Merging Partitions That Have Data Slices

Merging Partitions That Have Data Slices

When you merge partitions that have data slices specified in the Partition Wizard, the resulting partition can contain unexpected, incorrect data after it is processed. To prevent this, you can create a filter that specifies the data in the resulting partition.

For example, a cube containing information about three soft drink products has three partitions that use the same fact table. These partitions have data slices based on product. Partition 1 contains data about [ColaFull], Partition 2 contains data about [ColaDecaf], and Partition 3 contains data about [ColaDiet]. If Partition 3 is merged into Partition 2, the data in the resulting partition (Partition 2) is correct and the cube data is accurate. However, when Partition 2 is processed, if it does not have a filter, its content may be determined by the parent of the members at the product level. This parent, [SoftDrinks], also includes [ColaFull], the product in Partition 1. Processing Partition 2 loads the partition with data for all soft drinks, including [ColaFull]. The cube then contains duplicate data for [ColaFull] and returns incorrect data to end users.

In this example, after merging Partition 3 into Partition 2, you can provide a filter such as ("Product" = 'ColaDecaf' OR "Product" = 'ColaDiet') in the resulting Partition 2 to specify that only data about [ColaDecaf] and [ColaDiet] be extracted from the fact table, and that data pertaining to [ColaFull] be excluded. Alternatively, you can specify filters for Partition 2 and Partition 3 when they are created, and these filters will be combined during the merger process. In either case, after the partition is processed, the cube does not contain duplicate data.

See Also

Data Slice