Although the problem to solve is the nonapplicability of attributes, that is, the form
of heterogeneity illustrated in the dimension of Figure 4, the solution works for the
general case. The approach involves partitioning the original data cube into one
data cube for each dimension. A dimension that contains the structure shared by all
the resulting homogeneous dimensions, called the core dimension, is kept with a
data cube, called core data cube, which aggregate facts at all the bottom elements
of the original dimension.
Dimension. Constraints
In this section we describe the approach of modeling heterogeneity with integrity
constraints. We explain the framework of dimension constraints (Hurtado & Mendelzon,
2001; Hurtado et al., 2005), a class of integrity constraints for OLAP data
that provide semantics to the hierarchy schema so that it is turned into a better abstraction
to capture heterogeneity.
We first motivate dimension constraints. Then we explain how the homogeneous
structures mixed in the dimension, called frozen dimensions, can be computed from
the constraints, and explain inference of dimension constraints in this framework.
Finally, we show the application of dimension constraints to support the structural
adaptations explained in the previous section, and to reason about the correctness
of OLAP aggregate operators in heterogeneous dimensions.
Pages:
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121