Join keys determine how records are compared by the record assembler. For each data source feeding a join, you designate one or more properties or dimensions to function as the source's join key.
During the course of the join, the record assembler compares the values within each source's join key. Records that have the same values for their respective keys are considered equivalent for the purposes of the join. With two exceptions, all joins require a join key for each data source.
Comparisons are based solely on property and dimension values, not names. It is not a requirement, therefore, that the properties and dimensions you specify for your record keys have identical names.
Example 12. Example
As an example, consider the following left join with four record sources. Source 1 and Source 2 use Id as their join key. Source 3 and Source 4 use Pid as their join key. The other properties are not part of the join key for any of the sources.
For this data, we know:
The resulting left join looks like this:
In this example, the following occurred:
Note
Join keys rarely incorporate dimensions. One reason is that if you use dimensions in a key, the records must have previously been processed and mapped by Forge. That is, the records must have the dimensions tagged on them before the join begins.
Related links
In addition to a join key, you must also configure a record index for each data source that feeds a join. A record index is a key that indicates to the record assembler how it can identify records from that source.
A source's record index key must match its join key. In other words, the key that tells the record assembler how to find a source's records must be the same as the key that the record assembler uses to compare records from that source.
Note
There are two cases where join keys are not required for data sources and, hence, neither are record indexes.
