SQL finding distinct pairs

Question

I'm trying to figure out how to write an SQL query to find the distinct pairs of student ids that have the same quarter and year and class id and that have been in more than one classid together.

So the database looks like this

Studentid courseid quarter       year 
11035   1020       Fall          2012    
11035   1092       Fall          2012    
75234   3201       Winter        2012    
8871    1092       Fall          2013    
39077   1092       Fall          2013    
57923   9219       Winter        2013    
60973   9219       Winter        2013     
19992   3201       Winter        2013    
60973   8772       Spring        2013    
90421   8772       Spring        2013    
90421   2987       Spring        2013    
60973   2987       Spring        2013

the result that I am trying to get is:

Studentid student id  year quarter course id

60973   90421   2013    Spring  8772    
60973   90421   2013    Spring  2987

so far I have

SELECT s.studentid, 
       st.studentid, 
       st.year, 
       s.quarter, 
       st.courseid
  FROM enrolled s 
       LEFT JOIN enrolled st 
                 ON st.COURSEID = s.COURSEID
 WHERE s.QUARTER = st.quarter
   AND s.YEAR = st.year 
   AND s.studentid <> st.studentid

but it is giving me every combination that can be created

Answer 1

Self join twice - once for the match, and again for to find a different match:

SELECT s.studentid, st.studentid, s.year, s.quarter, s.courseid
FROM enrolled s
JOIN enrolled st ON st.COURSEID = s.COURSEID
  AND st.quarter = s.quarter
  AND st.year = s.year
  AND s.studentid < st.studentid
JOIN enrolled e2 ON e2.COURSEID != s.COURSEID
  AND e2.quarter = s.quarter
  AND e2.year = s.year
  AND e2.studentid = st.studentid

Output:

| STUDENTID | OTHER_STUDENTID | YEAR | QUARTER | COURSEID |
|-----------|-----------------|------|---------|----------|
|     60973 |           90421 | 2013 |  Spring |     2987 |
|     60973 |           90421 | 2013 |  Spring |     8772 |

See SQLFiddle

Notes:

• You want an (inner) join, not a left join
• Use s.studentid < st.studentid to stop:
  • students joining to themselves, and
  • both sides of pairs from displaying
• Use a different course for the second join but the same other student id

Answer 2

SELECT a.*,b.courseid
FROM (SELECT   a.year
                     ,a.quarter
                     ,a.studentid AS Student1
                     ,b.studentid AS Student2
             FROM enrolled a
             JOIN enrolled b 
                 ON    a.COURSEID = b.COURSEID
                   AND a.QUARTER = b.quarter
                   AND a.YEAR = b.year 
                   AND a.courseid = b.courseid 
                   AND a.studentid < b.studentid
             GROUP BY a.year, 
                      a.quarter, 
                      a.studentid, 
                      b.studentid
             HAVING COUNT(*) > 1) a
JOIN enrolled b
  ON a.year = b.year
  AND a.quarter = b.quarter
  AND a.Student1 = b.studentid

SQL Fiddle

I added a row in the table: (11035, 2987, 'Spring', 2013) That row breaks some answers that work with your sample data. You need a HAVING clause to ensure the pair of students have more than 1 class together in a given quarter.

Answer 3

Using a CTE for the aggregate:

WITH pairs AS (
        SELECT distinct e1.Studentid AS student1
        , e2.Studentid AS student2
        FROM enrolled e1
        JOIN enrolled e2 ON e1.courseid = e2.courseid AND e1.quarter = e2.quarter AND e1.year = e2.year
        WHERE e1.Studentid < e2.Studentid
        GROUP BY e1.Studentid,e2.Studentid
        HAVING COUNT(*) > 1
        )
SELECT e1.Studentid AS student1
        , e2.Studentid AS student2
        , e1.courseid
        , e1.quarter
        , e1.year
FROM pairs p
JOIN enrolled e1 ON e1.Studentid = p.student1
JOIN enrolled e2 ON e2.Studentid = p.student2
WHERE e1.courseid = e2.courseid AND e1.quarter = e2.quarter AND e1.year = e2.year
        ;