Need help about roll ups for average that contains logic

Question

--Hi everyone,
--I have posted this one on the forum before.
(how to do outer join select query for an APEX report )--I thought it works fine, but then yesterday, the tester check and said that the Average total is not correct
--the query is as below
DEFINE startmonth = "Aug 2012";
DEFINE endmonth   = "Oct 2012";
WITH  all_months  AS
   SELECT ADD_MONTHS(to_date('&startmonth','MON YYYY'), ROWNUM-1) AS which_month
   ,      ADD_MONTHS(to_date('&startmonth','MON YYYY'), ROWNUM  ) AS next_month
   from all_objects
   where
   rownum <= months_between(to_date('&endmonth','MON YYYY'), add_months(to_date('&startmonth','MON YYYY'), -1))
SELECT     ph.phase_number 
      ,TO_CHAR (am.which_month, 'Mon YYYY')  AS month
     , NVL(sum ( (CASE
                 WHEN ph.date_finished IS NULL OR ph.date_finished > last_day(am.which_month)
                 THEN last_day(am.which_month)
                 ELSE ph.date_finished
              END
             ) - ph.date_started + 1
           ) / count(def.def_id), 0) as avg_days
  FROM all_months am
       LEFT OUTER JOIN  a_phase_membership  ph  PARTITION BY (ph.phase_number)
          ON  am.which_month <= ph.date_started
          AND am.next_month  >  ph.date_started
          AND ph.date_started <= last_day(am.which_month)  -- May not be needed
          AND ph.active = 1
       LEFT OUTER JOIN  a_engagement  ce 
          ON  ph.mpm_eng_id = ce.engagement_id
          AND ce.court_name IS NOT NULL  -- or something involving LIKE
       LEFT OUTER join  a_defendant     def
          ON  ce.defendant_id = def.def_id
          AND def.active = 1
GROUP BY ROLLUP(phase_number,  am.which_month)
ORDER BY  ph.phase_number, am.which_month
--result is
PHASE_NUMBER                                       MONTH             AVG_DAYS              
PHASE ONE                                          Aug 2012          0                     
PHASE ONE                                          Sep 2012          12                    
PHASE ONE                                          Oct 2012          8                     
PHASE ONE                                                            11                    
PHASE THREE                                        Aug 2012          0                     
PHASE THREE                                        Sep 2012          1                     
PHASE THREE                                        Oct 2012          0                     
PHASE THREE                                                          1                     
PHASE TWO                                          Aug 2012          0                     
PHASE TWO                                          Sep 2012          9                     
PHASE TWO                                          Oct 2012          14                    
PHASE TWO                                                            11.5                  
                                                                     9.71428571428571428571428571428571428571
--And the tester is right, AVG days for phase one should be (12+8) /3 = 6.67 not 11, the same with phase two and three.
-- I tried to put a union select below the above which wrap it and do the average, but found that this select is too long and cumbersome.
--I have to ask for the Forum expert again, is there any way to make it look like
PHASE_NUMBER                                       MONTH             AVG_DAYS              
PHASE ONE                                          Aug 2012          0                     
PHASE ONE                                          Sep 2012          12                    
PHASE ONE                                          Oct 2012          8                     
PHASE ONE                                                            6.67                  
PHASE THREE                                        Aug 2012          0                     
PHASE THREE                                        Sep 2012          1                     
PHASE THREE                                        Oct 2012          0                     
PHASE THREE                                                          0.33                     
PHASE TWO                                          Aug 2012          0                     
PHASE TWO                                          Sep 2012          9                     
PHASE TWO                                          Oct 2012          14                    
PHASE TWO                                                            7.66                  
                                                                     4.896... (average of Phase one, phase two, and phase three)                                                                      --Here is the sample data structure (Database version is 11g)
CREATE TABLE "A_DEFENDANT"
    "DEF_ID"     NUMBER NOT NULL ENABLE,
    "FIRST_NAME" VARCHAR2(50 BYTE),
    "SURNAME"    VARCHAR2(20 BYTE) NOT NULL ENABLE,
    "DOB" DATE NOT NULL ENABLE,
    "ACTIVE" NUMBER(2,0) DEFAULT 1 NOT NULL ENABLE,
    CONSTRAINT "A_DEFENDANT_PK" PRIMARY KEY ("DEF_ID"));
Sample Data
Insert into A_DEFENDANT (DEF_ID,FIRST_NAME,SURNAME,DOB,ACTIVE) values (101,'Joe','Bloggs',to_date('12/12/99','DD/MM/RR'),1);
Insert into A_DEFENDANT (DEF_ID,FIRST_NAME,SURNAME,DOB,ACTIVE) values (102,'John','Smith',to_date('20/05/00','DD/MM/RR'),1);
Insert into A_DEFENDANT (DEF_ID,FIRST_NAME,SURNAME,DOB,ACTIVE) values (103,'Jane','Black',to_date('15/02/98','DD/MM/RR'),1);
Insert into A_DEFENDANT (DEF_ID,FIRST_NAME,SURNAME,DOB,ACTIVE) values (104,'Minnie','Mouse',to_date('13/12/88','DD/MM/RR'),0);
Insert into A_DEFENDANT (DEF_ID,FIRST_NAME,SURNAME,DOB,ACTIVE) values (105,'Daisy','Duck',to_date('05/08/00','DD/MM/RR'),1);
CREATE TABLE "A_ENGAGEMENT"
    "ENGAGEMENT_ID" NUMBER NOT NULL ENABLE,
    "COURT_NAME"    VARCHAR2(50 BYTE) NOT NULL ENABLE,
    "DATE_REFERRED" DATE,
    "DETERMINATION_HEARING_DATE" DATE,
    "DATE_JOINED_COURT" DATE,
    "DATE_TREATMENT_STARTED" DATE,
    "DATE_TERMINATED" DATE,
    "TERMINATION_TYPE" VARCHAR2(50 BYTE),
    "ACTIVE"           NUMBER(2,0) DEFAULT 1 NOT NULL ENABLE,
    "DEFENDANT_ID"     NUMBER,
    CONSTRAINT "A_ENGAGEMENT_PK" PRIMARY KEY ("ENGAGEMENT_ID"));
Insert into A_ENGAGEMENT (ENGAGEMENT_ID,COURT_NAME,DATE_REFERRED,DETERMINATION_HEARING_DATE,DATE_JOINED_COURT,DATE_TREATMENT_STARTED,DATE_TERMINATED,TERMINATION_TYPE,ACTIVE,DEFENDANT_ID) values (1,'AA',to_date('12/08/12','DD/MM/RR'),null,to_date('12/08/12','DD/MM/RR'),null,null,null,1,101);
Insert into A_ENGAGEMENT (ENGAGEMENT_ID,COURT_NAME,DATE_REFERRED,DETERMINATION_HEARING_DATE,DATE_JOINED_COURT,DATE_TREATMENT_STARTED,DATE_TERMINATED,TERMINATION_TYPE,ACTIVE,DEFENDANT_ID) values (2,'BB',to_date('01/09/12','DD/MM/RR'),null,to_date('02/09/12','DD/MM/RR'),null,null,null,1,102);
Insert into A_ENGAGEMENT (ENGAGEMENT_ID,COURT_NAME,DATE_REFERRED,DETERMINATION_HEARING_DATE,DATE_JOINED_COURT,DATE_TREATMENT_STARTED,DATE_TERMINATED,TERMINATION_TYPE,ACTIVE,DEFENDANT_ID) values (3,'AA',to_date('02/09/12','DD/MM/RR'),null,to_date('15/09/12','DD/MM/RR'),null,null,null,1,103);
Insert into A_ENGAGEMENT (ENGAGEMENT_ID,COURT_NAME,DATE_REFERRED,DETERMINATION_HEARING_DATE,DATE_JOINED_COURT,DATE_TREATMENT_STARTED,DATE_TERMINATED,TERMINATION_TYPE,ACTIVE,DEFENDANT_ID) values (4,'BB',to_date('01/10/12','DD/MM/RR'),null,to_date('02/10/12','DD/MM/RR'),null,null,null,1,105);
CREATE TABLE "A_PHASE_MEMBERSHIP"
    "MPM_ID"       NUMBER NOT NULL ENABLE,
    "MPM_ENG_ID"   NUMBER NOT NULL ENABLE,
    "PHASE_NUMBER" VARCHAR2(50 BYTE),
    "DATE_STARTED" DATE NOT NULL ENABLE,
    "DATE_FINISHED" DATE,
    "NOTES"  VARCHAR2(2000 BYTE),
    "ACTIVE" NUMBER(2,0) DEFAULT 1 NOT NULL ENABLE,
    CONSTRAINT "A_PHASE_MEMBERSHIP_PK" PRIMARY KEY ("MPM_ID"));
Insert into A_PHASE_MEMBERSHIP (MPM_ID,MPM_ENG_ID,PHASE_NUMBER,DATE_STARTED,DATE_FINISHED,NOTES,ACTIVE) values (1,1,'PHASE ONE',to_date('15/09/12','DD/MM/RR'),to_date('20/09/12','DD/MM/RR'),null,1);
Insert into A_PHASE_MEMBERSHIP (MPM_ID,MPM_ENG_ID,PHASE_NUMBER,DATE_STARTED,DATE_FINISHED,NOTES,ACTIVE) values (2,1,'PHASE TWO',to_date('21/09/12','DD/MM/RR'),to_date('29/09/12','DD/MM/RR'),null,1);
Insert into A_PHASE_MEMBERSHIP (MPM_ID,MPM_ENG_ID,PHASE_NUMBER,DATE_STARTED,DATE_FINISHED,NOTES,ACTIVE) values (3,2,'PHASE ONE',to_date('12/09/12','DD/MM/RR'),null,null,1);
Insert into A_PHASE_MEMBERSHIP (MPM_ID,MPM_ENG_ID,PHASE_NUMBER,DATE_STARTED,DATE_FINISHED,NOTES,ACTIVE) values (4,3,'PHASE ONE',to_date('20/09/12','DD/MM/RR'),to_date('01/10/12','DD/MM/RR'),null,1);
Insert into A_PHASE_MEMBERSHIP (MPM_ID,MPM_ENG_ID,PHASE_NUMBER,DATE_STARTED,DATE_FINISHED,NOTES,ACTIVE) values (5,3,'PHASE TWO',to_date('02/10/12','DD/MM/RR'),to_date('15/10/12','DD/MM/RR'),null,1);
Insert into A_PHASE_MEMBERSHIP (MPM_ID,MPM_ENG_ID,PHASE_NUMBER,DATE_STARTED,DATE_FINISHED,NOTES,ACTIVE) values (6,4,'PHASE ONE',to_date('03/10/12','DD/MM/RR'),to_date('10/10/12','DD/MM/RR'),null,1);
Insert into A_PHASE_MEMBERSHIP (MPM_ID,MPM_ENG_ID,PHASE_NUMBER,DATE_STARTED,DATE_FINISHED,NOTES,ACTIVE) values (7,3,'PHASE THREE',to_date('17/10/12','DD/MM/RR'),null,null,0);
Insert into A_PHASE_MEMBERSHIP (MPM_ID,MPM_ENG_ID,PHASE_NUMBER,DATE_STARTED,DATE_FINISHED,NOTES,ACTIVE) values (8,1,'PHASE THREE',to_date('30/09/12','DD/MM/RR'),to_date('16/10/12','DD/MM/RR'),null,1);--Probably after this crazy  project (reason: BA cannot make up her minds, and keep changing requirements about the reports, and then business want it done asap),
-- I need to buy a book and spend a lot of time to learn more about Oracle SQL Analytical function and Aggregation.
--Thanks a lot again, guys.
Ann

Answer 1

--Thanks  Jeneesh
--However, this time I found another bug relating filter the figures by court_name
DEFINE startmonth = "Aug 2012";
DEFINE endmonth   = "Oct 2012";
WITH  all_months_pre  AS
   SELECT ADD_MONTHS(to_date('&startmonth','MON YYYY'), ROWNUM-1) AS which_month
   ,      ADD_MONTHS(to_date('&startmonth','MON YYYY'), ROWNUM  ) AS next_month
   from all_objects
   where
   rownum <= months_between(to_date('&endmonth','MON YYYY'), add_months(to_date('&startmonth','MON YYYY'), -1))
all_months as
  select phase_number,which_month,next_month
  from all_months_pre,(select distinct phase_number from a_phase_membership)
total_data as
SELECT am.phase_number,
       TO_CHAR (am.which_month, 'Mon YYYY')  AS month,
       nvl((CASE
                 WHEN ph.date_finished IS NULL OR ph.date_finished > last_day(am.which_month)
                 THEN last_day(am.which_month)
                 ELSE ph.date_finished
              END
             ) - ph.date_started + 1,0) days,def_id
  FROM all_months am
       LEFT OUTER JOIN  a_phase_membership  ph  --PARTITION BY (ph.phase_number)
          ON  am.phase_number = ph.phase_number
            --I have found out that the Requirement don't need these, so I comment it out
          --and am.which_month <= ph.date_started
          --AND am.next_month  >  ph.date_started
          AND ph.date_started <= last_day(am.which_month)
          AND ph.active = 1
       LEFT JOIN  a_engagement  ce 
          ON  ph.mpm_eng_id = ce.engagement_id
          AND ce.court_name LIKE 'BB' 
       LEFT OUTER join  a_defendant     def
          ON  ce.defendant_id = def.def_id
          AND def.active = 1
avg_data as
    select phase_number,month,avg(days) avg_days
    from total_data
    group by phase_number,month
select phase_number,month,round(avg(avg_days),2) avg_days
from avg_data
group by rollup(phase_number,month)
order by 1
;--I run the  select for
AND court_name like 'BB'
--then run for
AND court_name LIKE 'AA'
or court_name LIKE '%'--They all return the same result which is
PHASE_NUMBER                                       MONTH             AVG_DAYS              
PHASE ONE                                          Aug 2012          0                     
PHASE ONE                                          Oct 2012          12.67                 
PHASE ONE                                          Sep 2012          7.2                   
PHASE ONE                                                            6.62                  
PHASE THREE                                        Aug 2012          0                     
PHASE THREE                                        Oct 2012          5.67                  
PHASE THREE                                        Sep 2012          0.33                  
PHASE THREE                                                          2                     
PHASE TWO                                          Aug 2012          0                     
PHASE TWO                                          Oct 2012          5.75                  
PHASE TWO                                          Sep 2012          3                     
PHASE TWO                                                            2.92                  
                                                                     3.85                  
13 rows selected --If I adjust the statement to
WITH  all_months_pre  AS
   SELECT ADD_MONTHS(to_date('&startmonth','MON YYYY'), ROWNUM-1) AS which_month
   ,      ADD_MONTHS(to_date('&startmonth','MON YYYY'), ROWNUM  ) AS next_month
   from all_objects
   where
   rownum <= months_between(to_date('&endmonth','MON YYYY'), add_months(to_date('&startmonth','MON YYYY'), -1))
all_months as
  select phase_number,which_month,next_month
  from all_months_pre,(select distinct phase_number from a_phase_membership)
total_data as
SELECT am.phase_number,
       TO_CHAR (am.which_month, 'Mon YYYY')  AS month,
       nvl((CASE
                 WHEN ph.date_finished IS NULL OR ph.date_finished > last_day(am.which_month)
                 THEN last_day(am.which_month)
                 ELSE ph.date_finished
              END
             ) - ph.date_started + 1,0) days,def_id
  FROM all_months am
       LEFT OUTER JOIN  a_phase_membership  ph  --PARTITION BY (ph.phase_number)
          ON  am.phase_number = ph.phase_number
          AND ph.date_started <= last_day(am.which_month)
          AND ph.active = 1
       INNER JOIN  a_engagement  ce 
          ON  ph.mpm_eng_id = ce.engagement_id
          AND ce.court_name LIKE 'BB' 
       LEFT OUTER join  a_defendant     def
          ON  ce.defendant_id = def.def_id
          AND def.active = 1
avg_data as
    select phase_number,month,avg(days) avg_days
    from total_data
    group by phase_number,month
select phase_number,month,round(avg(avg_days),2) avg_days
from avg_data
group by rollup(phase_number,month)
order by 1;--The result will be
PHASE_NUMBER                                       MONTH             AVG_DAYS              
PHASE ONE                                          Oct 2012          29                    
PHASE ONE                                          Sep 2012          19                    
PHASE ONE                                                            24                    
                                                                     24     --AND IF COURT_NAME = 'AA'
PHASE_NUMBER                                       MONTH             AVG_DAYS              
PHASE ONE                                          Oct 2012          9                     
PHASE ONE                                          Sep 2012          8.5                   
PHASE ONE                                                            8.75                  
PHASE THREE                                        Oct 2012          17                    
PHASE THREE                                        Sep 2012          1                     
PHASE THREE                                                          9                     
PHASE TWO                                          Oct 2012          11.5                  
PHASE TWO                                          Sep 2012          9                     
PHASE TWO                                                            10.25                 
                                                                     9.33   --It seems to me the figures now is correct. But then when there is no figures for the Phase, it will not display the phase.
-- is there any way to adjust the select statement so the result for ce.court_name LIKE 'BB' looks like
PHASE_NUMBER                                       MONTH             AVG_DAYS              
PHASE ONE                                          Oct 2012          29                    
PHASE ONE                                          Sep 2012          19                    
PHASE ONE                                                            24                                       
PHASE TWO                                          Oct 2012          0                  
PHASE TWO                                          Sep 2012          0                     
PHASE TWO                                                            0     
PHASE THREE                                        Oct 2012          0                   
PHASE THREE                                        Sep 2012          0                     
PHASE THREE                                                          0              
                                                                     8      The reason why I need it, because this is just one column that will be included in a union combined in the report that will looks like
PHASE_NUMBER                                       MONTH             SUM(AVG_IN_PHASE_ALL)  SUM(NUM_IN_PHASE_CURR)
Phase 1                                            Aug 2012          0                      0                     
                                                   Sep 2012          14.25                  0                     
                                                   Oct 2012          11.14                  1                     
                                                                     8.46                   1                     
Phase 2                                            Aug 2012          0                      0                     
                                                   Sep 2012          18                     0                     
                                                   Oct 2012          20.33                  2                     
                                                                     12.78                  2                     
Phase 3                                            Aug 2012          0                      0                     
                                                   Sep 2012          10                     0                     
                                                   Oct 2012          12                     0                     
                                                                     7.33                   0                     
                                                                     9.53                   3                     
13 rows selected And to put thing in big picture, this is the requirement from the Business Analyst
The requirements are:
The user must be able to request the extract for one or more calendar months, e.g.
May 2013
May 2013 – Sep 2013.
The file must contain a separate row for each calendar month in the requested range. Each row must contain the statistics computed for that calendar month.
The file must also include a row of totals.
The user must be able to request the extract for either AA or BB or Consolidated (both courts’ statistics accumulated).
Then the part that I am working on is
For each monitoring phase:
Phase name (e.g. “Phase One”)
Avg_time_in_phase_all_particip
for each phase name,
Add up days in each “phase name” Monitoring Phase, calculated as:
If Monitoring Phase.Date Finished is NULL or > month end date,
(*Month end date* Minus Monitoring Phase.Date Started Plus 1)
Otherwise (phase is complete)
(Monitoring Phase.Date Finished Minus Monitoring Phase.Date Started Plus 1.)
Divide by the numbers of all participants who have engaged in “phase name”.Thanks again for reading and spending time to help,
Ann
Edited by: Ann586341 on Nov 13, 2012 4:31 PM