description: Adding tables to the Data Controller is a matter of configuration, specifically the addition of a new record to `DATACTRL.MPE_TABLES`, and corresponding entries in `DATACTRL.MPE_SECURITY`.
Adding tables to the Data Controller is a matter of configuration, specifically the addition of a new record to the `DATACTRL.MPE_TABLES` table, and corresponding entries in the `DATACTRL.MPE_SECURITY` table.
The libref of the table. If not pre-assigned, and the serverType is SAS 9 (EBI), DC will assign it at runtime using the first definition found in metadata, using this [macro](https://core.sasjs.io/mm__assigndirectlib_8sas.html).
The dataset (table) name as visible when assigning a direct libref connection to `LIBREF`. If the target is a format catalog, it should have a "-FC" suffice (eg `FORMATS-FC`). More info on formats [here](formats.md).
This is an integer representing the number of approvals required before a table is updated. This mechanism lets you insist on, for example, 2 or 3 approvals before sensitive data is updated following a submission. Note that only one rejection is ever necessary to remove the submission.
- TXTEMPORAL. This signifies an SCD2 type load. For this type the validity fields (valid from, valid to) should be specified in the `VAR_TXFROM` and `VAR_TXTO` fields. The table itself should include `VAR_TXFROM` in the physical key. The remainder of the primary key fields (not including `VAR_TXFROM`) should be specified in `BUSKEY`.
- BITEMPORAL. These tables have two time dimensions - a version history, and a business history. The version history (SCD2) fields should be specified in `VAR_TXFROM` and `VAR_TXTO` and the business history fields should be specified in `VAR_BUSFROM` and `VAR_BUSTO`. Both the `VAR_TXFROM` and `VAR_BUSFROM` fields should be in the physical key of the actual table, but should NOT be specified in the `BUSKEY` field.
- REPLACE. This loadtype simply deletes all the rows and appends the staged data. Changes are NOT added to the audit table. In the diff screen, previous rows are displayed as deleted, and staged rows as new (modified values are not displayed). Can be useful for updating single-row tables.
The support for BITEMPORAL loads is restricted, in the sense it is only possible to load data at a single point in time (no support for loading _multiple_ business date ranges for a _specific_ BUSKEY). The workaround is simply to load each date range separately. As a result of this restriction, the EDIT page will only show the latest business date range for each key. To modify earlier values, a filter should be applied.
If your target table contains referential constraints (eg primary key values that are linked to a master table) then this will cause problems with the UPDATE and REPLACE load types. This is due to the fact these both involve delete operations. If removal of these constraints is not an option, the workaround would be to create a separate (mirror) table, and update that using PRE-EDIT and POST-APPROVE hook scripts. Please contact Data Controller support for advice / assistance.
The business (natural) key of the table. For SCD2 / Bitemporal, this does NOT include the validity dates. For Retained / Surrogate key tables, this contains the actual surrogate key - the underlying fields that are used to create the surrogate key are specified in [RK_UNDERLYING](#rk_underlying).
The SCD2 type validity dates, representing the point in time at which the field was created (`VAR_TXFROM`) and when it was closed out (`VAR_TXTO`) from a change or deletion. If the record is active, the `VAR_TXTO` field would contain a high value. `VAR_TXFROM` is a part of the physical key of the underlying table.
The BITEMPORAL _business_ dates which represent the reporting period to which the record is valid. Typically these contain _date_ values (rather than _datetime_ values). If variables are specified here, then the [LOADTYPE](#loadtype) should be `BITEMPORAL`.
By default, the Data Controller will only process the records that are part of a changeset. This means that records should be explicity marked for deletion. But what if you are performing a reload of a monthly batch, and the _absence_ of a record implies that it is no longer required? For this scenario, it is necessary to specify the range within a 'complete' load is expected. For instance, by reporting month, or month + product. When performing loads, the DC will then first extract a distinct list of values for this key and close them out in the target table, before performing the upload. The `CLOSE_VARS` are typically a subset of the [BUSKEY](#buskey) fields.
[Hook script](#hook-scripts) to execute _prior_ to an edit being made. This allows data to be modified before being presented for editing, or for display formats to be applied.
[Hook script](#hook-scripts) to execute _after_ an edit has been made. Useful when there is a need to augment data (derived / calculated columns), or perform advanced data quality checks prior to approval.
If your DQ check means that the program should not be submitted, then simply exit with `&syscc > 4`. You can even set a message to go back to the user by using the [mp_abort](https://core.sasjs.io/mp__abort_8sas.html) macro:
This [hook script](#hook-scripts) will execute twice during a typical workflow - firstly, before the approval diff is generated, and again after the approval (not rejection) and _before_ the change is applied.
This makes it a helpful place to prevent changes being made, eg in situations where the target table needs to be locked by alternative systems.
This [hook script](#hook-scripts) is `%inc`'d _after_ an approval is made. This is the most common type of hook script, and is useful for, say, running a SAS job after a mapping table is updated, or running a model after changing a parameter.
At the point of running this script, the data has already been loaded (successfully) to the target table. Therefore the target is typically the base libref.table (or format catalog) itself and can be referenced directly (YOURLIB.YOURDATASET), or using either of the following macro variable:
If you are making changes to the target table as part of the hook, then in order to prevent contention from other users making concurrent edits, you are advised to "LOCK" and "UNLOCK" it using the [mp_lockanytable](https://core.sasjs.io/mp__lockanytable_8sas.html) macro:
Used to determine a range (eg reporting month) to which a 'final version' can be marked. This allows a particular version of data to be marked as final, meaning that the data can continue to change afterwards (reports can simply query for the timestamp of the 'final' version of the data).
For retained / surrogate keys, an auto-incrementing field is used to represent each unique record. In this case, the RK (integer) field itself should be added in the [BUSKEY](#buskey) column, and the natural / underlying key should be added here.
If this field is blank (ie empty, missing), **every** change is captured in the [MPE_AUDIT](/tables/mpe_audit). This can result in large data volumes for frequently changing tables.
1. Enter a zero (`0`) to switch off audit logging completely
2. Enter a library.dataset reference of an alternative audit table in which to capture the change history.
For option 2, the base table structure can be generated using this macro: [https://core.sasjs.io/mddl__dc__difftable_8sas_source.html](https://core.sasjs.io/mddl__dc__difftable_8sas_source.html).
Do not place your hook scripts inside the Data Controller (logical) application folder, as they may be inadvertently lost during a deployment (eg in the case of a backup-and-deploy-new-instance approach).
