SQL to find if refund would be processed or not - mysql

Question

I have 2 tables, they can be created with the following query:
CREATE TABLE transactions(Id integer,ptime date, rtime date, sid text, itemid text, gtv integer);
/* Create few records in this table */
INSERT INTO transactions VALUES(3,'2019-09-19',null,'a','a1',58);
INSERT INTO transactions VALUES(12,'2019-12-10','2019-12-15','b','b2',475);
INSERT INTO transactions VALUES(3,'2020-09-01','2020-09-02','f','f9',33);
INSERT INTO transactions VALUES(2,'2020-04-30',null,'d','d3',250);
INSERT INTO transactions VALUES(1,'2020-10-22',null,'f','f2',91);
INSERT INTO transactions VALUES(8,'2020-04-16',null,'e','e7',24);
INSERT INTO transactions VALUES(5,'2019-09-23',null,'g','g6',61);
CREATE TABLE Items(sid text , itemid text, category text, name text);
/* Create a few records in this table */
INSERT INTO Items VALUES('a','a1','pants','denimpants');
INSERT INTO Items VALUES('a','a2','tops','blouse');
INSERT INTO Items VALUES('f','f1','table','coffee table');
INSERT INTO Items VALUES('f','f5','chair','loungechair');
INSERT INTO Items VALUES('f','f6','chair','armchair');
INSERT INTO Items VALUES('d','d2','jewelry','bracelet');
INSERT INTO Items VALUES('b','b4','earphone','airpods');
Select * from NAMES;
COMMIT;
Create a flag in the transaction items table indicating whether the refund can be processed or not. The condition for a refund to be processed is that it has to happen within 72 of Purchase time.
Expected Output: Only 1 of the three refunds would be processed in this case
Create a rank by buyer_id column in the transaction items table and filter for only the second purchase per
buyer. (Ignore refunds here)
Expected Output: Only the second purchase of buyer_id 3 should the output
How will you find the second transaction time per buyer (don’t use min/max; assume there were more
transactions per buyer in the table)
Expected Output: Only the second purchase of buyer_id along with a timestamp
I have been trying to wrap my head around this, but cant seem to understand where to start.

Answer 1

Here is a test script treating the questions. NB I am running MariaDB. There may be adjustments needed for mySQL
We set up the database, written to be reusable.
USE test;
DROP TABLE IF EXISTS transactions;
DROP TABLE IF EXISTS items;
CREATE TABLE IF NOT EXISTS transactions(Id integer,ptime date, rtime date, sid text, itemid text, gtv integer);
/* Create few records in this table */
INSERT INTO transactions VALUES(3,'2021-09-19',null,'a','a1',58);
INSERT INTO transactions VALUES(12,'2021-12-10','2022-03-15','b','b2',475);
INSERT INTO transactions VALUES(3,'2021-09-01','2021-09-02','f','f9',33);
INSERT INTO transactions VALUES(2,'2021-12-30',null,'d','d3',250);
INSERT INTO transactions VALUES(1,'2021-10-22',null,'f','f2',91);
INSERT INTO transactions VALUES(8,'2021-04-16',null,'e','e7',24);
INSERT INTO transactions VALUES(5,'2022-01-23',null,'g','g6',61);
CREATE TABLE IF NOT EXISTS items(sid text , itemid varchar(25) PRIMARY KEY, category text, name text);
/* Create a few records in this table */
INSERT INTO items VALUES('a','a1','pants','denimpants');
INSERT INTO items VALUES('a','a2','tops','blouse');
INSERT INTO items VALUES('f','f1','table','coffee table');
INSERT INTO items VALUES('f','f5','chair','loungechair');
INSERT INTO items VALUES('f','f6','chair','armchair');
INSERT INTO items VALUES('d','d2','jewelry','bracelet');
INSERT INTO items VALUES('b','b4','earphone','airpods');
COMMIT;
Select * from transactions;
Select * from items;
Here we add a virtual column which gives us the refund status using the sales date and the request date.
request date before sales date = error
no request date = Not requested
request date 0 to 72 days after sales date = accepted
request date more then 72 days after sales date = too late
ALTER TABLE transactions ADD COLUMN refund VARCHAR(20) AS (CASE WHEN rtime is NULL THEN 'Not requested' WHEN rtime < ptime THEN "Error" WHEN DATEDIFF(rtime,ptime) > 72 THEN "too late" ELSE "accepted" END);
Select ptime,rtime,datediff(ptime,rtime) dif,refund from transactions;
To get the 2nd purchase using rank we create a query assigning rank as follows:
SELECT ID,
ptime,
RANK() OVER(PARTITION BY ID ORDER BY ptime ASC ) Rank
FROM transactions
ORDER BY ID,
Rank;
Which we can then use as a CTE and use a where to only return the second record for each I'd.
WITH rankings AS (
SELECT ID,
ptime,
RANK() OVER(PARTITION BY ID ORDER BY ptime ASC ) Rank
FROM transactions
ORDER BY ID,
Rank )
SELECT ID,ptime
FROM rankings
WHERE Rank = 2;
This is all very well but the request was for a virtual column and not a query.
When we try to use aggregation functions in a virtual column we get errors because the definition of a virtual column should only reference fields in the same row, of columns already declared in the table definition. (These lines are commented out to avoid blocking the script.
Gives an error
#ALTER TABLE transactions ADD COLUMN rank_id INT AS ( RANK() OVER ( PARTITION BY ID ORDER BY ptime ASC ));
#ERROR 1901 (HY000) at line 38: Function or expression 'rank()' cannot be used in the GENERATED ALWAYS AS clause of `rank_id`
#ALTER TABLE transactions ADD COLUMN rank_id INT AS ( RANKX(transactions.id,sum(idl)));
#ERROR 1901 (HY000) at line 41: Function or expression 'sum()' cannot be used in the GENERATED ALWAYS AS clause of `rank_id`
#ALTER TABLE transactions ADD COLUMN rank_id INT AS ( RANKX(transactions.id,id));
#ERROR 1901 (HY000) at line 55: Function or expression '`RANKX`()' cannot be used in the GENERATED ALWAYS AS clause of `rank_id`
It seems that some aggregation functions can be used in virtual columns some SQL engines. I've tried a few but I have yet to find one which works here.
I am wondering whether this is a "trick" question to encourage you to research the principales of virtual columns?