Partitioning and column mapping in Microsoft Fabric OneLake
Partitioning improves query performance for Microsoft Fabric OneLake tables by limiting the amount of data scanned during queries.
Partitioning
A partitioned table organizes data based on the values in one or more columns. Queries that filter on the partition columns scan only the relevant segments instead of the full table.
You can select more than one column to partition on. On the Partition tab, select each column to include and set the order Data Integration applies them in. Column order matters: Data Integration organizes data by the first partition column, then the next, and so on.
Changing partitioning on an existing table
Microsoft Fabric does not support changing the partition columns or their order on a table that already contains data. To change either, recreate the target table.
Automatic optimization
When you use the Merge loading mode without a partition column, Data Integration automatically optimizes the table's layout based on your key columns. This optimization is not configurable, and does not apply if you select a partition column.
Column mapping
While loading your data, Data Integration matches source column types to Microsoft Fabric OneLake types. If a data type is not supported, the system maps it as a String type.
| Data Integration type | Microsoft Fabric OneLake type |
|---|---|
| STRING | STRING |
| INTEGER | INTEGER |
| SHORTDECIMAL / LONGDECIMAL | DECIMAL |
| BOOLEAN | BOOLEAN |
| DATE | DATE |
| TIMESTAMP | TIMESTAMP |
| OBJECT / BLOB | BINARY |
Known issues
- Schema mismatch on Merge: If a column's type narrows or becomes incompatible with the existing table, the load fails with a schema mismatch error. Widen the type at the source, or adjust the target table.
- Removed columns: If a source column is removed, Data Integration writes null values for that column in the target and logs a warning. It does not drop the column from the target table.
