As an
example, Figure 8 shows a homogeneous version of the dimension of Figure 4 with
null elements. Notice that in order for the dimension to be strict, the elements b1 and
b2 have been broken up into the null elements b11 , b12 and b21 , b22, respectively. Also
notice that the edges Brand ??° Category and Product ??° Category are not required in
this dimension. Now, the hierarchy schema only models the hierarchy path Product
??° Brand ??° Category ??° Department ??° All.
Although the addition of nulls turns hierarchy paths consistent, null elements may
cause meaningless aggregations inside rollup operations. As an example, consider
the dimension of Figure 4, and the fact table SalesAtProduct representing the sales
of the retailer. We may compute the sales and break them down by brands using
the following rollup operation:
ROLLUP Product TO Brand.
46 Hurtado & Gutierrez
Copyright ?© 2007, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of
Idea Group Inc. is prohibited.
Nothing will prevent the OLAP system from scanning the entire set of base facts to
compute the answer. However, not all of the base facts have brands; thus they are
not needed in the aggregation.
Null values may also interfere with the result of more complex aggregate queries
placed over facts at the data cube. An example is a class of queries that involves
multiple dependent aggregates called multifeature aggregates (Chatziantoniou &
Ross, 1996).
Pages:
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117