DB2 SQL Getting distinct value when grouping rows - sql

Question

BUSINESSTABLE looks like this:
HOTEL_CHAIN HOTEL_LOCATION HOTEL_OWNER
_____________________________________________________
Marriott Las Vegas Nelson
Best Western New York Richards
Best Western San Francisco Smith
Marriott New York Nelson
Hilton Boston James
I'm trying to execute an SQL statement in a DB2 database that groups these entries by HOTEL_CHAIN. If the rows that are grouped together contain the same HOTEL_LOCATION or HOTEL_OWNER, that info should be preserved. Otherwise, a value of 'NULL' should be displayed. For example, both Marriott hotels have the same owner, Nelson, so I want to display that information in the new table. However, each Marriott hotel is in a different location, so I'd like to display 'NULL' in that column.
The resulting table (HOTELTABLE) should look like this:
HOTEL_CHAIN HOTEL_LOCATION HOTEL_OWNER
_____________________________________________________
Marriott NULL Nelson
Best Western NULL NULL
Hilton Boston James
I'm trying to use the following SQL statement to accomplish this:
INSERT INTO HOTELTABLE(HOTEL_CHAIN,HOTEL_LOCATION,HOTEL_OWNER)
SELECT
HOTEL_CHAIN,
CASE COUNT(DISTINCT(HOTEL_LOCATION)) WHEN 1 THEN HOTEL_LOCATION ELSE 'NULL' END,
CASE COUNT(DISTINCT(HOTEL_OWNER)) WHEN 1 THEN HOTEL_OWNER ELSE 'NULL' END,
FROM BUSINESSTABLE GROUP BY HOTEL_CHAIN
I get an SQL error SQLCODE-119 A COLUMN OR EXPRESSION IN A HAVING CLAUSE IS NOT VALID. It seems to be complaining about the 2nd HOTEL_LOCATION and the 2nd HOTEL_OWNER within my case statements. I also tried using DISTINCT(HOTEL_LOCATION) and that threw another error. Can someone please explain the correct way to code this? Thank you!

Answer 1

Don't use COUNT(DISTINCT). Use MIN() and MAX():
INSERT INTO HOTELTABLE(HOTEL_CHAIN,HOTEL_LOCATION,HOTEL_OWNER)
SELECT HOTEL_CHAIN,
(CASE WHEN MIN(HOTEL_LOCATION) = MAX(HOTEL_LOCATION)
THEN MIN(HOTEL_LOCATION) ELSE 'NULL'
END),
(CASE WHEN MIN(HOTEL_OWNER) = MAX(HOTEL_OWNER)
THEN MIN(HOTEL_OWNER) ELSE 'NULL'
END)
FROM BUSINESSTABLE
GROUP BY HOTEL_CHAIN;
Notes:
Why not COUNT(DISTINCT)? It is generally much more expensive than MIN() and MAX() because it needs to maintain internal lists of all values.
I don't approve of a string value called 'NULL'. Seems like it is designed to foster confusion. Perhaps just NULL the value itself?

Answer 2

I agree Gordon for the null (gj Gordon).
other method
INSERT INTO HOTELTABLE(HOTEL_CHAIN,HOTEL_LOCATION,HOTEL_OWNER)
select distinct f1.HOTEL_CHAIN,
case when f2.HasDiffLocation is not null then 'NULL' else f1.HOTEL_LOCATION end as HOTEL_LOCATION,
case when f3.HasDiffOwner is not null then 'NULL' else f1.HOTEL_OWNER end as HOTEL_OWNER
from BUSINESSTABLE f1
left outer join lateral
(
select 1 HasDiffLocation from BUSINESSTABLE f2b
where f1.HOTEL_CHAIN=f2b.HOTEL_CHAIN and f1.HOTEL_LOCATION<>f2b.HOTEL_LOCATION
fetch first rows only
) f2 on 1=1
left outer join lateral
(
select 1 HasDiffOwner from BUSINESSTABLE f3b
where f1.HOTEL_CHAIN=f3b.HOTEL_CHAIN and f1.HOTEL_OWNER<>f3b.HOTEL_OWNER
fetch first rows only
) f3 on 1=1
or like this :
INSERT INTO HOTELTABLE(HOTEL_CHAIN,HOTEL_LOCATION,HOTEL_OWNER)
select distinct f1.HOTEL_CHAIN,
ifnull(f2.result, f1.HOTEL_LOCATION) as HOTEL_LOCATION,
ifnull(f3.result, f1.HOTEL_OWNER) as HOTEL_LOCATION,
from BUSINESSTABLE f1
left outer join lateral
(
select 'NULL' result from BUSINESSTABLE f2b
where f1.HOTEL_CHAIN=f2b.HOTEL_CHAIN and f1.HOTEL_LOCATION<>f2b.HOTEL_LOCATION
fetch first rows only
) f2 on 1=1
left outer join lateral
(
select 'NULL' result from BUSINESSTABLE f3b
where f1.HOTEL_CHAIN=f3b.HOTEL_CHAIN and f1.HOTEL_OWNER<>f3b.HOTEL_OWNER
fetch first rows only
) f3 on 1=1