我正在寻找有关我一直在努力的Oracle SQL查询的一些指导。
我有一个包含时间戳和帮助台票号的数据表...
每次技术人员更改故障单或记录备注时,相应的时间戳将与故障单号一起登录数据库。我想要完成的是计算每个系列中第一次出现的票号,我不能使用distinct,因为如果技术人员在当天晚些时候再次使用票证,它将不再计算票证。请参考附图。
http://s13.postimg.org/i5gx4l4nb/sample.jpg
使用滞后函数我可以在行级别标记票证,但我不能将其添加到SUM子句中以实现我的最终目标。
CASE WHEN LAG(IPCS_AUDIT_HISTORY.ENTITYID,1) over (ORDER BY IPCS_AUDIT_HISTORY.HISTORYDATE)=IPCS_AUDIT_HISTORY.ENTITYID THEN 0 ELSE 1 END
关于采取何种方向的任何想法或想法?
SELECT
UPPER(IPCS_PLANNER.PLANNERSURNAME || ', ' || IPCS_PLANNER.PLANNERFIRSTNAME) AS FULL_NAME,
COUNT(DISTINCT IPCS_AUDIT_HISTORY.ENTITYID) AS TICKETS_COUNT,
SUM(CASE WHEN LAG(IPCS_AUDIT_HISTORY.ENTITYID,1) over (ORDER BY IPCS_AUDIT_HISTORY.HISTORYDATE)=IPCS_AUDIT_HISTORY.ENTITYID THEN 0 ELSE 1 END) AS TOUCH_COUNT
FROM
IPCS_AUDIT_HISTORY,
IPCS_PLANNER
WHERE
IPCS_AUDIT_HISTORY.PLANNERID=IPCS_PLANNER.PLANNERID and
TRUNC(IPCS_AUDIT_HISTORY.HISTORYDATE, 'IW')='30-DEC-13'
GROUP BY
UPPER(IPCS_PLANNER.PLANNERSURNAME || ', ' || IPCS_PLANNER.PLANNERFIRSTNAME)
答案 0 :(得分:1)
我可能会误解您的问题,但我相信您只需要将ROWS BETWEEN
添加到SUM()
:
SUM(ColName) OVER(ORDER BY Col2 ROWS BETWEEN UNBOUNDED PRECEDING AND 1 PRECEDING)
答案 1 :(得分:0)
更新:以这种方式尝试
SELECT full_name,
COUNT(DISTINCT entityid) ticket_count,
COUNT(DISTINCT series) - COUNT(DISTINCT entityid) touch_count
FROM
(
SELECT plannerid, full_name, entityid,
rnum - ROW_NUMBER() OVER (PARTITION BY plannerid, entityid ORDER BY rnum) series
FROM
(
SELECT p.plannerid,
UPPER(p.plannersurname || ', ' || p.plannerfirstname) full_name,
h.historydate,
h.entityid,
ROW_NUMBER() OVER (PARTITION BY p.plannerid ORDER BY h.historydate) rnum
FROM ipcs_audit_history h JOIN ipcs_planner p
ON h.plannerid = p.plannerid
WHERE h.historydate >= TO_DATE('30-DEC-13')
AND h.historydate < TO_DATE('30-DEC-13') + 1
) q
) p
GROUP BY plannerid, full_name
示例输出:
| FULL_NAME | TICKET_COUNT | TOUCH_COUNT | |-----------|--------------|-------------| | DOE, JOHN | 4 | 1 |
这是 SQLFiddle 演示