Concepts → Schema Evolution
Overview
Before 2024.7.x, adding and removing physical columns required performing a full load of the respective physical table or materialized views (MVs). Starting 2024.7.x, Incorta has leveraged Delta Lake column mapping to support schema evolution, streamlining adding and removing columns without performing a full load of the respective objects.
Upgrade considerations
As a result of using Delta Lake column mapping, Incorta has modified the definition of columns in the schema metadata. After upgrading to 2024.7.x or later releases, the column definition will change with the next schema update.
When importing schemas from previous releases with the overwrite option, you can keep existing schema data. When the Keep Data option is selected,
- The imported schema XML is updated with the new table and column definitions.
- Incorta can use existing Parquet files.
- You do not need to perform a full load of the imported schema.
Column name considerations
Column names in Incorta are case-sensitive. Renaming a column (in the extraction query, for example) or changing its letter case results in a new column and typically does not require a full load starting 2024.7.x. Reverting the name changes results in a new column as well.
- The following actions still require a full load of the respective object:
- Adding, changing, or removing a key column
- Encrypting a column
- Changing the column data type
- You must revalidate the query of MVs referencing removed or updated columns.