Merge pull request #2037 from dudanogueira/fix/aggregate-empty-result-none-values

Fix aggregate metrics returning 0/0.0 instead of None for empty result sets

Author: dirkkul

integration/test_collection_aggregate.py

@@ -884,3 +884,66 @@ def test_hybrid_bm25_operators(collection_factory: CollectionFactory) -> None:
         object_limit=10,
     )
     assert res.total_count == 4
+
+
+def test_aggregate_empty_result_none_values(collection_factory: CollectionFactory) -> None:
+    """Test for issue #11219: aggregate metrics should be None for empty result sets, not 0/0.0."""
+    collection = collection_factory(
+        properties=[
+            Property(
+                name="bucket",
+                data_type=DataType.INT,
+                index_filterable=True,
+                index_range_filters=True,
+            ),
+            Property(name="intVal", data_type=DataType.INT),
+            Property(name="numberVal", data_type=DataType.NUMBER),
+        ],
+        vectorizer_config=Configure.Vectorizer.none(),
+        inverted_index_config=Configure.inverted_index(index_null_state=True),
+    )
+
+    if collection._connection._weaviate_version.is_lower_than(1, 29, 0):
+        pytest.skip("gRPC aggregates are only supported in versions 1.29.0 and higher")
+
+    # Insert one object with bucket=0
+    collection.data.insert({"bucket": 0, "intVal": 123, "numberVal": 456.78})
+
+    # Query with a filter that returns no results (bucket=99 doesn't exist)
+    res: AggregateReturn = collection.aggregate.over_all(
+        filters=Filter.by_property("bucket").equal(99),
+        total_count=True,
+        return_metrics=[
+            Metrics("intVal").integer(
+                count=True, maximum=True, mean=True, median=True, minimum=True, mode=True, sum_=True
+            ),
+            Metrics("numberVal").number(
+                count=True, maximum=True, mean=True, median=True, minimum=True, mode=True, sum_=True
+            ),
+        ],
+    )
+
+    # Verify total_count is 0
+    assert res.total_count == 0
+
+    # Verify integer metrics: count should be 0, all other metrics should be None (not 0)
+    int_metrics = res.properties["intVal"]
+    assert isinstance(int_metrics, AggregateInteger)
+    assert int_metrics.count == 0
+    assert int_metrics.maximum is None, "maximum should be None for empty result set"
+    assert int_metrics.mean is None, "mean should be None for empty result set"
+    assert int_metrics.median is None, "median should be None for empty result set"
+    assert int_metrics.minimum is None, "minimum should be None for empty result set"
+    assert int_metrics.mode is None, "mode should be None for empty result set"
+    assert int_metrics.sum_ is None, "sum_ should be None for empty result set"
+
+    # Verify number metrics: count should be 0, all other metrics should be None (not 0.0)
+    number_metrics = res.properties["numberVal"]
+    assert isinstance(number_metrics, AggregateNumber)
+    assert number_metrics.count == 0
+    assert number_metrics.maximum is None, "maximum should be None for empty result set"
+    assert number_metrics.mean is None, "mean should be None for empty result set"
+    assert number_metrics.median is None, "median should be None for empty result set"
+    assert number_metrics.minimum is None, "minimum should be None for empty result set"
+    assert number_metrics.mode is None, "mode should be None for empty result set"
+    assert number_metrics.sum_ is None, "sum_ should be None for empty result set"

weaviate/collections/aggregations/base_executor.py

@@ -269,38 +269,50 @@ def __parse_property_grpc(
         elif aggregation.HasField("int"):
             return AggregateInteger(
                 count=aggregation.int.count,
-                maximum=aggregation.int.maximum,
-                mean=aggregation.int.mean,
-                median=aggregation.int.median,
-                minimum=aggregation.int.minimum,
-                mode=aggregation.int.mode,
-                sum_=aggregation.int.sum,
+                maximum=aggregation.int.maximum if aggregation.int.HasField("maximum") else None,
+                mean=aggregation.int.mean if aggregation.int.HasField("mean") else None,
+                median=aggregation.int.median if aggregation.int.HasField("median") else None,
+                minimum=aggregation.int.minimum if aggregation.int.HasField("minimum") else None,
+                mode=aggregation.int.mode if aggregation.int.HasField("mode") else None,
+                sum_=aggregation.int.sum if aggregation.int.HasField("sum") else None,
             )
         elif aggregation.HasField("number"):
             return AggregateNumber(
                 count=aggregation.number.count,
-                maximum=aggregation.number.maximum,
-                mean=aggregation.number.mean,
-                median=aggregation.number.median,
-                minimum=aggregation.number.minimum,
-                mode=aggregation.number.mode,
-                sum_=aggregation.number.sum,
+                maximum=aggregation.number.maximum
+                if aggregation.number.HasField("maximum")
+                else None,
+                mean=aggregation.number.mean if aggregation.number.HasField("mean") else None,
+                median=aggregation.number.median if aggregation.number.HasField("median") else None,
+                minimum=aggregation.number.minimum
+                if aggregation.number.HasField("minimum")
+                else None,
+                mode=aggregation.number.mode if aggregation.number.HasField("mode") else None,
+                sum_=aggregation.number.sum if aggregation.number.HasField("sum") else None,
             )
         elif aggregation.HasField("boolean"):
             return AggregateBoolean(
                 count=aggregation.boolean.count,
-                percentage_false=aggregation.boolean.percentage_false,
-                percentage_true=aggregation.boolean.percentage_true,
-                total_false=aggregation.boolean.total_false,
-                total_true=aggregation.boolean.total_true,
+                percentage_false=aggregation.boolean.percentage_false
+                if aggregation.boolean.HasField("percentage_false")
+                else None,
+                percentage_true=aggregation.boolean.percentage_true
+                if aggregation.boolean.HasField("percentage_true")
+                else None,
+                total_false=aggregation.boolean.total_false
+                if aggregation.boolean.HasField("total_false")
+                else None,
+                total_true=aggregation.boolean.total_true
+                if aggregation.boolean.HasField("total_true")
+                else None,
             )
         elif aggregation.HasField("date"):
             return AggregateDate(
                 count=aggregation.date.count,
-                maximum=aggregation.date.maximum,
-                median=aggregation.date.median,
-                minimum=aggregation.date.minimum,
-                mode=aggregation.date.mode,
+                maximum=aggregation.date.maximum if aggregation.date.HasField("maximum") else None,
+                median=aggregation.date.median if aggregation.date.HasField("median") else None,
+                minimum=aggregation.date.minimum if aggregation.date.HasField("minimum") else None,
+                mode=aggregation.date.mode if aggregation.date.HasField("mode") else None,
             )
         elif aggregation.HasField("reference"):
             return AggregateReference(pointing_to=list(aggregation.reference.pointing_to))