How to fetch records that have an alternate entry - sql

Question

I need some help to fetch records having alternate set of entries associated with Unique value(ex: user_id)
I want output to be only (1111,2222,3333)
Here is the scenario:
user_id 1111 attended .net course from 2005-01-01 to 2006-12-31
he later attended java from 2007-01-01 to 2009-12-31
he later came back to .net
so i want to retrieve these kind of user_id's
user_id 4444 should not be in the output, because there is no alternative courses.
UPDATE: 4444 started his Java course from 2007 to 2009 he again
attended Java from 2010 - 2012 Later he attended .net but never came
back to Java so he must be excluded from output
If Group by is used, it will consider records irrespective of alternate course name.
We can create a procedure to accomplish this by looping and comparing the alternate course name but i want to know if a query can do this?

Answer 1

You can use two INNER JOIN operations:
SELECT DISTINCT user_id
FROM mytable AS t1
INNER JOIN mytable AS t2
ON t1.user_id = t2.user_id AND t1.id < t2.id AND t1.course_name <> t2.course_name
INNER JOIN mytable AS t3
ON t2.user_id = t3.user_id AND t2.id < t3.id AND t1.course_name = t3.course_name
I assume that id is an auto-increment field that reflects the order the rows have been inserted in the DB. Otherwise, you should use a date field in its place.

Answer 2

Same as Girogos Betsos' answer, only with select distinct to prevent duplicates.
SELECT DISTINCT user_id
FROM mytable AS t1
INNER JOIN mytable AS t2
ON t1.user_id = t2.user_id AND t1.Start_Date < t2.Start_Date AND
t1.course_name <> t2.course_name
INNER JOIN mytable AS t3
ON t2.user_id = t3.user_id AND t2.Start_Date < t3.Start_Date AND
t1.course_name = t3.course_name
EDIT: Using Start_Date since the answer has been updated and IDs are not necessarily sequential.

Answer 3

This is a version utilizing Windowed Aggregate Fuctions instead of multiple self joins:
SELECT DISTINCT user_id
FROM
(
SELECT user_id
,course_name
,start_date
,RANK() -- number all courses
OVER (PARTITION BY user_id
ORDER BY start_date)
-
RANK() -- number each course
OVER (PARTITION BY user_id, course_name
ORDER BY start_date) AS x
FROM tab
) dt
GROUP BY user_id, course_name
HAVING MIN(x) <> MAX(x) -- same course but another inbetween
If a user has a course multiple times in a series that x will stay the same, if there was another course inbetween it will change:
java 1 - 1 = 0
java 2 - 2 = 0 <--- min
.net 3 - 1 = 2
java 4 - 3 = 1 <--- max
java 1 - 1 = 0
java 2 - 2 = 0
.net 3 - 1 = 2
.net 4 - 2 = 2

Answer 4

Using a single table scan and does not rely on GROUP BY:
WITH table_name ( user_id, start_date, end_date, course_name, id ) AS (
SELECT 1111, DATE '2005-01-01', DATE '2006-12-31', '.net', 1 FROM DUAL UNION ALL
SELECT 1111, DATE '2007-01-01', DATE '2009-12-31', 'java', 2 FROM DUAL UNION ALL
SELECT 1111, DATE '2010-01-01', DATE '2020-12-31', '.net', 3 FROM DUAL UNION ALL
SELECT 2222, DATE '2005-01-01', DATE '2006-12-31', 'java', 4 FROM DUAL UNION ALL
SELECT 2222, DATE '2007-01-01', DATE '2008-12-31', '.net', 5 FROM DUAL UNION ALL
SELECT 2222, DATE '2009-01-01', DATE '2012-12-31', '.net', 6 FROM DUAL UNION ALL
SELECT 2222, DATE '2013-01-01', DATE '2016-12-31', 'java', 7 FROM DUAL UNION ALL
SELECT 3333, DATE '2005-01-01', DATE '2007-12-31', 'java', 8 FROM DUAL UNION ALL
SELECT 3333, DATE '2007-01-01', DATE '2008-12-31', '.net', 9 FROM DUAL UNION ALL
SELECT 3333, DATE '2009-01-01', DATE '2013-12-31', 'java', 10 FROM DUAL UNION ALL
SELECT 3333, DATE '2014-01-01', DATE '2016-12-31', '.net', 11 FROM DUAL UNION ALL
SELECT 4444, DATE '2007-01-01', DATE '2009-12-31', 'java', 12 FROM DUAL UNION ALL
SELECT 4444, DATE '2010-01-01', DATE '2012-12-31', 'java', 13 FROM DUAL UNION ALL
SELECT 4444, DATE '2013-01-01', DATE '2015-12-31', '.net', 14 FROM DUAL UNION ALL
SELECT 4444, DATE '2016-01-01', DATE '2016-12-31', '.net', 15 FROM DUAL
)
SELECT DISTINCT user_id
FROM (
SELECT user_id,
LEAD( course_name )
OVER ( PARTITION BY user_id, course_name ORDER BY start_date )
AS next_same_course,
LEAD( course_name )
OVER ( PARTITION BY user_id ORDER BY start_date )
AS next_course
FROM table_name
)
WHERE next_same_course IS NOT NULL
AND next_course <> next_same_course;