Expose Rollback Option in the Snowfakery Task

[aditya-balachander]

Background

The LoadData task already supports an enable_rollback option that, when set to True, performs a cleanup of all successfully-inserted records if a row-level error is encountered during a load. This is essential for keeping orgs in a consistent state when a data load partially fails.

The snowfakery task drives LoadData internally (via GenerateAndLoadDataFromYaml subtasks), but previously had no way to pass enable_rollback through to those subtasks — meaning users had no way to get rollback behaviour when using the Snowfakery task.

Changes

cumulusci/tasks/bulkdata/snowfakery.py

• Added enable_rollback to task_options with documentation.
• Parsed and stored the option in _validate_options.
• Raises TaskOptionsError if enable_rollback=True is combined with any run_until_* option — those run across many independent batches, each of which commits separately, so only the failing batch would be rolled back, which would be confusing behaviour.
• Threads the option through to all LoadData subtasks via both additional_load_options (for parallel batch loaders) and _run_generate_and_load_subtask (for the initial/validation batch).

cumulusci/tasks/bulkdata/load.py

• Added a warning in _init_options when enable_rollback=True and ignore_row_errors=True are both set, since row errors are suppressed before the rollback trigger is reached in that combination.