Question

I have a table like this (basic example, not the real thing):

FKEY  | NAME  | ATTRIBUTE_X
--------------------------
 1      '...'    42
 1      '...'    42
 1      '...'    42
 2      '...'    7
 2      '...'    7
 5      '...'    42
 5      '...'    42
 5      '...'    42
 5      '...'    42
 6      '...'    300
 6      '...'    300
 ....

Where – normally – each of the attribute_x values for a given fkey are all the same. (In my real data, I calculate attribute_x from some columns in the table and this property needs to be the same for all rows with the same fkey.

Now I have some rows where this property does not hold. I want to search the whole table to find all FKEYs with mismatched attribute_x values.

Example:

--------------------------
 145678973      '...'    23
 145678973      '...'    22 // Error, should also be 23
 145678973      '...'    23

My naive approach was:

SELECT distinct(TX1.FKEY)
FROM TABLEX TX1, TABLEX TX2
WHERE TX1.FKEY=TX2.FKEY
  AND TX1.ATTRIBUTE_X <> TX2.ATTRIBUTE_X
;

But with my real data this doesn’t complete (I ran of of temp tablespace and after the DBA increased the temp tablespace to 20 GIG the query ran for a few hours and then bailed out.)

Generally, is there a more efficient query for this?

---

I have a solution with PL/SQL where I just loop over the table sorted by FKEY, and if I find a different attribute_x vs. the last fetched record where the fkey stayed the same, I have found an erroneous fkey.

But this seems oh so primitve 🙂 Is there an efficient pure SQL solution?

Answer 1

Simplest way:

select fkey
from tablex
group by fkey
having count(distinct attribute_x) > 1