Question

We have a school district with 38 elementary schools. The kids took a test. The averages for the schools are widely dispersed, but I want to compare the averages of JUST THE TOP 10 students from each school.

Requirement: use temporary tables only.

I have done this in a very work-intensive, error-prone sort of way as follows.
(sch_code = e.g., 9043;
— schabbrev = e.g., “Carter”;
— totpct_stu = e.g., 61.3)

DROP TEMPORARY TABLE IF EXISTS avg_top10 ;
CREATE TEMPORARY TABLE avg_top10
   ( sch_code   VARCHAR(4),
     schabbrev  VARCHAR(75),
     totpct_stu DECIMAL(5,1)
   );

INSERT 
  INTO avg_top10
SELECT sch_code
     , schabbrev
     , totpct_stu
  FROM test_table
 WHERE sch_code IN ('5489')
 ORDER
    BY totpct_stu DESC
 LIMIT 10;

-- I do that last query for EVERY school, so the total 
-- length of the code is well in excess of 300 lines.  
-- Then, finally...

SELECT schabbrev, ROUND( AVG( totpct_stu ), 1 ) AS top10
  FROM avg_top10
 GROUP
    BY schabbrev
 ORDER
    BY top10 ;

-- OUTPUT:
-----------------------------------
schabbrev   avg_top10
----------  ---------
Goulding         75.4
Garth            77.7
Sperhead         81.4
Oak_P            83.7
Spring           84.9
-- etc...

Question: So this works, but isn’t there a lot better way to do it?

Thanks!

PS — Looks like homework, but this is, well…real.

Answer 1

Using this technique.

select sch_code,
       schabbrev,
       ROUND( AVG( totpct_stu ), 1 ) AS top10
from   (select sch_code,
               schabbrev,
               totpct_stu,
               @num := if(@group = sch_code, @num + 1, 1) as row_number,
               @group := sch_code as dummy
        from   test_table
        order by sch_code, totpct_stu desc) as x
where  row_number <= 10
GROUP BY sch_code,
       schabbrev