SQL query for GROUP BY, return groups that match the conditions of LEAST, COALESCE

问题

I want to write an SQL query calling for several columns with a bit complicated conditions. I'm working on R Studio using RMySQL package. My server is MySQL.

The table looks like this.

organisation    Tour_ID             A           B           C           D  
Ikea                  a    2018-04-01  2018-05-07  2018-05-09  2018-05-01
Ikea                  a    2018-06-01  2018-05-03  2018-05-29          NA   
Ikea                  a    2018-04-02  2018-05-01  2018-07-08  2018-05-26 
Ikea                  b    2018-06-02  2018-05-01          NA  2018-05-26
Ikea                  b    2018-06-02  2018-05-01          NA  2018-05-26
Ikea                  b            NA  2018-05-05  2018-08-02  2018-06-01
Ikea                  c    2018-06-01  2018-05-07  2018-05-09  2018-05-01
Ikea                  c    2018-06-01  2018-05-03          NA          NA   
Ikea                  c    2018-08-02  2018-05-09  2018-07-08  2018-05-26

This is what I want to do:

filter the rows where organisation = Ikea

groupby by Tour_ID like this:

organisation    Tour_ID             A           B           C           D  
Ikea                  a    2018-04-01  2018-05-07  2018-05-09  2018-05-01
Ikea                  a    2018-06-01  2018-05-03  2018-05-29          NA   
Ikea                  a    2018-04-02  2018-05-01  2018-07-08  2018-05-26 


Ikea                  b    2018-06-02  2018-05-01          NA  2018-05-26
Ikea                  b    2018-06-02  2018-05-01          NA  2018-05-26
Ikea                  b            NA  2018-05-05  2018-08-02  2018-06-01


Ikea                  c    2018-06-01  2018-05-07  2018-05-09  2018-05-01
Ikea                  c    2018-06-01  2018-05-03          NA          NA   
Ikea                  c    2018-08-02  2018-05-09  2018-07-08  2018-05-26

in each group of Tour_ID, look at the earliest date in columns A, B, C and D. If the earliest date among the four columns in the group is between 2018-05-01 and 2018-05-31, return the entire group. If a row contains NA values, I want to ignore the NAs and see what's the earliest date among the rest of the values. For example, for the group of Tour_ID = a, the earliest date is 2018-04-01 therefore it doesn't meet the criteria.

In conclusion, only the groups where Tour_ID = b and Tour_ID = c match the conditions. The result should be:

organisation    Tour_ID             A           B           C           D
Ikea                  b    2018-06-02  2018-05-01          NA  2018-05-26
Ikea                  b    2018-06-02  2018-05-01          NA  2018-05-26
Ikea                  b            NA  2018-05-05  2018-08-02  2018-06-01
Ikea                  c    2018-06-01  2018-05-07  2018-05-09  2018-05-01
Ikea                  c    2018-06-01  2018-05-03          NA          NA   
Ikea                  c    2018-08-02  2018-05-09  2018-07-08  2018-05-26

How should I write an SQL query? Here is my attempt, but I just don't know how to do groupby, and how to return the entire group not just the rows with the earliest date.

SELECT *
FROM myTable 
WHERE organisation LIKE 'Ikea' AND
GROUP BY 'Tour_ID' AND
LEAST(COALESCE(A, '2019-01-01'), COALESCE(B, '2019-01-01'), COALESCE(C, '2019-01-01'), COALESCE(D, '2019-01-01')) >= '2018-05-01' AND
LEAST(COALESCE(A, '2019-01-01'), COALESCE(B, '2019-01-01'), COALESCE(C, '2019-01-01'), COALESCE(D, '2019-01-01')) < '2018-06-01';

('2019-01-01' is to replace NAs)

Thank you for any kinds of help!

ADDED: Following the answer by Gordon, here I rewrote the SQL statement.

"SELECT t.* FROM myTable JOIN (SELECT organisation, Tour_ID 
                                                   FROM myTable
                                                   WHERE organisation LIKE 'Ikea' AND
                                                   GROUP BY organisation, Tour_ID
                                                   HAVING LEAST(COALESCE(MIN(A), '2119-01-01'), 
                                                                COALESCE(MIN(B), '2119-01-01'), 
                                                                COALESCE(MIN(C), '2119-01-01'), 
                                                                COALESCE(MIN(D), '2119-01-01')) >= '2018-05-01' AND
                                                          LEAST(COALESCE(MIN(A), '2119-01-01'), 
                                                                COALESCE(MIN(B), '2119-01-01'), 
                                                                COALESCE(MIN(C), '2119-01-01'), 
                                                                COALESCE(MIN(D), '2119-01-01')) < '2018-06-01'
                                                  ) tt
                                                  ON tt.Tour_ID = t.Tour_ID AND
                                                     tt.organisation = t.organisation"

And I ran dbGetQuery from RMySQL package. But I get the following error. I don't understand because GROUP BY part seems quite okay. Does anyone know why I'm getting this error?

dbGetQuery(connection = connection, statement = condition)

Error in .local(conn, statement, ...) : could not run statement: You have an error in your SQL syntax; check the manual that corresponds to your MySQL server version for the right syntax to use near 'GROUP BY organisation, Tour_ID HAVING LEAST(COALESCE(A' at line 1

回答1:

First get the tour_ids that match the conditions:

SELECT Tour_ID
FROM myTable 
WHERE organisation LIKE 'Ikea'
GROUP BY Tour_ID
HAVING LEAST(COALESCE(MIN(A), '2019-01-01'), COALESCE(MIN(B), '2019-01-01'), COALESCE(MIN(C), '2019-01-01'), COALESCE(MIN(D), '2019-01-01')) >= '2018-05-01' AND
       LEAST(COALESCE(MIN(A), '2019-01-01'), COALESCE(MIN(B), '2019-01-01'), COALESCE(MIN(C), '2019-01-01'), COALESCE(MIN(D), '2019-01-01')) < '2018-06-01';

Then put this into a query to get the original rows. Here is one way:

select t.*
from mytable t join
     (SELECT organisation, Tour_ID
      FROM myTable 
      WHERE organisation LIKE 'Ikea'
      GROUP BY organisation, Tour_ID
      HAVING LEAST(COALESCE(MIN(A), '2019-01-01'), COALESCE(MIN(B), '2019-01-01'), COALESCE(MIN(C), '2019-01-01'), COALESCE(MIN(D), '2019-01-01')) >= '2018-05-01' AND
             LEAST(COALESCE(MIN(A), '2019-01-01'), COALESCE(MIN(B), '2019-01-01'), COALESCE(MIN(C), '2019-01-01'), COALESCE(MIN(D), '2019-01-01')) < '2018-06-01'
     ) tt
     ON tt.tour_id = t.tour_id AND
        tt.organisation = t.organisation;

来源：https://stackoverflow.com/questions/51079281/sql-query-for-group-by-return-groups-that-match-the-conditions-of-least-coales

标签

mysql

sql

group-by

rmysql