如何根据某些条件合并记录

Question

我有下表，

id   id_concept  start_date  end_date
----------------------------------------------
100    282        20/06/2016  24/06/2016 
100    282        15/07/2016  18/07/2016
300    282        01/09/2016  02/09/2016

我需要合并具有相同 id、id_concept 的记录，并且一条记录的 END_DATE 和下一条记录的 START_DATE 之间的时间为 30 天或更短（

对于合并的记录，我需要将 start_date 作为记录的第一个 start_date，将 end_date 作为最后一条记录的 end_date

o/p 应该是这样的，

     id    id_concept start_date  end_date      count
    ---------------------------------------------------
    100    282        20/06/2016  18/07/2016      2
    300    282        01/09/2016  02/09/2016      1

Answer 1

SQL Fiddle

PostgreSQL 9.3 架构设置：

CREATE TABLE table_name ( id, id_concept, start_date, end_date ) AS
SELECT 100, 282, DATE '2016-06-20', DATE '2016-06-24' UNION ALL
SELECT 100, 282, DATE '2016-07-15', DATE '2016-07-18' UNION ALL
SELECT 300, 282, DATE '2016-09-01', DATE '2016-09-02';

查询 1：

SELECT id,
       id_concept,
       MIN( start_date ) AS start_date,
       MAX( end_date ) AS end_date,
       COUNT(*) AS "count"
FROM   (
  SELECT id,
         id_concept,
         start_date,
         end_date,
         SUM( diff ) OVER (
           PARTITION BY id, id_concept
           ORDER BY start_date, end_date
         ) AS grp
  FROM   (
    SELECT t.*,
           CASE
           WHEN LAG( end_date ) OVER (
                  PARTITION BY id, id_concept
                  ORDER BY start_date, end_date
                ) >= start_date - INTERVAL '30' DAY
           THEN 0
           ELSE 1
           END AS diff
    FROM   table_name t
  ) t
) t
GROUP BY id, id_concept, grp

Results：

|  id | id_concept |                  start_date |                    end_date | count |
|-----|------------|-----------------------------|-----------------------------|-------|
| 300 |        282 | September, 01 2016 00:00:00 | September, 02 2016 00:00:00 |     1 |
| 100 |        282 |      June, 20 2016 00:00:00 |      July, 18 2016 00:00:00 |     2 |