return combine(type, h);
        }

        case BasicJsonType::value_t::number_unsigned:
        {
-            const auto v = j.template get<number_unsigned_t>();
-            // Normalize to signed (matching operator== behavior for U-vs-I comparison)
-            const auto v_as_signed = static_cast<number_integer_t>(v);
-            // Use a shared numeric type tag so all numeric types that are equal hash the same
-            const auto numeric_type = static_cast<std::size_t>(BasicJsonType::value_t::number_float);
-
-            if (is_exactly_representable_as_float<BasicJsonType>(v_as_signed))
-            {
-                const auto h = std::hash<number_float_t> {}(static_cast<number_float_t>(v_as_signed));
-                return combine(numeric_type, h);
-            }
-
-            const auto h = std::hash<number_integer_t> {}(v_as_signed);
-            return combine(numeric_type, h);
+            const auto h = std::hash<number_unsigned_t> {}(j.template get<number_unsigned_t>());
+            return combine(type, h);
        }

        case BasicJsonType::value_t::number_float:
@@ -7347,7 +7267,7 @@ class wide_string_input_adapter

    // parsing binary with wchar doesn't make sense, but since the parsing mode can be runtime, we need something here
    template<class T>
-    std::size_t get_elements(T* /*dest*/, std::size_t /*count*/ = 1)
+    JSON_HEDLEY_NO_RETURN std::size_t get_elements(T* /*dest*/, std::size_t /*count*/ = 1)
    {
        JSON_THROW(parse_error::create(112, 1, "wide string type cannot be interpreted as binary data", nullptr));
    }
@@ -24707,17 +24627,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
        return *this == basic_json(rhs);
    }

-    /// @brief comparison: not equal
-    /// @sa https://json.nlohmann.me/api/basic_json/operator_ne/
-    bool operator!=(const_reference rhs) const noexcept
-    {
-        if (compares_unordered(rhs, true))
-        {
-            return false;
-        }
-        return !operator==(rhs);
-    }
-
    /// @brief comparison: 3-way
    /// @sa https://json.nlohmann.me/api/basic_json/operator_spaceship/
    std::partial_ordering operator<=>(const_reference rhs) const noexcept // *NOPAD*
@@ -24823,10 +24732,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
    /// @sa https://json.nlohmann.me/api/basic_json/operator_ne/
    friend bool operator!=(const_reference lhs, const_reference rhs) noexcept
    {
-        if (compares_unordered(lhs, rhs, true))
-        {
-            return false;
-        }
        return !(lhs == rhs);
    }

@@ -68,7 +68,11 @@ target_compile_options(test_main PUBLIC
    #       Disable warning C4566: character represented by universal-character-name '\uFF01'
    #                              cannot be represented in the current code page (1252)
    #       Disable warning C4996: 'nlohmann::basic_json<...>::operator <<': was declared deprecated
-    $<$<CXX_COMPILER_ID:MSVC>:/W4;/wd4566;/wd4996;$<$<CONFIG:Release>:/wd4702>>
+    #       Disable warning C4702: unreachable code; wide_string_input_adapter::get_elements()
+    #                              is annotated JSON_HEDLEY_NO_RETURN (it always throws), which
+    #                              makes MSVC flag the code following its call in binary_reader.hpp
+    #                              as unreachable for that instantiation, in both Debug and Release
+    $<$<CXX_COMPILER_ID:MSVC>:/W4;/wd4566;/wd4996;/wd4702>
    # https://github.com/nlohmann/json/issues/1114
    $<$<CXX_COMPILER_ID:MSVC>:/bigobj> $<$<BOOL:${MINGW}>:-Wa,-mbig-obj>

@@ -3,7 +3,18 @@ project(json_cuda LANGUAGES CUDA)

 add_executable(json_cuda json_cuda.cu)
 target_include_directories(json_cuda PRIVATE ../../include)
-target_compile_features(json_cuda PUBLIC cuda_std_11)
+
+# nvcc added C++20 support in CUDA 12.0 and C++17 in CUDA 11.0; pick the
+# newest standard the detected compiler actually supports (see #3907)
+# instead of hard-requiring one standard for every CUDA version.
+if(CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 12.0)
+    set(json_cuda_std 20)
+elseif(CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 11.0)
+    set(json_cuda_std 17)
+else()
+    set(json_cuda_std 11)
+endif()
+target_compile_features(json_cuda PUBLIC cuda_std_${json_cuda_std})
 set_target_properties(json_cuda PROPERTIES
    CUDA_EXTENSIONS OFF
    CUDA_STANDARD_REQUIRED ON
@@ -16,4 +16,20 @@ int main()
    // regression for #3013 (ordered_json::reset() compile error with nvcc)
    nlohmann::ordered_json metadata;
    metadata.erase("key");
+
+    // exercise comparisons (operator==/operator<=>, gated by
+    // JSON_HAS_THREE_WAY_COMPARISON, independent of JSON_HAS_RANGES) and
+    // range-based iteration (exercises iteration_proxy/ranges machinery
+    // beyond just the enable_borrowed_range specialization) — see #3907
+    nlohmann::json a = {1, 2, 3};
+    nlohmann::json b = {1, 2, 3};
+    static_cast<void>(a == b);
+#if JSON_HAS_THREE_WAY_COMPARISON
+    static_cast<void>(a <=> b); // *NOPAD*
+    static_cast<void>(a <=> 1); // *NOPAD*
+#endif
+    for (const auto& element : a)
+    {
+        static_cast<void>(element);
+    }
 }
@@ -24,7 +24,7 @@ struct bad_allocator : std::allocator<T>
    template<class U> bad_allocator(const bad_allocator<U>& /*unused*/) { }

    template<class... Args>
-    void construct(T* /*unused*/, Args&& ... /*unused*/) // NOLINT(cppcoreguidelines-missing-std-forward)
+    [[noreturn]] void construct(T* /*unused*/, Args&& ... /*unused*/) // NOLINT(cppcoreguidelines-missing-std-forward)
    {
        throw std::bad_alloc();
    }
@@ -369,6 +369,7 @@ TEST_CASE("lexicographical comparison operators")
        SECTION("comparison: not equal")
        {
            // check that two values compare unequal as expected
+            // operator!= now means exactly !(a==b) without special cases for NaN/discarded
            for (size_t i = 0; i < j_values.size(); ++i)
            {
                for (size_t j = 0; j < j_values.size(); ++j)
@@ -376,25 +377,12 @@ TEST_CASE("lexicographical comparison operators")
                    CAPTURE(i)
                    CAPTURE(j)

-                    if (json::compares_unordered(j_values[i], j_values[j], true))
-                    {
-                        // if two values compare unordered,
-                        // check that the boolean comparison result is always false
-                        CHECK_FALSE(j_values[i] != j_values[j]);
-                    }
-                    else
-                    {
-                        // otherwise, check that they compare according to their definition
-                        // as the inverse of equal
-                        CHECK((j_values[i] != j_values[j]) == !(j_values[i] == j_values[j]));
-                    }
+                    CHECK((j_values[i] != j_values[j]) == !(j_values[i] == j_values[j]));
                }
            }

            // compare with null pointer
            const json j_null;
-            CHECK((j_null != nullptr) == false);
-            CHECK((nullptr != j_null) == false);
            CHECK((j_null != nullptr) == !(j_null == nullptr));
            CHECK((nullptr != j_null) == !(nullptr == j_null));
        }
@@ -594,3 +582,34 @@ TEST_CASE("lexicographical comparison operators")
    }
 #endif
 }
+
+#if JSON_HAS_THREE_WAY_COMPARISON
+// JSON_HAS_CPP_20 (do not remove; see note at top of file)
+
+TEST_CASE("regression #3868 - heterogeneous comparisons compile under C++20 (P2468R2)")
+{
+    // Issue #3868: operator!= was preventing compiler from synthesizing reversed
+    // operator== candidates under C++20's P2468R2 rewritten candidate rules.
+    // Verify that heterogeneous comparisons now work.
+
+    SECTION("string vs json")
+    {
+        std::string s = "string";
+        json j = "string";
+        CHECK(s == j);
+        CHECK(j == s);
+        CHECK_FALSE(s != j);
+        CHECK_FALSE(j != s);
+    }
+
+    SECTION("other heterogeneous types")
+    {
+        int i = 42;
+        json j = 42;
+        CHECK(i == j);
+        CHECK(j == i);
+        CHECK_FALSE(i != j);
+        CHECK_FALSE(j != i);
+    }
+}
+#endif
@@ -278,7 +278,7 @@ TEST_CASE("constructors")
            const auto t = j.get<std::tuple<int, float, std::string>>();
            CHECK(std::get<0>(t) == j[0]);
            CHECK(std::get<1>(t) == j[1]);
-            // CHECK(std::get<2>(t) == j[2]); // commented out due to CI issue, see https://github.com/nlohmann/json/pull/3985 and https://github.com/nlohmann/json/issues/4025
+            CHECK(std::get<2>(t) == j[2]);
        }

        SECTION("std::tuple tie")
@@ -1761,16 +1761,27 @@ TEST_CASE("std::filesystem::path")
 }
 #endif

-#if !JSON_USE_IMPLICIT_CONVERSIONS
 TEST_CASE("std::optional")
 {
    SECTION("null")
    {
-        json j_null;
-        std::optional<std::string> opt_null;
+        const json j_null;
+        const std::optional<std::string> opt_null;

        CHECK(json(opt_null) == j_null);
        CHECK(j_null.get<std::optional<std::string>>() == std::nullopt);
+
+        // Constructing std::optional<T> directly from JSON null throws because
+        // std::optional's own converting constructor is chosen over basic_json's
+        // operator T(). This is a language-level limitation (std::optional<T> is
+        // constructible from T, and T is constructible from basic_json via the
+        // operator); there is no SFINAE path that distinguishes "call from inside
+        // std::optional's constructor" from "direct call". Use get<std::optional<T>>()
+        // or get_to() instead for correct null handling. See #4864 and #5246.
+        CHECK_THROWS_WITH_AS(std::optional<std::string>(j_null),
+                             "[json.exception.type_error.302] type must be string, but is null", json::type_error&);
+        CHECK_THROWS_WITH_AS(std::optional<int>(j_null),
+                             "[json.exception.type_error.302] type must be number, but is null", json::type_error&);
    }

    SECTION("string")
@@ -1819,7 +1830,6 @@ TEST_CASE("std::optional")
    }
 }
 #endif
-#endif

 #ifdef JSON_HAS_CPP_17
    #undef JSON_HAS_CPP_17
@@ -35,10 +35,10 @@ TEST_CASE("hash<nlohmann::json>")

    // number
    hashes.insert(std::hash<json> {}(json(0)));
-    hashes.insert(std::hash<json> {}(json(static_cast<unsigned>(0)))); // now same hash as json(0)
-    hashes.insert(std::hash<json> {}(json(0.0)));                       // now same hash as json(0)
+    hashes.insert(std::hash<json> {}(json(static_cast<unsigned>(0))));

    hashes.insert(std::hash<json> {}(json(-1)));
+    hashes.insert(std::hash<json> {}(json(0.0)));
    hashes.insert(std::hash<json> {}(json(42.23)));

    // array
@@ -60,16 +60,7 @@ TEST_CASE("hash<nlohmann::json>")
    // discarded
    hashes.insert(std::hash<json> {}(json(json::value_t::discarded)));

-    // Note: json(0), json(0U), and json(0.0) now hash to the same value
-    // (to satisfy the std::hash contract: equal values must hash equally)
-    // So we expect 19 distinct hashes instead of 21
-    CHECK(hashes.size() == 19);
-
-    // Verify the std::hash contract: equal values must hash equally
-    CHECK(std::hash<json> {}(json(0)) == std::hash<json> {}(json(static_cast<unsigned>(0))));
-    CHECK(std::hash<json> {}(json(0)) == std::hash<json> {}(json(0.0)));
-    CHECK(std::hash<json> {}(json(42)) == std::hash<json> {}(json(42u)));
-    CHECK(std::hash<json> {}(json(42)) == std::hash<json> {}(json(42.0)));
+    CHECK(hashes.size() == 21);
 }

 TEST_CASE("hash<nlohmann::ordered_json>")
@@ -93,10 +84,10 @@ TEST_CASE("hash<nlohmann::ordered_json>")

    // number
    hashes.insert(std::hash<ordered_json> {}(ordered_json(0)));
-    hashes.insert(std::hash<ordered_json> {}(ordered_json(static_cast<unsigned>(0)))); // now same hash as ordered_json(0)
-    hashes.insert(std::hash<ordered_json> {}(ordered_json(0.0)));                       // now same hash as ordered_json(0)
+    hashes.insert(std::hash<ordered_json> {}(ordered_json(static_cast<unsigned>(0))));

    hashes.insert(std::hash<ordered_json> {}(ordered_json(-1)));
+    hashes.insert(std::hash<ordered_json> {}(ordered_json(0.0)));
    hashes.insert(std::hash<ordered_json> {}(ordered_json(42.23)));

    // array
@@ -118,12 +109,5 @@ TEST_CASE("hash<nlohmann::ordered_json>")
    // discarded
    hashes.insert(std::hash<ordered_json> {}(ordered_json(ordered_json::value_t::discarded)));

-    // Note: ordered_json(0), ordered_json(0U), and ordered_json(0.0) now hash to the same value
-    CHECK(hashes.size() == 19);
-
-    // Verify the std::hash contract for ordered_json as well
-    CHECK(std::hash<ordered_json> {}(ordered_json(0)) == std::hash<ordered_json> {}(ordered_json(static_cast<unsigned>(0))));
-    CHECK(std::hash<ordered_json> {}(ordered_json(0)) == std::hash<ordered_json> {}(ordered_json(0.0)));
-    CHECK(std::hash<ordered_json> {}(ordered_json(42)) == std::hash<ordered_json> {}(ordered_json(42u)));
-    CHECK(std::hash<ordered_json> {}(ordered_json(42)) == std::hash<ordered_json> {}(ordered_json(42.0)));
+    CHECK(hashes.size() == 21);
 }
@@ -942,7 +942,7 @@ TEST_CASE("iterators 2")
                json j_expected{5, 4, 3, 2, 1};

                auto reversed = j | std::views::reverse;
-                CHECK(std::ranges::equal(reversed, j_expected));
+                CHECK(reversed == j_expected);
            }

            SECTION("transform")
Author	SHA1	Message	Date
Niels Lohmann	990ed2dc53	🚶 fix format Signed-off-by: Niels Lohmann <mail@nlohmann.me>	2026-07-10 13:38:52 +02:00
Niels Lohmann	40f3caad4d	📡 Fix documentation gaps found in a full GitHub Discussions review Reviewed all 1008 GitHub Discussions (2020-2026) for recurring questions that better or more visible documentation would have avoided. Adds/expands documentation for ~26 distinct gaps, including: - New "Debugging" page collecting natvis, GDB pretty printer, LLDB status, and JSON_DIAGNOSTICS pointers (previously scattered/undiscoverable) - Thread-safety and schema-validation FAQ entries - StringType's char-based requirement (no wstring/u16string/u32string) - Brace-initialization-yields-arrays warning directly on the constructor reference page (previously only in the FAQ, missed by users reading the constructor docs) - std::any exclusion from get<T>(), with a manual-dispatch example - Non-string-keyed std::map serializing as an array of pairs - ordered_json compatibility with NLOHMANN_DEFINE_TYPE_* macros (already worked, was undocumented) - std::array truncation on size-mismatched conversion (no exception) - static_cast vs. get<std::optional<T>>() divergence - Recipe for omitting a std::optional field instead of emitting null - No built-in nesting-depth limit during parsing + a callback-based workaround recipe - Recipe for streaming a large homogeneous array via parser callbacks - operator>> stream-position semantics for concatenated JSON values - JSON Pointer array-vs-object creation rule for non-existing paths - CMake target name (nlohmann_json_modules) needed to link C++20 modules - ESP-IDF/PlatformIO: no official package, link to a community fork - get(key, default) as the Python dict.get() equivalent - reserve() recipe for pre-allocating array capacity - JSONC as an alias for the existing ignore_comments/ignore_trailing_commas combination (distinct from the unsupported JSON5) - items() dereferenced-element type: decltype() idiom + detail-namespace stability caveat - Various macro/type-conversion limitations (MSGPACK_DEFINE_ARRAY equivalent, char-array round-tripping, ADL serializer macro gap) Signed-off-by: Niels Lohmann <mail@nlohmann.me>	2026-07-10 12:01:11 +02:00
Niels Lohmann	4d8e7a7210	🧛 fix build Signed-off-by: Niels Lohmann <mail@nlohmann.me>	2026-07-09 23:05:56 +02:00
Niels Lohmann	75e8fbac32	Documentation review: fix stale version-history placeholder in operator_ne.md (#5261 ) * 📡 Fix stale 3.12.x placeholder in operator_ne.md version history PR #5253 (removing the hand-written operator!= to fix #3868/P2468R2) merged after the earlier 3.12.x -> 3.13.0 global sweep, so its new version-history entries were written with the stale placeholder. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * 🚷 Fix stale twitter.com link in docset.json Signed-off-by: Niels Lohmann <mail@nlohmann.me> --------- Signed-off-by: Niels Lohmann <mail@nlohmann.me>	2026-07-09 20:57:41 +02:00
Niels Lohmann	631e667fe5	Document a duplicate-object-key rejection recipe (#5259 ) * 📡 Document a duplicate-object-key rejection recipe RFC 8259 leaves handling of duplicate object keys to the implementation; this library silently keeps only the last value for a repeated key. Discussion #5085 asked for an opt-in rejection mode. Decision: don't change library behavior, but document the existing parser-callback workaround instead. Adds a "Recipe: rejecting duplicate object keys" section to parser_callbacks.md, adapted from a community-contributed workaround. Fixed an off-by-one bug in the original snippet: object_start reports the depth of the object's parent, while key events inside that object report depth+1, so indexing the per-depth key set with the same depth in both places caused an out-of-bounds access on nested objects. Verified the published snippet compiles and behaves correctly for flat duplicates, nested duplicates, sibling objects sharing key names, and arrays of objects. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Cross-link the duplicate-key recipe with the existing object_t behavior docs object_t.md and features/types/index.md already document that duplicate object keys resolve to an unspecified value (RFC 8259 leaves this to the implementation). The new recipe's intro overstated this as a guaranteed "last value wins" rule, which isn't true in general -- parsing text keeps the last value, but constructing from an initializer list keeps the first. Reworded the recipe to point at object_t's "unspecified" behavior instead of asserting a specific rule, and added cross-links from both existing pages to the new recipe. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Turn the duplicate-key recipe into a standalone, compiled example Replace the inline code fence in the "rejecting duplicate object keys" recipe with a proper docs/mkdocs/docs/examples/*.cpp + .output pair, included via --8<-- like every other example on the site. The .output file was generated by running it through the project's actual example build (docs/Makefile: single_include, -std=c++11, -DJSON_USE_GLOBAL_UDLS=0) and cross-checked with `make check_output`, and the source passes the pinned astyle 3.4.13 formatting unchanged. Signed-off-by: Niels Lohmann <mail@nlohmann.me> --------- Signed-off-by: Niels Lohmann <mail@nlohmann.me>	2026-07-09 20:39:59 +02:00
Niels Lohmann	d0a43141ea	Fix #3868 : Remove operator!= to enable P2468R2 rewritten candidate synthesis (#5253 ) * Fix #3868: Remove operator!= to enable P2468R2 rewritten candidate synthesis Under C++20 P2468R2, a hand-written operator!= suppresses the compiler's rewritten-candidate synthesis for operator==, preventing heterogeneous comparisons like `std::string s; json j; s == j;` from compiling. Fix by removing the hand-written operator!=, allowing the compiler to synthesize != as !(a==b) in all language modes (C++20 member functions and pre-C++20 friend functions). Behavior change: operator!= now returns !(a==b) unconditionally, including for special values like NaN and discarded. This means: - NaN != NaN now returns true (matches IEEE-754 semantics) - discarded != x now returns true for any x (matches !(discarded == x)) This also fixes underlying defects in previously-working code: - Restores direct == comparison for views vs json (reverts std::ranges::equal workaround added in PR #3950 to dodge this bug) - Re-enables std::string == json comparisons (uncomments check in unit-constructor1.cpp) Fixes: #3868, #3979 Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Signed-off-by: Niels Lohmann <mail@nlohmann.me> * 🎓 fix warning Signed-off-by: Niels Lohmann <mail@nlohmann.me> --------- Signed-off-by: Niels Lohmann <mail@nlohmann.me> Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>	2026-07-09 20:38:39 +02:00
Niels Lohmann	ecff144b3a	📡 Document nvcc CUDA 12.0/12.1 JSON_HAS_RANGES exclusion (#5258 ) PR #5248 added a 5th JSON_HAS_RANGES exclusion branch to macro_scope.hpp (nvcc CUDA 12.0.x/12.1.x, fixed in 12.2, issue #3907) shortly after #5252 added the "Known compiler/stdlib exclusions" list to json_has_ranges.md, so the new branch was missing from the just-added doc section. Bring the list back to parity with the code (5 exclusion branches, 5 documented). Signed-off-by: Niels Lohmann <mail@nlohmann.me>	2026-07-09 20:19:22 +02:00
Niels Lohmann	855f511db4	Fix stale Clang -Weverything suppression comments; drop -Wno-missing-noreturn (#5250 ) * Fix stale Clang -Weverything suppression comments; eliminate -Wno-missing-noreturn cmake/clang_flags.cmake claimed -Wno-unsafe-buffer-usage was needed only for Doctest and that -Wno-missing-noreturn had "no way to silence... otherwise" (PR #4871, which never actually attempted a source fix). Neither held up under investigation (todo 130): - -Wno-unsafe-buffer-usage is pervasive (208 distinct sites across 19 files measured with clang trunk in silkeh/clang:dev), spanning the library's own low-level numeric/buffer code (to_chars, serializer, lexer, binary reader/writer, input adapters, json_pointer) as well as vendored Doctest itself (96 of the 208 sites). A source-level fix is not feasible at this scale; the comment now says so instead of blaming Doctest alone. - -Wno-missing-noreturn had exactly two real trigger sites, both genuinely and unconditionally non-returning: a test-only throwing allocator (tests/src/unit-allocator.cpp) and, previously undiscovered, wide_string_input_adapter::get_elements<T>() in include/nlohmann/detail/input/input_adapters.hpp. Verified this isn't a wider pattern by checking all 160 JSON_THROW call sites in the library for functions whose entire body is an unconditional throw. Annotated both ([[noreturn]] in the test file, since JSON_HEDLEY_NO_RETURN is #undef'd by the time test code runs; JSON_HEDLEY_NO_RETURN in the library file, its first real use anywhere in the codebase) and dropped the suppression entirely. single_include/nlohmann/json.hpp regenerated via `make amalgamate`; `make check-amalgamation` passes. Verified in Docker (silkeh/clang:dev, matching the ci_static_analysis_clang CI job): baseline builds clean, and the full 194-target test suite builds with zero warnings under the corrected CLANG_CXXFLAGS (-Wno-missing-noreturn no longer in the list). Also sanity-compiled and ran unit-allocator.cpp and unit-wstring.cpp on host Apple Clang to confirm behavior is unchanged. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix MSVC C4702 warning caused by JSON_HEDLEY_NO_RETURN on get_elements() PR #5250 annotated wide_string_input_adapter::get_elements<T>() with JSON_HEDLEY_NO_RETURN (it unconditionally throws). On MSVC this expands to __declspec(noreturn), and MSVC correctly determined that the code following its call in binary_reader.hpp is unreachable for that instantiation, firing C4702 under /W4 /WX in the msvc, msvc-vs2026, and msvc-arm64 Debug jobs. Clang doesn't flag this case, so the Docker verification for #5250 (which only checked Clang -Weverything) didn't catch it. This is the same warning class already tolerated for Release builds since PR #5216, where MSVC's optimizer independently found the same dead code after /Od was removed. Extend that existing /wd4702 suppression to Debug builds too, instead of reverting the noreturn annotation. Signed-off-by: Niels Lohmann <mail@nlohmann.me> --------- Signed-off-by: Niels Lohmann <mail@nlohmann.me>	2026-07-09 20:18:28 +02:00
Niels Lohmann	d0de6a9111	Document std::optional<T> direct construction limitation (#5247 ) * Document std::optional<T> direct-init/copy-init limitation with null Add regression test pinning current behavior (CHECK_THROWS_AS) in the null section of unit-conversions.cpp with detailed comment explaining the C++ language-level cause (std::optional's own converting constructor wins overload resolution over basic_json::operator T()). Add a warning callout in conversions.md documenting that direct construction/ assignment of std::optional<T> from JSON null throws type_error 302, with a clear workaround (use get<std::optional<T>>() or get_to() instead, which correctly produce std::nullopt). This is a limitation at the language level: there is no SFINAE path to distinguish "called from inside std::optional's own constructor" from "direct call", so fixing it would require breaking changes to operator ValueType(). A permanent fix belongs in the 4.0 type-strictness redesign (#3453). Signed-off-by: Niels Lohmann <mail@nlohmann.me> Co-Authored-By: Claude Code <noreply@anthropic.com> * Fix issue reference in std::optional test comment Update the comment in the null section test to reference #5246 instead of placeholder #XXXX, clarifying where the direct-init/copy-init limitation is tracked. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Use CHECK_THROWS_AS_WITH for std::optional test assertions Update the regression tests to use CHECK_THROWS_AS_WITH instead of CHECK_THROWS_AS to verify both the exception type and the error message. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix CI: use CHECK_THROWS_WITH_AS, the macro that actually exists CHECK_THROWS_AS_WITH is not a doctest macro; the correct one used throughout this test suite is CHECK_THROWS_WITH_AS(expr, message, exception_type&), with the message before the type and the type as a reference. The previous commit didn't catch this because it only compiled the file standalone with default settings; this TEST_CASE only compiles under `#if !JSON_USE_IMPLICIT_CONVERSIONS`, which is why ci_test_noimplicitconversions was the job that failed. Verified by building and running the test in that exact configuration (JSON_USE_IMPLICIT_CONVERSIONS=0): 14/14 assertions pass. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Run std::optional test under default implicit-conversions build too TEST_CASE("std::optional") was guarded by #if !JSON_USE_IMPLICIT_CONVERSIONS, so it only ever compiled in the non-default build with implicit conversions disabled. This traces back to commit `1d7688aef` (fixes #3859), which changed a previously dead #ifndef JSON_USE_IMPLICIT_CONVERSIONS guard (the macro is always defined by that point, so it never held) to #if !JSON_USE_IMPLICIT_CONVERSIONS -- making the test compile for the first time, but only in the disabled-conversions build. As a result, std::optional support had zero test coverage in the default configuration almost every user builds with. Verified the entire test case (all sections: null, string, bool, number, array, object) compiles and passes identically with JSON_USE_IMPLICIT_CONVERSIONS both on (default) and off -- nothing in it actually depends on the setting. Removing the guard closes the coverage gap with no behavior change: 285 assertions pass with implicit conversions on, 232 with them off (the difference comes from other, unrelated conditionally-compiled tests in this file). Signed-off-by: Niels Lohmann <mail@nlohmann.me> * 🎓 fix warning Signed-off-by: Niels Lohmann <mail@nlohmann.me> --------- Signed-off-by: Niels Lohmann <mail@nlohmann.me> Co-authored-by: Claude Code <noreply@anthropic.com>	2026-07-09 19:03:49 +02:00
Niels Lohmann	f8e99e856c	Fix nvcc CUDA 12.0/12.1 C++20 ranges parse error (#3907 ) (#5248 ) * Test ci_cuda_example against a CUDA version matrix at C++20 (#3907) The ci_cuda_example job compiled against the json-ci image's CUDA 11.0 toolkit at cuda_std_11, which cannot exercise #3907 (a c++20 parse error in iteration_proxy.hpp's enable_borrowed_range reported under nvcc). Switch the job to pull official nvidia/cuda devel images directly and matrix across CUDA 11.8-12.6 at cuda_std_20 so CI can empirically confirm which versions are actually affected before any source-level fix is attempted. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix nvcc CUDA 12.0/12.1 C++20 ranges parse error (#3907) The diagnostic matrix in this PR confirmed the affected range exactly: nvcc 12.0.1 and 12.1.1 both fail with "expected initializer before '<' token" on iteration_proxy.hpp's enable_borrowed_range variable template specialization at -std=c++20; 12.2.2 and newer already build cleanly. Guard JSON_HAS_RANGES off for that narrow nvcc version range, matching the existing GCC-11/libstdc++ carve-outs in the same ifdef chain, and regenerate single_include accordingly. Broaden the CUDA smoke test to also exercise comparisons (operator==/operator<=>, gated independently by JSON_HAS_THREE_WAY_COMPARISON) and range-based iteration, not just dump()/erase(), so the fix's actual scope is evidenced by CI rather than assumed from the single reported symptom. Have tests/cuda_example/CMakeLists.txt pick the newest C++ standard the detected nvcc version actually supports (20/17/11) instead of hard-requiring C++20, so older toolkits build at a lower standard instead of failing CMake configure outright. This is test-project-local only; the JSON_HAS_RANGES guard is what protects real client code, since a header can't control what -std= flag it's compiled with. Right-size the CI matrix from the 8-version diagnostic sweep down to 11.8.0 (C++17 fallback path) / 12.1.1 (permanent #3907 regression guard) / 12.6.3 (recent coverage), and update the compiler-version table in the quality assurance docs to match. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix ci_cuda_example CUDA 11.8 build after C++17 fallback (#3907) The 11.8.0 leg's graceful C++17 fallback (added in the previous commit) worked correctly, but the broadened smoke test used the <=> operator unconditionally, which isn't valid syntax pre-C++20 — nvcc rejected it with "expected an expression" once the CMake logic picked cuda_std_17 for the older toolkit. Gate those two lines behind JSON_HAS_THREE_WAY_COMPARISON like the library itself does internally. Sanity-compiled the file as plain C++ at both -std=c++17 (skips the guarded block) and -std=c++20 (includes it) locally; the actual nvcc build is verified via CI on PR #5248. Signed-off-by: Niels Lohmann <mail@nlohmann.me> --------- Signed-off-by: Niels Lohmann <mail@nlohmann.me>	2026-07-09 19:02:36 +02:00
Niels Lohmann	521a084827	Documentation review (#5257 ) * 📡 Fix documentation gaps for 3.13.0 release (todos 138-142) - Todo 138: Add "Known issues" section to modules.md with compiler-specific troubleshooting (GCC redefinition, MSVC symbol export). Add pointer note to quality_assurance.md. - Todo 139: Document CBOR/MessagePack half-precision float encoding for NaN/Infinity (0xF9/0xCA with exact byte sequences). Explain pre-3.13.0 double-precision bug mechanism without issue citations. - Todo 140: Document CBOR negative-integer-overflow rejection (parse_error.112) for magnitudes exceeding int64_t range (already implemented in rev 1). - Todo 141: Update version history in value.md and operator[].md with behavior-change details, removing issue citations per citation policy (prose is self-contained). - Todo 142: Global sed replace of 3.12.x → 3.13.0 placeholder across all 20 documentation files. Revision 2 incorporates feedback to reduce changelog-like issue citations. Only citations that add unique troubleshooting value are retained (#5103 for GCC workaround, #3970 for MSVC symbol export). "Known issues" section follows PR #5252's visual pattern (info admonition with bold-bullet format). Signed-off-by: Niels Lohmann <mail@nlohmann.me> * 📡 Document integer type selection, type_name() invalid value, and std::optional get() fix - number_handling.md: clarify that positive/negative integers select unsigned/signed storage based on the leading minus sign (todo 143). - type_name.md: document the new "invalid" return value for corrupted JSON values (todo 145). - get.md: note that get<std::optional<T>>() was unreachable in every configuration prior to 3.13.0 due to an internal macro-guard bug, unrelated to JSON_USE_IMPLICIT_CONVERSIONS's actual effect (todo 144). Signed-off-by: Niels Lohmann <mail@nlohmann.me> --------- Signed-off-by: Niels Lohmann <mail@nlohmann.me>	2026-07-09 17:24:19 +02:00
Niels Lohmann	ca91678af1	Document compiler/stdlib exclusions in macro_scope.hpp (#5252 ) * 📡 Document compiler/stdlib exclusions in macro_scope.hpp Add "Known compiler/stdlib exclusions" subsections to the public documentation for JSON_HAS_FILESYSTEM and JSON_HAS_RANGES, listing the exact compiler/stdlib versions that are silently excluded even when feature-test macros indicate support. Each exclusion references the originating issue. Also add a pointer note in the compiler compatibility section linking to these details. Signed-off-by: Niels Lohmann <mail@nlohmann.me> Co-Authored-By: Claude Code <noreply@anthropic.com> * 🧛 fix build Signed-off-by: Niels Lohmann <mail@nlohmann.me> --------- Signed-off-by: Niels Lohmann <mail@nlohmann.me> Co-authored-by: Claude Code <noreply@anthropic.com>	2026-07-09 17:15:57 +02:00
Niels Lohmann	ff34a3fd2f	Fix flaky ci_nvhpc job: pin nvc++ target to generic baseline (-tp=px) (#5254 )	2026-07-09 15:16:28 +02:00