Join Key defines a key that is used to join the records. The Join key attribute is a sequence of individual key expressions for the master and all of the slaves. The parts corresponding to particular input ports are separated from each other by hash. The order of these expressions must correspond to the order of the input ports starting with the master and continuing with the slaves. Driver (master) key is a sequence of driver (master) field names (each of them should be preceded by a dollar sign) separated by a colon, semicolon or pipe. Each slave key is a sequence of slave field names (each of them should be preceded by a dollar sign) separated by a colon, semicolon or pipe; e.g.:

$field1(a);$field2(d)#$hello(a);$olleh(d)

The Join key string in the example above contains sorting characters (a) (ascending) and (d) (descending). For more details on sorting, see Sort Key.

You must define the Join key. Records on the input ports must be sorted according to the corresponding parts of the Join key attribute. You can define Join key in the Join key dialog.

In it, you can see the tab for the driver (the Master key tab) and the tabs for all of the slave input ports (the Slave key tabs).

Master Key Tab

Figure 57.3. Join Key Wizard (Master Key Tab)

In the driver tab, there are two panes. The Fields pane on the left and the Master key pane on the right.

You need to select the driver expression by selecting the fields in the Fields pane on the left and moving them to the Master key pane on the right with the help of the Right arrow button.

To the selected Master key fields, the same number of fields should be mapped within each slave. Thus, the number of key fields is the same for all input ports (both the master and each slave). In addition, the driver (Master) key must be common for all slaves.

Slave Key Tab

Figure 57.4. Join Key Wizard (Slave Key Tab)

In each of the slave tab(s), there are two panes: Fields and Key mapping. The Fields pane is on the left. There you can see the list of the slave field names and their data types. The Key mapping pane is on the right. In the right pane you can see two columns: Master key field and Slave key field. The left column contains the selected field names of the driver input port.

If you want to map some driver field to some slave field, select the slave field in the left pane by clicking its item. Then push the left mouse button, drag the field to the Slave key field column in the right pane and release the button. The same must be done for each slave. Note that you can also use the Auto mapping button or other buttons in each tab.

Example 57.4. Join Key for ExtMergeJoin

$first_name;$last_name#$fname;$lname#$f_name;$l_name

Following is the part of Join key for the master data source (input port 0):

$first_name;$last_name

• Thus, these fields are joined with the two fields from the first slave data source (input port 1):

  $fname and $lname, respectively.

• And, these fields are also joined with the two fields from the second slave data source (input port 2):

  $f_name and $l_name, respectively.

Joining data in ExtMergeJoin works the following way. First of all, let us stress again that data on both the master and the slave have to be sorted.

The component takes the first record from the master and compares it to the first one from the slave (with respect to Join key). There are three possible comparison results:

Some input data contain sequences of same values. Then they are treated as one unit on the slave (a slave record knows the value of the following record), This happens only if Allow slave duplicates has been set to true. Moreover, the same-values unit gets stored in the memory. On the master, merging goes all the same by comparing one master record after another to the slave.

	Note
	In case there is a large number of duplicate values on the slave, they are stored on your disk.

A transformation in ExtMergeJoin is required.

Use one of the Transform, Transform URL and Transform class attributes to specify the transformation.

A transformation in ExtMergeJoin lets you define the transformation that sends records to the first output port. The unjoined master records sent to the second output cannot be modified within the ExtMergeJoin transformation.