3.4 Rising

Rising an n-dimensional cube is an operation that “lifts” the datacube to a higher dimension. This can occur in two important applications. (1) Two datacubes having the same number of dimensions and having the same dimensions can be merged together. Figure 8 displays an example.

PIC
Figure 8: Rising Identical Cubes

This case is especially interesting for comparative studies of two or several datacubes, comparing scenarios and outcoming of the same datacube. In LPL this is implemented as Multiple Snapshot Analysis. Rising implies to introduce an additional dimension into the resulting datacube. LPL automatically adds a set called _SNAP_. Hence the resulting datacube has then n + 1. dimensions.

(2) The second important application arises when the actual cube has to be partitioned into several groups, for example, along a time dimension, the months, one has to group the dimension along quarters; or along a product dimension, one has to group them into various product categories, etc (see Figure 9).

PIC
Figure 9: Rising Cubes by Extending and Merging Dimensions

The original cube is partitioned into the desired parts and build from each part a complete datacube of the original size – by getting eventually a very sparse cube. Then these cubes are risen by “merging” them along a new dimension, which implements the partition, see Figure 10.

PIC
Figure 10: Rising Cubes by Extending and Merging Dimensions

The idea behind this partition of a cube is that certain dimensions can be structured into hierarchies (Year – Month – Week – Day, Continent – Country – Region – State). The partition can be arbitrarily however, it can be even on several dimensions, that is, certain tuples of the cube may belong to one part and other tuples to another part. The two most important applications of this operation are grouping and hierarchy building in OLAP tool and grouping and subgrouping in reports.