About the “Prune” Data Source Action

Organizations need a way to back out a bad load. Deleting entities that were newly added by the bad load is easy. Backing out features that got added to pre-existing entities is a bit more complicated.

The “Prune” feature was introduced to remove specific features that got added during a bad load. The implementation also supports a more general maintenance activity should one want to remove old features from existing entities e.g., expiring old values beyond a certain age.

The Prune method is implemented via a data source action with associated dates as follows:

 <UMF_DOC>

               <OBS>

                              <DSRC_CODE>TEST</DSRC_CODE>

                              <DSRC_ACTION>P</DSRC_ACTION>

                              <OBS_SRC_KEY>001-01-NAME+F1</OBS_SRC_KEY>

                              <PRUNE_BEFORE>yyyy-mm-dd hh:mm:ss</PRUNE_BEFORE>

                              <PRUNE_AFTER>yyyy-mm-dd hh:mm:ss</PRUNE_AFTER>

                              <OBS_ENT>

                                            <ETYPE_CODE>PERSON</ETYPE_CODE>

                                             <ENT_SRC_KEY>001-01-NAME+F1</ENT_SRC_KEY>

                              </OBS_ENT>

               </OBS>

</UMF_DOC>

The Prune action requires either a PRUNE_BEFORE or PRUNE_AFTER tag. However, both could be supplied if desired. Upon receiving a Prune message, the system will remove any features meeting the PRUNE_BEFORE or PRUNE_AFTER date based on features last update date (obs_feat_ekey.sys_lstupd_dt).

To back out a load, the user can either create the Prune documents from their own data sources or from G2’s log_obs_ent table using the date the record was loaded as the PRUNE_AFTER date.