A data engineer needs downstream consumers to read row-level inserts, updates, and deletes from a Delta table incrementally, without scanning the full table on every run. Which combination correctly enables and then reads that change data?
Answer & explanation
Correct answer: B
WHY B: You enable CDF with delta.enableChangeDataFeed = true, then read it via the readChangeFeed option or table_changes(), which expose _change_type, _commit_version, and _commit_timestamp.
WHY NOT A: Time travel with VERSION AS OF returns table snapshots, not row-level _change_type records.
WHY NOT C: CDF is a table property, not a cluster flag, and you should not hand-parse _delta_log files to rebuild row changes.
WHY NOT D: The feed includes insert, update_preimage, update_postimage, and delete rows, not just post-images.
WHY NOT E: CDF is off by default and has to be enabled per table before table_changes() will work.