Sql: How can I select only one record which non present an id

问题

I use the Informix database where I have a table coord :

+--------------------+-------------+
| Field              | Type        |
+--------------------+-------------+
| cm_key_coord_code  | char(8)     |
| cm_t_coor          | int         |
| descr_coord        | char(30)    |
+--------------------+-------------+

which contains data such as those shown below:

+-cm_key_coord_code-+-cm_t_coor-+-descr_coord-+
+ 00000001          +    1      + coord01a    +
+ 00000001          +    2      + coord01b    +
+ 00000002          +    1      + coord02a    +
+ 00000002          +    2      + coord02b    +
+ 00000003          +    1      + coord03a    +
+ 00000004          +    2      + coord04a    +
+ 00000005          +    1      + coord05a    +
+-------------------+-----------+-------------+

The table has multiple record with same cm_key_coord_code, but different cm_t_coor. How can I select cm_key_coord_code that have only rows with cm_t_coor != 2 ?

I want to select only the following values:

+-cm_key_coord_code-+-cm_t_coor-+-descr_coord-+
+ 00000003          +    1      + coord03a    +
+ 00000005          +    1      + coord05a    +
+-------------------+-----------+-------------+

回答1:

SELECT t1.*
  FROM Table t1,
       ( SELECT cm_key_coord_code
           FROM Table
         GROUP BY cm_key_coord_code   
         HAVING COUNT(*) = 1      
        ) t2
 WHERE t1.cm_key_coord_code = t2.cm_key_coord_code
   AND t1.cm_t_coor <> 2;

回答2:

I would approach this using group by and having:

select c.cm_key_coord_code
from coord c
group by c.cm_key_coord_code
having sum(case when c.cm_t_coor = 2 then 1 else 0 end) = 0;

You can get the original rows in one of two ways. One method is to join the results together or (equivalently) using in or exists:

select c.*
from coord c
where c.cm_key_coord_code in (select c.cm_key_coord_code
                              from coord c
                              group by c.cm_key_coord_code
                              having sum(case when c.cm_t_coor = 2 then 1 else 0 end) = 0
                             );

Alternatively, if you know there is always only one row, you can use aggregation functions:

select c.cm_key_coord_code, min(cm_t_coor), min(descr_coord)
from coord c
group by c.cm_key_coord_code
having sum(case when c.cm_t_coor = 2 then 1 else 0 end) = 0

回答3:

You could do this with a NOT EXISTS clause as well. Try:

SELECT c.cm_key_coord_code
FROM coord c
WHERE NOT EXISTS
(SELECT 1
 FROM coord c2
 WHERE c2.cm_key_coord_code = c.cm_key_coord_code
 AND c2.cm_t_coor = 2)

来源：https://stackoverflow.com/questions/40640610/sql-how-can-i-select-only-one-record-which-non-present-an-id

标签

sql

select

informix