Databases Reference
In-Depth Information
An attribute's data is stored in multiple data structures, such as the key store, property
store, deleted member store, unary operator store, BLOB store, and a map store containing
attribute relationships. Across all the stores for an attribute, each record represents a
particular attribute member and the record's ordinal position is the
DataID
value for that
attribute member.
All attributes in a dimension are directly or indirectly related to each other, and such rela-
tionships define an attribute tree (as discussed in Chapter 5). For example, in the
Customer
dimension, the
City
and
Gender
attributes are related to the
Customer
attribute, and the
State
attribute is related to the
City
attribute. Figure 20.7 shows the schema of the
attribute tree of the
Customer
dimension.
State
Key Store
Related
Attributes
…
City
Gender
Key Store
Related
Attributes
Key Store
Related
Attributes
…
…
Customer
Key Store
Related
Attributes
…
FIGURE 20.7
All attributes in a dimension are directly or indirectly related to each other and
create an attribute tree.
In other words, using the terminology for relational data models, the dimension is a fully
normalized snowflake schema, the tables of which are joined with each other by the
virtual key—the ordinal position of the row.
Search WWH ::
Custom Search