feat: Support distinct count aggregation [#6116]

Signed-off-by: Nick Quinn <nicholas_quinn@apple.com>
feast-dev · franciscojavierarceo · Mar 17, 2026 · Mar 16, 2026 · Mar 16, 2026 · Mar 16, 2026
commit 04d55a03eca8a4a1f3b9ace50d60570b4e51aca0
diff --git a/sdk/python/feast/aggregation/__init__.py b/sdk/python/feast/aggregation/__init__.py
@@ -18,7 +18,7 @@ class Aggregation:
 
     Attributes:
         column: str  # Column name of the feature we are aggregating.
-        function: str  # Provided built in aggregations sum, max, min, count mean
+        function: str  # Provided built in aggregations sum, max, min, count, mean, count_distinct
         time_window: timedelta  # The time window for this aggregation.
         slide_interval: timedelta # The sliding window for these aggregations
         name: str  # Optional override for the output feature name (defaults to {function}_{column})
@@ -118,6 +118,11 @@ def resolved_name(self, time_window: Optional[timedelta] = None) -> str:
         return base
 
 
+_FUNCTION_ALIASES: Dict[str, str] = {
+    "count_distinct": "nunique",
+}
+
+
 def aggregation_specs_to_agg_ops(
     agg_specs: Iterable[Any],
     *,
@@ -128,7 +133,8 @@ def aggregation_specs_to_agg_ops(
         if getattr(agg, "time_window", None) is not None:
             raise ValueError(time_window_unsupported_error_message)
         alias = getattr(agg, "name", None) or f"{agg.function}_{agg.column}"
-        agg_ops[alias] = (agg.function, agg.column)
+        func_name = _FUNCTION_ALIASES.get(agg.function, agg.function)
+        agg_ops[alias] = (func_name, agg.column)
     return agg_ops
 
 

diff --git a/sdk/python/feast/aggregation/tiling/base.py b/sdk/python/feast/aggregation/tiling/base.py
@@ -86,6 +86,12 @@ def get_ir_metadata_for_aggregation(
             ),
         )
 
+    elif agg_type == "count_distinct":
+        raise ValueError(
+            "count_distinct does not support tiling. "
+            "Use enable_tiling=False or choose an algebraic aggregation (sum, count, min, max)."
+        )
+
     else:
         # Unknown aggregation: treat as algebraic
         return ([], IRMetadata(type="algebraic"))
diff --git a/sdk/python/feast/infra/compute_engines/backends/polars_backend.py b/sdk/python/feast/infra/compute_engines/backends/polars_backend.py
@@ -7,6 +7,8 @@
 
 
 class PolarsBackend(DataFrameBackend):
+    _POLARS_FUNC_MAP = {"nunique": "n_unique"}
+
     def columns(self, df: pl.DataFrame) -> list[str]:
         return df.columns
 
@@ -21,7 +23,7 @@ def join(self, left: pl.DataFrame, right: pl.DataFrame, on, how) -> pl.DataFrame
 
     def groupby_agg(self, df: pl.DataFrame, group_keys, agg_ops) -> pl.DataFrame:
         agg_exprs = [
-            getattr(pl.col(col), func)().alias(alias)
+            getattr(pl.col(col), self._POLARS_FUNC_MAP.get(func, func))().alias(alias)
             for alias, (func, col) in agg_ops.items()
         ]
         return df.groupby(group_keys).agg(agg_exprs)

diff --git a/sdk/python/feast/infra/compute_engines/ray/nodes.py b/sdk/python/feast/infra/compute_engines/ray/nodes.py
@@ -477,6 +477,8 @@ def _execute_standard_aggregation(self, dataset: Dataset) -> DAGValue:
                 agg_dict[feature_name] = (agg.column, "std")
             elif agg.function == "var":
                 agg_dict[feature_name] = (agg.column, "var")
+            elif agg.function == "count_distinct":
+                agg_dict[feature_name] = (agg.column, "nunique")
             else:
                 raise ValueError(f"Unknown aggregation function: {agg.function}.")
 
@@ -531,6 +533,8 @@ def _fallback_pandas_aggregation(self, dataset: Dataset, agg_dict: dict) -> Data
                     result = grouped[column].std()
                 elif function == "var":
                     result = grouped[column].var()
+                elif function == "nunique":
+                    result = grouped[column].nunique()
                 else:
                     raise ValueError(f"Unknown aggregation function: {function}.")
 

diff --git a/sdk/python/feast/infra/compute_engines/spark/nodes.py b/sdk/python/feast/infra/compute_engines/spark/nodes.py
@@ -371,8 +371,11 @@ def _execute_standard_aggregation(self, input_df: DataFrame) -> DAGValue:
         """Execute standard Spark aggregation (existing logic)."""
         agg_exprs = []
         for agg in self.aggregations:
-            func = getattr(F, agg.function)
-            expr = func(agg.column).alias(agg.resolved_name(agg.time_window))
+            if agg.function == "count_distinct":
+                func_expr = F.countDistinct(agg.column)
+            else:
+                func_expr = getattr(F, agg.function)(agg.column)
+            expr = func_expr.alias(agg.resolved_name(agg.time_window))
             agg_exprs.append(expr)
 
         if any(agg.time_window for agg in self.aggregations):

diff --git a/sdk/python/tests/unit/test_aggregation_ops.py b/sdk/python/tests/unit/test_aggregation_ops.py
@@ -1,8 +1,10 @@
 from datetime import timedelta
 
+import pandas as pd
 import pytest
 
 from feast.aggregation import Aggregation, aggregation_specs_to_agg_ops
+from feast.aggregation.tiling.base import get_ir_metadata_for_aggregation
 
 
 class DummyAggregation:
@@ -96,3 +98,40 @@ def test_aggregation_round_trip_with_name():
     restored = Aggregation.from_proto(proto)
     assert restored.name == "sum_seconds_watched_per_ad_1d"
     assert restored == agg
+
+
+def test_count_distinct_agg_ops():
+    """aggregation_specs_to_agg_ops maps count_distinct to the nunique pandas function."""
+    agg_specs = [DummyAggregation(function="count_distinct", column="item_id")]
+
+    agg_ops = aggregation_specs_to_agg_ops(
+        agg_specs,
+        time_window_unsupported_error_message="no windows",
+    )
+
+    assert agg_ops == {"count_distinct_item_id": ("nunique", "item_id")}
+
+
+def test_count_distinct_result():
+    """count_distinct via nunique returns the number of unique values per group."""
+    from feast.infra.compute_engines.backends.pandas_backend import PandasBackend
+
+    agg_specs = [DummyAggregation(function="count_distinct", column="item_id")]
+    agg_ops = aggregation_specs_to_agg_ops(
+        agg_specs,
+        time_window_unsupported_error_message="no windows",
+    )
+
+    df = pd.DataFrame({"user": ["A", "A", "B"], "item_id": [1, 2, 1]})
+    result = PandasBackend().groupby_agg(df, ["user"], agg_ops)
+    result = result.set_index("user")
+
+    assert result.loc["A", "count_distinct_item_id"] == 2
+    assert result.loc["B", "count_distinct_item_id"] == 1
+
+
+def test_count_distinct_tiling_raises():
+    """get_ir_metadata_for_aggregation raises ValueError for count_distinct."""
+    agg = Aggregation(column="item_id", function="count_distinct")
+    with pytest.raises(ValueError, match="count_distinct does not support tiling"):
+        get_ir_metadata_for_aggregation(agg, "count_distinct_item_id")