SQL Query to Collapse Duplicate Values By Date Range

Question

I have a table with the following structure: ID, Month, Year, Value with values for one entry per id per month, most months have the same value.

I would like to crea

Answer 1

I couldn't get the response from ngz to work when the input table contains multiple ids and date ranges that span years. I have a solution that does work, but with qualifications. It will only give you the correct answers if you know that you have a row for every month/year/id combination within the range. If there are "holes" it won't work. If you have holes, I know of know good way to do it other than writing some PL/SQL and using a cursor loop to create a new table in the format you want.

By the way, this is why data modeled this way is an abomination. You should always store stuff as start/from range records, not as discrete time period records. It's trivial to transform the former into the latter with a "multiplier" table, but it's almost impossible (as you've seen) to go the other direction.

SELECT ID
     , VALUE
     , start_date
     , end_date
  FROM (SELECT ID
             , VALUE
             , start_date
             , CASE
                  WHEN is_last = 0
                     THEN LEAD(end_date) OVER(PARTITION BY ID ORDER BY start_date)
                  ELSE end_date
               END end_date
             , is_first
          FROM (SELECT ID
                     , VALUE
                     , TO_CHAR(the_date, 'YYYY.MM') start_date
                     , TO_CHAR(NVL(LEAD(the_date - 31) OVER(PARTITION BY ID ORDER BY YEAR
                                  , MONTH), the_date), 'YYYY.MM') end_date
                     , is_first
                     , is_last
                  FROM (SELECT ID
                             , YEAR
                             , MONTH
                             , TO_DATE(TO_CHAR(YEAR) || '.' || TO_CHAR(MONTH) || '.' || '15', 'YYYY.MM.DD') the_date
                             , VALUE
                             , ABS(SIGN(VALUE -(NVL(LAG(VALUE) OVER(PARTITION BY ID ORDER BY YEAR
                                                   , MONTH), VALUE - 1)))) is_first
                             , ABS(SIGN(VALUE -(NVL(LEAD(VALUE) OVER(PARTITION BY ID ORDER BY YEAR
                                                   , MONTH), VALUE - 1)))) is_last
                          FROM test_table)
                 WHERE is_first = 1
                    OR is_last = 1))
 WHERE is_first = 1