ORACLE SQL: correlated update issue

Question

hi guys i have this kind of problem concerning correlated update on oracle.

consider i have a table params.

id_s    id_p    value    desc
-----------------------------------------------
10064     9      aaa     r
10064     8      bbb     t
10064     4      ccc     t
10064     4      ttt     y
11119     9      ddd     f
11119     8      eee     g 
11119     4      fff     b
11119     4      kkk     x

so i want to update params to get the following

id_s    id_p    value    desc
-----------------------------------------------
10064     9      aaa     r
10064     8      bbb     t
10064     4      ccc     t
10064     4      ttt     y
11119     9      aaa     r
11119     8      bbb     t 
11119     4      ccc     t
11119     4      ttt     y

i write update like this

update params p1
   set (p1.value, p1.desc) = (
           select p2.value
                , p2.desc 
             from params p2
            where p2.id_s = 10064
              and p2.id_p = p1.id_p
       )
 where p1.id_s = 11119
     ;

execution returns error 'ORA01427:Single-row subquery return more than one row'

how can i make this update work?

Answer 1

you have to add an additional condition to your subquery:

update params p1
    set (p1.value, p1.desc) = (
            select p2.value
                 , p2.desc 
              from params p2
             where p2.id_s = 10064
               and p2.id_p = p1.id_p
               and p1.id_s = 11119
        )
  where p1.id_s = 11119
      ;

edit: matters are further complicated by the update of the op's spec, which effectively amounts to a partial pk update (as far as the table excerpts given are complete wrt their columns).

a possible solution implements the following basic idea: the result set for the master combination of id_p, id_s is paired with the result set of for the slave combination of the same columns according to the rank in 2 orderings. the orderings are simply the orderings of the result set according to the columns valueand desc, but of course any other would do as well (note that in particular 2 different orderings were feasible).

the said pairing is then correlated with the update resultset.

in oracle sql:

update params p1
   set (p1.value, p1.desc) = (
           select emb.value
                , emb.desc 
             from (
                    select p2.value
                         , p2.desc
                         , p2.id_p
                         , rownum     rn 
                      from params p2
                     where p2.id_s = 10064
                  order by p2.value
                         , p2.desc
                  ) emb
             join (
                    select pm.value
                         , pm.desc
                         , pm.id_p
                         , rownum     rn 
                      from params pm
                     where pm.id_s = 11119
                  order by pm.value
                         , pm.desc
                  ) emb_master
                       ON (     emb_master.id_p = emb.id_p
                            AND emb_master.rn   = emb.rn   )
            where p1.id_s = 11119
              and emb_master.id_p     = p1.id_p
              and emb_master.value    = p1.value
              and emb_master.desc     = p1.desc
       )
 where p1.id_s = 11119
     ;

this scheme's viability hinges on the assumption that the result sets modulo each value of the tuple (id_p, id_s) be of the same cardibality. if thy don't, updates will be incomplete.

tested on ora 11g2 based on the excerpts given.

Answer 2

You don't specify what to do when there is more than one row returned. You can just select an arbitrary row using rownum = 1:

update params p1
   set (p1.value, p1.desc) = (
           select p2.value, p2.desc 
           from params p2
           where p2.id_s = 10064 and p2.id_p = p1.id_p and rownum = 1
       )
where p1.id_s = 11119;