T-SQL to "Merge" two rows, or "Rekey" all FK relationships - sql

Question

I have a production database where occasionally redundant rows in a single table need to be "Merged".
Let's assume that both rows in this table have identical values, except their IDs.
Table "PrimaryStuff"
ID | SomeValue
1 | "I have value"
2 | "I have value"
3 | "I am different"
Let's also assume that a number of related tables exist. Because duplicates were created in the "PrimaryStuff" table, often rows are created in these child tables that SHOULD all be related to a single record on the PrimaryStuff table. The number and names of these tables are not under my control and should be considered dynamically at runtime. IE: I don't know the names or even the number of related records, as other people may edit the database without my knowledge.
Table "ForeignStuff"
ID | PrimaryStuffId | LocalValue
1| 1| "I have the correct FK"
2| 1| "I have the correct FK"
3| 2| "I should get pointed to an FK of 1"
To resolve the duplication of PrimaryStuff's row 1 and 2, I wish to have ALL related tables change their FK's to 1s and then delete the PrimaryStuff's row 2. This SHOULD be trivial, as if PrimaryStuff's row 1 didn't exist, I could just update the Primary Key on Row 2 to 1, and the changes would cascade out. I cannot do this because that would be a duplicate key in the PrimaryStuff's unique index.
Feel free to ask questions and I'll try to clear up anything that's confusing.

Answer 1

First lets get a list of the rows that need to be updated (as I understand it you want the lowest ID to replace all the higher IDs)
SELECT MIN(ID) OVER (PARTITION BY SomeValue ORDER BY SomeValue, ID ASC) AS FirstID,
ID,
SOMEVALUE
FROM PrimaryStuff
We can remove the ones where FirstID and ID match, these don't matter
SELECT FirstID, ID FROM
(
SELECT MIN(ID) OVER (PARTITION BY SomeValue ORDER BY SomeValue, ID ASC) AS FirstID,
ID,
SOMEVALUE
FROM PrimaryStuff
) T
WHERE FirstID != ID
Now we have a change list. We can use this in an update statement, put it in a temp table (or a CTE as I did below):
WITH ChangeList AS
(
SELECT FirstID, ID FROM
(
SELECT MIN(ID) OVER (PARTITION BY SomeValue ORDER BY SomeValue, ID ASC) AS FirstID,
ID
FROM PrimaryStuff
) T
WHERE FirstID != ID
)
UPDATE ForeignStuff
SET PrimaryStuffId = ChangeList.FirstID
FROM ForeignStuff
JOIN ChangeList ON ForeignStuff.ID = ChangeList.ID
NB - Code not tested, might have typos.

Answer 2

Could you be more proactive and either use the existing ID when SomeValue already exists and enforce a unique constraint on PrimaryStuff.SomeValue, or why not make SomeValue the primary key of PrimaryStuff. With it as the PrimaryKey then you would only ever add a record to PrimaryStuff if SomeValue did not already exist in it.
Lastly, and most simply, if SomeValue is always arbitrarily defined by others and you take whatever they give you, why not just drop PrimaryStuff altogether and let users enter whatever they wish into ForeignStuff? If you need a unique listing for SomeValue, create a view based on your main table. If you need to speed up querying then add an index to ForeignStuff.SomeValue field.
Here's an (untested) view when there are multiple tables like ForeignStuff:
-- dynamically generate a distinct list of values of interest
select SomeValue from ForeignStuffA
union select SomeValue from ForeignStuffB
union select SomeValue from ForeignStuffC
-- and so on, the union applies distinct