CODEWITHSUNDEEP
sqloracle-database
JenPartridge
JenPartridge

Reputation: 95

How can I find duplicate consecutive values in this table?

Say I have a table which I query like so:

select date, value from mytable order by date

and this gives me results:

date                  value
02/26/2009 14:03:39   1                
02/26/2009 14:10:52   2          (a)
02/26/2009 14:27:49   2          (b)
02/26/2009 14:34:33   3
02/26/2009 14:48:29   2          (c)
02/26/2009 14:55:17   3
02/26/2009 14:59:28   4

I'm interested in the rows of this result set where the value is the same as the one in the previous or next row, like row b which has value=2 the same as row (a). I don't care about rows like row (c) which has value=2 but does not come directly after a row with value=2. How can I query the table to give me all rows like (a) and (b) only? This is on Oracle, if it matters.

Upvotes: 8

Views: 5758

Answers (3)

Janek Bogucki
Janek Bogucki

Reputation: 5123

Use the lead and lag analytic functions.

create table t3 (d number, v number);
insert into t3(d, v) values(1, 1);
insert into t3(d, v) values(2, 2);
insert into t3(d, v) values(3, 2);
insert into t3(d, v) values(4, 3);
insert into t3(d, v) values(5, 2);
insert into t3(d, v) values(6, 3);
insert into t3(d, v) values(7, 4);

select d, v, case when v in (prev, next) then '*' end match, prev, next from (
  select
    d,
    v,
    lag(v, 1) over (order by d) prev,
    lead(v, 1) over (order by d) next
  from
    t3
)
order by
  d
;

Matching neighbours are marked with * in the match column,

alt text

Upvotes: 13

Janek Bogucki
Janek Bogucki

Reputation: 5123

This is a simplified version of @Bob Jarvis' answer, the main difference being the use of just one subquery instead of four,

with f as (select row_number() over(order by d) rn, d, v from t3)
select
  a.d, a.v,
  case when a.v in (prev.v, next.v) then '*' end match
from
  f a
    left join
  f prev
    on a.rn = prev.rn + 1
    left join
  f next
    on a.rn = next.rn - 1
order by a.d
;

Upvotes: 3

Bob Jarvis - Слава Україні
Bob Jarvis - Слава Україні

Reputation: 50067

As @Janek Bogucki has pointed out LEAD and LAG are probably the easiest way to accomplish this - but just for fun let's try to do it by using only basic join operations:

SELECT mydate, VALUE FROM
  (SELECT a.mydate, a.value,
          CASE WHEN a.value = b.value THEN '*' ELSE NULL END AS flag1,
          CASE WHEN a.value = c.value THEN '*' ELSE NULL END AS flag2
     FROM
       (SELECT ROWNUM AS outer_rownum, mydate, VALUE
         FROM mytable
         ORDER BY mydate) a
     LEFT OUTER JOIN
       (select ROWNUM-1 AS inner_rownum, mydate, VALUE
         from mytable
         order by myDATE) b
       ON b.inner_rownum = a.outer_rownum
     LEFT OUTER JOIN
       (select ROWNUM+1 AS inner_rownum, mydate, VALUE
         from mytable
         order by myDATE) c
       ON c.inner_rownum = a.outer_rownum
     ORDER BY a.mydate)
  WHERE flag1 = '*' OR
        flag2 = '*';

Share and enjoy.

Upvotes: 1

Related Questions