Question

我如何获得每个月NSR和DSR的差异，然后获得％的增加/减少百分比以下sql结果：

type        period      typecount   typetotal
----------------------------------------------
DSR         2014-10     88          117
NSR         2014-10     29          117
DSR         2014-09     139         363
NSR         2014-09     224         363
DSR         2014-08     226         439
NSR         2014-08     213         439
DSR         2014-07     181         409
NSR         2014-07     228         409
DSR         2014-06     103         321
NSR         2014-06     218         321
DSR         2014-05     334         552
NSR         2014-05     218         552
DSR         2014-04     188         398
NSR         2014-04     210         398
DSR         2014-03     199         447
NSR         2014-03     248         447
DSR         2014-02     166         505
NSR         2014-02     339         505
DSR         2014-01     294         559
NSR         2014-01     265         559

Answer 1

我建议使用条件聚合来获取一行中的所有内容，然后在该结构中完成工作：

select period,
       max(case when type = 'DSR' then typecount end) as typecount_dsr,
       max(case when type = 'NSR' then typecount end) as typecount_nsr,
       max(typetotal) as typetotal
from table t
group by period;

然后你可以很容易地进行计算，例如：

with p as (
      select period,
             max(case when type = 'DSR' then typecount end) as typecount_dsr,
             max(case when type = 'NSR' then typecount end) as typecount_nsr,
             max(typetotal) as typetotal
      from table t
      group by period
     )
select p.*, (typecount_dsr - typecount_nsr) as diff,
       (typecount_dsr - typecount_nsr) / lag(typecount_dsr - typecount_nsr) over (order by period) as diffinc
from p;

Answer 2

感谢@ Gordon，@ Mihai这是我尝试过的，现在正在努力。

SELECT DISTINCT SUBSTR(cmse_application_number,1,3) App_Type,
  TO_CHAR(cmse_created_date,'YYYY-MM') Period,
  COUNT(SUBSTR(cmse_application_number,1,3)) APP_TYPE_COUNT,
  SUM(COUNT(SUBSTR(cmse_application_number,1,3))) OVER(PARTITION BY TO_CHAR(cmse_created_date,'YYYY-MM') ORDER BY TO_CHAR(cmse_created_date,'YYYY-MM')) App_MONTH_TOTAL,
  COUNT(SUBSTR(cmse_application_number,1,3))-lag(COUNT(SUBSTR(cmse_application_number,1,3))) over(PARTITION BY TO_CHAR(cmse_created_date,'YYYY-MM') ORDER BY TO_CHAR(cmse_created_date,'YYYY-MM'))  AS NSRDiff ,
  COUNT(SUBSTR(cmse_application_number,1,3))-LEAD(COUNT(SUBSTR(cmse_application_number,1,3))) over(PARTITION BY TO_CHAR(cmse_created_date,'YYYY-MM') ORDER BY TO_CHAR(cmse_created_date,'YYYY-MM')) AS DSRDiff
FROM
  ( SELECT DISTINCT cmse_service_number,
    CMSE_APPLICATION_NUMBER,
    CMSE_CREATED_DATE,
    cmse_ss_uid
  FROM cmse_service_enquiry a ,
    cmlt_line_type l,
    subd_detail s,
    bst_type t -- links to subd to eliminate failed applications
  WHERE A.CMSE_CMLT_UID                    =L.CMLT_UID
  AND A.CMSE_BST_UID                       =T.BST_UID
  AND A.CMSE_SERVICE_NUMBER                = S.SUBD_SERVICE_NAME
  AND SUBSTR(cmse_application_number,1,3) IN ('NSR','DSR')
    -- AND TO_CHAR(cmse_created_date,'YYYY-MM')='2014-10'  -- filter to check specific period
  AND cmse_ss_uid=58 -- makes sure workflow completes
  )
GROUP BY SUBSTR(cmse_application_number,1,3) ,
  TO_CHAR(cmse_created_date,'YYYY-MM')
ORDER BY 2 DESC ;

聚合和分析功能

2 个答案: