Fall back to plain-object encoding when _ArrayType_ is not a string

write_bjdata_ndarray() looked up _ArrayType_ by calling get<string_t>()
directly, which throws type_error.302 when the annotation is not a
string (e.g. a number, null, boolean, array, or object). Per the
documented BJData ndarray contract, an object only qualifies for the
compact ndarray encoding if _ArrayType_ names a known type; anything
else must fall back to plain-object encoding, the same way an unknown
type-name string already does.

Add an is_string() check before the get<string_t>() call so a
non-string _ArrayType_ takes the existing "unrecognized type name"
fallback path instead of throwing.

Fixes #5398.

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

docs/mkdocs/docs/api/basic_json/insert.md

@@ -88,8 +88,6 @@ Strong exception safety: if an exception occurs, the original value stays intact
       do not belong to the same JSON value; example: `"iterators do not fit"`
     - Throws [`invalid_iterator.211`](../../home/exceptions.md#jsonexceptioninvalid_iterator211) if `first` or `last`
       are iterators into container for which insert is called; example: `"passed iterators may not belong to container"`
-    - Throws [`invalid_iterator.202`](../../home/exceptions.md#jsonexceptioninvalid_iterator202) if `first` or `last`
-      do not point to an array; example: `"iterators first and last must point to arrays"`
 4. The function can throw the following exceptions:
     - Throws [`type_error.309`](../../home/exceptions.md#jsonexceptiontype_error309) if called on JSON values other than
       arrays; example: `"cannot use insert() with string"`

docs/mkdocs/docs/home/exceptions.md

@@ -933,17 +933,6 @@ BSON stores the length of documents, arrays, strings, and binary values in a sig
     [`to_bson`](../api/basic_json/to_bson.md) produced documents with negative length prefixes that
     [`from_bson`](../api/basic_json/from_bson.md) rejected.
 
-### json.exception.out_of_range.413
-
-MessagePack's ext type and BSON's binary subtype are each stored in a single byte. This exception is thrown when serializing a
-[`byte_container_with_subtype`](../api/byte_container_with_subtype/index.md) whose subtype exceeds 255.
-
-!!! failure "Example message"
-
-    ```
-    [json.exception.out_of_range.413] subtype 70000 is too large for the MessagePack ext type (max 255)
-    ```
-
 ## Further exceptions
 
 This exception is thrown in case of errors that cannot be classified with the

include/nlohmann/detail/input/json_sax.hpp

@@ -8,11 +8,10 @@
 
 #pragma once
 
-#include <algorithm> // find_if
 #include <cstddef>
 #include <string> // string
 #include <type_traits> // enable_if_t
-#include <utility> // move, pair
+#include <utility> // move
 #include <vector> // vector
 
 #include <nlohmann/detail/exceptions.hpp>
@@ -250,7 +249,7 @@ class json_sax_dom_parser
 
         if (JSON_HEDLEY_UNLIKELY(len != detail::unknown_size() && len > ref_stack.back()->max_size()))
         {
-            return parse_error(0, "", out_of_range::create(408, concat("excessive object size: ", std::to_string(len)), ref_stack.back()));
+            JSON_THROW(out_of_range::create(408, concat("excessive object size: ", std::to_string(len)), ref_stack.back()));
         }
 
         return true;
@@ -299,7 +298,7 @@ class json_sax_dom_parser
 
         if (JSON_HEDLEY_UNLIKELY(len != detail::unknown_size() && len > ref_stack.back()->max_size()))
         {
-            return parse_error(0, "", out_of_range::create(408, concat("excessive array size: ", std::to_string(len)), ref_stack.back()));
+            JSON_THROW(out_of_range::create(408, concat("excessive array size: ", std::to_string(len)), ref_stack.back()));
         }
 
         return true;
@@ -569,7 +568,7 @@ class json_sax_dom_callback_parser
             // check object limit
             if (JSON_HEDLEY_UNLIKELY(len != detail::unknown_size() && len > ref_stack.back()->max_size()))
             {
-                return parse_error(0, "", out_of_range::create(408, concat("excessive object size: ", std::to_string(len)), ref_stack.back()));
+                JSON_THROW(out_of_range::create(408, concat("excessive object size: ", std::to_string(len)), ref_stack.back()));
             }
         }
         return true;
@@ -586,17 +585,7 @@ class json_sax_dom_callback_parser
         // add discarded value at the given key and store the reference for later
         if (keep && ref_stack.back())
         {
-            auto& obj = *ref_stack.back()->m_data.m_value.object;
-            const auto it = obj.find(val);
-            if (it != obj.end())
-            {
-                // this is a duplicate key (legal in JSON); remember its
-                // current value so it can be restored later if the new
-                // value is rejected by the callback, instead of being
-                // erased together with the discarded placeholder
-                duplicate_key_stash.emplace_back(&(it->second), it->second);
-            }
-            object_element = &(obj[val] = discarded);
+            object_element = &(ref_stack.back()->m_data.m_value.object->operator[](val) = discarded);
         }
 
         return true;
@@ -608,18 +597,13 @@ class json_sax_dom_callback_parser
         {
             if (!callback(static_cast<int>(ref_stack.size()) - 1, parse_event_t::object_end, *ref_stack.back()))
             {
-                // discard object, unless this slot holds a duplicate key's
-                // previous value pending restoration, in which case that
-                // value is restored instead of being discarded
-                if (!resolve_duplicate_key_stash(ref_stack.back(), true))
-                {
-                    *ref_stack.back() = discarded;
+                // discard object
+                *ref_stack.back() = discarded;
 
 #if JSON_DIAGNOSTIC_POSITIONS
-                    // Set start/end positions for discarded object.
-                    handle_diagnostic_positions_for_json_value(*ref_stack.back());
+                // Set start/end positions for discarded object.
+                handle_diagnostic_positions_for_json_value(*ref_stack.back());
 #endif
-                }
             }
             else
             {
@@ -633,10 +617,6 @@ class json_sax_dom_callback_parser
 #endif
 
                 ref_stack.back()->set_parents();
-                // this object is finally, definitively kept; drop any
-                // pending duplicate-key stash entry for its slot since it
-                // can no longer be restored
-                resolve_duplicate_key_stash(ref_stack.back(), false);
             }
         }
 
@@ -679,7 +659,7 @@ class json_sax_dom_callback_parser
             // check array limit
             if (JSON_HEDLEY_UNLIKELY(len != detail::unknown_size() && len > ref_stack.back()->max_size()))
             {
-                return parse_error(0, "", out_of_range::create(408, concat("excessive array size: ", std::to_string(len)), ref_stack.back()));
+                JSON_THROW(out_of_range::create(408, concat("excessive array size: ", std::to_string(len)), ref_stack.back()));
             }
         }
 
@@ -706,25 +686,16 @@ class json_sax_dom_callback_parser
 #endif
 
                 ref_stack.back()->set_parents();
-                // this array is finally, definitively kept; drop any
-                // pending duplicate-key stash entry for its slot since it
-                // can no longer be restored
-                resolve_duplicate_key_stash(ref_stack.back(), false);
             }
             else
             {
-                // discard array, unless this slot holds a duplicate key's
-                // previous value pending restoration, in which case that
-                // value is restored instead of being discarded
-                if (!resolve_duplicate_key_stash(ref_stack.back(), true))
-                {
-                    *ref_stack.back() = discarded;
+                // discard array
+                *ref_stack.back() = discarded;
 
 #if JSON_DIAGNOSTIC_POSITIONS
-                    // Set start/end positions for discarded array.
-                    handle_diagnostic_positions_for_json_value(*ref_stack.back());
+                // Set start/end positions for discarded array.
+                handle_diagnostic_positions_for_json_value(*ref_stack.back());
 #endif
-                }
             }
         }
 
@@ -838,48 +809,14 @@ class json_sax_dom_callback_parser
     }
 #endif
 
-    /// if there is a pending duplicate-key stash entry for this exact slot,
-    /// remove it from the stash; if restore_value is true, the stashed
-    /// previous value is moved back into the slot first (use this when the
-    /// new value at that slot was rejected); otherwise the stash entry is
-    /// simply dropped (use this when the new value was accepted, so it
-    /// correctly supersedes the old one and no restore should ever happen
-    /// for this slot again)
-    /// @return whether a matching stash entry was found (and processed)
-    bool resolve_duplicate_key_stash(BasicJsonType* slot, bool restore_value)
-    {
-        const auto it = std::find_if(duplicate_key_stash.begin(), duplicate_key_stash.end(),
-                                     [slot](const std::pair<BasicJsonType*, BasicJsonType>& entry)
-        {
-            return entry.first == slot;
-        });
-
-        if (it == duplicate_key_stash.end())
-        {
-            return false;
-        }
-
-        if (restore_value)
-        {
-            *slot = std::move(it->second);
-        }
-        duplicate_key_stash.erase(it);
-        return true;
-    }
-
-    /// remove the discarded value the callback rejected from its parent,
-    /// unless it is a duplicate key's slot with a stashed previous value,
-    /// in which case that previous value is restored instead
-    void remove_discarded_value(BasicJsonType& parent)
+    /// remove the discarded value the callback rejected from its parent
+    static void remove_discarded_value(BasicJsonType& parent)
     {
         for (auto it = parent.begin(); it != parent.end(); ++it)
         {
             if (it->is_discarded())
             {
-                if (!resolve_duplicate_key_stash(&(*it), true))
-                {
-                    parent.erase(it);
-                }
+                parent.erase(it);
                 break;
             }
         }
@@ -977,16 +914,6 @@ class json_sax_dom_callback_parser
 
         JSON_ASSERT(object_element);
         *object_element = std::move(value);
-        if (!skip_callback)
-        {
-            // this scalar value finally, definitively replaces whatever was
-            // at this slot; drop any pending duplicate-key stash entry for
-            // it since it can no longer be restored (a container value at
-            // this slot is resolved later, in end_object()/end_array(),
-            // since skip_callback is true for the placeholder handling that
-            // happens here for those)
-            resolve_duplicate_key_stash(object_element, false);
-        }
         return {true, object_element};
     }
 
@@ -1000,12 +927,6 @@ class json_sax_dom_callback_parser
     std::vector<bool> key_keep_stack {}; // NOLINT(readability-redundant-member-init)
     /// helper to hold the reference for the next object element
     BasicJsonType* object_element = nullptr;
-    /// stash of (slot pointer, previous value) for object members that
-    /// already existed when key() was called again for the same key
-    /// (duplicate keys); used to restore the previous value if the new
-    /// value is later rejected by the callback, instead of erasing the
-    /// member entirely
-    std::vector<std::pair<BasicJsonType*, BasicJsonType>> duplicate_key_stash {};
     /// whether a syntax error occurred
     bool errored = false;
     /// callback function

include/nlohmann/detail/output/binary_writer.hpp

@@ -688,11 +688,6 @@ class binary_writer
                 // step 1.5: if this is an ext type, write the subtype
                 if (use_ext)
                 {
-                    if (JSON_HEDLEY_UNLIKELY(j.m_data.m_value.binary->subtype() > (std::numeric_limits<std::uint8_t>::max)()))
-                    {
-                        JSON_THROW(out_of_range::create(413, concat("subtype ", std::to_string(j.m_data.m_value.binary->subtype()), " is too large for the MessagePack ext type (max 255)"), &j));
-                    }
-
                     write_number(static_cast<std::int8_t>(j.m_data.m_value.binary->subtype()));
                 }
 
@@ -1192,12 +1187,6 @@ class binary_writer
         write_bson_entry_header(name, 0x05);
 
         write_number<std::int32_t>(to_bson_length(value.size()), true);
-
-        if (value.has_subtype() && JSON_HEDLEY_UNLIKELY(value.subtype() > (std::numeric_limits<std::uint8_t>::max)()))
-        {
-            JSON_THROW(out_of_range::create(413, concat("subtype ", std::to_string(value.subtype()), " is too large for the BSON binary subtype (max 255)"), nullptr));
-        }
-
         write_number(value.has_subtype() ? static_cast<std::uint8_t>(value.subtype()) : static_cast<std::uint8_t>(0x00));
 
         oa->write_characters(reinterpret_cast<const CharType*>(value.data()), value.size());
@@ -1658,6 +1647,20 @@ class binary_writer
         return 'D';  // float 64
     }
 
+    /*!
+    @brief checks whether a JSON number fits into @a TargetType
+    @param[in] el a JSON number of either the signed or unsigned integer kind
+    @return whether @a el's value can be represented by @a TargetType without
+            wrapping, regardless of which of the two kinds it is stored as
+    */
+    template<typename TargetType>
+    static bool bjdata_ndarray_value_in_range(const BasicJsonType& el)
+    {
+        return el.is_number_unsigned()
+               ? value_in_range_of<TargetType>(el.template get<std::uint64_t>())
+               : value_in_range_of<TargetType>(el.template get<std::int64_t>());
+    }
+
     /*!
     @return false if the object is successfully converted to a bjdata ndarray, true if the type or size is invalid
     */
@@ -1669,6 +1672,16 @@ class binary_writer
         };
 
         string_t key = "_ArrayType_";
+        // the type name is looked up as a string below; a non-string
+        // annotation (e.g. a number, null, or an array) cannot name a known
+        // dtype, so it is treated the same as an unrecognized type name and
+        // falls back to a plain object encoding instead of throwing
+        // type_error.302 out of get<string_t>()
+        if (!value.at(key).is_string())
+        {
+            return true;
+        }
+
         // use get<string_t>() instead of static_cast<string_t> to avoid an
         // ambiguous conversion under explicit instantiation on C++17 (see #4825)
         auto it = bjdtype.find(value.at(key).template get<string_t>());
@@ -1678,6 +1691,16 @@ class binary_writer
         }
         CharType dtype = it->second;
 
+        // the 'B' (byte) marker is only defined by BJData Draft 3; emitting it
+        // under the default Draft 2 mode would produce a stream that Draft 2
+        // readers reject, so such an object falls back to a plain object
+        // encoding instead (see the "Binary values" section of the BJData
+        // documentation)
+        if (dtype == 'B' && bjdata_version != bjdata_version_t::draft3)
+        {
+            return true;
+        }
+
         key = "_ArraySize_";
         // the dimensions are written verbatim as the header length below, so a
         // value that is not an array cannot produce a valid one: null emits 'Z'
@@ -1742,6 +1765,60 @@ class binary_writer
             }
         }
 
+        // every element is cast to the (possibly narrower) C++ type matching
+        // dtype below; a value that does not fit that type would silently
+        // wrap (integers) or overflow to infinity (the "single" precision
+        // float) instead of being reported, so such an object falls back to
+        // a plain object encoding as well
+        for (const auto& el : value.at(key))
+        {
+            bool in_range = true;
+            switch (dtype)
+            {
+                case 'U':
+                case 'C':
+                case 'B':
+                    in_range = bjdata_ndarray_value_in_range<std::uint8_t>(el);
+                    break;
+                case 'i':
+                    in_range = bjdata_ndarray_value_in_range<std::int8_t>(el);
+                    break;
+                case 'u':
+                    in_range = bjdata_ndarray_value_in_range<std::uint16_t>(el);
+                    break;
+                case 'I':
+                    in_range = bjdata_ndarray_value_in_range<std::int16_t>(el);
+                    break;
+                case 'm':
+                    in_range = bjdata_ndarray_value_in_range<std::uint32_t>(el);
+                    break;
+                case 'l':
+                    in_range = bjdata_ndarray_value_in_range<std::int32_t>(el);
+                    break;
+                case 'M':
+                    in_range = bjdata_ndarray_value_in_range<std::uint64_t>(el);
+                    break;
+                case 'L':
+                    in_range = bjdata_ndarray_value_in_range<std::int64_t>(el);
+                    break;
+                case 'd':
+                {
+                    const auto dval = el.template get<double>();
+                    in_range = !std::isfinite(dval) ||
+                               (dval >= static_cast<double>(std::numeric_limits<float>::lowest()) &&
+                                dval <= static_cast<double>((std::numeric_limits<float>::max)()));
+                    break;
+                }
+                default:
+                    // 'D' (double) already spans the full range of number_float_t
+                    break;
+            }
+            if (!in_range)
+            {
+                return true;
+            }
+        }
+
         oa->write_character('[');
         oa->write_character('$');
         oa->write_character(dtype);

include/nlohmann/json.hpp

@@ -3425,12 +3425,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
             JSON_THROW(invalid_iterator::create(211, "passed iterators may not belong to container", this));
         }
 
-        // passed iterators must belong to arrays
-        if (JSON_HEDLEY_UNLIKELY(!first.m_object->is_array()))
-        {
-            JSON_THROW(invalid_iterator::create(202, "iterators first and last must point to arrays", this));
-        }
-
         // insert to array and return iterator
         return insert_iterator(pos, first.m_it.array_iterator, last.m_it.array_iterator);
     }
@@ -3579,7 +3573,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
         {
             using std::swap;
             swap(*(m_data.m_value.array), other);
-            set_parents();
         }
         else
         {
@@ -3596,7 +3589,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
         {
             using std::swap;
             swap(*(m_data.m_value.object), other);
-            set_parents();
         }
         else
         {
@@ -5165,139 +5157,34 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
 
             case value_t::object:
             {
-                // first pass: record, for every source key, whether it is
-                // common to both objects (in source's iteration order) or
-                // was deleted (i.e., in source but not in target) -- this is
-                // a by-product of the target.find() call already needed to
-                // tell the two cases apart, so it adds no extra lookups. The
-                // "remove" ops themselves are emitted later, interleaved
-                // with the recursive per-key diffs in the fast path below,
-                // to match source's original iteration order (as the
-                // original, pre-reordering-aware implementation did) instead
-                // of grouping all removes before all recursive diffs.
-                std::vector<typename object_t::key_type> common_keys_source_order;
+                // first pass: traverse this object's elements
                 for (auto it = source.cbegin(); it != source.cend(); ++it)
                 {
+                    // escape the key name to be used in a JSON patch
+                    const auto path_key = detail::concat<string_t>(path, '/', detail::escape(it.key()));
+
                     if (target.find(it.key()) != target.end())
                     {
-                        common_keys_source_order.push_back(it.key());
-                    }
-                }
-
-                // second pass: find keys that were added (i.e., in target but
-                // not in source), and record the keys common to both, in
-                // target's iteration order -- again a by-product of the
-                // source.find() call already needed to detect added keys. At
-                // the same time, determine whether every added key comes
-                // after every common key in target's order (a precondition
-                // for the fast path below, which only ever appends new keys
-                // at the very end): for an object_t whose iteration order is
-                // a pure function of the key set (e.g. the default std::map,
-                // which always iterates in sorted key order), the order
-                // check further below is always true and this whole
-                // mechanism is effectively a no-op; it only matters for a
-                // reorderable object_t such as the one backing `ordered_json`.
-                // patch ops for keys that were added (i.e., in target but not
-                // in source); built here so the fast path below can reuse
-                // them without a second source.find() per target key. Only
-                // used by the fast path -- the slow (reordering) path
-                // rebuilds "add" ops for every key itself.
-                std::vector<typename object_t::key_type> common_keys_target_order;
-                basic_json added_ops(value_t::array);
-                bool new_keys_form_suffix = true;
-                bool seen_new_key = false;
-                for (auto it = target.cbegin(); it != target.cend(); ++it)
-                {
-                    if (source.find(it.key()) == source.end())
-                    {
-                        seen_new_key = true;
-                        const auto path_key = detail::concat<string_t>(path, '/', detail::escape(it.key()));
-                        added_ops.push_back(
-                        {
-                            {"op", "add"}, {"path", path_key},
-                            {"value", it.value()}
-                        });
+                        // recursive call to compare object values at key it
+                        auto temp_diff = diff(it.value(), target[it.key()], path_key);
+                        result.insert(result.end(), temp_diff.begin(), temp_diff.end());
                     }
                     else
                     {
-                        common_keys_target_order.push_back(it.key());
-                        if (seen_new_key)
-                        {
-                            new_keys_form_suffix = false;
-                        }
-                    }
-                }
-
-                if (common_keys_source_order == common_keys_target_order && new_keys_form_suffix)
-                {
-                    // fast path: order of common keys already matches (or the
-                    // object_t's iteration order does not depend on
-                    // insertion history), so a plain per-key recursive diff
-                    // is correct and minimal, as before. common_keys_source_order
-                    // is, by construction, the subsequence of source's keys
-                    // that are common to both objects, in source's iteration
-                    // order -- so it can be walked in lockstep with `source`
-                    // using a cheap key comparison instead of another lookup.
-                    // Deleted keys (those source keys not in common_keys_source_order)
-                    // are interleaved here too, in source's original order, to
-                    // match the historical (pre-reordering-aware) output order.
-                    auto common_it = common_keys_source_order.cbegin();
-                    for (auto it = source.cbegin(); it != source.cend(); ++it)
-                    {
-                        if (common_it != common_keys_source_order.cend() && it.key() == *common_it)
-                        {
-                            const auto path_key = detail::concat<string_t>(path, '/', detail::escape(it.key()));
-                            auto temp_diff = diff(it.value(), target[it.key()], path_key);
-                            result.insert(result.end(), temp_diff.begin(), temp_diff.end());
-                            ++common_it;
-                        }
-                        else
-                        {
-                            // found a key that is not in target -> remove it
-                            const auto path_key = detail::concat<string_t>(path, '/', detail::escape(it.key()));
-                            result.push_back(object(
-                            {
-                                {"op", "remove"}, {"path", path_key}
-                            }));
-                        }
-                    }
-
-                    // append the "add" ops for brand-new keys collected above
-                    // during the pass over target -- no second source.find()
-                    // per target key needed
-                    result.insert(result.end(), added_ops.begin(), added_ops.end());
-                }
-                else
-                {
-                    // slow path: the common keys are in a different relative
-                    // order in source and target (only possible for a
-                    // reorderable object_t like ordered_map). Building a
-                    // minimal reordering patch is a nontrivial (LCS-like)
-                    // problem; instead, remove every source key -- both
-                    // deleted keys (which must be removed regardless) and
-                    // common keys (removed so they can be re-added in
-                    // target's order) -- and re-add every key that should
-                    // remain, with its final target value, in target's
-                    // order. basic_json::patch()'s "add" operation on an
-                    // object uses operator[], which appends at the end for a
-                    // vector-backed insertion-ordered map when the key does
-                    // not already exist -- so removing a key and then adding
-                    // it moves it to the end, fixing its position.
-                    for (auto it = source.cbegin(); it != source.cend(); ++it)
-                    {
-                        const auto path_key = detail::concat<string_t>(path, '/', detail::escape(it.key()));
+                        // found a key that is not in o -> remove it
                         result.push_back(object(
                         {
                             {"op", "remove"}, {"path", path_key}
                         }));
                     }
+                }
 
-                    // add every key that is either common (just removed
-                    // above) or brand new, in target's iteration order, so
-                    // that the final order after applying the patch matches
-                    // target exactly
-                    for (auto it = target.cbegin(); it != target.cend(); ++it)
+                // second pass: traverse other object's elements
+                for (auto it = target.cbegin(); it != target.cend(); ++it)
+                {
+                    if (source.find(it.key()) == source.end())
                     {
+                        // found a key that is not in this -> add it
                         const auto path_key = detail::concat<string_t>(path, '/', detail::escape(it.key()));
                         result.push_back(
                         {

single_include/nlohmann/json.hpp

@@ -7768,11 +7768,10 @@ NLOHMANN_JSON_NAMESPACE_END
 
 
 
-#include <algorithm> // find_if
 #include <cstddef>
 #include <string> // string
 #include <type_traits> // enable_if_t
-#include <utility> // move, pair
+#include <utility> // move
 #include <vector> // vector
 
 // #include <nlohmann/detail/exceptions.hpp>
@@ -9778,7 +9777,7 @@ class json_sax_dom_parser
 
         if (JSON_HEDLEY_UNLIKELY(len != detail::unknown_size() && len > ref_stack.back()->max_size()))
         {
-            return parse_error(0, "", out_of_range::create(408, concat("excessive object size: ", std::to_string(len)), ref_stack.back()));
+            JSON_THROW(out_of_range::create(408, concat("excessive object size: ", std::to_string(len)), ref_stack.back()));
         }
 
         return true;
@@ -9827,7 +9826,7 @@ class json_sax_dom_parser
 
         if (JSON_HEDLEY_UNLIKELY(len != detail::unknown_size() && len > ref_stack.back()->max_size()))
         {
-            return parse_error(0, "", out_of_range::create(408, concat("excessive array size: ", std::to_string(len)), ref_stack.back()));
+            JSON_THROW(out_of_range::create(408, concat("excessive array size: ", std::to_string(len)), ref_stack.back()));
         }
 
         return true;
@@ -10097,7 +10096,7 @@ class json_sax_dom_callback_parser
             // check object limit
             if (JSON_HEDLEY_UNLIKELY(len != detail::unknown_size() && len > ref_stack.back()->max_size()))
             {
-                return parse_error(0, "", out_of_range::create(408, concat("excessive object size: ", std::to_string(len)), ref_stack.back()));
+                JSON_THROW(out_of_range::create(408, concat("excessive object size: ", std::to_string(len)), ref_stack.back()));
             }
         }
         return true;
@@ -10114,17 +10113,7 @@ class json_sax_dom_callback_parser
         // add discarded value at the given key and store the reference for later
         if (keep && ref_stack.back())
         {
-            auto& obj = *ref_stack.back()->m_data.m_value.object;
-            const auto it = obj.find(val);
-            if (it != obj.end())
-            {
-                // this is a duplicate key (legal in JSON); remember its
-                // current value so it can be restored later if the new
-                // value is rejected by the callback, instead of being
-                // erased together with the discarded placeholder
-                duplicate_key_stash.emplace_back(&(it->second), it->second);
-            }
-            object_element = &(obj[val] = discarded);
+            object_element = &(ref_stack.back()->m_data.m_value.object->operator[](val) = discarded);
         }
 
         return true;
@@ -10136,18 +10125,13 @@ class json_sax_dom_callback_parser
         {
             if (!callback(static_cast<int>(ref_stack.size()) - 1, parse_event_t::object_end, *ref_stack.back()))
             {
-                // discard object, unless this slot holds a duplicate key's
-                // previous value pending restoration, in which case that
-                // value is restored instead of being discarded
-                if (!resolve_duplicate_key_stash(ref_stack.back(), true))
-                {
-                    *ref_stack.back() = discarded;
+                // discard object
+                *ref_stack.back() = discarded;
 
 #if JSON_DIAGNOSTIC_POSITIONS
-                    // Set start/end positions for discarded object.
-                    handle_diagnostic_positions_for_json_value(*ref_stack.back());
+                // Set start/end positions for discarded object.
+                handle_diagnostic_positions_for_json_value(*ref_stack.back());
 #endif
-                }
             }
             else
             {
@@ -10161,10 +10145,6 @@ class json_sax_dom_callback_parser
 #endif
 
                 ref_stack.back()->set_parents();
-                // this object is finally, definitively kept; drop any
-                // pending duplicate-key stash entry for its slot since it
-                // can no longer be restored
-                resolve_duplicate_key_stash(ref_stack.back(), false);
             }
         }
 
@@ -10207,7 +10187,7 @@ class json_sax_dom_callback_parser
             // check array limit
             if (JSON_HEDLEY_UNLIKELY(len != detail::unknown_size() && len > ref_stack.back()->max_size()))
             {
-                return parse_error(0, "", out_of_range::create(408, concat("excessive array size: ", std::to_string(len)), ref_stack.back()));
+                JSON_THROW(out_of_range::create(408, concat("excessive array size: ", std::to_string(len)), ref_stack.back()));
             }
         }
 
@@ -10234,25 +10214,16 @@ class json_sax_dom_callback_parser
 #endif
 
                 ref_stack.back()->set_parents();
-                // this array is finally, definitively kept; drop any
-                // pending duplicate-key stash entry for its slot since it
-                // can no longer be restored
-                resolve_duplicate_key_stash(ref_stack.back(), false);
             }
             else
             {
-                // discard array, unless this slot holds a duplicate key's
-                // previous value pending restoration, in which case that
-                // value is restored instead of being discarded
-                if (!resolve_duplicate_key_stash(ref_stack.back(), true))
-                {
-                    *ref_stack.back() = discarded;
+                // discard array
+                *ref_stack.back() = discarded;
 
 #if JSON_DIAGNOSTIC_POSITIONS
-                    // Set start/end positions for discarded array.
-                    handle_diagnostic_positions_for_json_value(*ref_stack.back());
+                // Set start/end positions for discarded array.
+                handle_diagnostic_positions_for_json_value(*ref_stack.back());
 #endif
-                }
             }
         }
 
@@ -10366,48 +10337,14 @@ class json_sax_dom_callback_parser
     }
 #endif
 
-    /// if there is a pending duplicate-key stash entry for this exact slot,
-    /// remove it from the stash; if restore_value is true, the stashed
-    /// previous value is moved back into the slot first (use this when the
-    /// new value at that slot was rejected); otherwise the stash entry is
-    /// simply dropped (use this when the new value was accepted, so it
-    /// correctly supersedes the old one and no restore should ever happen
-    /// for this slot again)
-    /// @return whether a matching stash entry was found (and processed)
-    bool resolve_duplicate_key_stash(BasicJsonType* slot, bool restore_value)
-    {
-        const auto it = std::find_if(duplicate_key_stash.begin(), duplicate_key_stash.end(),
-                                     [slot](const std::pair<BasicJsonType*, BasicJsonType>& entry)
-        {
-            return entry.first == slot;
-        });
-
-        if (it == duplicate_key_stash.end())
-        {
-            return false;
-        }
-
-        if (restore_value)
-        {
-            *slot = std::move(it->second);
-        }
-        duplicate_key_stash.erase(it);
-        return true;
-    }
-
-    /// remove the discarded value the callback rejected from its parent,
-    /// unless it is a duplicate key's slot with a stashed previous value,
-    /// in which case that previous value is restored instead
-    void remove_discarded_value(BasicJsonType& parent)
+    /// remove the discarded value the callback rejected from its parent
+    static void remove_discarded_value(BasicJsonType& parent)
     {
         for (auto it = parent.begin(); it != parent.end(); ++it)
         {
             if (it->is_discarded())
             {
-                if (!resolve_duplicate_key_stash(&(*it), true))
-                {
-                    parent.erase(it);
-                }
+                parent.erase(it);
                 break;
             }
         }
@@ -10505,16 +10442,6 @@ class json_sax_dom_callback_parser
 
         JSON_ASSERT(object_element);
         *object_element = std::move(value);
-        if (!skip_callback)
-        {
-            // this scalar value finally, definitively replaces whatever was
-            // at this slot; drop any pending duplicate-key stash entry for
-            // it since it can no longer be restored (a container value at
-            // this slot is resolved later, in end_object()/end_array(),
-            // since skip_callback is true for the placeholder handling that
-            // happens here for those)
-            resolve_duplicate_key_stash(object_element, false);
-        }
         return {true, object_element};
     }
 
@@ -10528,12 +10455,6 @@ class json_sax_dom_callback_parser
     std::vector<bool> key_keep_stack {}; // NOLINT(readability-redundant-member-init)
     /// helper to hold the reference for the next object element
     BasicJsonType* object_element = nullptr;
-    /// stash of (slot pointer, previous value) for object members that
-    /// already existed when key() was called again for the same key
-    /// (duplicate keys); used to restore the previous value if the new
-    /// value is later rejected by the callback, instead of erasing the
-    /// member entirely
-    std::vector<std::pair<BasicJsonType*, BasicJsonType>> duplicate_key_stash {};
     /// whether a syntax error occurred
     bool errored = false;
     /// callback function
@@ -17775,11 +17696,6 @@ class binary_writer
                 // step 1.5: if this is an ext type, write the subtype
                 if (use_ext)
                 {
-                    if (JSON_HEDLEY_UNLIKELY(j.m_data.m_value.binary->subtype() > (std::numeric_limits<std::uint8_t>::max)()))
-                    {
-                        JSON_THROW(out_of_range::create(413, concat("subtype ", std::to_string(j.m_data.m_value.binary->subtype()), " is too large for the MessagePack ext type (max 255)"), &j));
-                    }
-
                     write_number(static_cast<std::int8_t>(j.m_data.m_value.binary->subtype()));
                 }
 
@@ -18279,12 +18195,6 @@ class binary_writer
         write_bson_entry_header(name, 0x05);
 
         write_number<std::int32_t>(to_bson_length(value.size()), true);
-
-        if (value.has_subtype() && JSON_HEDLEY_UNLIKELY(value.subtype() > (std::numeric_limits<std::uint8_t>::max)()))
-        {
-            JSON_THROW(out_of_range::create(413, concat("subtype ", std::to_string(value.subtype()), " is too large for the BSON binary subtype (max 255)"), nullptr));
-        }
-
         write_number(value.has_subtype() ? static_cast<std::uint8_t>(value.subtype()) : static_cast<std::uint8_t>(0x00));
 
         oa->write_characters(reinterpret_cast<const CharType*>(value.data()), value.size());
@@ -18745,6 +18655,20 @@ class binary_writer
         return 'D';  // float 64
     }
 
+    /*!
+    @brief checks whether a JSON number fits into @a TargetType
+    @param[in] el a JSON number of either the signed or unsigned integer kind
+    @return whether @a el's value can be represented by @a TargetType without
+            wrapping, regardless of which of the two kinds it is stored as
+    */
+    template<typename TargetType>
+    static bool bjdata_ndarray_value_in_range(const BasicJsonType& el)
+    {
+        return el.is_number_unsigned()
+               ? value_in_range_of<TargetType>(el.template get<std::uint64_t>())
+               : value_in_range_of<TargetType>(el.template get<std::int64_t>());
+    }
+
     /*!
     @return false if the object is successfully converted to a bjdata ndarray, true if the type or size is invalid
     */
@@ -18756,6 +18680,16 @@ class binary_writer
         };
 
         string_t key = "_ArrayType_";
+        // the type name is looked up as a string below; a non-string
+        // annotation (e.g. a number, null, or an array) cannot name a known
+        // dtype, so it is treated the same as an unrecognized type name and
+        // falls back to a plain object encoding instead of throwing
+        // type_error.302 out of get<string_t>()
+        if (!value.at(key).is_string())
+        {
+            return true;
+        }
+
         // use get<string_t>() instead of static_cast<string_t> to avoid an
         // ambiguous conversion under explicit instantiation on C++17 (see #4825)
         auto it = bjdtype.find(value.at(key).template get<string_t>());
@@ -18765,6 +18699,16 @@ class binary_writer
         }
         CharType dtype = it->second;
 
+        // the 'B' (byte) marker is only defined by BJData Draft 3; emitting it
+        // under the default Draft 2 mode would produce a stream that Draft 2
+        // readers reject, so such an object falls back to a plain object
+        // encoding instead (see the "Binary values" section of the BJData
+        // documentation)
+        if (dtype == 'B' && bjdata_version != bjdata_version_t::draft3)
+        {
+            return true;
+        }
+
         key = "_ArraySize_";
         // the dimensions are written verbatim as the header length below, so a
         // value that is not an array cannot produce a valid one: null emits 'Z'
@@ -18829,6 +18773,60 @@ class binary_writer
             }
         }
 
+        // every element is cast to the (possibly narrower) C++ type matching
+        // dtype below; a value that does not fit that type would silently
+        // wrap (integers) or overflow to infinity (the "single" precision
+        // float) instead of being reported, so such an object falls back to
+        // a plain object encoding as well
+        for (const auto& el : value.at(key))
+        {
+            bool in_range = true;
+            switch (dtype)
+            {
+                case 'U':
+                case 'C':
+                case 'B':
+                    in_range = bjdata_ndarray_value_in_range<std::uint8_t>(el);
+                    break;
+                case 'i':
+                    in_range = bjdata_ndarray_value_in_range<std::int8_t>(el);
+                    break;
+                case 'u':
+                    in_range = bjdata_ndarray_value_in_range<std::uint16_t>(el);
+                    break;
+                case 'I':
+                    in_range = bjdata_ndarray_value_in_range<std::int16_t>(el);
+                    break;
+                case 'm':
+                    in_range = bjdata_ndarray_value_in_range<std::uint32_t>(el);
+                    break;
+                case 'l':
+                    in_range = bjdata_ndarray_value_in_range<std::int32_t>(el);
+                    break;
+                case 'M':
+                    in_range = bjdata_ndarray_value_in_range<std::uint64_t>(el);
+                    break;
+                case 'L':
+                    in_range = bjdata_ndarray_value_in_range<std::int64_t>(el);
+                    break;
+                case 'd':
+                {
+                    const auto dval = el.template get<double>();
+                    in_range = !std::isfinite(dval) ||
+                               (dval >= static_cast<double>(std::numeric_limits<float>::lowest()) &&
+                                dval <= static_cast<double>((std::numeric_limits<float>::max)()));
+                    break;
+                }
+                default:
+                    // 'D' (double) already spans the full range of number_float_t
+                    break;
+            }
+            if (!in_range)
+            {
+                return true;
+            }
+        }
+
         oa->write_character('[');
         oa->write_character('$');
         oa->write_character(dtype);
@@ -24943,12 +24941,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
             JSON_THROW(invalid_iterator::create(211, "passed iterators may not belong to container", this));
         }
 
-        // passed iterators must belong to arrays
-        if (JSON_HEDLEY_UNLIKELY(!first.m_object->is_array()))
-        {
-            JSON_THROW(invalid_iterator::create(202, "iterators first and last must point to arrays", this));
-        }
-
         // insert to array and return iterator
         return insert_iterator(pos, first.m_it.array_iterator, last.m_it.array_iterator);
     }
@@ -25097,7 +25089,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
         {
             using std::swap;
             swap(*(m_data.m_value.array), other);
-            set_parents();
         }
         else
         {
@@ -25114,7 +25105,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
         {
             using std::swap;
             swap(*(m_data.m_value.object), other);
-            set_parents();
         }
         else
         {
@@ -26683,139 +26673,34 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
 
             case value_t::object:
             {
-                // first pass: record, for every source key, whether it is
-                // common to both objects (in source's iteration order) or
-                // was deleted (i.e., in source but not in target) -- this is
-                // a by-product of the target.find() call already needed to
-                // tell the two cases apart, so it adds no extra lookups. The
-                // "remove" ops themselves are emitted later, interleaved
-                // with the recursive per-key diffs in the fast path below,
-                // to match source's original iteration order (as the
-                // original, pre-reordering-aware implementation did) instead
-                // of grouping all removes before all recursive diffs.
-                std::vector<typename object_t::key_type> common_keys_source_order;
+                // first pass: traverse this object's elements
                 for (auto it = source.cbegin(); it != source.cend(); ++it)
                 {
+                    // escape the key name to be used in a JSON patch
+                    const auto path_key = detail::concat<string_t>(path, '/', detail::escape(it.key()));
+
                     if (target.find(it.key()) != target.end())
                     {
-                        common_keys_source_order.push_back(it.key());
-                    }
-                }
-
-                // second pass: find keys that were added (i.e., in target but
-                // not in source), and record the keys common to both, in
-                // target's iteration order -- again a by-product of the
-                // source.find() call already needed to detect added keys. At
-                // the same time, determine whether every added key comes
-                // after every common key in target's order (a precondition
-                // for the fast path below, which only ever appends new keys
-                // at the very end): for an object_t whose iteration order is
-                // a pure function of the key set (e.g. the default std::map,
-                // which always iterates in sorted key order), the order
-                // check further below is always true and this whole
-                // mechanism is effectively a no-op; it only matters for a
-                // reorderable object_t such as the one backing `ordered_json`.
-                // patch ops for keys that were added (i.e., in target but not
-                // in source); built here so the fast path below can reuse
-                // them without a second source.find() per target key. Only
-                // used by the fast path -- the slow (reordering) path
-                // rebuilds "add" ops for every key itself.
-                std::vector<typename object_t::key_type> common_keys_target_order;
-                basic_json added_ops(value_t::array);
-                bool new_keys_form_suffix = true;
-                bool seen_new_key = false;
-                for (auto it = target.cbegin(); it != target.cend(); ++it)
-                {
-                    if (source.find(it.key()) == source.end())
-                    {
-                        seen_new_key = true;
-                        const auto path_key = detail::concat<string_t>(path, '/', detail::escape(it.key()));
-                        added_ops.push_back(
-                        {
-                            {"op", "add"}, {"path", path_key},
-                            {"value", it.value()}
-                        });
+                        // recursive call to compare object values at key it
+                        auto temp_diff = diff(it.value(), target[it.key()], path_key);
+                        result.insert(result.end(), temp_diff.begin(), temp_diff.end());
                     }
                     else
                     {
-                        common_keys_target_order.push_back(it.key());
-                        if (seen_new_key)
-                        {
-                            new_keys_form_suffix = false;
-                        }
-                    }
-                }
-
-                if (common_keys_source_order == common_keys_target_order && new_keys_form_suffix)
-                {
-                    // fast path: order of common keys already matches (or the
-                    // object_t's iteration order does not depend on
-                    // insertion history), so a plain per-key recursive diff
-                    // is correct and minimal, as before. common_keys_source_order
-                    // is, by construction, the subsequence of source's keys
-                    // that are common to both objects, in source's iteration
-                    // order -- so it can be walked in lockstep with `source`
-                    // using a cheap key comparison instead of another lookup.
-                    // Deleted keys (those source keys not in common_keys_source_order)
-                    // are interleaved here too, in source's original order, to
-                    // match the historical (pre-reordering-aware) output order.
-                    auto common_it = common_keys_source_order.cbegin();
-                    for (auto it = source.cbegin(); it != source.cend(); ++it)
-                    {
-                        if (common_it != common_keys_source_order.cend() && it.key() == *common_it)
-                        {
-                            const auto path_key = detail::concat<string_t>(path, '/', detail::escape(it.key()));
-                            auto temp_diff = diff(it.value(), target[it.key()], path_key);
-                            result.insert(result.end(), temp_diff.begin(), temp_diff.end());
-                            ++common_it;
-                        }
-                        else
-                        {
-                            // found a key that is not in target -> remove it
-                            const auto path_key = detail::concat<string_t>(path, '/', detail::escape(it.key()));
-                            result.push_back(object(
-                            {
-                                {"op", "remove"}, {"path", path_key}
-                            }));
-                        }
-                    }
-
-                    // append the "add" ops for brand-new keys collected above
-                    // during the pass over target -- no second source.find()
-                    // per target key needed
-                    result.insert(result.end(), added_ops.begin(), added_ops.end());
-                }
-                else
-                {
-                    // slow path: the common keys are in a different relative
-                    // order in source and target (only possible for a
-                    // reorderable object_t like ordered_map). Building a
-                    // minimal reordering patch is a nontrivial (LCS-like)
-                    // problem; instead, remove every source key -- both
-                    // deleted keys (which must be removed regardless) and
-                    // common keys (removed so they can be re-added in
-                    // target's order) -- and re-add every key that should
-                    // remain, with its final target value, in target's
-                    // order. basic_json::patch()'s "add" operation on an
-                    // object uses operator[], which appends at the end for a
-                    // vector-backed insertion-ordered map when the key does
-                    // not already exist -- so removing a key and then adding
-                    // it moves it to the end, fixing its position.
-                    for (auto it = source.cbegin(); it != source.cend(); ++it)
-                    {
-                        const auto path_key = detail::concat<string_t>(path, '/', detail::escape(it.key()));
+                        // found a key that is not in o -> remove it
                         result.push_back(object(
                         {
                             {"op", "remove"}, {"path", path_key}
                         }));
                     }
+                }
 
-                    // add every key that is either common (just removed
-                    // above) or brand new, in target's iteration order, so
-                    // that the final order after applying the patch matches
-                    // target exactly
-                    for (auto it = target.cbegin(); it != target.cend(); ++it)
+                // second pass: traverse other object's elements
+                for (auto it = target.cbegin(); it != target.cend(); ++it)
+                {
+                    if (source.find(it.key()) == source.end())
                     {
+                        // found a key that is not in this -> add it
                         const auto path_key = detail::concat<string_t>(path, '/', detail::escape(it.key()));
                         result.push_back(
                         {

tests/src/unit-bjdata.cpp

@@ -2586,7 +2586,12 @@ TEST_CASE("BJData")
                 CHECK(json::to_bjdata(json::from_bjdata(v_d), true, true) == v_d);
                 CHECK(json::to_bjdata(json::from_bjdata(v_D), true, true) == v_D);
                 CHECK(json::to_bjdata(json::from_bjdata(v_C), true, true) == v_C);
-                CHECK(json::to_bjdata(json::from_bjdata(v_B), true, true) == v_B);
+                // v_B uses the Draft-3-only 'B' marker, so it round-trips only when
+                // Draft 3 is explicitly selected (see GitHub issue #5404); the
+                // default Draft 2 falls back to a plain object instead, covered by
+                // the "ndarray with _ArrayType_ "byte" is gated by the BJData draft
+                // version" section below
+                CHECK(json::to_bjdata(json::from_bjdata(v_B), true, true, json::bjdata_version_t::draft3) == v_B);
             }
 
             SECTION("ndarray with data not matching _ArrayType_ is written as an object")
@@ -2629,8 +2634,10 @@ TEST_CASE("BJData")
                 // the C++ API stores an int literal as number_integer, so _ArrayType_
                 // names the wire type rather than the storage. Both storages have to
                 // produce the same typed array for every type.
+                // "byte" is checked separately below since it additionally requires
+                // BJData Draft 3 to be selected explicitly (see GitHub issue #5404).
                 for (const char* type :
-                        {"uint8", "int8", "uint16", "int16", "uint32", "int32", "uint64", "int64", "char", "byte"
+                        {"uint8", "int8", "uint16", "int16", "uint32", "int32", "uint64", "int64", "char"
                         })
                 {
                     CAPTURE(type);
@@ -2641,6 +2648,14 @@ TEST_CASE("BJData")
                     CHECK(from_text == json::to_bjdata(json({{"_ArrayType_", type}, {"_ArraySize_", {2, 3}}, {"_ArrayData_", {1, 2, 3, 4, 5, 6}}})));
                 }
 
+                {
+                    const std::string text = R"({"_ArrayType_":"byte","_ArraySize_":[2,3],"_ArrayData_":[1,2,3,4,5,6]})";
+                    const auto from_text = json::to_bjdata(json::parse(text), true, true, json::bjdata_version_t::draft3);
+                    CHECK(from_text.at(0) == '[');
+                    CHECK(from_text == json::to_bjdata(json({{"_ArrayType_", "byte"}, {"_ArraySize_", {2, 3}}, {"_ArrayData_", {1, 2, 3, 4, 5, 6}}}),
+                    true, true, json::bjdata_version_t::draft3));
+                }
+
                 // negative values under a signed type behave the same way
                 const auto from_neg = json::to_bjdata(json::parse(R"({"_ArrayType_":"int32","_ArraySize_":[2],"_ArrayData_":[-5,7]})"));
                 CHECK(from_neg.at(0) == '[');
@@ -2731,6 +2746,39 @@ TEST_CASE("BJData")
                 CHECK(json::from_bjdata(json::to_bjdata(j_size), true, true) == j_size);
             }
 
+            SECTION("ndarray whose _ArrayType_ is not a string stays as object")
+            {
+                // the type name is looked up as a string below the annotation
+                // check; a non-string _ArrayType_ cannot name a known dtype,
+                // so calling get<string_t>() on it would throw type_error.302
+                // instead of falling back like an unrecognized type name
+                // already does (see GitHub issue #5398)
+                json const j_number = json({{"_ArrayType_", 1}, {"_ArraySize_", {2}}, {"_ArrayData_", {1, 2}}});
+                const auto out_number = json::to_bjdata(j_number);
+                CHECK(out_number.at(0) == '{');
+                CHECK(json::from_bjdata(out_number) == j_number);
+
+                json const j_null = json({{"_ArrayType_", nullptr}, {"_ArraySize_", {2}}, {"_ArrayData_", {1, 2}}});
+                const auto out_null = json::to_bjdata(j_null);
+                CHECK(out_null.at(0) == '{');
+                CHECK(json::from_bjdata(out_null) == j_null);
+
+                json const j_bool = json({{"_ArrayType_", true}, {"_ArraySize_", {2}}, {"_ArrayData_", {1, 2}}});
+                const auto out_bool = json::to_bjdata(j_bool);
+                CHECK(out_bool.at(0) == '{');
+                CHECK(json::from_bjdata(out_bool) == j_bool);
+
+                json const j_array = json({{"_ArrayType_", {"uint8"}}, {"_ArraySize_", {2}}, {"_ArrayData_", {1, 2}}});
+                const auto out_array = json::to_bjdata(j_array);
+                CHECK(out_array.at(0) == '{');
+                CHECK(json::from_bjdata(out_array) == j_array);
+
+                json const j_object = json({{"_ArrayType_", {{"a", 1}}}, {"_ArraySize_", {2}}, {"_ArrayData_", {1, 2}}});
+                const auto out_object = json::to_bjdata(j_object);
+                CHECK(out_object.at(0) == '{');
+                CHECK(json::from_bjdata(out_object) == j_object);
+            }
+
             SECTION("ndarray whose dimensions overflow stays as object")
             {
                 // the product of the dimensions wraps around std::size_t to 0
@@ -2776,6 +2824,83 @@ TEST_CASE("BJData")
                 CHECK(out_num.at(0) == '{');
                 CHECK(json::from_bjdata(out_num) == j_num);
             }
+
+            SECTION("ndarray with out-of-range _ArrayData_ elements stays as object")
+            {
+                // each element is cast to the (possibly narrower) C++ type
+                // named by _ArrayType_ before being written; a value that
+                // does not fit that type would silently wrap instead of
+                // being reported, so such an object falls back to a plain
+                // object encoding that still round-trips (see GitHub issue #5403)
+
+                // an unsigned element that does not fit uint8
+                json const j_uint8 = json({{"_ArrayType_", "uint8"}, {"_ArraySize_", {2}}, {"_ArrayData_", {1, 256}}});
+                const auto out_uint8 = json::to_bjdata(j_uint8);
+                CHECK(out_uint8.at(0) == '{');
+                CHECK(json::from_bjdata(out_uint8) == j_uint8);
+
+                // a signed element that does not fit int8
+                json const j_int8 = json({{"_ArrayType_", "int8"}, {"_ArraySize_", {2}}, {"_ArrayData_", {1, 200}}});
+                const auto out_int8 = json::to_bjdata(j_int8);
+                CHECK(out_int8.at(0) == '{');
+                CHECK(json::from_bjdata(out_int8) == j_int8);
+
+                // a negative element is likewise out of range for an
+                // unsigned _ArrayType_
+                json const j_uint16_neg = json({{"_ArrayType_", "uint16"}, {"_ArraySize_", {2}}, {"_ArrayData_", {1, -1}}});
+                const auto out_uint16_neg = json::to_bjdata(j_uint16_neg);
+                CHECK(out_uint16_neg.at(0) == '{');
+                CHECK(json::from_bjdata(out_uint16_neg) == j_uint16_neg);
+
+                // a double element that overflows to infinity when narrowed
+                // to the "single" (float) precision named by _ArrayType_
+                json const j_single = json({{"_ArrayType_", "single"}, {"_ArraySize_", {2}}, {"_ArrayData_", {1.5, 1e40}}});
+                const auto out_single = json::to_bjdata(j_single);
+                CHECK(out_single.at(0) == '{');
+                CHECK(json::from_bjdata(out_single) == j_single);
+
+                // in-range boundary values still use the compact ndarray encoding
+                json const j_uint8_ok = json({{"_ArrayType_", "uint8"}, {"_ArraySize_", {2}}, {"_ArrayData_", {0, 255}}});
+                CHECK(json::to_bjdata(j_uint8_ok) == std::vector<uint8_t>({'[', '$', 'U', '#', '[', 'i', 2, ']', 0, 255}));
+
+                json const j_int8_ok = json({{"_ArrayType_", "int8"}, {"_ArraySize_", {2}}, {"_ArrayData_", {-128, 127}}});
+                CHECK(json::to_bjdata(j_int8_ok) == std::vector<uint8_t>({'[', '$', 'i', '#', '[', 'i', 2, ']', 0x80, 0x7F}));
+
+                json const j_single_ok = json({{"_ArrayType_", "single"}, {"_ArraySize_", {1}}, {"_ArrayData_", {1.5}}});
+                const auto out_single_ok = json::to_bjdata(j_single_ok);
+                CHECK(out_single_ok.at(0) == '[');
+                CHECK(json::from_bjdata(out_single_ok) == json({1.5f}));
+            }
+
+            SECTION("ndarray with _ArrayType_ \"byte\" is gated by the BJData draft version")
+            {
+                // the 'B' (byte) marker used by _ArrayType_ "byte" is only defined
+                // by BJData Draft 3; Draft 2 (the default) has no such marker, so
+                // emitting it unconditionally produced a stream that a Draft 2
+                // reader could not parse as intended (see GitHub issue #5404).
+                // Two dimensions are used so that a successfully written ndarray
+                // round-trips back into the annotated object (a single dimension
+                // is, by the BJData ndarray convention, read back as a plain
+                // binary value rather than the annotated object, same as every
+                // other single-dimension ndarray of a non-"byte" type is read
+                // back as a plain array instead of the annotated object).
+                json const j_byte = json({{"_ArrayType_", "byte"}, {"_ArraySize_", {2, 3}}, {"_ArrayData_", {1, 2, 3, 4, 5, 6}}});
+
+                // default (Draft 2): falls back to a plain object and round-trips
+                const auto out_draft2 = json::to_bjdata(j_byte);
+                CHECK(out_draft2.at(0) == '{');
+                CHECK(json::from_bjdata(out_draft2) == j_byte);
+
+                // explicit Draft 2: same as the default
+                const auto out_draft2_explicit = json::to_bjdata(j_byte, true, true, json::bjdata_version_t::draft2);
+                CHECK(out_draft2_explicit.at(0) == '{');
+                CHECK(json::from_bjdata(out_draft2_explicit) == j_byte);
+
+                // Draft 3 explicitly selected: still uses the compact 'B' ndarray encoding
+                const auto out_draft3 = json::to_bjdata(j_byte, true, true, json::bjdata_version_t::draft3);
+                CHECK(out_draft3 == std::vector<uint8_t>({'[', '$', 'B', '#', '[', '$', 'i', '#', 'i', 2, 2, 3, 1, 2, 3, 4, 5, 6}));
+                CHECK(json::from_bjdata(out_draft3) == j_byte);
+            }
         }
     }
 

tests/src/unit-diagnostics.cpp

@@ -273,36 +273,5 @@ TEST_CASE("Regression tests for extended diagnostics")
         CHECK(j1["numbers"]["two"] == 2);
         CHECK(j1["string"] == "t");
     }
-
-    SECTION("Regression test - swap(array_t&)/swap(object_t&) must update JSON_DIAGNOSTICS parent pointers")
-    {
-        // swap(array_t&)
-        {
-            json j = json::array();
-            json::array_t arr = {json::array({1})};
-            j.swap(arr);
-
-            // parent pointers of the moved-in elements must point into j, not
-            // into the now-defunct free-standing array_t
-            CHECK_THROWS_WITH_AS(j[0][0].get<std::string>(), "[json.exception.type_error.302] (/0/0) type must be string, but is number", json::type_error);
-
-            // must not trigger assert_invariant() in a debug/assert-enabled build
-            json const k = j;
-            CHECK(k == j);
-        }
-
-        // swap(object_t&)
-        {
-            json o = json::object();
-            json::object_t obj = {{"a", json::array({1})}};
-            o.swap(obj);
-
-            CHECK_THROWS_WITH_AS(o["a"][0].get<std::string>(), "[json.exception.type_error.302] (/a/0) type must be string, but is number", json::type_error);
-
-            // must not trigger assert_invariant() in a debug/assert-enabled build
-            json const p = o;
-            CHECK(p == o);
-        }
-    }
 }
 

tests/src/unit-modifiers.cpp

@@ -641,20 +641,6 @@ TEST_CASE("modifiers")
                 CHECK_THROWS_WITH_AS(j_array.insert(j_array.end(), j_other_array.begin(), j_other_array2.end()), "[json.exception.invalid_iterator.210] iterators do not fit",
                                      json::invalid_iterator&);
             }
-
-            SECTION("iterators not pointing into an array")
-            {
-                json j_object2 = {{"k", 1}, {"l", 2}};
-                json j_primitive = 5;
-                json j_null;
-
-                CHECK_THROWS_WITH_AS(j_array.insert(j_array.begin(), j_object2.begin(), j_object2.end()), "[json.exception.invalid_iterator.202] iterators first and last must point to arrays",
-                                     json::invalid_iterator&);
-                CHECK_THROWS_WITH_AS(j_array.insert(j_array.begin(), j_primitive.begin(), j_primitive.end()), "[json.exception.invalid_iterator.202] iterators first and last must point to arrays",
-                                     json::invalid_iterator&);
-                CHECK_THROWS_WITH_AS(j_array.insert(j_array.begin(), j_null.begin(), j_null.end()), "[json.exception.invalid_iterator.202] iterators first and last must point to arrays",
-                                     json::invalid_iterator&);
-            }
         }
 
         SECTION("range for object")

tests/src/unit-ordered_json.cpp

@@ -81,84 +81,3 @@ TEST_CASE("regression test for issue #3732 - iteration_proxy_value<iter_impl<ord
     };
     static_cast<void>(fn);
 }
-
-TEST_CASE("regression test - diff() must account for ordered_json member order")
-{
-    SECTION("pure reorder, no value changes")
-    {
-        ordered_json a = {{"a", 1}, {"b", 2}};
-        ordered_json b = {{"b", 2}, {"a", 1}};
-        CHECK(a != b); // order-sensitive equality
-        CHECK(a.patch(ordered_json::diff(a, b)) == b);
-    }
-
-    SECTION("new key must land at the front")
-    {
-        ordered_json c = {{"b", 2}};
-        ordered_json e = {{"a", 1}, {"b", 2}};
-        CHECK(c.patch(ordered_json::diff(c, e)) == e);
-    }
-
-    SECTION("reorder plus a value change on one of the reordered keys")
-    {
-        ordered_json a = {{"a", 1}, {"b", 2}};
-        ordered_json b = {{"b", 20}, {"a", 1}};
-        CHECK(a != b);
-        CHECK(a.patch(ordered_json::diff(a, b)) == b);
-    }
-
-    SECTION("reorder plus a deleted key")
-    {
-        ordered_json a = {{"a", 1}, {"b", 2}, {"c", 3}};
-        ordered_json b = {{"b", 2}, {"a", 1}};
-        CHECK(a != b);
-        CHECK(a.patch(ordered_json::diff(a, b)) == b);
-    }
-
-    SECTION("reorder plus a nested value that itself needs a recursive diff")
-    {
-        ordered_json a = {{"a", {{"x", 1}, {"y", 2}}}, {"b", 2}};
-        ordered_json b = {{"b", 2}, {"a", {{"x", 1}, {"y", 99}}}};
-        CHECK(a != b);
-        CHECK(a.patch(ordered_json::diff(a, b)) == b);
-    }
-
-    SECTION("three or more keys shuffled into a different order")
-    {
-        ordered_json a = {{"a", 1}, {"b", 2}, {"c", 3}, {"d", 4}};
-        ordered_json b = {{"d", 4}, {"b", 2}, {"a", 1}, {"c", 3}};
-        CHECK(a != b);
-        CHECK(a.patch(ordered_json::diff(a, b)) == b);
-    }
-
-    SECTION("matching order still produces a minimal patch (fast path unaffected)")
-    {
-        ordered_json a = {{"a", 1}, {"b", 2}, {"c", 3}};
-        ordered_json b = {{"a", 1}, {"b", 20}, {"c", 3}};
-        auto p = ordered_json::diff(a, b);
-        // only the changed value should be touched, not a wholesale remove+add
-        CHECK(p.size() == 1);
-        CHECK(p[0]["op"] == "replace");
-        CHECK(p[0]["path"] == "/b");
-        CHECK(a.patch(p) == b);
-    }
-
-    SECTION("plain json (std::map-backed) is unaffected by same-key-different-insertion-order")
-    {
-        json a;
-        a["b"] = 2;
-        a["a"] = 1;
-
-        json b;
-        b["a"] = 1;
-        b["b"] = 2;
-
-        // std::map iteration is always sorted by key, so a == b regardless of
-        // insertion order, and diff() must still produce the same minimal
-        // (empty) result as before this fix
-        CHECK(a == b);
-        auto p = json::diff(a, b);
-        CHECK(p.empty());
-        CHECK(a.patch(p) == b);
-    }
-}

tests/src/unit-regression2.cpp

@@ -1566,132 +1566,4 @@ TEST_CASE("issue #5402 - update(merge_objects=true) overwrites a primitive with
     CHECK(mixed == json({{"keep", {{"a", 1}, {"b", 2}}}, {"replace", {{"x", 2}}}}));
 }
 
-TEST_CASE("regression test - parser callback must not lose a duplicate key's prior value")
-{
-    // a callback that rejects only the scalar value 2
-    const json::parser_callback_t drop_value_2 = [](int /*depth*/, json::parse_event_t ev, json & v) noexcept
-    {
-        return !(ev == json::parse_event_t::value && v == 2);
-    };
-
-    SECTION("duplicate key, second (scalar) value rejected - prior value is restored")
-    {
-        const json j = json::parse(R"({"a":1,"a":2})", drop_value_2);
-        CHECK(j.dump() == "{\"a\":1}");
-    }
-
-    SECTION("duplicate key, second value is an object rejected at object_end - prior value is restored")
-    {
-        const json j = json::parse(R"({"a":1,"a":{"x":2}})",
-                                   [](int depth, json::parse_event_t ev, json& /*parsed*/) noexcept
-        {
-            return !(ev == json::parse_event_t::object_end && depth == 1);
-        });
-        CHECK(j.dump() == "{\"a\":1}");
-    }
-
-    SECTION("duplicate key, second value is an array rejected at array_end - prior value is restored")
-    {
-        const json j = json::parse(R"({"a":1,"a":[9,9]})",
-                                   [](int depth, json::parse_event_t ev, json& /*parsed*/) noexcept
-        {
-            return !(ev == json::parse_event_t::array_end && depth == 1);
-        });
-        CHECK(j.dump() == "{\"a\":1}");
-    }
-
-    SECTION("duplicate key, second value accepted (scalar) - last value wins")
-    {
-        const json j = json::parse(R"({"a":1,"a":2})", [](int, json::parse_event_t, json&) noexcept
-        {
-            return true;
-        });
-        CHECK(j.dump() == "{\"a\":2}");
-    }
-
-    SECTION("duplicate key, second value accepted (object) - last value wins")
-    {
-        const json j = json::parse(R"({"a":1,"a":{"x":2}})", [](int, json::parse_event_t, json&) noexcept
-        {
-            return true;
-        });
-        CHECK(j.dump() == "{\"a\":{\"x\":2}}");
-    }
-
-    SECTION("brand new (non-duplicate) key, value rejected - member is fully absent")
-    {
-        const json j = json::parse(R"({"a":1,"b":2})", drop_value_2);
-        CHECK(j.dump() == "{\"a\":1}");
-    }
-
-    SECTION("duplicate key nested two levels deep")
-    {
-        const json j = json::parse(R"({"outer":{"a":1,"a":2}})", drop_value_2);
-        CHECK(j.dump() == "{\"outer\":{\"a\":1}}");
-    }
-
-    SECTION("three occurrences of the same key - middle rejected, last accepted")
-    {
-        const json j = json::parse(R"({"k":1,"k":2,"k":3})", drop_value_2);
-        CHECK(j.dump() == "{\"k\":3}");
-    }
-}
-
-TEST_CASE("regression test - excessive binary container size honors allow_exceptions=false")
-{
-    // CBOR array with declared length 2^63
-    const std::vector<std::uint8_t> cbor = {0x9b, 0x80, 0, 0, 0, 0, 0, 0, 0};
-    // CBOR map with declared length 2^63
-    const std::vector<std::uint8_t> cbor_m = {0xbb, 0x80, 0, 0, 0, 0, 0, 0, 0};
-    // UBJSON array with declared length 2^63-1
-    const std::vector<std::uint8_t> ubj = {'[', '#', 'L', 0x7f, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
-    // BJData array with declared length 2^63-1 (little endian)
-    const std::vector<std::uint8_t> bjd = {'[', '#', 'L', 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x7f};
-
-    // allow_exceptions=false must report failure instead of throwing/aborting
-    CHECK(json::from_cbor(cbor, true, false).is_discarded());
-    CHECK(json::from_cbor(cbor_m, true, false).is_discarded());
-    CHECK(json::from_ubjson(ubj, true, false).is_discarded());
-    CHECK(json::from_bjdata(bjd, true, false).is_discarded());
-
-    // allow_exceptions=true (the default) must still throw exactly as before.
-    // The exact message text is not checked here: on platforms where
-    // std::size_t is 32-bit, the CBOR reader's own length-narrowing check
-    // (get_cbor_container_size(), unrelated to this fix) intercepts a
-    // declared length of 2^63 before it ever reaches the check this test
-    // targets, with different (but equally valid, and already correct)
-    // wording -- see unit-cbor.cpp for coverage of that message.
-    json _;
-    CHECK_THROWS_AS(_ = json::from_cbor(cbor), json::out_of_range);
-
-    // regression guard: a genuinely truncated CBOR input must remain discarded
-    CHECK(json::from_cbor(std::vector<std::uint8_t> {0x9b, 0, 0, 0, 0, 0, 0, 0, 0x02}, true, false).is_discarded());
-}
-
-TEST_CASE("regression test - MessagePack/BSON writers reject binary subtypes that don't fit their wire format")
-{
-    // MessagePack: subtype 0-255 must still round-trip correctly (regression guard, pre-existing behavior)
-    CHECK(json::from_msgpack(json::to_msgpack(json::binary({1, 2}, 0))).get_binary().subtype() == 0);
-    CHECK(json::from_msgpack(json::to_msgpack(json::binary({1, 2}, 200))).get_binary().subtype() == 200);
-    CHECK(json::from_msgpack(json::to_msgpack(json::binary({1, 2}, 255))).get_binary().subtype() == 255);
-
-    // MessagePack: subtype > 255 must now throw instead of silently truncating
-    CHECK_THROWS_AS(json::to_msgpack(json::binary({1, 2}, 256)), json::out_of_range);
-    CHECK_THROWS_AS(json::to_msgpack(json::binary({1, 2}, 70000)), json::out_of_range);
-    CHECK_THROWS_WITH_AS(json::to_msgpack(json::binary({1, 2}, 70000)), "[json.exception.out_of_range.413] subtype 70000 is too large for the MessagePack ext type (max 255)", json::out_of_range);
-
-    // BSON: same pattern
-    json doc255 = {{"b", json::binary({1, 2}, 255)}};
-    CHECK(json::from_bson(json::to_bson(doc255))["b"].get_binary().subtype() == 255);
-    CHECK_THROWS_AS(json::to_bson(json{{"b", json::binary({1, 2}, 256)}}), json::out_of_range);
-    CHECK_THROWS_AS(json::to_bson(json{{"b", json::binary({1, 2}, 300)}}), json::out_of_range);
-    CHECK_THROWS_WITH_AS(json::to_bson(json{{"b", json::binary({1, 2}, 300)}}), "[json.exception.out_of_range.413] subtype 300 is too large for the BSON binary subtype (max 255)", json::out_of_range);
-
-    // CBOR must remain unaffected (full 64-bit subtype range already supported correctly) - regression guard
-    CHECK(json::from_cbor(json::to_cbor(json::binary({1, 2}, 70000)), true, true, json::cbor_tag_handler_t::store).get_binary().subtype() == 70000);
-
-    // a binary value with NO subtype at all must be completely unaffected by this change
-    CHECK(json::from_msgpack(json::to_msgpack(json::binary({1, 2}))).get_binary().has_subtype() == false);
-}
-
 DOCTEST_CLANG_SUPPRESS_WARNING_POP