Determine rows based on their 'grouping'

Question

Ok assume there is a table (TableA) in this format
col1     col2      col3               col4     col5     col6
R1     route1     route1Description1     AA     BB     CC     
R1     route1     route1Description1     AA     CC     
R1     route1     route1Description1     CC     BB     
R2     route2     route1Description2     GG     KK     LL     
R2     route2     route1Description2     GG     LL     
R2     route2     route1Description2     LL     KK
R3     route3     route1Description3     DD     EE     FF     
R3     route3     route1Description3     DD     FF     
R3     route3     route1Description3     MM     NN     ZZ
R4     route4     route4Description4     FF     SS     XX|PP|OO
R4     route4     route4Description4     FF     XX
R4     route4     route4Description4     XX     PP
R4     route4     route4Description4     PP     OO
R4     route4     route4Description4     OO     SS
R5     route5     route5Description5     RR     TT     UU|VV
     The data in the table was imported from a csv file and there is a relationship between the rows. Each combination of col1, col2 and col3 describes a full route of a journey. The row with an entry in col6 describes the full route and the other rows describes each leg in the route.
For example, for R1, the route is AA to BB via CC.
Another example for R4 the route is FF to SS via XX, PP, and OO.
What i would like to do is missing a route. For example the route for R3 is DD to EE via FF. There is an entry for DD to FF but is missing an entry for FF to EE.
The results should return the following rows which are incomplete
R3     route3     route1Description3     DD     EE     FF     
R3     route3     route1Description3     DD     FF     
R5     route5     route5Description5     RR     TT     UU|VVWhat is the best way to do this?
Here is what i have come up with but it doesnt quite returned the correct result.
select * from tableA a
Where not exists(
     select 1 from tableA b
     where instr(col6,col4,1)>0 and instr(col6,col1,1)>0)
     And a.col1=b.col1
     And a.col2=b.col2
     And a.col3=b.col3
     )Is there an easier way to achieve this?
Edited by: ziggy on Jul 9, 2010 1:39 PM

Answer 1

Hi,
Here is a solution that show the road which are incomplete:
SELECT col1,road,'Route not completed, expected end point:'||endpoint remark
  FROM (WITH r AS
             (SELECT col1, col4, col5, NVL2 (col6, NULL, col1 || col4) ca,
                     NVL2 (col6, col1 || col4, col1 || col5) cb, col6
                FROM routes)
        SELECT     LEVEL, r.*,
                      col1
                   || ':'
                   || SYS_CONNECT_BY_PATH (NVL2 (col6, col4, col5), '/') road,
                   CONNECT_BY_ISLEAF thenend,
                   nvl2(col6,col5,null) endpoint
              FROM r
        CONNECT BY PRIOR cb = ca
        START WITH col6 IS NOT NULL) p
WHERE thenend = 1 AND SUBSTR (road, -2) != col5
COL1,ROAD,REMARK
R3,R3:/MM,Route not completed, expected end point:NN
R5,R5:/RR,Route not completed, expected end point:TT
Routes completed:
SELECT col1,road,'Route  completed!' remark
  FROM (WITH r AS
             (SELECT col1, col4, col5, NVL2 (col6, NULL, col1 || col4) ca,
                     NVL2 (col6, col1 || col4, col1 || col5) cb, col6
                FROM routes)
        SELECT     LEVEL, r.*,
                      col1
                   || ':'
                   || SYS_CONNECT_BY_PATH (NVL2 (col6, col4, col5), '/') road,
                   CONNECT_BY_ISLEAF thenend,
                   nvl2(col6,col5,null) endpoint
              FROM r
        CONNECT BY PRIOR cb = ca
        START WITH col6 IS NOT NULL) p
WHERE thenend = 1 AND SUBSTR (road, -2) = col5
COL1,ROAD,REMARK
R1,R1:/AA/CC/BB,Route  completed!
R2,R2:/GG/LL/KK,Route  completed!
R3,R3:/DD/FF,Route  completed!
R4,R4:/FF/XX/PP/OO/SS,Route  completed!Edited by: user11268895 on Jul 9, 2010 11:44 PM