Question

I am trying to produce a report which identifies client cases which were open during each week of a year. Currently I have the following SQL which returns all clients with an indicator on whether their case was open during week 1 of our calendar. A client has two aspects which identifies if their case is open – their MOV_START_DATE and their ESU_START DATE should be greater than end date of the period, and their MOV_END_DATE/ESU_START DATE should be either null or greater than the start date of the period.

The below code works, but I thought I could just copy the left join WK1 and rename it WK2 to return information for week 2 but I’m getting an error relating to ambiguously named columns. Additionally, I’m guessing that having 52 (one for each week) left joins on a report isn’t particularly advisable, so again I’m wondering if there is a better way of achieving this?

    SELECT
A.ESU_PER_GRO_ID,
A.ESU_ID,
A.STATUS,
B.MOV_ID,
B.MOV_START_DATE,
B.MOV_END_DATE,
A.ESU_START_DATE,
A.ESU_END_DATE,
LS.CLS_DESC,
nvl2(wk1.PRD_PERIOD_NUM,'Y','N') as "Week1"

FROM
A

LEFT JOIN B ON B.MOV_PER_GRO_ID = A.ESU_PER_GRO_ID

LEFT JOIN LS ON LS.CLS_CODE = A.STATUS

LEFT JOIN O_PERIODS WK1 ON B.MOV_START_DATE < WK1.PRD_END_DATE
AND (B.MOV_END_DATE IS NULL OR B.MOV_END_DATE > WK1.PRD_START_DATE)
AND A.ESU_START_DATE  < WK1.PRD_END_DATE
AND (A.ESU_END_DATE IS NULL OR A.ESU_END_DATE > WK1.PRD_START_DATE)
AND PRD_CAL_ID = 'E1190' AND WK1.PRD_PERIOD_NUM = 1 AND WK1.PRD_YEAR = 2012

WHERE
B.MOV_START_DATE  Is Not Null  
AND A.STATUS <> ('X') 

Hopefully I have provided enough information, but if not, I am happy to answer questions. Thanks!

Sample Data (Produced by above query)

 P ID    ESU_ID STATUS  MOV_ID  M_START     M_END   DESC    Week1
 1      ESU1       New      1M  01/01/2012           Boo    Y
 2      ESU2       New     2M   01/03/2012           Boo    N

Desired output (Week1 – Week 52)

 P ID    ESU_ID STATUS  MOV_ID  M_START     M_END   DESC    Week1 Week2
 1      ESU1       New     1M   01/01/2012           Boo    Y     Y
 2      ESU2       New     2M   01/03/2012           Boo    N     N

Answer 1

I suspect that the reason creating a WK2 join like WK1 didn’t work was that the column PRD_CAL_ID didn’t have a table alias on it. However, as you guessed, 52 joins is probably not going to perform very well. Try the following:

SELECT A.ESU_PER_GRO_ID,    
       A.ESU_ID,    
       A.STATUS,    
       B.MOV_ID,    
       B.MOV_START_DATE,    
       B.MOV_END_DATE,    
       A.ESU_START_DATE,    
       A.ESU_END_DATE,    
       LS.CLS_DESC,    
       'Week' || TRIM(TO_CHAR(pd.PRD_PERIOD_NUM)) WEEK_DESC
  FROM A
  LEFT JOIN B
    ON B.MOV_PER_GRO_ID = A.ESU_PER_GRO_ID    
  LEFT JOIN LS
    ON LS.CLS_CODE = A.STATUS    
  LEFT JOIN O_PERIODS pd
    ON B.MOV_START_DATE < pd.PRD_END_DATE AND
       (B.MOV_END_DATE IS NULL OR
        B.MOV_END_DATE > pd.PRD_START_DATE) AND
       A.ESU_START_DATE  < pd.PRD_END_DATE AND
       (A.ESU_END_DATE IS NULL OR
        A.ESU_END_DATE > pd.PRD_START_DATE)
WHERE B.MOV_START_DATE Is Not Null AND
      A.STATUS <> ('X') AND
      pd.PRD_CAL_ID = 'E1190' AND
      pd.PRD_YEAR = 2012
ORDER BY WEEK_DESC

This produces slightly different results than your original query, having a WEEK_DESC instead of trying to create 52 different columns, one for each week, but I think it will perform better.

Share and enjoy.