When you add data to
a cube, you must select a table that contains the new data for the
cube. This might be new data for existing members or new members with
new data. The table must be defined in the metadata repository and
it must contain all the columns for the measures. Depending whether
you are incrementally updating the cube or rebuilding the cube, you
must manage the input data accordingly. If you are updating the cube
incrementally, then your table does not need to contain all the cumulative
data. However, if you are rebuilding the cube, then your input tables
must contain all the data from the original build and all the incremental
updates.
If you are rebuilding
a cube, it is very important that the data used as input to the rebuilt
cube contain all the incremental updates that have been made since
the original cube was created. This can be done one of two ways:
-
The input base or fact
table definition in the SAS metadata should be a view that combines
the original input table plus all subsequent update tables into a
single logical table. This is the recommended approach.
-
The input base or fact
table itself is physically updated to add data from all update tables.
It is not possible for
PROC OLAP or SAS OLAP Cube Studio to verify the correctness of the
input data. This is the responsibility of the administrator or cube
builder. Depending on what data source the cube is loaded from, some
requirements apply:
Input Table Considerations
|
|
|
All the level columns
must be in the new data table.
|
|
The foreign key columns
must be in the new data table. However, the level columns must be
in the dimension table along with the unique key columns.
For star schema cubes,
you do not have to select a source table. If a source table is not
selected, then only the dimension tables will be processed and only
if the UPDATE_DIMENSION option is set to MEMBERS or MEMBERS_AND_PROPERTIES
for the dimension. No new data is added to the cube. In this case,
the ADD_DATA option should not be added to the PROC OLAP statement.
You can, however, still
specify new dimension tables and specify a value for the UPDATE_DIMENSION
option. This is valid only for dimensions in a star schema that originate
from a dimension table. Dimensions in a star schema that come from
the fact table can be updated only if a source table is selected and
ADD_DATA is specified in the PROC OLAP statement.
|
|
The cube cannot be updated.
|
When you add data to
a cube, the cube and its corresponding tables must be kept in sync.
When you add new members to the cube using the Incremental Update
process, the original tables are re-read by PROC OLAP unless a different
table is provided.
Note: If the original dimension
table for a star schema is updated with the new members, you do not
need to select a table for a dimension.