UPSERT feature in ODI

Question

Hi Gurus/Experts,
I have a requirement to put different transformation logic for a column based on the value of another columns in target and source table. If the value is already there in source then there is a different logic than the rows not having values in source. Hope I am clear with the requirement.
Please let me know how this to be implemented in ODI?
Regards,
Vipin

Answer 1

When you say "put different transformation logic", do you mean :
if your source row exists on the target, the target column A will receive the mapping XXX (and the row will be updated, of course)
if your source row doesn't exists on the target, the target column A will receive the mapping YYY (and the row will be inserted, of course)
If so, it will be difficult.
Because if you take for example the IKM Incremental Update, the different transformations occurs during the insertion in I$ table (if the mapping is on "work area" zone) or during insertion in the C$ table (if the mapping is on "source" zone). But at this time, ODI hasn't already checked if the data exists on the target. The step "flag row for updated" is done later and will give you this answer thanks to the "IND_UPDATE" field.
So, you cannot use the IND_UPDATE field in your mappings that occurs on "work area" zone or "source zone".
But... if your transformation can be processed in "target zone" (that means they don't depend on source data, but depends on variable, sequence or static data), you will be able to do it.
Example, you can write for column A :
"CASE WHEN IND_UPDATE = I THEN '#variable1' ELSE '#variable2' END'
In this example : XXX = variable2 and YYY = variable1.
another example :
"CASE WHEN IND_UPDATE = I THEN 'data_inserted' ELSE 'data_updated' END'
If your transformations need source data, then they must be set on work area zone or source zone. In this case, you won't be able to use IND_UPDATE field. For your requirement, you should create 2 interfaces : one for inserted data, and one for updated data. Or 1 interface with 2 unioned datasets (odi 11g only).