Fix nan correctness of sum aggregator and remove as much branching as…

… possible

cpp/arcticdb/processing/aggregation.hpp

@@ -226,26 +226,35 @@ template<typename T>
 class SumBucketAggregator {
 public:
     void push(T value) {
+        has_value_ = true;
         if constexpr (std::is_floating_point_v<T>) {
             if (ARCTICDB_LIKELY(!std::isnan(value))) {
-                sum_ = sum_.value_or(0) + value;
+                sum_ += value;
             }
         } else {
-            sum_ = sum_.value_or(0) + value;
+            sum_ += value;
         }
     }
 
     T finalize() {
-        auto res = sum_.value_or(0);
-        sum_.reset();
+        T res;
+        if (ARCTICDB_LIKELY(has_value_)) {
+            res = sum_;
+        } else {
+            res = default_value_;
+        }
+        sum_ = 0;
+        has_value_ = false;
         return res;
     }
 
     [[nodiscard]] bool has_value() const {
-        return sum_.has_value();
+        return has_value_;
     }
 private:
-    std::optional<T> sum_{std::nullopt};
+    bool has_value_{false};
+    static constexpr T default_value_{0};
+    T sum_{0};
 };
 
 template<typename T>

python/tests/unit/arcticdb/version_store/test_resample.py

@@ -75,8 +75,36 @@ def test_resampling(lmdb_version_store_v1, freq, date_range, closed, label):
     )
 
 
-def test_resampling_with_nans(lmdb_version_store_v1):
-    lib = lmdb_version_store_v1
+def test_tmp(lmdb_version_store_tiny_segment):
+    lib = lmdb_version_store_tiny_segment
+    sym = "test_tmp"
+    idx = [0, 1]
+    idx = np.array(idx, dtype="datetime64[ns]")
+    col = np.arange(2, dtype=np.float64)
+    col[0] = np.nan
+    col[1] = np.nan
+
+    df = pd.DataFrame({"col": col}, index=idx)
+    lib.write(sym, df)
+
+    expected = df.resample("us").agg(
+        sum=pd.NamedAgg("col", "sum"),
+    )
+    expected = expected.reindex(columns=sorted(expected.columns))
+
+    q = QueryBuilder()
+    q = q.resample("us").agg(
+        {
+            "sum": ("col", "sum"),
+        }
+    )
+    received = lib.read(sym, query_builder=q).data
+    received = received.reindex(columns=sorted(received.columns))
+    assert_frame_equal(expected, received)
+
+
+def test_resampling_with_nans(lmdb_version_store_tiny_segment):
+    lib = lmdb_version_store_tiny_segment
     sym = "test_resampling_with_nans"
     # Create 5 buckets worth of data, each containing 3 values:
     # - No nans