Question

I have a table Stores, and a table Schools. This is a one to many relationship– multiple schools can be served by the same store, but not vice-versa.

Earlier on in development, I made the mistake of repeating the same store multiple times in the Stores database. I inserted rows like:

Store_ID| Store_URL
1       | http://sameurl.com
2       | http://sameurl.com

And then if two different schools were at that same store, I’d be referencing 1 in one school row, and 2 in another.

I’m able to identify duplicates quite easily by using GROUP BY on Store_URL and using COUNT() to identify duplicates.

The difficult task ahead of me is making all the Schools point to non-duplicate Stores. If I simply delete duplicate Stores, I’ll have Schools which point to nonexistent rows.

What can I do to eliminate duplicates and make schools that share the same store point to the same Store row?

Note: there are thousands of schools and stores. Manual solutions don’t work.

Answer 1

Assuming your School table has a store_ID from what you’ve said.

I would start by figuring out for each duplicate, which store_ID you want to keep. I will also assume that you want it to be the lowest ID value. I would then update the Schools’ store_ID to be the MIN(store_ID) for the current URL they have. You should then be free to delete the extra store_ID records

This is how I would go about the update:

UPDATE sch
SET sch.Store_ID = matcher.store_ID
FROM Schools AS sch
INNER JOIN Stores AS st ON sch.store_ID = st.store_ID
INNER JOIN
(
   SELECT MIN(st.store_id) AS store_ID, store_url
   FROM Schools AS sch
   INNER JOIN Stores AS st ON sch.store_ID = st.store_ID
   GROUP BY Store_URL
) AS matcher ON st.Store_URL = matcher.Store_Url
   AND st.Store_ID != matcher.store_ID

If you are able to delete stores that do not have an associated school, the following query will remove the extra rows:

DELETE FROM st
FROM Stores AS st
LEFT JOIN Schools AS sch ON st.Store_ID = sch.Store_Id
WHERE sch.Store_id IS NULL

If you only want to delete the Store’s duplicate records, I would look at this query instead of the above:

DELETE FROM st
FROM Stores AS st
INNER JOIN
(
   SELECT MIN(st.store_ID) store_Id, st.Store_Url
   FROM Stores AS st
   GROUP BY st.Store_URL
) AS useful ON st.Store_Url = useful.Store_URL
WHERE st.Store_ID != useful.store_Id