Distinct values in each column

Question

Hi
I have a table with three columns and I would like to display distinct values in each column. Could any one please help me to get the output.
col1     col2     col3
====     ====     ====
a     1     x     
a     2     y
a     3      z
a     4      u
b     5      v
b     6      x
b     7      x
b     9      x
b     10     y
b     11     y
b     12     y
b     13     y
b     14     x
b     15     y
b     16     z
b     17     u
b     18     v
b     19     x
b     20     x
c     21     x
c     22     y
c     23     y
c     24     y
c     25     y
c     26     x
c     27     y
c     28     z
Output needed
===========
col1     col2     col3
====     ====     ====
a     1     x
b     2     y
c     3     z
     4     u
     5     v
     6
     7
     9
     10
     11
     12
     13
     14
     15
     16
     17
     18
     19
     20
     21
     22
     23
     24
     25
     26
     27
     28
Thanks
Vasanth

Answer 1

but the only drawback - you need to know the column, comprising the biggest amount of distinct values beforehand.No. There is no any difference in which column will be chosen as "based".
For demo purposes I've added extra row number column to show full generality and symmetry of the approach:
exec dbms_random.seed(0)
with t as ( select chr(ascii('a') + dbms_random.value(0, 3)) as c1,
                   trunc(dbms_random.value(1, 20))           as c2,
                   chr(ascii('u') + dbms_random.value(0, 6)) as c3
              from dual connect by level <= 20
   tt1 as ( select lag(          null,  1, c1) over (partition by c1 order by null) as c1,
                   lag(to_number(null), 1, c2) over (partition by c2 order by null) as c2,
                   lag(          null,  1, c3) over (partition by c3 order by null) as c3
              from t
   tt2 as ( select tt1.*,
                   row_number() over (order by c1) as rn1,
                   row_number() over (order by c2) as rn2,
                   row_number() over (order by c3) as rn3,
                   rownum rn
              from tt1
   tt3 as ( select case when rn1 > rn
                        then last_value(c1) over (order by rn1 range between abs(rn1 - rn) preceding
                                                                         and abs(rn1 - rn) preceding)
                        else last_value(c1) over (order by rn1 range between abs(rn - rn1) following
                                                                         and abs(rn - rn1) following)
                   end as c1,
                   case when rn2 > rn
                        then last_value(c2) over (order by rn2 range between abs(rn2 - rn) preceding
                                                                         and abs(rn2 - rn) preceding)
                        else last_value(c2) over (order by rn2 range between abs(rn - rn2) following
                                                                         and abs(rn - rn2) following)
                   end as c2,
                   case when rn3 > rn
                        then last_value(c3) over (order by rn3 range between abs(rn3 - rn) preceding
                                                                         and abs(rn3 - rn) preceding)
                        else last_value(c3) over (order by rn3 range between abs(rn - rn3) following
                                                                         and abs(rn - rn3) following)
                   end as c3
              from tt2
select c1, c2, c3 from tt3
  where c1 || c2 || c3 is not null
  order by c1, c2, c3
C1            C2 C3
a              1 u
b              3 v
c              4 w
               5 x
               7 y
               8 z
               9
              11
              13
              15
              16
              17
              19
13 rows selected.P.S. Thanks to [url http://www.sql.ru/forum/actualthread.aspx?bid=3&tid=482506&hl=over+range#4785373]Vladimir Sitnikov for the demo of the approach.