Rollup and Display Non Existing Rows

Question

This query produces the results except that I would like to have a final total and overal average (excluding the "No GPA"). Shown as XXX in the output sample. Can this be done in one query or do I need to union for the summary?
Additionally, if possible, how can I write some statements so that rows that may not exist can be displayed on the output? Example: there are no GPAs in the range of 0.1 thru 0.9. My query results do not show the first two rows, however, the users would like to see these rows displayed regardless of the existence or not.
GPARange            Count  Avg
0.1 - 0.6  (F )            0      0.00
0.7 - 0.9  (D-)            0      0.00
1.0 - 1.2  (D )            1     
1.3 - 1.6  (D+)            1    
1.7 - 1.9  (C-)            4     
2.0 - 2.2  (C )           20   
2.3 - 2.6  (C+)           81
2.7 - 2.9  (B-)          189
3.0 - 3.2  (B )          387
3.3 - 3.6  (B+)          452
3.7 - 4.9  (A )          316
NO GPA                   271
TOTAL                   XXXX    X.XX
SELECT academic_period,
       stu_population,
       College, 
        CASE
            WHEN ((MBA_GPA)) <= 0.6 then '0.1 - 0.6 (F )'
            WHEN ((MBA_GPA)) <= 0.9 then '0.7 - 0.9 (D-)'
            WHEN ((MBA_GPA)) <= 1.2 then '1.0 - 1.2 (D )'
            WHEN ((MBA_GPA)) <= 1.6 then '1.3 - 1.6 (D+)'
            WHEN ((MBA_GPA)) <= 1.9 then '1.7 - 1.9 (C-)'
            WHEN ((MBA_GPA)) <= 2.2 then '2.0 - 2.2 (C )'
            WHEN ((MBA_GPA)) <= 2.6 then '2.3 - 2.6 (C+)'
            WHEN ((MBA_GPA)) <= 2.9 then '2.7 - 2.9 (B-)'
            WHEN ((MBA_GPA)) <= 3.2 then '3.0 - 3.2 (B )'
            WHEN ((MBA_GPA)) <= 3.6 then '3.3 - 3.6 (B+)'
            WHEN ((MBA_GPA)) >  3.6 then '3.7 - 4.9 (A-)'
            WHEN ((MBA_GPA)) is null then 'No GPA'
        ELSE
            'What Happened'
       END as GPARange,
       count(id), round(avg(mba_gpa),2) AVG_GPA
FROM gpa_stat
where stu_population='F'
and academic_period ='200940'
GROUP BY  stu_population, 
          academic_period,
        CASE
            WHEN ((MBA_GPA)) <= 0.6 then '0.1 - 0.6 (F )'
            WHEN ((MBA_GPA)) <= 0.9 then '0.7 - 0.9 (D-)'
            WHEN ((MBA_GPA)) <= 1.2 then '1.0 - 1.2 (D )'
            WHEN ((MBA_GPA)) <= 1.6 then '1.3 - 1.6 (D+)'
            WHEN ((MBA_GPA)) <= 1.9 then '1.7 - 1.9 (C-)'
            WHEN ((MBA_GPA)) <= 2.2 then '2.0 - 2.2 (C )'
            WHEN ((MBA_GPA)) <= 2.6 then '2.3 - 2.6 (C+)'
            WHEN ((MBA_GPA)) <= 2.9 then '2.7 - 2.9 (B-)'
            WHEN ((MBA_GPA)) <= 3.2 then '3.0 - 3.2 (B )'
            WHEN ((MBA_GPA)) <= 3.6 then '3.3 - 3.6 (B+)'
            WHEN ((MBA_GPA)) >  3.6 then '3.7 - 4.9 (A-)'
            WHEN ((MBA_GPA)) is null then 'No GPA'
        ELSE
            'What Happened'
       END,
          rollup(college)
order by 1, 2, 3, 4;

Answer 1

Hi,
user1069723 wrote:
This query produces the results except that I would like to have a final total and overal average (excluding the "No GPA"). Shown as XXX in the output sample. Can this be done in one query or do I need to union for the summary?Yes, you can do that in one query, using ROLLUP. You don't need a UNION.
Additionally, if possible, how can I write some statements so that rows that may not exist can be displayed on the output? Example: there are no GPAs in the range of 0.1 thru 0.9. My query results do not show the first two rows, however, the users would like to see these rows displayed regardless of the existence or not.That's exactly what outer joins do: they display data from one table with matching data from a second table, if any matching data is found. The data from the first table is displayed whether or not there is a match. If there is no match, all columns from the second table are NULL. If you want to treat those NULLs as 0's, use NVL, as I did when displayijng the average, below.
In this problem, what is the first table, the table that has one row per grade range, and will be displayed even if no row in the result set falls into that grade range?
You ought to have such a table. The information about where the ranges begin and end, what letter grades are associated with each, and so on, is data. Data belongs in tables, not hard-coded into queries. If, next year, they decide to change where the boundary between D+ and C- falls, or to add an A group, an authorized user can make the change to one table. No matter how many queries use grade ranges, no programmer will have to change any query.
Here's how you might create such a table:
CREATE TABLE     GPA_Range
AS
SELECT  0.0 AS low_gpa, 0.7 AS high_gpa, '0.1 - 0.6 (F)'  AS display_txt, 'F'  AS letter_txt     FROM dual     UNION ALL
SELECT  0.7 AS low_gpa, 1.0 AS high_gpa, '0.7 - 0.9 (D-)' AS display_txt, 'D-' AS letter_txt     FROM dual     UNION ALL
SELECT  1.0 AS low_gpa, 1.3 AS high_gpa, '1.0 - 1.2 (D)'  AS display_txt, 'D'  AS letter_txt     FROM dual     UNION ALL
SELECT  1.3 AS low_gpa, 1.7 AS high_gpa, '1.3 - 1.6 (D+)' AS display_txt, 'D+' AS letter_txt     FROM dual     UNION ALL
SELECT  1.7 AS low_gpa, 2.0 AS high_gpa, '1.7 - 1.9 (C-)' AS display_txt, 'C-' AS letter_txt     FROM dual     UNION ALL
SELECT  2.0 AS low_gpa, 2.3 AS high_gpa, '2.0 - 2.2 (C)'  AS display_txt, 'C'  AS letter_txt     FROM dual     UNION ALL
SELECT  2.3 AS low_gpa, 2.7 AS high_gpa, '2.3 - 2.6 (C+)' AS display_txt, 'C+' AS letter_txt     FROM dual     UNION ALL
SELECT  2.7 AS low_gpa, 3.0 AS high_gpa, '2.7 - 2.9 (B-)' AS display_txt, 'B-' AS letter_txt     FROM dual     UNION ALL
SELECT  3.0 AS low_gpa, 3.3 AS high_gpa, '3.0 - 3.2 (B)'  AS display_txt, 'B'  AS letter_txt     FROM dual     UNION ALL
SELECT  3.3 AS low_gpa, 3.7 AS high_gpa, '3.3 - 3.6 (B+)' AS display_txt, 'B+' AS letter_txt     FROM dual     UNION ALL
SELECT  3.7 AS low_gpa, 999 AS high_gpa, '3.7 - 4.9 (A-)' AS display_txt, 'A-' AS letter_txt     FROM dual     UNION ALL
SELECT  NULL,            NULL,             'No GPA',                         NULL             FROM dual;As always, document everything, especially non-intuitive or potentially misleading things:
COMMENT ON COLUMN gpa_range.low_gpa IS 'Lowest numerical grade inside this range.';
COMMENT ON COLUMN gpa_range.low_gpa IS 'Lowest numerical grade in THE NEXT range.  This value is OUTSIDE of the range';You might prefer to assign impossible low_gpa and high_gpa values (like -2 and -1) to the 'No GPA' row. It's completely arbitrary, non-intutive and potentially confusing, but it would make joining easier.
If you can't create such a table, you can use a sub-query almost identical to the CREATE TABLE statement above in every query that needs it.
Here's the main query to do what you requested:
WITH      interesting_gpa_stat     AS
     SELECT     college
     ,     mba_gpa
     FROM     gpa_stat
     WHERE     stu_population     = 'F'
     AND     academic_period     = '200940'
,     all_colleges     AS
     SELECT DISTINCT     college
     FROM             interesting_gpa_stat
SELECT        c.college
,        r.display_txt           AS gparange
,        COUNT (s.college)             AS count
,        NVL ( AVG (s.mba_gpa)
            , 0
            )               AS avg
FROM                all_colleges          c
CROSS JOIN        gpa_range          r
LEFT OUTER JOIN        interesting_gpa_stat     s     ON (     c.college     =  s.college
                                            AND     r.low_gpa     <= s.mba_gpa
                                          AND     r.high_gpa     >  s.mba_gpa
                              OR (     c.college     =  s.college
                                 AND     r.low_gpa     IS NULL
                                       AND     s.mba_gpa     IS NULL
GROUP BY  ROLLUP ( c.college
                  , r.display_txt
ORDER BY  c.college
,            r.display_txtWithout any sample data and the results you want from that data, I had to make several assumptions.
For example, I assumed you wanted to have several colleges in the same result set. For each college, there will be one row of output for every grade range in the gpa_range table, regardless of whether any matches are found in the gpa_stat table. It's okay if there happens to be only one college if the result set.
The query includes a total for each college, and a grand total for the entire result set.
Any filtering (such as stu_population='F') should be done in the interesting_gpa_stat sub-query. You could get by without this sub-query, but any filtering conditions that would be in the WHERE clause would have to be added to outer-join conditions.
NULL values are ignored in all aggregate funcitons, so nothing special has to be done to keep rows in the 'No GPA' group from distorting the averages. The NVL function that presents 0 as the average when no grades were found in the other ranges always causes this average to apperar as 0 in the 'No GPA' rows, too. If you would prefer NULL in that column, then put the NVL inside a CASE expression.
Edited by: Frank Kulash on Jun 28, 2009 5:03 AM