How to use SQL DISTINCT to remove duplicates from multiple columns? - sql

Question

Let's say I have a table with lots of duplicated values. I want to remove the duplicates for each column individually. Using DISTINCT removes duplicate combinations of columns so other columns still contain duplicated values.
Original table is:
Col1 | Col2 | Col3
-----+------+------
a1 | b1 | c1
a1 | b2 | c1
a2 | b1 | NULL
a2 | b2 | c1
a3 | b1 | c1
a3 | NULL | NULL
My desire result is:
Col1 | Col2 | Col3
-----+------+------
a1 | b1 | c1
a2 | b2 | NULL
a3 | NULL | NULL
I can get this result by several query separately:
SELECT DISTINCT Col1
FROM TABLE
SELECT DISTINCT Col2
FROM TABLE
SELECT DISTINCT Col3
FROM TABLE
But how can I do it in a singe query and return result in one result set?
Thanks

Answer 1

I'd use a group by...
;WITH c1 AS (
SELECT col1
, ROW_NUMBER() OVER (ORDER BY col1) AS [r]
FROM #foo
WHERE col1 IS NOT NULL
GROUP BY col1
)
, c2 AS (
SELECT col2
, ROW_NUMBER() OVER (ORDER BY col2) as [r]
FROM #foo
WHERE col2 IS NOT NULL
GROUP BY col2
)
, c3 AS (
SELECT col3
, ROW_NUMBER() OVER (ORDER BY col3) as [r]
FROM #foo
WHERE col3 IS NOT NULL
GROUP BY col3
)
select c1.col1
, c2.col2
, c3.col3
from c1 LEFT join c2
on c1.r = c2.r
left join c3
on c1.r = c3.r
ORDER BY c1.r ASC;
I wasn't quite sure from the problem description what you wanted. I crafted this based on the ideal-output provided.
Here is the sample data set I used.
CREATE TABLE #foo (
col1 char(2)
, col2 char(2)
, col3 char(2)
);
INSERT INTO #foo (col1, col2, col3)
VALUES ('a1', 'b2', null)
, ('a1', 'b1', 'c1')
, ('a2', Null, 'c1')
, ('a2', 'b1', null)
, ('a3', null, 'c1')
GO
Here is the dataset and output from the query:
Hope this helps!

Answer 2

You can UNION those three queries together:
SELECT DISTINCT Col1 FROM TABLE
UNION
SELECT DISTINCT Col2 FROM TABLE
UNION
SELECT DISTINCT Col3 FROM TABLE
This requires that all three fields be of the same type (can't mix numbers and strings and dates).
This smells of bad design though. If you find yourself unioning these often then perhaps change your table to look like the UNION'd results.