在 MySQL 中，根据条件从多行中选择一行，否则选择第一行并排除其余行

Question

I am interested in MySQL of writing a query that looks through a list consisting of IDs and locations. Each ID represents a unique person who can be tied to multiple locations.

I have the following table to simplify things:

+----+----------+
| ID | Location |
+----+----------+
|  1 | Bldg#1   |
|  1 | Bldg#2   |
|  2 | Bldg#3   |
+----+----------+

I am looking to deduplicate the above table to only end up with ONE row per ID, but I would also like to add a conditional that preferences Bldg#1 for any given ID. In other words, given a table of multiple rows with potentially the same ID and multiple locations, I would like to write a query that outputs 1 row per ID, and if any of the rows associated with that ID also have a location of Bldg#1, I want to keep that row and drop the rest. Otherwise, I just want to keep one arbitrary location row for that ID.

For the above table, I would like the following as output:

+----+----------+
| ID | Location |
+----+----------+
|  1 | Bldg#1   |
|  2 | Bldg#3   |
+----+----------+

Answer 1

You can group by id and use conditional aggregation:

select id,
  case 
    when max(location = 'Bldg#1') then 'Bldg#1' 
    else any_value(location)
  end location 
from tablename
group by id

See the demo .
Results:

| id  | location |
| --- | -------- |
| 1   | Bldg#1   |
| 2   | Bldg#3   |

Answer 2

You can use row_number() with a case expression:

select id, location
from (select t.*,
             row_number() over (partition by id
                                order by (case location when 'Bldg#1' then 1 when 'Bldg#2' then 2 when 'Bldg#3' then 3 else 4 end)
                               ) as seqnum
      from t
     ) t
where seqnum = 1;

This does not assume any particular ordering -- such as alphabetical ordering.

Answer 3

Is this you looking for?

Exmaple:

Query:

DROP TABLE IF EXISTS #TEST
CREATE TABLE #TEST (ID INT, Location NVARCHAR(10))

INSERT INTO #TEST 
SELECT 1,'Bldg#1'
UNION
SELECT 1,'Bldg#2'
UNION
SELECT 2,'Bldg#3'

SELECT ID,Location FROM (
SELECT ID,Location, ROW_NUMBER() OVER(PARTITION BY ID ORDER BY ID) AS RNM
FROM #TEST) T
WHERE RNM = 1

Answer 4

inner query will make sure the location is in order so that Bldg#1 is always the first for each id, so then the outer group by will pick the first record

SELECT * FROM 
(
    SELECT id, location
    FROM location
    ORDER BY id, location ASC
)a 
GROUP BY id