Delta table - version number change on merge

[SKC01]

I am running a merge with pyspark on a delta table in which nothing is getting updated in the target table. Still target table version is incremented when I check the table history. Is that expected behavior?

[Kaniz · Accepted solution]

Hi @SKC01 ,

Yes, this is the expected behaviour. Even if no rows are updated in the target Delta table during a merge operation, the version of the table will still be incremented. This is because a merge operation in Delta Lake is considered a transaction that can potentially modify the table, and every transaction in Delta Lake, whether it modifies the data or not, results in a new version of the table. 

In the provided information, the merge operation is used to upsert data from a source table into a target Delta table. Even if no rows match the condition and no updates are made, this operation is still considered a transaction and will increment the version of the target table.

[Sidhant07]

Yes, this is the expected behavior. In Delta Lake, every operation, including MERGE, is atomic. This means that each operation is a transaction that can either succeed completely or fail; it cannot have partial success. Even if the MERGE operation doesn't result in any changes to the target table, it is still considered a transaction and thus increments the table version.When a MERGE operation is performed, Delta Lake performs several steps:1. It identifies the rows in the source data that match the condition specified in the MERGE statement.
2. It applies the update, delete, or insert actions to the matched and not-matched source and target rows.
3. It writes out the result as a new version of the target Delta table.Even if no rows are updated, deleted, or inserted, these steps are still performed, and the result is written out as a new version of the target Delta table. This is why the table version is incremented even if nothing is updated in the target table.