Speed up array-to-container from_json (#5833)

* Speed up array-to-container from_json

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

* Hint the end when converting an object to a standard container

from_json for map-like containers called emplace() without a hint, which
searches the tree for every key. The keys of the source object are sorted, so
inserting with the end as the hint is constant time, as it was in 3.12.0 via
std::inserter. get<std::map<...>>() of a 25,000 entry object: 4.2 ms -> 3.25 ms
(3.12.0: 3.7 ms).

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

* Fall back to emplace() in object from_json for containers without emplace_hint()

ordered_map has no hinted insert, so the insert(end(), ...) call broke
from_json into types holding an ordered_map (regression test #5122).
Use emplace_hint() where available and plain emplace() otherwise.

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

---------

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

include/nlohmann/detail/conversions/from_json.hpp

@@ -266,6 +266,22 @@ template<typename ConstructibleArrayType>
 void from_json_array_reserve(ConstructibleArrayType& /*arr*/, std::size_t /*size*/, priority_tag<0> /*unused*/)
 {}
 
+// append through push_back() where available; containers without it (such as
+// std::set) fall back to insert(end(), ...), which is what std::inserter did
+template<typename ConstructibleArrayType, typename ValueType>
+auto from_json_array_append(ConstructibleArrayType& arr, ValueType&& value, priority_tag<1> /*unused*/)
+-> decltype(arr.push_back(std::forward<ValueType>(value)), void())
+{
+    arr.push_back(std::forward<ValueType>(value));
+}
+
+template<typename ConstructibleArrayType, typename ValueType>
+void from_json_array_append(ConstructibleArrayType& arr, ValueType&& value, priority_tag<0> /*unused*/)
+{
+    using std::end;
+    arr.insert(end(arr), std::forward<ValueType>(value));
+}
+
 template<typename BasicJsonType, typename ConstructibleArrayType,
          enable_if_t<
              std::is_assignable<ConstructibleArrayType&, ConstructibleArrayType>::value,
@@ -275,17 +291,15 @@ auto from_json_array_impl(const BasicJsonType& j, ConstructibleArrayType& arr, p
     j.template get<typename ConstructibleArrayType::value_type>(),
     void())
 {
-    using std::end;
-
     ConstructibleArrayType ret;
-    from_json_array_reserve(ret, j.size(), priority_tag<1> {});
-    std::transform(j.begin(), j.end(),
-                   std::inserter(ret, end(ret)), [](const BasicJsonType & i)
+    const auto* inner_array = j.template get_ptr<const typename BasicJsonType::array_t*>();
+    from_json_array_reserve(ret, inner_array->size(), priority_tag<1> {});
+    for (const auto& elem : *inner_array)
     {
         // get<BasicJsonType>() returns *this, this won't call a from_json
         // method when value_type is BasicJsonType
-        return i.template get<typename ConstructibleArrayType::value_type>();
-    });
+        from_json_array_append(ret, elem.template get<typename ConstructibleArrayType::value_type>(), priority_tag<1> {});
+    }
     arr = std::move(ret);
 }
 
@@ -371,6 +385,23 @@ template<typename ConstructibleObjectType>
 inline void from_json_object_reserve(ConstructibleObjectType& /*obj*/, std::size_t /*size*/, priority_tag<0> /*unused*/)
 {}
 
+// the keys of the source are sorted for the usual object types, so hinting the
+// end makes every insertion constant-time (for containers that order their keys
+// the same way); containers without emplace_hint() (such as ordered_map) fall
+// back to emplace()
+template<typename ConstructibleObjectType, typename Key, typename Value>
+auto from_json_object_emplace(ConstructibleObjectType& obj, Key&& key, Value&& value, priority_tag<1> /*unused*/)
+-> decltype(obj.emplace_hint(obj.end(), std::forward<Key>(key), std::forward<Value>(value)), void())
+{
+    obj.emplace_hint(obj.end(), std::forward<Key>(key), std::forward<Value>(value));
+}
+
+template<typename ConstructibleObjectType, typename Key, typename Value>
+void from_json_object_emplace(ConstructibleObjectType& obj, Key&& key, Value&& value, priority_tag<0> /*unused*/)
+{
+    obj.emplace(std::forward<Key>(key), std::forward<Value>(value));
+}
+
 template<typename BasicJsonType, typename ConstructibleObjectType,
          enable_if_t<is_constructible_object_type<BasicJsonType, ConstructibleObjectType>::value, int> = 0>
 inline void from_json(const BasicJsonType& j, ConstructibleObjectType& obj)
@@ -385,7 +416,7 @@ inline void from_json(const BasicJsonType& j, ConstructibleObjectType& obj)
     from_json_object_reserve(ret, inner_object->size(), priority_tag<1> {});
     for (const auto& p : *inner_object)
     {
-        ret.emplace(p.first, p.second.template get<typename ConstructibleObjectType::mapped_type>());
+        from_json_object_emplace(ret, p.first, p.second.template get<typename ConstructibleObjectType::mapped_type>(), priority_tag<1> {});
     }
     obj = std::move(ret);
 }

single_include/nlohmann/json.hpp

@@ -5970,6 +5970,22 @@ template<typename ConstructibleArrayType>
 void from_json_array_reserve(ConstructibleArrayType& /*arr*/, std::size_t /*size*/, priority_tag<0> /*unused*/)
 {}
 
+// append through push_back() where available; containers without it (such as
+// std::set) fall back to insert(end(), ...), which is what std::inserter did
+template<typename ConstructibleArrayType, typename ValueType>
+auto from_json_array_append(ConstructibleArrayType& arr, ValueType&& value, priority_tag<1> /*unused*/)
+-> decltype(arr.push_back(std::forward<ValueType>(value)), void())
+{
+    arr.push_back(std::forward<ValueType>(value));
+}
+
+template<typename ConstructibleArrayType, typename ValueType>
+void from_json_array_append(ConstructibleArrayType& arr, ValueType&& value, priority_tag<0> /*unused*/)
+{
+    using std::end;
+    arr.insert(end(arr), std::forward<ValueType>(value));
+}
+
 template<typename BasicJsonType, typename ConstructibleArrayType,
          enable_if_t<
              std::is_assignable<ConstructibleArrayType&, ConstructibleArrayType>::value,
@@ -5979,17 +5995,15 @@ auto from_json_array_impl(const BasicJsonType& j, ConstructibleArrayType& arr, p
     j.template get<typename ConstructibleArrayType::value_type>(),
     void())
 {
-    using std::end;
-
     ConstructibleArrayType ret;
-    from_json_array_reserve(ret, j.size(), priority_tag<1> {});
-    std::transform(j.begin(), j.end(),
-                   std::inserter(ret, end(ret)), [](const BasicJsonType & i)
+    const auto* inner_array = j.template get_ptr<const typename BasicJsonType::array_t*>();
+    from_json_array_reserve(ret, inner_array->size(), priority_tag<1> {});
+    for (const auto& elem : *inner_array)
     {
         // get<BasicJsonType>() returns *this, this won't call a from_json
         // method when value_type is BasicJsonType
-        return i.template get<typename ConstructibleArrayType::value_type>();
-    });
+        from_json_array_append(ret, elem.template get<typename ConstructibleArrayType::value_type>(), priority_tag<1> {});
+    }
     arr = std::move(ret);
 }
 
@@ -6075,6 +6089,23 @@ template<typename ConstructibleObjectType>
 inline void from_json_object_reserve(ConstructibleObjectType& /*obj*/, std::size_t /*size*/, priority_tag<0> /*unused*/)
 {}
 
+// the keys of the source are sorted for the usual object types, so hinting the
+// end makes every insertion constant-time (for containers that order their keys
+// the same way); containers without emplace_hint() (such as ordered_map) fall
+// back to emplace()
+template<typename ConstructibleObjectType, typename Key, typename Value>
+auto from_json_object_emplace(ConstructibleObjectType& obj, Key&& key, Value&& value, priority_tag<1> /*unused*/)
+-> decltype(obj.emplace_hint(obj.end(), std::forward<Key>(key), std::forward<Value>(value)), void())
+{
+    obj.emplace_hint(obj.end(), std::forward<Key>(key), std::forward<Value>(value));
+}
+
+template<typename ConstructibleObjectType, typename Key, typename Value>
+void from_json_object_emplace(ConstructibleObjectType& obj, Key&& key, Value&& value, priority_tag<0> /*unused*/)
+{
+    obj.emplace(std::forward<Key>(key), std::forward<Value>(value));
+}
+
 template<typename BasicJsonType, typename ConstructibleObjectType,
          enable_if_t<is_constructible_object_type<BasicJsonType, ConstructibleObjectType>::value, int> = 0>
 inline void from_json(const BasicJsonType& j, ConstructibleObjectType& obj)
@@ -6089,7 +6120,7 @@ inline void from_json(const BasicJsonType& j, ConstructibleObjectType& obj)
     from_json_object_reserve(ret, inner_object->size(), priority_tag<1> {});
     for (const auto& p : *inner_object)
     {
-        ret.emplace(p.first, p.second.template get<typename ConstructibleObjectType::mapped_type>());
+        from_json_object_emplace(ret, p.first, p.second.template get<typename ConstructibleObjectType::mapped_type>(), priority_tag<1> {});
     }
     obj = std::move(ret);
 }

tests/benchmarks/src/benchmarks.cpp

@@ -9,7 +9,10 @@
 #include <benchmark/benchmark.h>
 #include <nlohmann/json.hpp>
 #include <fstream>
+#include <array>
+#include <deque>
 #include <numeric>
+#include <set>
 #include <vector>
 #include <test_data.hpp>
 
@@ -621,4 +624,27 @@ static void FromCborChunkedString(benchmark::State& state, const std::size_t chu
 
 BENCHMARK_CAPTURE(FromCborChunkedString, 10000 chunks, 10000);
 
+// conversion of a JSON array into standard containers
+template<typename Container>
+static void GetArray(benchmark::State& state)
+{
+    json j = json::array();
+    for (int i = 0; i < 1000; ++i)
+    {
+        j.push_back(i);
+    }
+
+    for (auto _ : state)
+    {
+        auto c = j.get<Container>();
+        benchmark::DoNotOptimize(c);
+    }
+    state.SetItemsProcessed(state.iterations() * 1000);
+}
+
+BENCHMARK_TEMPLATE(GetArray, std::vector<int>);
+BENCHMARK_TEMPLATE(GetArray, std::deque<int>);
+BENCHMARK_TEMPLATE(GetArray, std::set<int>);
+BENCHMARK_TEMPLATE(GetArray, std::array<int, 1000>);
+
 BENCHMARK_MAIN();