fix: allow full replacement of bq schema with table_schema when if_ex…

…ists="replace" (googleapis#670)

A previous issue (googleapis#315) noted that when using `if_exists="append"`, pandas should retrieve the existing schema from BQ and update the user-supplied `table_schema` to ensure that all column modes (REQUIRED/NULLABLE) match those of the existing table.

The commit that fixed this issue applies this policy to *all* write_dispositions, but it is only applicable when appending. When replacing, the user should be able to overwrite with new datatypes.

pandas_gbq/gbq.py

@@ -1205,12 +1205,14 @@ def to_gbq(
         )
         table_connector.create(table_id, table_schema)
     else:
-        # Convert original schema (the schema that already exists) to pandas-gbq API format
-        original_schema = pandas_gbq.schema.to_pandas_gbq(table.schema)
-
-        # Update the local `table_schema` so mode (NULLABLE/REQUIRED)
-        # matches. See: https://github.com/pydata/pandas-gbq/issues/315
-        table_schema = pandas_gbq.schema.update_schema(table_schema, original_schema)
+        if write_disposition == "WRITE_APPEND":
+            # If appending, ensure types and modes of existing columns match supplied schema
+            # Convert original schema (the schema that already exists) to pandas-gbq API format
+            original_schema = pandas_gbq.schema.to_pandas_gbq(table.schema)
+
+            # Update the local `table_schema` so mode (NULLABLE/REQUIRED)
+            # matches. See: https://github.com/pydata/pandas-gbq/issues/315
+            table_schema = pandas_gbq.schema.update_schema(table_schema, original_schema)
 
     if dataframe.empty:
         # Create the table (if needed), but don't try to run a load job with an