[DRAFT] feat: expose rows_affected in adapter_response

[ericvinicius]

Description

Add rows_affected to DatabricksAdapterResponse so downstream consumers (monitoring, validation pipelines) can read the number of rows affected by DML operations (INSERT/UPDATE/DELETE/MERGE) directly from run_results.json.

Changes

• Added rows_affected: Optional[int] field to DatabricksAdapterResponse dataclass
• Read cursor.rowcount from the Databricks SQL connector, using getattr with defensive type checking
• Added unit tests to verify the new field is populated correctly
• Updated CHANGELOG.md with the new feature

Why cursor.rowcount?

The value is sourced from the public cursor.rowcount attribute, which the databricks-sql-connector populates after executing DML statements.

In databricks-sql-connector (client.py, around line 1384-1390), the connector surfaces the affected-row count from the backend:

# Surface the affected-row count for DML (INSERT/UPDATE/DELETE/MERGE) as
# cursor.rowcount instead of the hardcoded -1. num_modified_rows is None
# for SELECT (and statements the server does not report a count for) →
# leave rowcount at its -1 default.
num_modified_rows = getattr(self.active_result_set, "num_modified_rows", None)
if num_modified_rows is not None:
    self.rowcount = num_modified_rows

Implementation Details

• Reads cursor.rowcount after execution (no additional query required)
• Uses getattr with defensive type checking to handle edge cases (missing attribute, wrong types)
• Returns None when rowcount is -1 (default for SELECT/DDL) or not available
• Minimal and focused on the specific use case

Related Issues