Understanding Dimensions

A dimension is an OLAP metadata entity. This means that it is a logical object, identified by name and owner, within the OLAP 2 Catalog (CWM2).

OLAP dimensions provide the dimensionality of OLAP measures. A measure represents data, stored in a fact table, that can be accessed by specifying values for its dimensions. For example, a measure representing sales data might be dimensioned by time, product, and location. This means that the sales data can be accessed for a given time period, a given product, and a given location. A set of measures with the same dimensionality can be represented as an OLAP cube.

OLAP dimensions represent columns in your data warehouse dimension tables as levels and attributes of levels. Each level defines a set of dimension members that share the same level of aggregation. For example, within a time dimension, the levels might be months, quarters, and years. Each level attribute holds descriptive information about dimension members within a given level. For example, the level attribute END DATE holds the final date for each of the months in the month level, for each of the quarters in the quarter level, and so on.

Dimensions usually have hierarchies, which define parent/child relationships between levels, and dimension attributes, which define sets of level attributes.

Note: Dimensions in the OLAP 2 Catalog map directly to columns in dimension tables and have no relationship to Oracle database dimension objects.

Dimension Table Requirements

The dimension tables that underlie OLAP dimensions must be organized in levels. The dimension members for each level are stored in one or more columns.

If your dimension tables are in parent-child format instead of level-based format, you must call CREATE_SCRIPT in the CWM2_OLAP_PC_TRANSFORM package to generate level-based views for your parent-child tables. Then you can call procedures in CWM2_OLAP_DIMENSION and related packages to create dimensions in the OLAP 2 Catalog.

If your source dimensions are stored in an analytic workspace within the database, you must call procedures in the CWM2_OLAP_AW_ACCESS package to create relational views that reference the workspace. Then you can call procedures in the CWM2_OLAP_DIMENSION package to create dimensions in the OLAP 2 Catalog.

Normal Dimensions and Time Dimensions

There are two types of dimensions: normal dimensions and time dimensions. The default type is normal.

A time dimension is a normal dimension with two mandatory dimension attributes: END DATE and TIME SPAN. Time dimensions typically contain levels for day, month, quarter, and year and one or more hierarchies defining the parent/child relationships between the levels. The END DATE and TIME SPAN attributes must be mapped for all the levels of a time dimension.

Dependencies Between Dimensional Entities

While a dimension is itself an OLAP metadata entity, it must have at least one child entity in order to be valid. This child entity must represent a level with a valid mapping to one or more columns in a dimension table.

The following table shows the parent/child dependencies between the metadata entities that make up a dimension.

Table 17-1 Hierarchical Relationships Between OLAP Dimensional Metadata Entities

Parent Entity	Child Entities
dimension	dimension attribute, hierarchy, level
dimension attribute	level attribute
hierarchy	level
level	level attribute