chore: fix sqlglot compiler misses the columns of SqlDataSource

Author: chelsea-lin

bigframes/core/compile/sqlglot/compiler.py

@@ -202,11 +202,13 @@ def compile_readlocal(
 @_compile_node.register
 def compile_readtable(node: sql_nodes.SqlDataSource, child: sqlglot_ir.SQLGlotIR):
     table_obj = node.source.table
+    columns = () if node.is_star_selection else node.source.schema.names
     return sqlglot_ir.SQLGlotIR.from_table(
         table_obj.project_id,
         table_obj.dataset_id,
         table_obj.table_id,
         uid_gen=child.uid_gen,
+        columns=columns,
         sql_predicate=node.source.sql_predicate,
         system_time=node.source.at_time,
     )

bigframes/core/compile/sqlglot/sqlglot_ir.py

@@ -178,6 +178,7 @@ def from_table(
         dataset_id: str,
         table_id: str,
         uid_gen: guid.SequentialUIDGenerator,
+        columns: typing.Sequence[str] = (),
         sql_predicate: typing.Optional[str] = None,
         system_time: typing.Optional[datetime.datetime] = None,
     ) -> SQLGlotIR:
@@ -187,9 +188,8 @@ def from_table(
             project_id (str): The project ID of the BigQuery table.
             dataset_id (str): The dataset ID of the BigQuery table.
             table_id (str): The table ID of the BigQuery table.
-            col_names (typing.Sequence[str]): The names of the columns to select.
-            alias_names (typing.Sequence[str]): The aliases for the selected columns.
             uid_gen (guid.SequentialUIDGenerator): A generator for unique identifiers.
+            columns (typing.Sequence[str]): The names of the columns to select.
             sql_predicate (typing.Optional[str]): An optional SQL predicate for filtering.
             system_time (typing.Optional[str]): An optional system time for time-travel queries.
         """
@@ -210,14 +210,22 @@ def from_table(
             version=version,
             alias=sql.identifier(table_alias),
         )
-        if sql_predicate:
+
+        if not columns and not sql_predicate:
+            return cls.from_expr(expr=table_expr, uid_gen=uid_gen)
+
+        if len(columns) > 0:
+            select_cols = [sql.identifier(col) for col in columns]
+            select_expr = sge.Select().select(*select_cols).from_(table_expr)
+        else:
             select_expr = sge.Select().select(sge.Star()).from_(table_expr)
+
+        if sql_predicate:
             select_expr = select_expr.where(
                 sg.parse_one(sql_predicate, dialect=sql.base.DIALECT), append=False
             )
-            return cls.from_expr(expr=select_expr, uid_gen=uid_gen)
 
-        return cls.from_expr(expr=table_expr, uid_gen=uid_gen)
+        return cls.from_expr(expr=select_expr, uid_gen=uid_gen)
 
     @classmethod
     def from_cte_ref(

bigframes/core/sql_nodes.py

@@ -45,6 +45,12 @@ def fields(self) -> Sequence[nodes.Field]:
             for source_id in self.source.schema.names
         )
 
+    @property
+    def is_star_selection(self) -> bool:
+        return tuple(self.source.schema.names) == tuple(
+            field.name for field in self.source.table.physical_schema
+        )
+
     @property
     def variables_introduced(self) -> int:
         # This operation only renames variables, doesn't actually create new ones

tests/unit/core/compile/sqlglot/snapshots/test_compile_readtable/test_compile_readtable_w_columns_filters/out.sql

@@ -1,5 +1,7 @@
 SELECT
-  *
+  `rowindex`,
+  `int64_col`,
+  `string_col`
 FROM `bigframes-dev`.`sqlglot_test`.`scalar_types` AS `bft_0`
 WHERE
   `rowindex` > 0 AND `string_col` IN ('Hello, World!')