SQL : get PK IDs that are in a CSV list (or field) that aren't in another CSV list - sql

Question

(Edited to add info on the context)
I have 2 fields in Table A containing CSV lists of IDs of records in 2 other tables. The "USERS" field contains a CSV list of records in USERS_TABLE; the "CONTACTS" field contains a CSV list of records in CONTACTS_TABLE:
USERS_FIELD: "1,2,3,4,5,6"
CONTACTS_FIELD: "2,4,6,8"
I want to find all records that are in USERS_FIELD list but not in CONTACTS_FIELD list. In this instance I want records 1,3,5. The lists can be anywhere from 1 ID to hundreds.
The solution has to run in the WHERE clause of a query. My environment is a VBScript-based scripting language inside a COTS product: MicroFocus/Serena SBM running on MS Windows Server and SQL Server 2012. The scripting language allows me to specify the WHERE and ORDERBY clauses and it does the query and returns the results. The storage of multiple record IDs as CSV is built into the product. I can't do anything about it, nor can I create SQL temp tables or define SQL functions. The implementation of the host scriping language removed arrays and the "Split" function. While I can parse the CSV to a Dictionary object, iterating a pair of those, each with several hundred elements is not fast. This is all happening while the end-user is waiting for a web page to complete. Again, that's how the product was designed.
Can I use a UNION type operator and do something like:
Select ID from USERS_TABLE Where ID in USERS_FIELD
MINUS
Select ID from CONTACTS_TABLE Where ID in CONTACTS_FIELD

Answer 1

Not sure I follow the requirement for the solution needing to run in the WHERE clause. If you're using SQL Server 2017, you can take advantage of the STRING_SPLIT (also available in SQL Server 2016) and STRING_AGG functions.
DROP TABLE IF EXISTS #A;
CREATE TABLE #A (id INT PRIMARY KEY IDENTITY, users VARCHAR(MAX), contacts VARCHAR(MAX));
INSERT INTO #A (users, contacts)
VALUES
('1,2,3,4,5,6', '2,4,6,8'),
('3,5,6', '4,6,9'),
('2,4,7,9', '2,4,9');
SELECT
A.id,
A.users,
A.contacts,
STRING_AGG(B.value, ',') intersection
FROM #A A
CROSS APPLY STRING_SPLIT(users, ',') B
WHERE NOT EXISTS (SELECT * FROM STRING_SPLIT(A.contacts, ',') X1 WHERE B.value = X1.value) -- where user is not in contacts
GROUP BY
A.id,
A.users,
A.contacts;