You can always let a SAS
Data Integration Studio transformation perform a simple load of its
output table that drops and replaces the table. However, you can also
add a Table Loader transformation to a permanent output table. Then,
you can use the options in the Table Load transformation to control
how data is loaded into the target table. In fact, a separate Table
Loader transformation might be desirable under the following conditions:
-
loading a DBMS table with any technique
other than drop and replace.
-
loading tables that contain rows
that must be updated upon load (instead of dropping and recreating
the table each time the job is executed).
-
creating primary keys, foreign
keys, or column constraints.
-
performing operations on constraints
before or after the loading of the output table.
-
performing operations on indexes
other than after the loading of the output table.
-
supporting the pushdown feature
that enables you to process relational database tables directly on
the appropriate relational database server. For more information,
see
Pushing ELT Job Code Down to a Database.
The Table
Loader transformation generates code that reads a single source table
(or view) and updates, replaces, or appends it to a permanent target
table. Supported target types include SAS, Excel, and a wide variety
of DBMS types. For data types that support constraints such as not-null
and primary, unique, and foreign keys, a Table Loader can be set to
generate the appropriate code to add or remove constraints. Constraint
actions can be set independently for before and after the load. Likewise,
the adding and removing of indexes can be controlled in the same way.
Choosing
the Load Style and Technique is critical to getting the Table Loader
to perform the correct task for the job efficiently. User requirements
control which style (Update, Replace, or Append) to select. Once the
style has been selected, a number of possible techniques to accomplish
the task are presented. Choosing the correct technique is often a
matter of deciding which technique will likely result in the best
performance for the job when it later runs in production. The exact
number and types of available styles and techniques depend on the
target’s data type. Some data types support clearing old rows
by using a technique known as Truncate, while others do not. Some
data types support a special Upsert technique, which updates rows
that match on a specific key and appends the other rows to the master.
Some support direct access; for those, the DATA step Modify technique
is a choice. For more information about all the available techniques,
see the Help topic for the Load Technique.
Once the
technique is chosen, additional options that are associated with the
selected technique should be reviewed to determine whether any option
values should be changed from their defaults. Also, with performance
in mind, you should consider any special handling of constraints and
indexes.
It is
important to know that non-loader transformations can load data directly
into a permanent table if it has no constraints, in effect doing a
Replace Entire table without using a Table Loader. This
is done in the
Job Editor by replacing the
non-loader’s output WORK table with a registered table. This
technique is not supported by all transformations for all data types.
A new
Replace Simulating truncate load style has been added
for SAS targets. This choice empties the output table by using a DATA
step with SET and STOP statements. This actually recreates the target
table with no rows before data from the source is appended. Original
data is physically deleted, not just logically deleted as with
Replace All rows using delete. Constraints are restored
as they were on the physical table before the load.
CAUTION:
When using this load style, the new table structure
is derived from the physical table (assuming it pre-existed) and not
from metadata. This load style does not reflect changes to the column,
index, or constraint metadata after the creation of the table.
One feature
that is available for SAS tables with
Replace Simulating
truncate, but not available with other Replace types,
is the ability to use generation data sets. Generation data sets are
a way of automatically saving a specified number of backups of the
target. In SAS, this feature is enabled by adding the data set option
GENMAX=#.