SQL query to compare multiple columns in same table in oracle

Question

I have a requirement to find emplid having data difference in same table. Table consist of 50-60 columns.. I need to check if any column has change in data from previous row, emplidshould get pick up as well as if any new employee get add that also needs to pick up..

I have created a basic query and it is working but need some way to achieve same purpose as I do not want to write every column name.

My query:

select 
    emplid 
from 
    ps_custom_tbl t, ps_custom_tbl prev_t 
where 
    prev_t.emplid = t.emplid 
    and t.effdt = (select max effdt from ps_custom_tbl t2 
                   where t2.emplid = t.emplid) 
    and prev_t.effdt = (select max(effdt) from ps_custom_tbl prev_t2 
                        where emplid = prev_t.emplid and effdt < t.effdt)
    and (t.first_name prev_t.first_name Or t.last_name prev_t.last_name …. 50 columns);

Can you please suggest another way to achieve same thing?

Bad habits to kick : using old-style JOINs - that old-style comma-separated list of tables style was replaced with the proper ANSI JOIN syntax in the ANSI-92 SQL Standard (more than 25 years ago) and its use is discouraged — marc_s
– marc_s, Commented Oct 13, 2020 at 5:01
You'll have to write all column names. You can, however, get these from a query, e.g. select 'or (decode(t.' || column_name || ', prev_t.' || column_name || ', ''same'', ''different'') = ''different'')' from user_tab_cols where table_name = 'PS_CUSTOM_TBL'. This also means you can write a PL/SQL function building the query dynamically to retrieve the differences. — Thorsten Kettner
– Thorsten Kettner, Commented Oct 13, 2020 at 12:10

Monika Lewandowska · Accepted Answer · 2020-10-17 18:55:12Z

1

You can use MINUS.

if no_data then both are the same, if there are some records - mean that there is a difference between

create table emp as select * from hr.employees;

insert into emp select employee_id+1000, first_name, last_name, email, phone_number, hire_date, job_id, salary, commission_pct, manager_id, 
 decode(department_id ,30,70, department_id)
 from hr.employees;

select first_name, last_name, email, phone_number, hire_date, job_id, salary, commission_pct, manager_id, department_id  
from emp where employee_id <= 1000
minus
select first_name, last_name, email, phone_number, hire_date, job_id, salary, commission_pct, manager_id, department_id  
from emp where employee_id > 1000;

But you have to list all columns, because if you have eg different dates or ids - they will be compared too. But it's easier to list columns in SELECT clause then write for everyone WHERE condition. Maybe it will help.

-- or if different tables and want to compare all cols simply do

drop table emp;
create table emp as select * from hr.employees;
    
create table emp2  as 
select employee_id, first_name, last_name, email, phone_number, hire_date, job_id, salary, commission_pct, manager_id, 
 decode(department_id ,30,70, department_id) department_id
 from hr.employees;

select * from emp
minus 
select * from emp2;

---- ADD DATE CRITERIA

-- yes, you can add date criteria and using analytical functions check which -- is newer and which is

older and then compare one to another. like below:

drop table emp;
create table emp as select * from hr.employees;
        
insert into emp 
        select
           employee_id, 
           first_name, 
           last_name, 
           email, 
           phone_number, 
           hire_date+1, 
           job_id, 
           salary, 
           commission_pct, 
           manager_id, 
           decode(department_id ,30,70, department_id)
        from hr.employees;
        
     with data as   --- thanks to WITH you retrieve data only once
            (select employee_id, first_name, last_name, email, phone_number, 
                hire_date, 
                row_number() over(partition by employee_id order by hire_date desc) rn  -- distinguish newer and older record, 
               job_id, salary, commission_pct, manager_id, department_id  
        from emp)
        select employee_id, first_name, last_name, email, phone_number, department_id from data where rn  = 1
        MIUNUS--- find the differences
        select employee_id, first_name, last_name, email, phone_number, department_id from data where rn = 2;

edited Oct 17, 2020 at 18:55

answered Oct 13, 2020 at 11:23

Monika Lewandowska

4862 silver badges7 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

angel Over a year ago

Regarding minus operator, is it possible to apply date criteria.. Every run going to load employee data so I have this date field unquie but if for particular employee ID no rows added.. Or say 1sep employee a loaded 2 sep employee a not loaded but b and c loaded.. 3 sep again employee a loaded.. Will this be possible.. Like we used to do effdt criteria..

Monika Lewandowska Over a year ago

yes, you can add date criteria. check my answer, I edited it. You can use WITH to get data only once, you can use ANALYTICAL function to distinguish newer and older date, and at the and use MINUS to find differences

Connor McDonald · Accepted Answer · 2020-10-13 03:57:16Z

1

You will have to write all columns in some sense no matter what you do.

In terms of comparing current and previous, you might find this easier

select 
  col1,
  col2,
  ...
  lag(col1) over ( partition by empid order by effdt ) as prev_col1,
  lag(col2) over ( partition by empid order by effdt ) as prev_col2
  ...

and then you comparison will be along the lines of

select *
from (  <query above >
where
  decode(col1,prev_col1,0,1) = 1 or 
  decode(col2,prev_col2,0,1) = 1 or 
  ...

The use of DECODE in this way handles the issues of nulls.

answered Oct 13, 2020 at 3:57

Connor McDonald

11.8k1 gold badge14 silver badges21 bronze badges

1 Comment

angel Over a year ago

Thank you for the reply.. What I understand I need to write all 50 column names even if I use lag function.. Will except function work.. I just need to know if there is change in two rows..

eurokous · Accepted Answer · 2021-11-13 15:31:51Z

My requirement is to send out data to managers, they change any/all/none of the data in the columns, and send back to me. I then have to identify each column that has a difference from what I sent, and mark those columns as changed for a central office reviewer to visually scan and approve/deny the changes for integration back into the central data set.

This solution may not fit your needs of course, but a template structure is offered here that you can augment to meet your needs no matter the number of columns. In the case of your question, 50-60 columns will make this SQL query huge, but I've written heinously long queries in the past with great success. Add columns a few at a time rather than all wholesale according to this template and see if they work along the way.

You could easily write pl/sql to write this query for you for the tables in question. This would get very cumbersome if you had to compare columns from 3 or more tables or bi-directional changes. I only care about single direction changes. Did the person change my original row columns or not. If so, what columns did they change, and what was my before value and what is their after value, and show me nothing else please.

In other words, only show me rows with columns that have changes with their before values and nothing else.

create table thing1 (id number, firstname varchar2(10), lastname varchar2(10));
create table thing2 (id number, firstname varchar2(10), lastname varchar2(10));

insert into thing1 values (1,'Buddy', 'Slacker');
insert into thing2 values (1,'Buddy', 'Slacker');
insert into thing1 values (2,'Mary', 'Slacker');
insert into thing2 values (2,'Mary', 'Slacke');
insert into thing1 values (3,'Timmy', 'Slacker');
insert into thing2 values (3,'Timm', 'Slacker');
insert into thing1 values (4,'Missy', 'Slacker');
insert into thing2 values (4,'Missy', 'Slacker');
commit;

Un-comment commented select * queries one at a time after each data set to understand what is in each data set at each stage of the refinement process.

with rowdifferences as
(
select
id
,firstname
,lastname
from thing2
minus
select
id
,firstname
,lastname
from thing1
)
--select * from rowdifferences
,thing1matches as
(
select
t1.id
,t1.firstname
,t1.lastname
from thing1 t1
join rowdifferences rd on t1.id = rd.id
)
--select * from thing1matches
, col1differences as
(
select
id
,firstname
from rowdifferences
minus
select
id
,firstname
from thing1matches
)
--select * from col1differences
, col2differences as
(
select
id
,lastname
from rowdifferences
minus
select
id
,lastname
from thing1matches
)
--select * from col2differences
,truedifferences as
(
select
case when c1.id is not null then c1.id
when c2.id is not null then c2.id
end id
,c1.firstname
,c2.lastname
from col1differences c1
full join col2differences c2 on c1.id = c2.id
)
--select * from truedifferences
select
t1m.id
,case when td.firstname is not null then t1m.firstname end beforefirstname
,td.firstname afterfirstname
,case when td.lastname is not null then t1m.lastname end beforelastname
,td.lastname afterlastname
from thing1matches t1m
join truedifferences td on t1m.id = td.id
;

Collectives™ on Stack Overflow

SQL query to compare multiple columns in same table in oracle

3 Answers 3

2 Comments

1 Comment

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

2 Comments

1 Comment

Comments

Your Answer

Sign up or log in

Post as a guest

Related