fix(bigframes): Fix bugs compiling ambiguous ids and in subqueries

Author: TrevorBergeron

packages/bigframes/bigframes/core/compile/compiled.py

@@ -381,6 +381,7 @@ def isin_join(
             new_column = (
                 (left_table[conditions[0]])
                 .isin((right_table[conditions[1]]))
+                .fillna(False)
                 .name(indicator_col)
             )
 

packages/bigframes/bigframes/session/_io/bigquery/__init__.py

@@ -516,19 +516,19 @@ def to_query(
 ) -> str:
     """Compile query_or_table with conditions(filters, wildcards) to query."""
     if is_query(query_or_table):
-        sub_query = f"({query_or_table})"
+        from_item = f"({query_or_table})"
     else:
         # Table ID can have 1, 2, 3, or 4 parts. Quoting all parts to be safe.
         # See: https://cloud.google.com/bigquery/docs/reference/standard-sql/lexical#identifiers
         parts = query_or_table.split(".")
-        sub_query = ".".join(f"`{part}`" for part in parts)
+        from_item = ".".join(f"`{part}`" for part in parts)
 
     # TODO(b/338111344): Generate an index based on DefaultIndexKind if we
     # don't have index columns specified.
     if columns:
         # We only reduce the selection if columns is set, but we always
         # want to make sure index_cols is also included.
-        select_clause = "SELECT " + ", ".join(f"`{column}`" for column in columns)
+        select_clause = "SELECT " + ", ".join(f"`_bf_source`.`{column}`" for column in columns)
     else:
         select_clause = "SELECT *"
 
@@ -545,7 +545,7 @@ def to_query(
 
     return (
         f"{select_clause} "
-        f"FROM {sub_query}"
+        f"FROM {from_item} AS _bf_source"
         f"{time_travel_clause}{where_clause}{limit_clause}"
     )
 

packages/bigframes/bigframes/session/polars_executor.py

@@ -122,7 +122,7 @@ def _is_node_polars_executable(node: nodes.BigFrameNode):
         return False
     for expr in node._node_expressions:
         if isinstance(expr, agg_expressions.Aggregation):
-            if not type(expr.op) in _COMPATIBLE_AGG_OPS:
+            if type(expr.op) not in _COMPATIBLE_AGG_OPS:
                 return False
         if isinstance(expr, expression.Expression):
             if not set(map(type, _get_expr_ops(expr))).issubset(_COMPATIBLE_SCALAR_OPS):

packages/bigframes/third_party/bigframes_vendored/sqlglot/parser.py

@@ -4469,6 +4469,14 @@ def _parse_table(
         if schema:
             return self._parse_schema(this=this)
 
+        # see: https://docs.cloud.google.com/bigquery/docs/reference/standard-sql/query-syntax#from_clause
+        # from_item, then alias, then time travel, then sample.
+        alias = self._parse_table_alias(
+            alias_tokens=alias_tokens or self.TABLE_ALIAS_TOKENS
+        )
+        if alias:
+            this.set("alias", alias)
+
         version = self._parse_version()
 
         if version:
@@ -4477,11 +4485,6 @@ def _parse_table(
         if self.dialect.ALIAS_POST_TABLESAMPLE:
             this.set("sample", self._parse_table_sample())
 
-        alias = self._parse_table_alias(
-            alias_tokens=alias_tokens or self.TABLE_ALIAS_TOKENS
-        )
-        if alias:
-            this.set("alias", alias)
 
         if self._match(TokenType.INDEXED_BY):
             this.set("indexed", self._parse_table_parts())