如何在SQL中每列值仅选择一行？

Question

So, I have the following schema build_tasks:

id|building|queue_time|start_time|completion_time|status|createdAt|updatedAt|baseId|

I'm trying to get only the build tasks that have status 'pending', where theres no build_task with the same baseId in status 'in-progress'.

So far, I managed to get a table with all the pending build tasks where there's no in-progress build task. This is the query:

select * from (select build_tasks.* from build_tasks
            where status = 'pending') as p
left join in_progress_build_tasks ipbt on p."baseId" = ipbt."baseId"
      where ipbt."baseId" is null;

where in_progress_build_tasks is the view:

CREATE OR REPLACE VIEW "public".in_progress_build_tasks AS
 SELECT DISTINCT build_tasks."baseId"
   FROM build_tasks
  WHERE build_tasks.status = 'in-progress'::enum_build_tasks_status;

Which for the table:

id |building            |queue_time          |start_time          |completion_time     |status      |createdAt           |updatedAt           |baseId |
---|--------------------|--------------------|--------------------|--------------------|------------|--------------------|--------------------|-------|
7  |resource01_refinery |2018-02-04 14:09:49 |                    |                    |pending     |2018-02-04 14:09:49 |2018-02-04 14:09:49 |1      |
10 |resource01_refinery |2018-02-04 14:45:07 |                    |                    |pending     |2018-02-04 14:45:07 |2018-02-04 14:45:07 |1      |
6  |resource01_refinery |2018-02-04 14:07:32 |2018-02-04 14:07:58 |2018-02-04 14:08:08 |in-progress |2018-02-04 14:07:32 |2018-02-04 14:08:09 |1      |
12 |resource01_refinery |2018-02-04 14:46:04 |2018-02-04 14:46:04 |2018-02-04 14:46:04 |successful  |2018-02-04 14:46:04 |2018-02-04 14:58:28 |2      |
8  |resource01_refinery |2018-02-04 14:10:29 |2018-02-04 14:10:29 |2018-02-04 14:10:39 |successful  |2018-02-04 14:10:29 |2018-02-04 14:10:39 |2      |
9  |resource01_refinery |2018-02-04 14:11:38 |                    |                    |pending     |2018-02-04 14:11:38 |2018-02-04 14:11:38 |2      |
11 |resource01_refinery |2018-02-04 14:45:14 |                    |                    |pending     |2018-02-04 14:45:14 |2018-02-04 14:45:14 |2      |
13 |resource01_refinery |2018-02-04 15:11:16 |                    |                    |pending     |2018-02-04 15:11:16 |2018-02-04 15:11:16 |3      |
15 |resource01_refinery |2018-02-04 15:11:19 |                    |                    |pending     |2018-02-04 15:11:19 |2018-02-04 15:11:19 |3      |
14 |resource01_refinery |2018-02-04 15:11:18 |                    |                    |pending     |2018-02-04 15:11:18 |2018-02-04 15:11:18 |3      |

Gives me the output:

id |building            |queue_time          |start_time |completion_time |status  |createdAt           |updatedAt           |baseId |baseId |
---|--------------------|--------------------|-----------|----------------|--------|--------------------|--------------------|-------|-------|
9  |resource01_refinery |2018-02-04 14:11:38 |           |                |pending |2018-02-04 14:11:38 |2018-02-04 14:11:38 |2      |       |
11 |resource01_refinery |2018-02-04 14:45:14 |           |                |pending |2018-02-04 14:45:14 |2018-02-04 14:45:14 |2      |       |
13 |resource01_refinery |2018-02-04 15:11:16 |           |                |pending |2018-02-04 15:11:16 |2018-02-04 15:11:16 |3      |       |
14 |resource01_refinery |2018-02-04 15:11:18 |           |                |pending |2018-02-04 15:11:18 |2018-02-04 15:11:18 |3      |       |
15 |resource01_refinery |2018-02-04 15:11:19 |           |                |pending |2018-02-04 15:11:19 |2018-02-04 15:11:19 |3      |       |

How can I reduce the results to only 1 row per base_id, selected based on the lowest queue_time?

Answer 1

I am unclear what you want for output. But if you want to identify the base ids that meet your conditions, you can use aggregation:

select bt.baseid
from build_tasks bt
group by bt.baseid
having sum( (bt.status = 'pending'::enum_build_tasks_status)::int) > 0 and
       sum( (bt.status = 'in-progress'::enum_build_tasks_status)::int) = 0 ;

I am not sure what else you want in the output. Likely you can get what you want using aggregation. Alternatively, a join , in , or exists can get what you want.

However, you do not a need a view to accomplish what you are doing.

Answer 2

You may apply DISTINCT ON (baseId) above your output.

SELECT * FROM
(

SELECT DISTINCT ON (baseId)  youroutput.*
  FROM youroutput ORDER BY baseId,updatedAt
) as a;