Querying with DATE PART

.changes/unreleased/Features-20230911-190924.yaml

@@ -0,0 +1,6 @@
+kind: Features
+body: Enable DATE PART aggregation for time dimensions
+time: 2023-09-11T19:09:24.960342-07:00
+custom:
+  Author: courtneyholcomb
+  Issue: "770"

metricflow/dag/id_generation.py

@@ -35,6 +35,7 @@
 SQL_EXPR_IS_NULL_PREFIX = "isn"
 SQL_EXPR_CAST_TO_TIMESTAMP_PREFIX = "ctt"
 SQL_EXPR_DATE_TRUNC = "dt"
+SQL_EXPR_EXTRACT = "ex"
 SQL_EXPR_RATIO_COMPUTATION = "rc"
 SQL_EXPR_BETWEEN_PREFIX = "betw"
 SQL_EXPR_WINDOW_FUNCTION_ID_PREFIX = "wfnc"

metricflow/dataset/convert_semantic_model.py

@@ -38,6 +38,7 @@
     SqlColumnReferenceExpression,
     SqlDateTruncExpression,
     SqlExpressionNode,
+    SqlExtractExpression,
     SqlStringExpression,
 )
 from metricflow.sql.sql_plan import (
@@ -46,6 +47,7 @@
     SqlSelectStatementNode,
     SqlTableFromClauseNode,
 )
+from metricflow.time.date_part import DatePart
 
 logger = logging.getLogger(__name__)
 
@@ -102,12 +104,14 @@ def _create_time_dimension_instance(
         time_dimension: Dimension,
         entity_links: Tuple[EntityReference, ...],
         time_granularity: TimeGranularity = DEFAULT_TIME_GRANULARITY,
+        date_part: Optional[DatePart] = None,
     ) -> TimeDimensionInstance:
         """Create a time dimension instance from the dimension object from a semantic model in the model."""
         time_dimension_spec = TimeDimensionSpec(
             element_name=time_dimension.reference.element_name,
             entity_links=entity_links,
             time_granularity=time_granularity,
+            date_part=date_part,
         )
 
         return TimeDimensionInstance(
@@ -219,6 +223,11 @@ def _convert_dimensions(
         select_columns = []
 
         for dimension in dimensions or []:
+            dimension_select_expr = SemanticModelToDataSetConverter._make_element_sql_expr(
+                table_alias=table_alias,
+                element_name=dimension.reference.element_name,
+                element_expr=dimension.expr,
+            )
             if dimension.type == DimensionType.CATEGORICAL:
                 dimension_instance = self._create_dimension_instance(
                     semantic_model_name=semantic_model_name,
@@ -228,11 +237,7 @@ def _convert_dimensions(
                 dimension_instances.append(dimension_instance)
                 select_columns.append(
                     SqlSelectColumn(
-                        expr=SemanticModelToDataSetConverter._make_element_sql_expr(
-                            table_alias=table_alias,
-                            element_name=dimension.reference.element_name,
-                            element_expr=dimension.expr,
-                        ),
+                        expr=dimension_select_expr,
                         column_alias=dimension_instance.associated_column.column_name,
                     )
                 )
@@ -251,11 +256,7 @@ def _convert_dimensions(
                 time_dimension_instances.append(time_dimension_instance)
                 select_columns.append(
                     SqlSelectColumn(
-                        expr=SemanticModelToDataSetConverter._make_element_sql_expr(
-                            table_alias=table_alias,
-                            element_name=dimension.reference.element_name,
-                            element_expr=dimension.expr,
-                        ),
+                        expr=dimension_select_expr,
                         column_alias=time_dimension_instance.associated_column.column_name,
                     )
                 )
@@ -274,16 +275,31 @@ def _convert_dimensions(
                         select_columns.append(
                             SqlSelectColumn(
                                 expr=SqlDateTruncExpression(
-                                    time_granularity=time_granularity,
-                                    arg=SemanticModelToDataSetConverter._make_element_sql_expr(
-                                        table_alias=table_alias,
-                                        element_name=dimension.reference.element_name,
-                                        element_expr=dimension.expr,
-                                    ),
+                                    time_granularity=time_granularity, arg=dimension_select_expr
                                 ),
                                 column_alias=time_dimension_instance.associated_column.column_name,
                             )
                         )
+
+                # Add all date part options for easy query resolution
+                for date_part in DatePart:
+                    if date_part.to_int() >= defined_time_granularity.to_int():
+                        time_dimension_instance = self._create_time_dimension_instance(
+                            semantic_model_name=semantic_model_name,
+                            time_dimension=dimension,
+                            entity_links=entity_links,
+                            time_granularity=defined_time_granularity,
+                            date_part=date_part,
+                        )
+                        time_dimension_instances.append(time_dimension_instance)
+
+                        select_columns.append(
+                            SqlSelectColumn(
+                                expr=SqlExtractExpression(date_part=date_part, arg=dimension_select_expr),
+                                column_alias=time_dimension_instance.associated_column.column_name,
+                            )
+                        )
+
             else:
                 assert False, f"Unhandled dimension type: {dimension.type}"
 

metricflow/dataset/dataset.py

@@ -1,14 +1,15 @@
 from __future__ import annotations
 
 import logging
-from typing import Sequence
+from typing import Optional, Sequence
 
 from dbt_semantic_interfaces.references import TimeDimensionReference
 from dbt_semantic_interfaces.type_enums.time_granularity import TimeGranularity
 from dbt_semantic_interfaces.validations.unique_valid_name import MetricFlowReservedKeywords
 
 from metricflow.instances import InstanceSet, TimeDimensionInstance
 from metricflow.specs.specs import TimeDimensionSpec
+from metricflow.time.date_part import DatePart
 
 logger = logging.getLogger(__name__)
 
@@ -48,12 +49,15 @@ def metric_time_dimension_name() -> str:
         return DataSet.metric_time_dimension_reference().element_name
 
     @staticmethod
-    def metric_time_dimension_spec(time_granularity: TimeGranularity) -> TimeDimensionSpec:
+    def metric_time_dimension_spec(
+        time_granularity: TimeGranularity, date_part: Optional[DatePart] = None
+    ) -> TimeDimensionSpec:
         """Spec that corresponds to DataSet.metric_time_dimension_reference."""
         return TimeDimensionSpec(
             element_name=DataSet.metric_time_dimension_reference().element_name,
             entity_links=(),
             time_granularity=time_granularity,
+            date_part=date_part,
         )
 
     def __repr__(self) -> str:  # noqa: D

metricflow/engine/metricflow_engine.py

@@ -456,6 +456,7 @@ def _create_execution_plan(self, mf_query_request: MetricFlowQueryRequest) -> Me
                 query_spec = self._query_parser.parse_and_validate_query(
                     metric_names=mf_query_request.metric_names,
                     group_by_names=mf_query_request.group_by_names,
+                    group_by=mf_query_request.group_by,
                     limit=mf_query_request.limit,
                     time_constraint_start=mf_query_request.time_constraint_start,
                     time_constraint_end=mf_query_request.time_constraint_end,

metricflow/naming/linkable_spec_name.py

@@ -6,6 +6,8 @@
 
 from dbt_semantic_interfaces.type_enums.time_granularity import TimeGranularity
 
+from metricflow.time.date_part import DatePart
+
 DUNDER = "__"
 
 logger = logging.getLogger(__name__)
@@ -24,6 +26,7 @@ class StructuredLinkableSpecName:
     entity_link_names: Tuple[str, ...]
     element_name: str
     time_granularity: Optional[TimeGranularity] = None
+    date_part: Optional[DatePart] = None
 
     @staticmethod
     def from_name(qualified_name: str) -> StructuredLinkableSpecName:
@@ -32,7 +35,7 @@ def from_name(qualified_name: str) -> StructuredLinkableSpecName:
 
         # No dunder, e.g. "ds"
         if len(name_parts) == 1:
-            return StructuredLinkableSpecName((), name_parts[0])
+            return StructuredLinkableSpecName(entity_link_names=(), element_name=name_parts[0])
 
         associated_granularity = None
         granularity: TimeGranularity
@@ -44,18 +47,30 @@ def from_name(qualified_name: str) -> StructuredLinkableSpecName:
         if associated_granularity:
             #  e.g. "ds__month"
             if len(name_parts) == 2:
-                return StructuredLinkableSpecName((), name_parts[0], associated_granularity)
+                return StructuredLinkableSpecName(
+                    entity_link_names=(), element_name=name_parts[0], time_granularity=associated_granularity
+                )
             # e.g. "messages__ds__month"
-            return StructuredLinkableSpecName(tuple(name_parts[:-2]), name_parts[-2], associated_granularity)
+            return StructuredLinkableSpecName(
+                entity_link_names=tuple(name_parts[:-2]),
+                element_name=name_parts[-2],
+                time_granularity=associated_granularity,
+            )
+
         # e.g. "messages__ds"
         else:
-            return StructuredLinkableSpecName(tuple(name_parts[:-1]), name_parts[-1])
+            return StructuredLinkableSpecName(entity_link_names=tuple(name_parts[:-1]), element_name=name_parts[-1])
 
     @property
     def qualified_name(self) -> str:
-        """Return the full name form. e.g. ds or listing__ds__month."""
+        """Return the full name form. e.g. ds or listing__ds__month.
+
+        If date_part is specified, don't include granularity in qualified_name since it will not impact the result.
+        """
         items = list(self.entity_link_names) + [self.element_name]
-        if self.time_granularity:
+        if self.date_part:
+            items.append(self.date_part_suffix(date_part=self.date_part))
+        elif self.time_granularity:
             items.append(self.time_granularity.value)
         return DUNDER.join(items)
 
@@ -66,3 +81,8 @@ def entity_prefix(self) -> Optional[str]:
             return DUNDER.join(self.entity_link_names)
 
         return None
+
+    @staticmethod
+    def date_part_suffix(date_part: DatePart) -> str:
+        """Suffix used for names with a date_part."""
+        return f"extract_{date_part.value}"

metricflow/plan_conversion/column_resolver.py

@@ -55,6 +55,7 @@ def visit_time_dimension_spec(self, time_dimension_spec: TimeDimensionSpec) -> C
             entity_link_names=tuple(x.element_name for x in time_dimension_spec.entity_links),
             element_name=time_dimension_spec.element_name,
             time_granularity=time_dimension_spec.time_granularity,
+            date_part=time_dimension_spec.date_part,
         ).qualified_name
 
         return ColumnAssociation(

metricflow/plan_conversion/instance_converters.py

@@ -47,6 +47,7 @@
     SqlFunctionExpression,
 )
 from metricflow.sql.sql_plan import SqlSelectColumn
+from metricflow.time.date_part import DatePart
 
 logger = logging.getLogger(__name__)
 
@@ -262,6 +263,7 @@ class _DimensionValidityParams:
 
     dimension_name: str
     time_granularity: TimeGranularity
+    date_part: Optional[DatePart] = None
 
 
 class CreateValidityWindowJoinDescription(InstanceSetTransform[Optional[ValidityWindowJoinDescription]]):
@@ -324,12 +326,16 @@ def transform(self, instance_set: InstanceSet) -> Optional[ValidityWindowJoinDes
             start_specs = [
                 spec
                 for spec in specs
-                if spec.element_name == start_dim.dimension_name and spec.time_granularity == start_dim.time_granularity
+                if spec.element_name == start_dim.dimension_name
+                and spec.time_granularity == start_dim.time_granularity
+                and spec.date_part == start_dim.date_part
             ]
             end_specs = [
                 spec
                 for spec in specs
-                if spec.element_name == end_dim.dimension_name and spec.time_granularity == end_dim.time_granularity
+                if spec.element_name == end_dim.dimension_name
+                and spec.time_granularity == end_dim.time_granularity
+                and spec.date_part == end_dim.date_part
             ]
             linkless_start_specs = {spec.without_entity_links for spec in start_specs}
             linkless_end_specs = {spec.without_entity_links for spec in end_specs}
@@ -401,6 +407,7 @@ def transform(self, instance_set: InstanceSet) -> InstanceSet:  # noqa: D
                     + time_dimension_instance.spec.entity_links
                 ),
                 time_granularity=time_dimension_instance.spec.time_granularity,
+                date_part=time_dimension_instance.spec.date_part,
             )
             time_dimension_instances_with_additional_link.append(
                 TimeDimensionInstance(