Introduction
You have created shallow clones of Delta tables where deletion vectors (DVs) are enabled on the source. You notice the cloned tables also have DVs enabled.
Since shallow clones do not physically duplicate data, they inherit any file-level characteristics, such as DVs, from the source table.
If the source table has DVs enabled, the clone references the same physical Parquet files, including those that may contain logical deletions or updates encoded as DVs.
You want to create a cloned table without the DVs enabled.
Instructions
In the cloned table, disable DVs. Then if necessary, remove references to DV files.
Disable DVs in the cloned table
Disabling DVs in the cloned table prevents new operations from generating further DV files.
Run the following code in a notebook.
ALTER TABLE <your-cloned-table> SET TBLPROPERTIES (
delta.enableDeletionVectors = false
);
Disabling does not, however, remove any existing DV-encoded files that the table may already reference.
Remove references to DV files in the cloned table
Note
If the source Delta table has not undergone row-level delete or update operations, and if the cloned table also has not had any such operations applied before disabling DVs, no further action is required. Disabling DVs on the cloned table is sufficient.
If the source Delta table has undergone row-level operations like DELETE, UPDATE, or MERGE, it is possible some of its Parquet files may already contain DVs in the cloned table.
To remove references to DV files in the cloned table, rewrite the files in your cloned table containing DVs or soft-deleted columns to replace them with fully materialized files. Run the following command in your notebook.
REORG TABLE <your-cloned-table> APPLY (PURGE)
After this rewrite, your cloned table should no longer reference any DV files.