As an example, the rollup relation
?“[Brand,Category] can be obtained by the following query:
SELECT Brand,Category FROM Table_Product.
The translation between graph and snowflake dimensions is straightforward.
The problems reported in this chapter also appear in star and snowflake dimensions.
We next explain further implications of heterogeneity for them.
Allowing Heterogeneity
In a relational table each element of dimension table needs one entry for each attribute.
Thus in order to allow structural heterogeneity in star and snowflake schemas,
null values should be allowed in the tables. It is not easy to do this since functional
dependencies should be interpreted in presence of null values. In order to allow
nulls, Lehner et al. (1998) propose weak functional dependencies, that is, functional
dependencies A ?†’ B that do not constrain tuples when they have null values in
the attribute B. The attributes that participate in the right sides of weak functional
dependencies are treated outside the hierarchy schema as descriptive attributes for
the categories. Weak functional dependencies can be used in snowflake dimensions.
However, in star dimensions we may also need to interpret functional dependencies
when nulls appear in the left side of the functional dependence. An additional
problem is that due to the denormalized nature of star dimensions, heterogeneity
may lead to a proliferation of null values in the table.
Pages:
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113