Unmerge

If the data for a merged i2 Analyze record is determined to no longer represent the same real-world object, the i2 Analyze record can be unmerged into two i2 Analyze records. For the Information Store to unmerge records, the correlation identifier or implicit discriminators of the data associated with that record must be changed.

Assuming that the implicit discriminators are compatible, the unmerge operation occurs when the correlation identifier of a row in the staging table is different from the correlation identifier on the merged record that it is currently associated with by its origin identifier.

After the unmerge operation, the following statements are true for the existing i2 Analyze record:

• The piece of provenance for the source information that caused the operation is unmerged from the record.

• The property values for the record are taken from the source information associated with the provenance that has the most recent value for source_last_updated.

  If only one piece of provenance for a record has a value for the source_last_updated column, the property values from the source information that is associated with that provenance are used. Otherwise, the property values to use are determined by the ascending order of the origin identifier keys that are associated with the record. The piece of provenance that is last in the order is chosen. To ensure data consistency, update your existing records with a value for the source_last_updated column before you start to use correlation, and continue to update the value.

  If the default behavior does not match the requirements of your deployment, you can change the method for defining property values for merged records. For more information, see Define how property values of merged records are calculated.

• All notes remain on the record.

• The last updated time of the record is updated to the time that the unmerge operation occurred.

• If the existing record was an entity record at the end of any links, any links to the unmerged piece of provenance are updated to reference the record that now contains the provenance.

After the unmerge operation, depending on the change in correlation identifier that is presented to the Information Store, either a new i2 Analyze record is inserted or a merge operation is completed. During ingestion, this process is reported in the unmerge_count, insert_count, and merge_count columns of the ingestion report.

The following diagrams demonstrate the unmerge operation. In each diagram, the data that is ingested from the staging table contains a different correlation identifier to the one on the record that it is associated with by its origin identifier. One unmerge operation results in a new record, and one results in a merge operation.

In the first example of an unmerge, the correlation identifier of the provenance that is unmerged from an existing record (a) does not match with another correlation identifier in the staging table or the Information Store. A new i2 Analyze record (b) is created with the property values from the staging table.

In the diagram, the data in the staging table has a different correlation identifier to the record (a) that it is currently associated with. This causes an unmerge operation. The provenance is unmerged from the existing record (a). The existing record (a) only contains the provenance for the origin identifier OI.12 and the property values from the source information that is associated with that provenance. The correlation identifier of the staging table data does not match with any others in the Information Store, so a new record (b) is inserted.

In the second example of an unmerge, if the correlation identifier of the provenance that is unmerged from an existing record (a) now matches the correlation identifier of another record (b), a merge operation is performed.

In the diagram, the data in the staging table that is ingested has a different correlation identifier to the record (a) that it is currently associated with. This causes the unmerge operation. The origin identifier and provenance are unmerged from the existing record (a). The existing record (a) now only contains the provenance for the origin identifier OI.12 and the property values from the source information that is associated with that provenance.

The correlation identifier of the staging table data now matches with another record (b) in the Information Store, so a merge operation occurs. In this example, it is assumed that the staging table data is more recent than the existing data. As part of the merge, the property values from the staging table row are used. This results in a change to the value for the first name property from "Jon" to "John". The merged i2 Analyze record (b) now contains two pieces of provenance, one for the origin identifier OI.32 and one for the new data, OI.22.

For more information about the behavior of a merge operation, see Merge.