Question

I have a table MYTABLE

N_REC | MYFIELD |
1     | foo     |
2     | foo     |
3     | bar     |

where N_REC is the primary key and MYFIELD is a non-unique field.
I need to query this table on MYFIELD and extract the associated N_REC, but only if there is only one occurrence of MYFIELD; otherwise I need no records returned.
So if I go with MYFIELD=’bar’ I will get 3, if I go with MYFIELD=’foo’ I will get no records.

I went with the following query

select * from 
   ( 
   select 
     n_rec, 
     ( select count(*) from mytable where mycolumn=my.mycolumn ) as counter 
   from mytable my where mycolumn=? 
   ) 
where counter=1

While it gives me the desired result I feel like I’m running the same query twice.
Are there better ways to achieve what I’m doing?

Answer 1

You might also try the analytic or windowing version of count(*) and compare plans to the other options:

select n_rec, my_field
from (select n_rec, my_field
        , count(*) over (partition by my_field) as Counter
    from myTable
    where my_field = ?)
where Counter = 1