我如何为附加的报告编写SQL？

Question

We're moving from data system to another. I'm having to rewrite some reports the new system has a custom reporting module where you can write a query and paste it into a window and push it out to the end user. I have one report that all of users have been requesting but I can't figure out how to write it.

It breaks it down the school population by grade level, gender and race. see attached below.

I wrote this which breaks down the population by grade level but not sure where to go from here.

SELECT
gl.title as Grade,
COUNT (s.student_id)
FROM
students s,
student_enrollment se,
school_gradelevels gl
WHERE
s.student_id = se.student_id
AND se.school_id=gl.school_id
AND se.grade_id=gl."id"
AND se.syear =2012
AND se.end_date IS NULL
AND se.school_id =10

group by gl.title

order by Grade;

For reference gender is s.gender and race is s.race. I'm wondering if were going to have to purchase a reporting software like crystal reports.

Answer 1

There is no data example so would presum that the gender and race information is captured against the student.

What you want to do is get an initial table that would look like that:

GradeLevel   |   Category     |  NumOfStudents
------------------------------------------------
PK           |     B - Male   |  10
PK           |     B - Female |  12
0            |     B - Male   |  5
0            |     B - Female |  6

This is easy to achieve by doing something like that:

SELECT
gl.title as GradeLevel,
s.Race + '-' +s.Gender as Category
COUNT (s.student_id) as NumOfStudents
FROM 
           students s,
inner join student_enrollment se on s.student_id = se.student_id,
inner join school_gradelevels gl on se.school_id=gl.school_id AND se.grade_id=gl."id"
WHERE
    se.syear =2012
AND se.end_date IS NULL
AND se.school_id =10

group by gl.title, s.Race + '-' +s.Gender

order by Grade;

next you would want to make a Pivot out of that data by using your reporting tool weather its SSRS or Crystal or anything else that supports matrix reports.

if there is no report tool available that knows how to make a matrix use the SQL Pivot clause: http://msdn.microsoft.com/en-us/library/ms177410(v=sql.105).aspx

SELECT GradeLevel, [B-Male], [B-Female] ....
FROM 
  (
      SELECT
      gl.title as GradeLevel,
      s.Race + '-' +s.Gender as Category
      COUNT (s.student_id) as NumOfStudents
      FROM 
                 students s,
      inner join student_enrollment se on s.student_id = se.student_id,
      inner join school_gradelevels gl on se.school_id=gl.school_id AND se.grade_id=gl."id"
      WHERE
          se.syear =2012
      AND se.end_date IS NULL
      AND se.school_id =10
      group by gl.title, s.Race + '-' +s.Gender   
  ) rawData
  PIVOT
  (
  SUM(NumOfStudents)
  FOR Category IN ( [B-Male], [B-Female], ... )
) AS pvt 

Answer 2

Fundamentally, you must group by everything you're counting, which appears to be grade, race, and sex. Do that first: get everything in rows, summed up the way you need.

What may be confusing you is that the report has two levels of grouping: one by grade, race, and sex, and one by grade & race. In effect, it looks like this:

create view V as 
SELECT   gl.title as Grade
       , s.sex
       , s.race 
       , COUNT (s.student_id) as Q
FROM students as s
JOIN student_enrollment as se
ON s.student_id = se.student_id
JOIN school_gradelevels gl
ON se.school_id = gl.school_id
AND se.grade_id = gl.id
WHERE
    se.syear = 2012
AND se.end_date IS NULL
AND se.school_id = 10
group by gl.title, s.sex, s.race

select Grade, race, sex, Q
from V
UNION
select Grade, race, 'Z', sum(Q) as Q
from V
group by Grade, race

How you turn that into a cross-tab report depends on the tools at your disposal. It's possible in SQL, using the pivot operator or a series of case s in your select . The trick is to notice you need three columns for every race: male, female, and both. It's a little tedious, here's a sample

select   Grade, 1 as sort_order  
       , max(case race when 1 then 
                  case sex when 'M' then Q end end) as r1mq
       , NULL as r1mb
       , max(case race when 1 then 
                  case sex when 'F' then Q end end) as r1fq
       , max(case race when 2 then 
                  case sex when 'M' then Q end end) as r2mq
       , NULL as r2mb
       , max(case race when 2 then 
                  case sex when 'F' then Q end end) as r2fq
       ...
from V group by Grade
UNION
select   Grade, 2 as sort_order  
       , NULL as r1mq
       , max(case race when 1 then 
                  case sex when 'Z' then Q end end) as r1bq
       , NULL as r1fq
       , NULL as r2mq
       , max(case race when 2 then 
                  case sex when 'Z' then Q end end) as r2bq
       , NULL as r2fq
       ...
from V group by Grade
order by Grade, sort_order

That produces two rows per grade, with the combined student totals in the second row for each one.

You get the idea. You'll also need queries to count students grouped by grade and sex, and by grade alone. Those could be joined to the output of your pivot query. And totals by race, for the bottom row, unioned on with grade of 99 or something to make it sort at the bottom. The result would match the row & column layout of your report.

Answer 3

with detail as (
    select
        gl.title as grade, s.race, s.gender, count(s.student_id) total
    from
        students s
        inner join
        student_enrollment se using (student_id)
        inner join
        school_gradelevels gl on se.school_id = gl.school_id and se.grade = gl.id
    where
        and se.syear = 2012
        and se.end_date is null
        and se.school_id = 10
    group by 1, 2, 3
    order by 1, 2, 3
) agg as (
    select grade, array_agg(total) total
    from detail
    group by grade
)
select
    grade,
    (select sum(e) from unnest(total) s(e)) total_grade,
    total[1,1] + total[2,1] + total[3,1] + total[4,1] + total[5,1] + total[6,1] total_F,
    total[1,2] + total[2,2] + total[3,2] + total[4,2] + total[5,2] + total[6,2] total_M,
    total[1,1] total_A_F,
    total[1,2] total_A_M,
    total[1,1] + total[1,2] total_A,
    total[2,1] total_B_F,
    total[2,2] total_B_M,
    total[2,1] + total[2,2] total_B,
    total[3,1] total_H_F,
    total[3,2] total_H_M,
    total[3,1] + total[3,2] total_H,
    total[4,1] total_I_F,
    total[4,2] total_I_M,
    total[4,1] + total[4,2] total_I,
    total[5,1] total_M_F,
    total[5,2] total_M_M,
    total[5,1] + total[5,2] total_MM,
    total[6,1] total_W_F,
    total[6,2] total_W_M,
    total[6,1] + total[6,2] total_W,
from agg