Create a copy when reusing SELECT nodes.

metricflow/plan_conversion/dataflow_to_sql.py

@@ -385,7 +385,9 @@ def _make_time_spine_data_set(
     def visit_source_node(self, node: ReadSqlSourceNode) -> SqlDataSet:
         """Generate the SQL to read from the source."""
         return SqlDataSet(
-            sql_select_node=node.data_set.checked_sql_select_node,
+            # This visitor is assumed to create a unique SELECT node for each dataflow node, so create a copy.
+            # The column pruner relies on this assumption to keep track of what columns are required at each node.
+            sql_select_node=node.data_set.checked_sql_select_node.create_copy(),
             instance_set=node.data_set.instance_set,
         )
 

metricflow/sql/sql_plan.py

@@ -227,6 +227,21 @@ def nearest_select_columns(
     ) -> Optional[Sequence[SqlSelectColumn]]:
         return self.select_columns
 
+    def create_copy(self) -> SqlSelectStatementNode:  # noqa: D102
+        return SqlSelectStatementNode.create(
+            description=self.description,
+            select_columns=self.select_columns,
+            from_source=self.from_source,
+            from_source_alias=self.from_source_alias,
+            cte_sources=self.cte_sources,
+            join_descs=self.join_descs,
+            group_bys=self.group_bys,
+            order_bys=self.order_bys,
+            where=self.where,
+            limit=self.limit,
+            distinct=self.distinct,
+        )
+
 
 @dataclass(frozen=True, eq=False)
 class SqlTableNode(SqlQueryPlanNode):