Integration table always inserting not updating

Question

Hi,
I put a sequence in an integration layer target table (which is the result of 2 joined tables). I put the option INSERT/UPDATE, but the target table seems to be in insert mode, she never updates ! Any idea what I could look for?
Thank's

Answer 1

Hi folks,
I finally got to read this entire thing. There are 2 issues here:
1) How do you use the attributes on a target to make sure there is an update possible when you use surrogate keys (sequence values)?
2) How does OWB generate sequence numbers?
The answer to 1) is best shown with an example. Suppose I have a table T1 with 3 columns:
ID
CODE
NAME
ID is the primary key, filled with the sequence and plays the role of surrogate key. Code is the actual code as found in the source system (let's say customer code), is is the PK on the source and is unique. Name is the name of the customer and may not be unique.
Scenario1: No keys are set on T1 and therefore no uniqueness is enforced. How do I do insert/update?
Answer1: You set the following attributes:
ID:
Load column when inserting row = YES
Load column when updating row = NO
Match column when updating row = NO
CODE:
Load column when inserting row = YES
Load column when updating row = NO
Match column when updating row = YES
NAME:
Load column when inserting row = YES
Load column when updating row = YES
Match column when updating row = NO
This means that any update on the NAME or any other thing that changes will update based on the CODE column, which is the pk of the source and therefore the correct match. The ID will NOT be updated in this update case. New records will get a new id, code and name as per source.
Scenario2: A PK/UK is set on T1, column ID. How do I do insert/update?
Answer1: You set the following attributes:
Match by constraint = NO CONSTRAINT
ID:
Load column when inserting row = YES
Load column when updating row = NO
Match column when updating row = NO
CODE:
Load column when inserting row = YES
Load column when updating row = NO
Match column when updating row = YES
NAME:
Load column when inserting row = YES
Load column when updating row = YES
Match column when updating row = NO
You must avoid matching on the contraint/PK/UK because that is a generated number and will never match with the source. So you must make sure to match on the source attribute, aka the CODE attribute. Then you merge will perfectly work, and it is very fast!
#2 is a bit different. OWB (there is a bug on this) generates sequences and discards them if not needed. So you may see gaps.
Now here are 2 things to consider. Oracle RDBMS caches sequence numbers by default and therefore by default you will/may see gaps anyways. A sequence (as documented) is not intended to guarantee sequential numbers without gaps!!! So use you own function if you need a meaningful number with no gaps.
Now here is the other thing. IDs are in classic DWH supposed to be meaningles. In other words it does not matter by definition what the number is, as long as it is unique. The number should not be used to base anything on, it's there to optimize storage in big fact tables. No other reason.
Hope that explains the lot, and maybe this obsoletes the "-1" solution as you can do this out of the box.
Jean-Pierre