Better approach for checking column values between two different rows

Question

My requirement is to find best Approach to find difference between column values of two different rows.Below I've mentioned two different
approaches I'm thinking of,but I'd like to know is there any other better approach.
version details
SQL> SELECT *
  2  FROM V$VERSION;
BANNER
Oracle Database 11g Enterprise Edition Release 11.1.0.7.0 - 64bit Production
PL/SQL Release 11.1.0.7.0 - Production
CORE    11.1.0.7.0      Production
TNS for Solaris: Version 11.1.0.7.0 - Production
NLSRTL Version 11.1.0.7.0 - ProductionTable creation script
CREATE TABLE R_DUMMY
   (CA_ID VARCHAR2(16) NOT NULL ENABLE,
     CA_VER_NUM NUMBER(4,2) NOT NULL ENABLE,
     BRWR_SHORT_NAME VARCHAR2(25 CHAR),
     sic_code     number,
     FAC_ID VARCHAR2(10) NOT NULL ENABLE
/ insert script
insert into r_dummy (CA_ID, CA_VER_NUM, BRWR_SHORT_NAME, sic_code, FAC_ID)
values ('CA2001/11/0002', 2.00, 'Nandu',1234, 'FA000008');
insert into r_dummy (CA_ID, CA_VER_NUM, BRWR_SHORT_NAME, sic_code, FAC_ID)
values ('CA2001/11/0002', 3.00, 'SHIJU',456, 'FA000008');Desired O/P :
ca_id               fac_id          column_name          previous name          after_modification
CA2001/11/0002          FA000008     BRWR_SHORT_NAME          Nandu               SHIJU
CA2001/11/0002          FA000008     sic_code          1234               456My approach
select      ca_id,fac_id,column_name,
     decode(column_name,'BRWR_SHORT_NAME',lg_brwr,lg_sic) previous_name ,
     decode(column_name,'BRWR_SHORT_NAME',ld_brwr,ld_sic) after_modification
from
     select
               case
                    when ld_brwr != lg_brwr then
                    'BRWR_SHORT_NAME'
                    when ld_brwr != lg_brwr then
                    'sic_code'
               end
          ) column_name,ca_id,fac_id,lg_brwr,ld_brwr,ld_sic,lg_sic
     from     (
          select ca_id,fac_id,lag_brwr,ld_brwr,ld_sic,lag_sic
          from
                    Select      lead(brwr_short_name,1) over(partition by ca_id,fac_id) ld_brwr,
                         lag(brwr_short_name,1) over(partition by ca_id,fac_id) lg_brwr,
                         lead(sic_code,1) over(partition by ca_id,fac_id) ld_sic,
                         lag(sic_code,1) over(partition by ca_id,fac_id) lg_sic,
                         ca_id,fac_id
                    from r_dummy
          where (ld_brwr != lg_brwr or ld_sic != lg_sic)
)2nd Approach :
=============
select      ca_id,fac_id,column_name,
     decode(column_name,'BRWR_SHORT_NAME',lg_brwr,lg_sic) previous_name ,
     decode(column_name,'BRWR_SHORT_NAME',ld_brwr,ld_sic) after_modification
from
     select
               case
                    when ld_brwr != lg_brwr then
                    'BRWR_SHORT_NAME'
                    when ld_brwr != lg_brwr then
                    'sic_code'
               end
          ) column_name,ca_id,fac_id,lg_brwr,ld_brwr,ld_sic,lg_sic
     from     (
          select ca_id,fac_id,brwr_short_name,sic_code
          from
                    Select      ca_id,fac_id,brwr_short_name lg_brwr,sic_code lg_sic
                    from     r_dummy
                    where     ca_ver_num = '2.00'
               )o,(
                    Select      ca_id,fac_id,brwr_short_name ld_brwr,sic_code ld_sic
                    from     r_dummy
                    where     ca_ver_num = '3.00'
                          )n
          where      0.ca_id = n.ca_id
               and 0.fac_id = n.fac_id
               and (ld_brwr != lg_brwr or ld_sic != lg_sic)
)Hi Experts,
     I've provided sample data where I'm checking for just two columns viz brwr_short_name ,sic_code,but in real time
I've to check for 8 more columns so please suggest me with a better approach.
I appreciate your precious suggestions.

Answer 1

Hi,
Thanks for posting the CREATE TABLE and INSERT statements; that really helps!
Here's one wa. Like your 2nd approach, this uses a self-join:
WITH     got_r_num     AS
     SELECT  ca_id
     ,     ROW_NUMBER () OVER ( PARTITION BY  ca_id
                               ,                    fac_id
                         ORDER BY        ca_ver_num
                             )    AS r_num
     ,     brwr_short_name
     ,     TO_CHAR (sic_code)     AS sic_code
     ,     fac_id
--     ,     ...     -- Other columns (using TO_CHAR if needed)
     FROM     r_dummy
,     unpivoted_data     AS
     SELECT     *
     FROM     got_r_num
     UNPIVOT     INCLUDE NULLS
          (    txt
          FOR  column_name IN ( brwr_short_name          AS 'BRWR_SHORT_NAME'
                        , sic_code               AS 'SIC_CODE'
--                        , ...     -- Other columns
SELECT       p.ca_id
,       p.fac_id
,       p.column_name
,       p.txt          AS previous_name
,       a.txt          AS after_modification
FROM       unpivoted_data   p
JOIN       unpivoted_data   a  ON  p.ca_id     = a.ca_id
                       AND p.fac_id     = a.fac_id
                     AND p.column_name     = a.column_name
                     AND p.r_num      = a.r_num - 1
                     AND p.txt || 'X' != a.txt || 'X'
ORDER BY  a.r_num
;To include other columns, add them in the 2 places where I put the comment "Other columns".
Ca_ver_num can have any values, not just 2.00 and 3.00.
This will show cases where a value in one of the columns changed to NULL, or where NULL changed to a value.
There ought to be a way to do this without a separate sub-query like got_r_num. According to the SQL Language manual, you can put expressions in the UPIVOT ... IN list, but when I tried
...     UNPIVOT     INCLUDE NULLS
          (    txt
          FOR  column_name IN ( brwr_short_name          AS 'BRWR_SHORT_NAME'
                             , TO_CHAR (sic_code)     AS 'SIC_CODE'
          )I got the error "ORA_00917: missing comma" right after TO_CHAR. Perhaps someone else can show how to eliminate one of the sub-queries.