Handling bigquery dialect when previewing data (apache#5655)

* Handling bigquery dialect when previewing data

* review comments

* lint

superset/db_engine_specs.py

@@ -311,6 +311,10 @@ def where_latest_partition(
             cls, table_name, schema, database, qry, columns=None):
         return False
 
+    @classmethod
+    def _get_fields(cls, cols):
+        return [sqla.column(c.get('name')) for c in cols]
+
     @classmethod
     def select_star(cls, my_db, table_name, engine, schema=None, limit=100,
                     show_cols=False, indent=True, latest_partition=True,
@@ -321,7 +325,7 @@ def select_star(cls, my_db, table_name, engine, schema=None, limit=100,
             cols = my_db.get_columns(table_name, schema)
 
         if show_cols:
-            fields = [sqla.column(c.get('name')) for c in cols]
+            fields = cls._get_fields(cols)
         quote = engine.dialect.identifier_preparer.quote
         if schema:
             full_table_name = quote(schema) + '.' + quote(table_name)
@@ -1410,6 +1414,19 @@ def fetch_data(cls, cursor, limit):
             data = [r.values() for r in data]
         return data
 
+    @classmethod
+    def _get_fields(cls, cols):
+        """
+        BigQuery dialect requires us to not use backtick in the fieldname which are
+        nested.
+        Using literal_column handles that issue.
+        http://docs.sqlalchemy.org/en/latest/core/tutorial.html#using-more-specific-text-with-table-literal-column-and-column
+        Also explicility specifying column names so we don't encounter duplicate
+        column names in the result.
+        """
+        return [sqla.literal_column(c.get('name')).label(c.get('name').replace('.', '__'))
+                for c in cols]
+
 
 class ImpalaEngineSpec(BaseEngineSpec):
     """Engine spec for Cloudera's Impala"""