Database Administrators Stack Exchange is a question and answer site for database professionals who wish to improve their database skills and learn from others in the community. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I have two queries which are doing basically the same thing, but have different grouping, the first query (query 1) is used to populate a chart, and the second to populate a table.

Query 1:

SELECT key_id
FROM table
WHERE emp_id = 1
GROUP BY key_id, created_at

Query 2:

SELECT key_id
       ,count(*) OVER() AS full_count
FROM table
WHERE emp_id = 1
GROUP BY key_id

I have created a function that returns those two queries in a json format, chart: [...], table: [...]. The problem is that I need to query the same table twice, because of my grouping. Is there any way of dealing with this situations?

share|improve this question
If the number of rows coming back is small enough, it's completely legitimate to do query 1 and then do a second roll up in code/memory instead of query 2. – jpmc26 Jan 26 at 0:06
up vote 4 down vote accepted

The new 9.5 version has added CUBE, GROUPING SETS and ROLLUP extensions to GROUP BY which can be used for such queries. If you have to do this in previous versions, one way is using a CTE, with something like (note that it's not the same exact output as your query, since I didn't know the type of created_at. I guess it's a timestamp and didn't want to mix it with the integer count):

  ( SELECT key_id,
           sum(salary) AS salary,
           sum(bonus)  AS bonus,
           count(*)    AS full_count
    FROM table
    WHERE emp_id = 1
    GROUP BY key_id, created_at
FROM grp


SELECT key_id, 
FROM grp
GROUP BY key_id 

ORDER BY key_id, created_at NULLS LAST;
share|improve this answer
@AndriyM: I think so, too. It's the difference I tried to address with count_distinct_created_at_per_key vs. count_rows_per_key in my query. – Erwin Brandstetter Jan 25 at 19:11
@Erwin and AndriyM thanx, I missed that. – ypercubeᵀᴹ Jan 25 at 19:58

Depending on how you want to present your data, you could do it in a single query, a single query-level, even:

SELECT key_id
     , created_at
     , sum(salary) AS sum_salary  -- numbers per (key_id, created_at)
     , sum(bonus)  AS sum_bonus
     , count(*)    AS count_rows
     , sum(sum(salary)) OVER w AS total_salary_per_key  -- totals per (key_id)
     , sum(sum(bonus))  OVER w AS total_bonus_per_key
     , count(*)         OVER w AS count_distinct_created_at_per_key
     , sum(count(*))    OVER w AS count_rows_per_key
FROM   tbl
WHERE  emp_id = 1
GROUP  BY key_id, created_at

This lists totals per key (repeatedly) for every aggregated row per (key_id, created_at).

Detailed explanation for the technique:

If you actually want to implement ROLLUP see @ypercube's answer or this related one:

share|improve this answer

Since each row's values are aggregated twice, you could also try duplicating each row in a manner that would allow you to perform both level's aggregations in one pass. The query below uses LATERAL for that purpose:

  SUM(salary)   AS total_salary,
  SUM(bonus)    AS total_bonus,
  COUNT(*)      AS full_count
  table AS t,
      (t.key_id, t.created_at, t.salary, t.bonus),
      (t.key_id, NULL,         t.salary, t.bonus)
    ) AS v
  v.key_id      ASC,
  v.created_at  ASC NULLS LAST

As you can see, one of the copies has NULL instead of created_at and the query groups by the copied key_id and created_at rather then the original ones. The effect of this is that one half of the row set effectively gets grouped by key_id alone, and so you get aggregated results for both levels in one go.

I have to admit, though, that I have no PostgreSQL instance to test this solution and thus no idea how it will perform compared to the other suggestions posted.

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.