Remove is_bjd_excluded_optimized_type(), which has no caller left; fix two comments

is_valid_optimized_type() rejects the markers BJData excludes, so the reader
no longer needs its own list; the test now checks is_valid_optimized_type()
of a BJData reader against the writer's list. The BSON size comment now says
that open_bson_document() rejects a size below 5.

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

include/nlohmann/detail/input/binary_reader.hpp

@@ -344,8 +344,8 @@ class binary_reader
     nested document could declare a length that disagrees with where its
     terminator actually falls and quietly hand the bytes in between to the
     enclosing document. A well-formed document is at least 5 bytes (the prefix
-    plus the terminator); the equality also rejects those impossible sizes,
-    since at least 5 bytes are always consumed.
+    plus the terminator); open_bson_document() rejects a smaller size before
+    it is stored, so @a document_size is never negative here.
 
     @param[in] document_start  value of chars_read before the size prefix
     @param[in] document_size   the declared document size
@@ -353,7 +353,7 @@ class binary_reader
     */
     bool check_bson_document_size(const std::size_t document_start, const std::int32_t document_size)
     {
-        if (JSON_HEDLEY_UNLIKELY(document_size < 0 || static_cast<std::size_t>(document_size) != chars_read - document_start))
+        if (JSON_HEDLEY_UNLIKELY(static_cast<std::size_t>(document_size) != chars_read - document_start))
         {
             return bson_document_size_error(document_size, chars_read - document_start);
         }
@@ -3084,16 +3084,13 @@ class binary_reader
         {
             if (size_and_type.second != 0)
             {
-                if (size_and_type.second != 'N')
+                for (std::size_t i = 0; i < size_and_type.first; ++i)
                 {
-                    for (std::size_t i = 0; i < size_and_type.first; ++i)
+                    if (JSON_HEDLEY_UNLIKELY(!get_ubjson_size_value(dimlen, no_ndarray, size_and_type.second)))
                     {
-                        if (JSON_HEDLEY_UNLIKELY(!get_ubjson_size_value(dimlen, no_ndarray, size_and_type.second)))
-                        {
-                            return false;
-                        }
-                        dim.push_back(dimlen);
+                        return false;
                     }
+                    dim.push_back(dimlen);
                 }
             }
             else
@@ -3295,11 +3292,9 @@ class binary_reader
                     // follow the documented _ArrayType_, _ArraySize_, _ArrayData_ order
                     if (ndarray_dtype != 0)
                     {
+                        // every optimized type marker is_valid_optimized_type() accepts for BJData has a name
                         const char* type_name = bjd_type_name(ndarray_dtype);
-                        if (JSON_HEDLEY_UNLIKELY(type_name == nullptr))
-                        {
-                            return unexpected_byte("invalid byte", "type");
-                        }
+                        JSON_ASSERT(type_name != nullptr);
 
                         string_t type_key = "_ArrayType_";
                         string_t type = type_name; // sax->string() takes a reference
@@ -3382,12 +3377,6 @@ class binary_reader
         if (current == '$')
         {
             result.second = get();  // must not ignore 'N', because 'N' maybe the type
-            if (input_format == input_format_t::bjdata
-                    && JSON_HEDLEY_UNLIKELY(is_bjd_excluded_optimized_type(result.second)))
-            {
-                return optimized_type_error();
-            }
-
             if (JSON_HEDLEY_UNLIKELY(!unexpect_eof("type")))
             {
                 return false;
@@ -3415,7 +3404,7 @@ class binary_reader
             // an ndarray was read here only if the flag flipped; when it was
             // seeded true, get_ubjson_size_value() already rejected the nested
             // dimension vector
-            if (input_format == input_format_t::bjdata && is_ndarray && !inside_ndarray)
+            if (is_ndarray && !inside_ndarray)
             {
                 result.second |= (1 << 8); // use bit 8 to indicate ndarray, all UBJSON and BJData markers should be ASCII letters
             }
@@ -3425,7 +3414,7 @@ class binary_reader
         if (current == '#')
         {
             const bool is_error = get_ubjson_size_value(result.first, is_ndarray);
-            if (input_format == input_format_t::bjdata && is_ndarray && !inside_ndarray)
+            if (is_ndarray && !inside_ndarray)
             {
                 return last_byte_error(exception_id::unexpected_byte, "ndarray requires both type and size", "size");
             }
@@ -3599,7 +3588,6 @@ class binary_reader
     JSON_HEDLEY_NEVER_INLINE
     bool parse_ubjson_typed_run(std::size_t& remaining, const char_int_type type, std::true_type /*bulk*/)
     {
-        const bool bjdata = input_format == input_format_t::bjdata;
         switch (type)
         {
             case 'U':
@@ -3613,11 +3601,11 @@ class binary_reader
             case 'L':
                 return parse_ubjson_typed_run<std::int64_t>(remaining);
             case 'u':
-                return !bjdata || parse_ubjson_typed_run<std::uint16_t>(remaining);
+                return parse_ubjson_typed_run<std::uint16_t>(remaining);
             case 'm':
-                return !bjdata || parse_ubjson_typed_run<std::uint32_t>(remaining);
+                return parse_ubjson_typed_run<std::uint32_t>(remaining);
             case 'M':
-                return !bjdata || parse_ubjson_typed_run<std::uint64_t>(remaining);
+                return parse_ubjson_typed_run<std::uint64_t>(remaining);
             case 'd':
                 return parse_ubjson_typed_run<float>(remaining);
             case 'D':
@@ -3698,7 +3686,7 @@ class binary_reader
         // if bit-8 of size_and_type.second is set to 1, encode bjdata ndarray as an object in JData annotated array format (https://github.com/NeuroJSON/jdata):
         // {"_ArrayType_" : "typeid", "_ArraySize_" : [n1, n2, ...], "_ArrayData_" : [v1, v2, ...]}
 
-        if (input_format == input_format_t::bjdata && size_and_type.first != npos && (size_and_type.second & (1 << 8)) != 0)
+        if ((size_and_type.second & (1 << 8)) != 0)
         {
             size_and_type.second &= ~(static_cast<char_int_type>(1) << 8);  // use bit 8 to indicate ndarray, here we remove the bit to restore the type marker
 
@@ -3757,7 +3745,7 @@ class binary_reader
         }
 
         // If BJData type marker is 'B' decode as binary
-        if (input_format == input_format_t::bjdata && size_and_type.first != npos && size_and_type.second == 'B')
+        if (size_and_type.first != npos && size_and_type.second == 'B')
         {
             binary_t result;
             return get_binary(size_and_type.first, result) && sax->binary(result);
@@ -3806,7 +3794,7 @@ class binary_reader
         }
 
         // do not accept ND-array size in objects in BJData
-        if (input_format == input_format_t::bjdata && size_and_type.first != npos && (size_and_type.second & (1 << 8)) != 0)
+        if ((size_and_type.second & (1 << 8)) != 0)
         {
             return last_byte_error(exception_id::unexpected_byte, "BJData object does not support ND-array size in optimized format", "object");
         }
@@ -5087,8 +5075,9 @@ class binary_reader
   JSON_PRIVATE_UNLESS_TESTED:
     /*!
     @brief check whether @a marker is a type marker that may follow '$'
-    UBJSON: Z N T F i U I l L d D H C S [ {; BJData: the numeric markers and C
-    (BJData's excluded markers are rejected separately).
+    UBJSON: Z N T F i U I l L d D H C S [ {; BJData: the numeric markers and C.
+    The markers BJData excludes (F H N S T Z [ {) are the ones
+    binary_writer::is_bjdata_excluded_type_marker() lists; keep the two in sync.
     */
     bool is_valid_optimized_type(const char_int_type marker) const noexcept
     {
@@ -5123,20 +5112,6 @@ class binary_reader
         }
     }
 
-    /*!
-    @brief whether @a marker is excluded from BJData's optimized ND-array types
-    @return whether @a marker is one of 'F', 'H', 'N', 'S', 'T', 'Z', '[', '{'
-
-    Mirrors binary_writer's @ref binary_writer::is_bjdata_excluded_type_marker
-    "is_bjdata_excluded_type_marker()`, which encodes the same list the other
-    way; keep the two in sync.
-    */
-    static constexpr bool is_bjd_excluded_optimized_type(const char_int_type marker) noexcept
-    {
-        return marker == '[' || marker == '{' || marker == 'S' || marker == 'H'
-               || marker == 'T' || marker == 'F' || marker == 'N' || marker == 'Z';
-    }
-
     /*!
     @brief look up the ND-array element type name for a BJData dtype marker
     @return the type name ("uint8", "int8", ...), or nullptr if @a marker does

include/nlohmann/detail/output/binary_writer.hpp

@@ -3129,22 +3129,19 @@ class binary_writer
                 // negative fixnum
                 write_number(static_cast<std::int8_t>(j.m_data.m_value.number_integer));
             }
-            else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int8_t>::min)() &&
-                     j.m_data.m_value.number_integer <= (std::numeric_limits<std::int8_t>::max)())
+            else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int8_t>::min)())
             {
                 // int 8
                 oa.write_character(to_char_type(0xD0));
                 write_number(static_cast<std::int8_t>(j.m_data.m_value.number_integer));
             }
-            else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int16_t>::min)() &&
-                     j.m_data.m_value.number_integer <= (std::numeric_limits<std::int16_t>::max)())
+            else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int16_t>::min)())
             {
                 // int 16
                 oa.write_character(to_char_type(0xD1));
                 write_number(static_cast<std::int16_t>(j.m_data.m_value.number_integer));
             }
-            else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int32_t>::min)() &&
-                     j.m_data.m_value.number_integer <= (std::numeric_limits<std::int32_t>::max)())
+            else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int32_t>::min)())
             {
                 // int 32
                 oa.write_character(to_char_type(0xD2));
@@ -4060,7 +4057,7 @@ class binary_writer
     Containers, strings, high-precision numbers, booleans and null cannot be
     declared as the single type of an optimized container in BJData; such a
     container is written unoptimized. The reader rejects them with the same
-    list (binary_reader::is_bjd_excluded_optimized_type()).
+    list (binary_reader::is_valid_optimized_type()).
     */
     static constexpr bool is_bjdata_excluded_type_marker(const CharType marker) noexcept
     {

include/nlohmann/json.hpp

@@ -543,10 +543,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
     ////////////////////////
 
   JSON_PRIVATE_UNLESS_TESTED:
-    // the type and value of a basic_json without its base class (defined
-    // below); the destruction helpers of json_value use it as a stand-in
-    struct data;
-
     /*!
     @brief a JSON value
 
@@ -723,28 +719,27 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
         explicit json_value(array_t* value) noexcept : array(value) {}
 
 private:
-        // raw, allocation-free transfer of the type and value from src to
-        // dst: no set_parents()/assert_invariant() (the former is
-        // O(#children) per call under JSON_DIAGNOSTICS, which would make the
-        // walk below quadratic); dst takes ownership, src is left as
-        // value_t::null.
-        static void take(data& dst, data& src) noexcept
+        // raw, allocation-free transfer of m_data from src to dst: no
+        // set_parents()/assert_invariant() (the former is O(#children) per
+        // call under JSON_DIAGNOSTICS, which would make the walk below
+        // quadratic); dst takes ownership, src is left as value_t::null.
+        static void take(basic_json& dst, basic_json& src) noexcept
         {
-            dst.m_type = src.m_type;
-            dst.m_value = src.m_value;
-            src.m_type = value_t::null;
+            dst.m_data.m_type = src.m_data.m_type;
+            dst.m_data.m_value = src.m_data.m_value;
+            src.m_data.m_type = value_t::null;
         }
 
         // true if v is not an array/object, or is an already-empty one
-        static bool has_no_children(const data& v) noexcept
+        static bool has_no_children(const basic_json& v) noexcept
         {
-            if (v.m_type == value_t::array)
+            if (v.m_data.m_type == value_t::array)
             {
-                return v.m_value.array->empty();
+                return v.m_data.m_value.array->empty();
             }
-            if (v.m_type == value_t::object)
+            if (v.m_data.m_type == value_t::object)
             {
-                return v.m_value.object->empty();
+                return v.m_data.m_value.object->empty();
             }
             return true;
         }
@@ -753,21 +748,21 @@ private:
         // its own, i.e. v can be freed in one go (its destructor then
         // destroys scalars and empty containers only, so it recurses at most
         // one level deep); stops at the first structured child it meets
-        static bool all_children_leaves(const data& v) noexcept
+        static bool all_children_leaves(const basic_json& v) noexcept
         {
             // array_t only needs forward iterators (see custom_array_type)
-            if (v.m_type == value_t::array)
+            if (v.m_data.m_type == value_t::array)
             {
-                const auto& a = *v.m_value.array;
+                const auto& a = *v.m_data.m_value.array;
                 return std::all_of(a.begin(), a.end(), [](const basic_json & el)
                 {
-                    return has_no_children(el.m_data);
+                    return has_no_children(el);
                 });
             }
-            const auto& o = *v.m_value.object;
+            const auto& o = *v.m_data.m_value.object;
             return std::all_of(o.begin(), o.end(), [](const typename object_t::value_type & el)
             {
-                return has_no_children(el.second.m_data);
+                return has_no_children(el.second);
             });
         }
 
@@ -799,30 +794,30 @@ private:
 
         // the child of a non-empty array/object v that the walk in
         // destroy_container() continues with (see walk_child_it() above)
-        static data& walk_child(data& v)
+        static basic_json& walk_child(basic_json& v)
         {
-            if (v.m_type == value_t::array)
+            if (v.m_data.m_type == value_t::array)
             {
-                return v.m_value.array->back().m_data;
+                return v.m_data.m_value.array->back();
             }
-            JSON_ASSERT(v.m_type == value_t::object);
-            return walk_child_it(*v.m_value.object)->second.m_data;
+            JSON_ASSERT(v.m_data.m_type == value_t::object);
+            return walk_child_it(*v.m_data.m_value.object)->second;
         }
 
         // removes walk_child(v) from a non-empty array/object v; this never
         // allocates, and since it is only ever called when that child is a
         // scalar or an already-empty array/object, destroying it never
         // recurses more than one level deep (see destroy() below)
-        static void pop_walk_child(data& v)
+        static void pop_walk_child(basic_json& v)
         {
-            if (v.m_type == value_t::array)
+            if (v.m_data.m_type == value_t::array)
             {
-                v.m_value.array->pop_back();
+                v.m_data.m_value.array->pop_back();
             }
             else
             {
-                JSON_ASSERT(v.m_type == value_t::object);
-                v.m_value.object->erase(walk_child_it(*v.m_value.object));
+                JSON_ASSERT(v.m_data.m_type == value_t::object);
+                v.m_data.m_value.object->erase(walk_child_it(*v.m_data.m_value.object));
             }
         }
 
@@ -831,23 +826,23 @@ private:
         // recurses at most one level deep); this is the same allocator-based
         // free the old recursive implementation used, just factored out so
         // every level of the walk in destroy() can share it
-        static void free_container(data& v) noexcept
+        static void free_container(basic_json& v) noexcept
         {
             JSON_ASSERT(has_no_children(v) || all_children_leaves(v));
-            if (v.m_type == value_t::array)
+            if (v.m_data.m_type == value_t::array)
             {
                 AllocatorType<array_t> alloc;
-                std::allocator_traits<decltype(alloc)>::destroy(alloc, v.m_value.array);
-                std::allocator_traits<decltype(alloc)>::deallocate(alloc, v.m_value.array, 1);
+                std::allocator_traits<decltype(alloc)>::destroy(alloc, v.m_data.m_value.array);
+                std::allocator_traits<decltype(alloc)>::deallocate(alloc, v.m_data.m_value.array, 1);
             }
             else
             {
-                JSON_ASSERT(v.m_type == value_t::object);
+                JSON_ASSERT(v.m_data.m_type == value_t::object);
                 AllocatorType<object_t> alloc;
-                std::allocator_traits<decltype(alloc)>::destroy(alloc, v.m_value.object);
-                std::allocator_traits<decltype(alloc)>::deallocate(alloc, v.m_value.object, 1);
+                std::allocator_traits<decltype(alloc)>::destroy(alloc, v.m_data.m_value.object);
+                std::allocator_traits<decltype(alloc)>::deallocate(alloc, v.m_data.m_value.object, 1);
             }
-            v.m_type = value_t::null; // avoid a double free if v is later destructed
+            v.m_data.m_type = value_t::null; // avoid a double free if v is later destructed
         }
 
 public:
@@ -878,16 +873,14 @@ public:
 
             // This json_value is not itself a basic_json, so the top-level
             // container is first moved into a local stand-in ("cur"); a
-            // default-constructed data has a null pointer in its m_value
-            // (see data::m_value's initializer), so swapping it with *this
-            // leaves this union's own pointer null, and it is never looked
-            // at or freed a second time. The stand-ins are data, not
-            // basic_json, so that destroying a value never constructs a
-            // CustomBaseClass, whose constructor may allocate (#5995).
-            data cur;
-            cur.m_type = t;
+            // default-constructed basic_json has a null pointer in its
+            // m_value (see data::m_value's initializer), so swapping it
+            // with *this leaves this union's own pointer null, and it is
+            // never looked at or freed a second time.
+            basic_json cur;
+            cur.m_data.m_type = t;
             using std::swap;
-            swap(cur.m_value, *this);
+            swap(cur.m_data.m_value, *this);
 
             destroy_bounded(cur, 0);
         }
@@ -896,7 +889,7 @@ public:
         // The first recursion_depth_limit() levels are destroyed by plain
         // recursion, which is the fastest way and cannot throw; anything
         // nested deeper is handed to destroy_walk(), which uses no call stack.
-        static void destroy_bounded(data& v, std::size_t depth) noexcept
+        static void destroy_bounded(basic_json& v, std::size_t depth) noexcept
         {
             if (JSON_HEDLEY_UNLIKELY(depth >= detail::recursion_depth_limit()))
             {
@@ -904,23 +897,23 @@ public:
                 return;
             }
 
-            if (v.m_type == value_t::array)
+            if (v.m_data.m_type == value_t::array)
             {
-                for (auto& child : *v.m_value.array)
+                for (auto& child : *v.m_data.m_value.array)
                 {
                     if (child.m_data.m_type == value_t::array || child.m_data.m_type == value_t::object)
                     {
-                        destroy_bounded(child.m_data, depth + 1);
+                        destroy_bounded(child, depth + 1);
                     }
                 }
             }
             else
             {
-                for (auto& child : *v.m_value.object)
+                for (auto& child : *v.m_data.m_value.object)
                 {
                     if (child.second.m_data.m_type == value_t::array || child.second.m_data.m_type == value_t::object)
                     {
-                        destroy_bounded(child.second.m_data, depth + 1);
+                        destroy_bounded(child.second, depth + 1);
                     }
                 }
             }
@@ -947,15 +940,18 @@ public:
         // array/object, which neither allocates nor recurses more
         // than one level deep.
         //
-        // Like destroy_container(), the walk keeps the containers it
-        // holds in data stand-ins (cur, prev, gp, tmp), which never
-        // construct a CustomBaseClass.
-        static void destroy_walk(data& top) noexcept
+        // This json_value is not itself a basic_json, so the
+        // top-level container is first moved into a local stand-in
+        // ("cur"); a default-constructed basic_json has a null
+        // pointer in its m_value (see data::m_value's initializer),
+        // so swapping it with *this leaves this union's own pointer
+        // null, and it is never looked at or freed a second time.
+        static void destroy_walk(basic_json& top) noexcept
         {
-            data cur;
+            basic_json cur;
             take(cur, top);
 
-            data prev; // value_t::null: no parent
+            basic_json prev; // value_t::null: no parent
 
             // true right after cur was entered: its children have not been
             // looked at yet. If none of them has children of its own, cur is
@@ -969,7 +965,7 @@ public:
                 fresh = false;
                 if (has_no_children(cur) || (fresh_now && all_children_leaves(cur)))
                 {
-                    if (prev.m_type == value_t::null)
+                    if (prev.m_data.m_type == value_t::null)
                     {
                         free_container(cur);
                         return; // back at the top with nothing left to do
@@ -978,7 +974,7 @@ public:
                     // ascend: detach the grandparent link from prev's
                     // walk_child() slot, drop that (now null) slot, free cur
                     // (it is empty), then move up one level
-                    data gp;
+                    basic_json gp;
                     take(gp, walk_child(prev));
                     pop_walk_child(prev);
 
@@ -989,7 +985,7 @@ public:
                     continue;
                 }
 
-                data& cur_child_ref = walk_child(cur);
+                basic_json& cur_child_ref = walk_child(cur);
 
                 if (has_no_children(cur_child_ref))
                 {
@@ -1001,7 +997,7 @@ public:
                 // descend into the non-empty child, reversing the
                 // link: its slot takes over prev, and the child becomes
                 // the new cur
-                data tmp;
+                basic_json tmp;
                 take(tmp, cur_child_ref);
                 take(cur_child_ref, prev);
                 take(prev, cur);

single_include/nlohmann/json.hpp

@@ -18351,8 +18351,8 @@ class binary_reader
     nested document could declare a length that disagrees with where its
     terminator actually falls and quietly hand the bytes in between to the
     enclosing document. A well-formed document is at least 5 bytes (the prefix
-    plus the terminator); the equality also rejects those impossible sizes,
-    since at least 5 bytes are always consumed.
+    plus the terminator); open_bson_document() rejects a smaller size before
+    it is stored, so @a document_size is never negative here.
 
     @param[in] document_start  value of chars_read before the size prefix
     @param[in] document_size   the declared document size
@@ -18360,7 +18360,7 @@ class binary_reader
     */
     bool check_bson_document_size(const std::size_t document_start, const std::int32_t document_size)
     {
-        if (JSON_HEDLEY_UNLIKELY(document_size < 0 || static_cast<std::size_t>(document_size) != chars_read - document_start))
+        if (JSON_HEDLEY_UNLIKELY(static_cast<std::size_t>(document_size) != chars_read - document_start))
         {
             return bson_document_size_error(document_size, chars_read - document_start);
         }
@@ -21091,16 +21091,13 @@ class binary_reader
         {
             if (size_and_type.second != 0)
             {
-                if (size_and_type.second != 'N')
+                for (std::size_t i = 0; i < size_and_type.first; ++i)
                 {
-                    for (std::size_t i = 0; i < size_and_type.first; ++i)
+                    if (JSON_HEDLEY_UNLIKELY(!get_ubjson_size_value(dimlen, no_ndarray, size_and_type.second)))
                     {
-                        if (JSON_HEDLEY_UNLIKELY(!get_ubjson_size_value(dimlen, no_ndarray, size_and_type.second)))
-                        {
-                            return false;
-                        }
-                        dim.push_back(dimlen);
+                        return false;
                     }
+                    dim.push_back(dimlen);
                 }
             }
             else
@@ -21302,11 +21299,9 @@ class binary_reader
                     // follow the documented _ArrayType_, _ArraySize_, _ArrayData_ order
                     if (ndarray_dtype != 0)
                     {
+                        // every optimized type marker is_valid_optimized_type() accepts for BJData has a name
                         const char* type_name = bjd_type_name(ndarray_dtype);
-                        if (JSON_HEDLEY_UNLIKELY(type_name == nullptr))
-                        {
-                            return unexpected_byte("invalid byte", "type");
-                        }
+                        JSON_ASSERT(type_name != nullptr);
 
                         string_t type_key = "_ArrayType_";
                         string_t type = type_name; // sax->string() takes a reference
@@ -21389,12 +21384,6 @@ class binary_reader
         if (current == '$')
         {
             result.second = get();  // must not ignore 'N', because 'N' maybe the type
-            if (input_format == input_format_t::bjdata
-                    && JSON_HEDLEY_UNLIKELY(is_bjd_excluded_optimized_type(result.second)))
-            {
-                return optimized_type_error();
-            }
-
             if (JSON_HEDLEY_UNLIKELY(!unexpect_eof("type")))
             {
                 return false;
@@ -21422,7 +21411,7 @@ class binary_reader
             // an ndarray was read here only if the flag flipped; when it was
             // seeded true, get_ubjson_size_value() already rejected the nested
             // dimension vector
-            if (input_format == input_format_t::bjdata && is_ndarray && !inside_ndarray)
+            if (is_ndarray && !inside_ndarray)
             {
                 result.second |= (1 << 8); // use bit 8 to indicate ndarray, all UBJSON and BJData markers should be ASCII letters
             }
@@ -21432,7 +21421,7 @@ class binary_reader
         if (current == '#')
         {
             const bool is_error = get_ubjson_size_value(result.first, is_ndarray);
-            if (input_format == input_format_t::bjdata && is_ndarray && !inside_ndarray)
+            if (is_ndarray && !inside_ndarray)
             {
                 return last_byte_error(exception_id::unexpected_byte, "ndarray requires both type and size", "size");
             }
@@ -21606,7 +21595,6 @@ class binary_reader
     JSON_HEDLEY_NEVER_INLINE
     bool parse_ubjson_typed_run(std::size_t& remaining, const char_int_type type, std::true_type /*bulk*/)
     {
-        const bool bjdata = input_format == input_format_t::bjdata;
         switch (type)
         {
             case 'U':
@@ -21620,11 +21608,11 @@ class binary_reader
             case 'L':
                 return parse_ubjson_typed_run<std::int64_t>(remaining);
             case 'u':
-                return !bjdata || parse_ubjson_typed_run<std::uint16_t>(remaining);
+                return parse_ubjson_typed_run<std::uint16_t>(remaining);
             case 'm':
-                return !bjdata || parse_ubjson_typed_run<std::uint32_t>(remaining);
+                return parse_ubjson_typed_run<std::uint32_t>(remaining);
             case 'M':
-                return !bjdata || parse_ubjson_typed_run<std::uint64_t>(remaining);
+                return parse_ubjson_typed_run<std::uint64_t>(remaining);
             case 'd':
                 return parse_ubjson_typed_run<float>(remaining);
             case 'D':
@@ -21705,7 +21693,7 @@ class binary_reader
         // if bit-8 of size_and_type.second is set to 1, encode bjdata ndarray as an object in JData annotated array format (https://github.com/NeuroJSON/jdata):
         // {"_ArrayType_" : "typeid", "_ArraySize_" : [n1, n2, ...], "_ArrayData_" : [v1, v2, ...]}
 
-        if (input_format == input_format_t::bjdata && size_and_type.first != npos && (size_and_type.second & (1 << 8)) != 0)
+        if ((size_and_type.second & (1 << 8)) != 0)
         {
             size_and_type.second &= ~(static_cast<char_int_type>(1) << 8);  // use bit 8 to indicate ndarray, here we remove the bit to restore the type marker
 
@@ -21764,7 +21752,7 @@ class binary_reader
         }
 
         // If BJData type marker is 'B' decode as binary
-        if (input_format == input_format_t::bjdata && size_and_type.first != npos && size_and_type.second == 'B')
+        if (size_and_type.first != npos && size_and_type.second == 'B')
         {
             binary_t result;
             return get_binary(size_and_type.first, result) && sax->binary(result);
@@ -21813,7 +21801,7 @@ class binary_reader
         }
 
         // do not accept ND-array size in objects in BJData
-        if (input_format == input_format_t::bjdata && size_and_type.first != npos && (size_and_type.second & (1 << 8)) != 0)
+        if ((size_and_type.second & (1 << 8)) != 0)
         {
             return last_byte_error(exception_id::unexpected_byte, "BJData object does not support ND-array size in optimized format", "object");
         }
@@ -23094,8 +23082,9 @@ class binary_reader
   JSON_PRIVATE_UNLESS_TESTED:
     /*!
     @brief check whether @a marker is a type marker that may follow '$'
-    UBJSON: Z N T F i U I l L d D H C S [ {; BJData: the numeric markers and C
-    (BJData's excluded markers are rejected separately).
+    UBJSON: Z N T F i U I l L d D H C S [ {; BJData: the numeric markers and C.
+    The markers BJData excludes (F H N S T Z [ {) are the ones
+    binary_writer::is_bjdata_excluded_type_marker() lists; keep the two in sync.
     */
     bool is_valid_optimized_type(const char_int_type marker) const noexcept
     {
@@ -23130,20 +23119,6 @@ class binary_reader
         }
     }
 
-    /*!
-    @brief whether @a marker is excluded from BJData's optimized ND-array types
-    @return whether @a marker is one of 'F', 'H', 'N', 'S', 'T', 'Z', '[', '{'
-
-    Mirrors binary_writer's @ref binary_writer::is_bjdata_excluded_type_marker
-    "is_bjdata_excluded_type_marker()`, which encodes the same list the other
-    way; keep the two in sync.
-    */
-    static constexpr bool is_bjd_excluded_optimized_type(const char_int_type marker) noexcept
-    {
-        return marker == '[' || marker == '{' || marker == 'S' || marker == 'H'
-               || marker == 'T' || marker == 'F' || marker == 'N' || marker == 'Z';
-    }
-
     /*!
     @brief look up the ND-array element type name for a BJData dtype marker
     @return the type name ("uint8", "int8", ...), or nullptr if @a marker does
@@ -31101,22 +31076,19 @@ class binary_writer
                 // negative fixnum
                 write_number(static_cast<std::int8_t>(j.m_data.m_value.number_integer));
             }
-            else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int8_t>::min)() &&
-                     j.m_data.m_value.number_integer <= (std::numeric_limits<std::int8_t>::max)())
+            else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int8_t>::min)())
             {
                 // int 8
                 oa.write_character(to_char_type(0xD0));
                 write_number(static_cast<std::int8_t>(j.m_data.m_value.number_integer));
             }
-            else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int16_t>::min)() &&
-                     j.m_data.m_value.number_integer <= (std::numeric_limits<std::int16_t>::max)())
+            else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int16_t>::min)())
             {
                 // int 16
                 oa.write_character(to_char_type(0xD1));
                 write_number(static_cast<std::int16_t>(j.m_data.m_value.number_integer));
             }
-            else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int32_t>::min)() &&
-                     j.m_data.m_value.number_integer <= (std::numeric_limits<std::int32_t>::max)())
+            else if (j.m_data.m_value.number_integer >= (std::numeric_limits<std::int32_t>::min)())
             {
                 // int 32
                 oa.write_character(to_char_type(0xD2));
@@ -32032,7 +32004,7 @@ class binary_writer
     Containers, strings, high-precision numbers, booleans and null cannot be
     declared as the single type of an optimized container in BJData; such a
     container is written unoptimized. The reader rejects them with the same
-    list (binary_reader::is_bjd_excluded_optimized_type()).
+    list (binary_reader::is_valid_optimized_type()).
     */
     static constexpr bool is_bjdata_excluded_type_marker(const CharType marker) noexcept
     {
@@ -38373,10 +38345,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
     ////////////////////////
 
   JSON_PRIVATE_UNLESS_TESTED:
-    // the type and value of a basic_json without its base class (defined
-    // below); the destruction helpers of json_value use it as a stand-in
-    struct data;
-
     /*!
     @brief a JSON value
 
@@ -38553,28 +38521,27 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
         explicit json_value(array_t* value) noexcept : array(value) {}
 
 private:
-        // raw, allocation-free transfer of the type and value from src to
-        // dst: no set_parents()/assert_invariant() (the former is
-        // O(#children) per call under JSON_DIAGNOSTICS, which would make the
-        // walk below quadratic); dst takes ownership, src is left as
-        // value_t::null.
-        static void take(data& dst, data& src) noexcept
+        // raw, allocation-free transfer of m_data from src to dst: no
+        // set_parents()/assert_invariant() (the former is O(#children) per
+        // call under JSON_DIAGNOSTICS, which would make the walk below
+        // quadratic); dst takes ownership, src is left as value_t::null.
+        static void take(basic_json& dst, basic_json& src) noexcept
         {
-            dst.m_type = src.m_type;
-            dst.m_value = src.m_value;
-            src.m_type = value_t::null;
+            dst.m_data.m_type = src.m_data.m_type;
+            dst.m_data.m_value = src.m_data.m_value;
+            src.m_data.m_type = value_t::null;
         }
 
         // true if v is not an array/object, or is an already-empty one
-        static bool has_no_children(const data& v) noexcept
+        static bool has_no_children(const basic_json& v) noexcept
         {
-            if (v.m_type == value_t::array)
+            if (v.m_data.m_type == value_t::array)
             {
-                return v.m_value.array->empty();
+                return v.m_data.m_value.array->empty();
             }
-            if (v.m_type == value_t::object)
+            if (v.m_data.m_type == value_t::object)
             {
-                return v.m_value.object->empty();
+                return v.m_data.m_value.object->empty();
             }
             return true;
         }
@@ -38583,21 +38550,21 @@ private:
         // its own, i.e. v can be freed in one go (its destructor then
         // destroys scalars and empty containers only, so it recurses at most
         // one level deep); stops at the first structured child it meets
-        static bool all_children_leaves(const data& v) noexcept
+        static bool all_children_leaves(const basic_json& v) noexcept
         {
             // array_t only needs forward iterators (see custom_array_type)
-            if (v.m_type == value_t::array)
+            if (v.m_data.m_type == value_t::array)
             {
-                const auto& a = *v.m_value.array;
+                const auto& a = *v.m_data.m_value.array;
                 return std::all_of(a.begin(), a.end(), [](const basic_json & el)
                 {
-                    return has_no_children(el.m_data);
+                    return has_no_children(el);
                 });
             }
-            const auto& o = *v.m_value.object;
+            const auto& o = *v.m_data.m_value.object;
             return std::all_of(o.begin(), o.end(), [](const typename object_t::value_type & el)
             {
-                return has_no_children(el.second.m_data);
+                return has_no_children(el.second);
             });
         }
 
@@ -38629,30 +38596,30 @@ private:
 
         // the child of a non-empty array/object v that the walk in
         // destroy_container() continues with (see walk_child_it() above)
-        static data& walk_child(data& v)
+        static basic_json& walk_child(basic_json& v)
         {
-            if (v.m_type == value_t::array)
+            if (v.m_data.m_type == value_t::array)
             {
-                return v.m_value.array->back().m_data;
+                return v.m_data.m_value.array->back();
             }
-            JSON_ASSERT(v.m_type == value_t::object);
-            return walk_child_it(*v.m_value.object)->second.m_data;
+            JSON_ASSERT(v.m_data.m_type == value_t::object);
+            return walk_child_it(*v.m_data.m_value.object)->second;
         }
 
         // removes walk_child(v) from a non-empty array/object v; this never
         // allocates, and since it is only ever called when that child is a
         // scalar or an already-empty array/object, destroying it never
         // recurses more than one level deep (see destroy() below)
-        static void pop_walk_child(data& v)
+        static void pop_walk_child(basic_json& v)
         {
-            if (v.m_type == value_t::array)
+            if (v.m_data.m_type == value_t::array)
             {
-                v.m_value.array->pop_back();
+                v.m_data.m_value.array->pop_back();
             }
             else
             {
-                JSON_ASSERT(v.m_type == value_t::object);
-                v.m_value.object->erase(walk_child_it(*v.m_value.object));
+                JSON_ASSERT(v.m_data.m_type == value_t::object);
+                v.m_data.m_value.object->erase(walk_child_it(*v.m_data.m_value.object));
             }
         }
 
@@ -38661,23 +38628,23 @@ private:
         // recurses at most one level deep); this is the same allocator-based
         // free the old recursive implementation used, just factored out so
         // every level of the walk in destroy() can share it
-        static void free_container(data& v) noexcept
+        static void free_container(basic_json& v) noexcept
         {
             JSON_ASSERT(has_no_children(v) || all_children_leaves(v));
-            if (v.m_type == value_t::array)
+            if (v.m_data.m_type == value_t::array)
             {
                 AllocatorType<array_t> alloc;
-                std::allocator_traits<decltype(alloc)>::destroy(alloc, v.m_value.array);
-                std::allocator_traits<decltype(alloc)>::deallocate(alloc, v.m_value.array, 1);
+                std::allocator_traits<decltype(alloc)>::destroy(alloc, v.m_data.m_value.array);
+                std::allocator_traits<decltype(alloc)>::deallocate(alloc, v.m_data.m_value.array, 1);
             }
             else
             {
-                JSON_ASSERT(v.m_type == value_t::object);
+                JSON_ASSERT(v.m_data.m_type == value_t::object);
                 AllocatorType<object_t> alloc;
-                std::allocator_traits<decltype(alloc)>::destroy(alloc, v.m_value.object);
-                std::allocator_traits<decltype(alloc)>::deallocate(alloc, v.m_value.object, 1);
+                std::allocator_traits<decltype(alloc)>::destroy(alloc, v.m_data.m_value.object);
+                std::allocator_traits<decltype(alloc)>::deallocate(alloc, v.m_data.m_value.object, 1);
             }
-            v.m_type = value_t::null; // avoid a double free if v is later destructed
+            v.m_data.m_type = value_t::null; // avoid a double free if v is later destructed
         }
 
 public:
@@ -38708,16 +38675,14 @@ public:
 
             // This json_value is not itself a basic_json, so the top-level
             // container is first moved into a local stand-in ("cur"); a
-            // default-constructed data has a null pointer in its m_value
-            // (see data::m_value's initializer), so swapping it with *this
-            // leaves this union's own pointer null, and it is never looked
-            // at or freed a second time. The stand-ins are data, not
-            // basic_json, so that destroying a value never constructs a
-            // CustomBaseClass, whose constructor may allocate (#5995).
-            data cur;
-            cur.m_type = t;
+            // default-constructed basic_json has a null pointer in its
+            // m_value (see data::m_value's initializer), so swapping it
+            // with *this leaves this union's own pointer null, and it is
+            // never looked at or freed a second time.
+            basic_json cur;
+            cur.m_data.m_type = t;
             using std::swap;
-            swap(cur.m_value, *this);
+            swap(cur.m_data.m_value, *this);
 
             destroy_bounded(cur, 0);
         }
@@ -38726,7 +38691,7 @@ public:
         // The first recursion_depth_limit() levels are destroyed by plain
         // recursion, which is the fastest way and cannot throw; anything
         // nested deeper is handed to destroy_walk(), which uses no call stack.
-        static void destroy_bounded(data& v, std::size_t depth) noexcept
+        static void destroy_bounded(basic_json& v, std::size_t depth) noexcept
         {
             if (JSON_HEDLEY_UNLIKELY(depth >= detail::recursion_depth_limit()))
             {
@@ -38734,23 +38699,23 @@ public:
                 return;
             }
 
-            if (v.m_type == value_t::array)
+            if (v.m_data.m_type == value_t::array)
             {
-                for (auto& child : *v.m_value.array)
+                for (auto& child : *v.m_data.m_value.array)
                 {
                     if (child.m_data.m_type == value_t::array || child.m_data.m_type == value_t::object)
                     {
-                        destroy_bounded(child.m_data, depth + 1);
+                        destroy_bounded(child, depth + 1);
                     }
                 }
             }
             else
             {
-                for (auto& child : *v.m_value.object)
+                for (auto& child : *v.m_data.m_value.object)
                 {
                     if (child.second.m_data.m_type == value_t::array || child.second.m_data.m_type == value_t::object)
                     {
-                        destroy_bounded(child.second.m_data, depth + 1);
+                        destroy_bounded(child.second, depth + 1);
                     }
                 }
             }
@@ -38777,15 +38742,18 @@ public:
         // array/object, which neither allocates nor recurses more
         // than one level deep.
         //
-        // Like destroy_container(), the walk keeps the containers it
-        // holds in data stand-ins (cur, prev, gp, tmp), which never
-        // construct a CustomBaseClass.
-        static void destroy_walk(data& top) noexcept
+        // This json_value is not itself a basic_json, so the
+        // top-level container is first moved into a local stand-in
+        // ("cur"); a default-constructed basic_json has a null
+        // pointer in its m_value (see data::m_value's initializer),
+        // so swapping it with *this leaves this union's own pointer
+        // null, and it is never looked at or freed a second time.
+        static void destroy_walk(basic_json& top) noexcept
         {
-            data cur;
+            basic_json cur;
             take(cur, top);
 
-            data prev; // value_t::null: no parent
+            basic_json prev; // value_t::null: no parent
 
             // true right after cur was entered: its children have not been
             // looked at yet. If none of them has children of its own, cur is
@@ -38799,7 +38767,7 @@ public:
                 fresh = false;
                 if (has_no_children(cur) || (fresh_now && all_children_leaves(cur)))
                 {
-                    if (prev.m_type == value_t::null)
+                    if (prev.m_data.m_type == value_t::null)
                     {
                         free_container(cur);
                         return; // back at the top with nothing left to do
@@ -38808,7 +38776,7 @@ public:
                     // ascend: detach the grandparent link from prev's
                     // walk_child() slot, drop that (now null) slot, free cur
                     // (it is empty), then move up one level
-                    data gp;
+                    basic_json gp;
                     take(gp, walk_child(prev));
                     pop_walk_child(prev);
 
@@ -38819,7 +38787,7 @@ public:
                     continue;
                 }
 
-                data& cur_child_ref = walk_child(cur);
+                basic_json& cur_child_ref = walk_child(cur);
 
                 if (has_no_children(cur_child_ref))
                 {
@@ -38831,7 +38799,7 @@ public:
                 // descend into the non-empty child, reversing the
                 // link: its slot takes over prev, and the child becomes
                 // the new cur
-                data tmp;
+                basic_json tmp;
                 take(tmp, cur_child_ref);
                 take(cur_child_ref, prev);
                 take(prev, cur);

tests/src/unit-bjdata.cpp

@@ -78,23 +78,25 @@ TEST_CASE("BJData")
 {
     SECTION("binary_reader BJData lookup tables")
     {
-        // both lookups are static member functions
+        // bjd_type_name() is static; is_valid_optimized_type() needs a BJData reader
         std::vector<std::uint8_t> const data;
         using reader_t = nlohmann::detail::binary_reader<json, decltype(nlohmann::detail::input_adapter(data))>;
 
-        // the excluded optimized-type markers must match binary_writer's
-        // is_bjdata_excluded_type_marker(), which encodes the same 8 markers
+        // the markers BJData excludes as optimized types must match
+        // binary_writer's is_bjdata_excluded_type_marker(), which encodes the
+        // same 8 markers
+        reader_t reader(nlohmann::detail::input_adapter(data), json::input_format_t::bjdata);
         for (const char marker :
-                {'[', '{', 'S', 'H', 'T', 'F', 'N', 'Z'
+                {'[', '{', 'S', 'H', 'T', 'F', 'N', 'Z', 'x'
                 })
         {
-            CHECK(reader_t::is_bjd_excluded_optimized_type(static_cast<unsigned char>(marker)));
+            CHECK(!reader.is_valid_optimized_type(static_cast<unsigned char>(marker)));
         }
         for (const char marker :
-                {'U', 'i', 'u', 'I', 'm', 'l', 'M', 'L', 'd', 'D', 'C', 'B', 'x'
+                {'U', 'i', 'u', 'I', 'm', 'l', 'M', 'L', 'd', 'D', 'C', 'B', 'h'
                 })
         {
-            CHECK(!reader_t::is_bjd_excluded_optimized_type(static_cast<unsigned char>(marker)));
+            CHECK(reader.is_valid_optimized_type(static_cast<unsigned char>(marker)));
         }
 
         // every dtype marker must round-trip to its ND-array type name

tests/src/unit-custom-base-class.cpp

@@ -7,7 +7,6 @@
 // SPDX-License-Identifier: MIT
 
 #include <algorithm>
-#include <memory>
 #include <set>
 #include <sstream>
 #include <string>
@@ -496,92 +495,3 @@ TEST_CASE("Regression test for issue #5674 - copy construction must not require
         CHECK(*c == 1);
     }
 }
-
-// counts the constructions and destructions of the base class (#5995)
-struct counting_base
-{
-    static std::size_t constructions;
-    static std::size_t destructions;
-
-    counting_base() noexcept
-    {
-        ++constructions;
-    }
-    counting_base(const counting_base& /*unused*/) noexcept
-    {
-        ++constructions;
-    }
-    counting_base(counting_base&& /*unused*/) noexcept
-    {
-        ++constructions;
-    }
-    counting_base& operator=(const counting_base& /*unused*/) noexcept = default;
-    counting_base& operator=(counting_base&& /*unused*/) noexcept = default;
-    ~counting_base()
-    {
-        ++destructions;
-    }
-
-    static void reset() noexcept
-    {
-        constructions = 0;
-        destructions = 0;
-    }
-};
-
-std::size_t counting_base::constructions = 0;
-std::size_t counting_base::destructions = 0;
-
-using json_with_counting_base = nlohmann::json::with_base_class_t<counting_base>;
-
-TEST_CASE("destroying a value does not construct the custom base class")
-{
-    // regression test for #5995: destroying an array or object used a
-    // default-constructed basic_json as a stand-in, which constructed a base
-    // class that belonged to no value (and could allocate in the noexcept
-    // destructor); now every destroyed value destroys exactly its own base
-    // object, and none is constructed
-    const auto check_destroy = [](std::unique_ptr<json_with_counting_base> j, std::size_t values)
-    {
-        counting_base::reset();
-        j = nullptr;
-        CHECK(counting_base::constructions == 0);
-        CHECK(counting_base::destructions == values);
-    };
-
-    SECTION("scalar")
-    {
-        check_destroy(std::unique_ptr<json_with_counting_base>(new json_with_counting_base(42)), 1);
-    }
-
-    SECTION("array and object")
-    {
-        check_destroy(std::unique_ptr<json_with_counting_base>(new json_with_counting_base(json_with_counting_base::parse(R"({"a":[1,2,3],"b":{"c":true}})"))), 7);
-        check_destroy(std::unique_ptr<json_with_counting_base>(new json_with_counting_base(json_with_counting_base::array())), 1);
-    }
-
-    SECTION("nested deeper than the recursion depth limit")
-    {
-        // alternating arrays and objects, so both kinds of container are
-        // walked by destroy_walk()
-        constexpr std::size_t depth = 1000;
-        std::unique_ptr<json_with_counting_base> j(new json_with_counting_base(json_with_counting_base::array({1, 2})));
-        for (std::size_t i = 0; i < depth; ++i)
-        {
-            json_with_counting_base outer = (i % 2 == 0) ? json_with_counting_base::object() : json_with_counting_base::array();
-            if (outer.is_object())
-            {
-                outer["x"] = std::move(*j);
-                outer["y"] = nullptr;
-            }
-            else
-            {
-                outer.push_back(std::move(*j));
-                outer.push_back(true);
-            }
-            *j = std::move(outer);
-        }
-        // the innermost array with its two elements, and two values per level
-        check_destroy(std::move(j), 3 + (2 * depth));
-    }
-}