Is there a way to return all unique values within a given row - sql

Question

I am working in Oracle SQL and I have a data set with these headers
In this data set card number is unique, however, the customer can have duplicates in their Suggestion fields. Rather than going through and writing case statements, is there a way to keep only the unique values within the given row?
Please note, some customers will be left with more "unique" suggestions than others
For example:
My goal would be for my final output to look like this
As I have mentioned, previously, I would just write case statements saying
SELECT DISTINCT CARD_NUMBER
,SUGGESTION_1
,CASE
WHEN SUGGESTION_2 != SUGGESTION_1
THEN SUGGESTION_2
WHEN SUGGESTION_3 != SUGGESTION_1
THEN SUGGESTION_3
WHEN SUGGESTION_4 != SUGGESTION_1
THEN SUGGESTION_4
WHEN SUGGESTION_5 != SUGGESTION_1
THEN SUGGESTION_5
END AS SUGGESTION_2
CASE
WHEN SUGGESTION_2 != SUGGESTION_1
AND SUGGESTION_3 != SUGGESTION_1
AND SUGGESTION_3 != SUGGESTION_2
THEN SUGGESTION_3
`
I would do this until all unique values are left, and there just has to be an easier way
Any help would be EXTREMELY appreciated, thank you!

Answer 1

You can use union all and conditional aggregation. Here is the idea that puts the results in a single column:
select card, listagg(suggestion, ', ') within group (order by which) as suggestions
from (select card, suggestion, min(which) as which
from ((select card, 1 as which, suggestion_1 as suggestion from t) union all
(select card, 2, suggestion_2 from t) union all
(select card, 3, suggestion_3 from t) union all
(select card, 4, suggestion_4 from t) union all
(select card, 5, suggestion_5 from t)
) t
group by card, suggestion
) t
group by card;
You can do something similar with conditional aggregation if you want the values in separate columns.

Answer 2

I would try to pivot the table to long, and then back to wide
Setup:
create table testtbl
(
CARD_NUMBER int
,SUGGESTION_1 varchar2(100)
,SUGGESTION_2 varchar2(100)
,SUGGESTION_3 varchar2(100)
,SUGGESTION_4 varchar2(100)
,SUGGESTION_5 varchar2(100)
);
insert into testtbl values (1234,'G11','G4','G3','G2','G6');
insert into testtbl values (5678,'G4','G6','G6','G11','G6');
insert into testtbl values (9101,'G1','G3','G11','G4','G11');
Then the Query itself, first the pivoting to long. Here I use a function just to return the numbers from 1 to 5 - this is instead of joining the table 5 times to itself, this way it should only pass through the test table once.
I then use the analytical functionrow_number to sort the unique values according to their first placement.
The second select uses MAX to pivot back to wide
with cte AS
(
SELECT
CARD_NUMBER
,MIN(n.column_value ) n
,CASE n.column_value
WHEN 1 THEN f.SUGGESTION_1
WHEN 2 THEN f.SUGGESTION_2
WHEN 3 THEN f.SUGGESTION_3
WHEN 4 THEN f.SUGGESTION_4
WHEN 5 THEN f.SUGGESTION_5
END Suggestion
,ROW_NUMBER() OVER (PARTITION BY f.CARD_NUMBER ORDER BY MIN(n.column_value)) rn
FROM testtbl f
CROSS JOIN table(sys.odcinumberlist(1,2,3,4,5)) n
GROUP BY f.CARD_NUMBER,CASE n.column_value
WHEN 1 THEN f.SUGGESTION_1
WHEN 2 THEN f.SUGGESTION_2
WHEN 3 THEN f.SUGGESTION_3
WHEN 4 THEN f.SUGGESTION_4
WHEN 5 THEN f.SUGGESTION_5
END
)
SELECT
CARD_NUMBER
,MAX(CASE WHEN rn=1 THEN Suggestion ELSE '' end)SUGGESTION_1
,MAX(CASE WHEN rn=2 THEN Suggestion ELSE '' end)SUGGESTION_2
,MAX(CASE WHEN rn=3 THEN Suggestion ELSE '' end)SUGGESTION_3
,MAX(CASE WHEN rn=4 THEN Suggestion ELSE '' end)SUGGESTION_4
,MAX(CASE WHEN rn=5 THEN Suggestion ELSE '' end)SUGGESTION_5
FROM cte
GROUP BY CARD_NUMBER
ORDER BY CARD_NUMBER