Oracle nested correlated subquery problem

Consider table1 and table2 with a one-to-many relationship (table1 is the master table and table2 is the detail table). I want to get records from table1 where some value ('XXX') is the value of the most recent record in table2 of the detail records correlated to table1. What I want to do is this:

select t1.pk_id
  from table1 t1
 where 'XXX' = (select a_col
                  from (  select a_col
                            from table2 t2
                           where t2.fk_id = t1.pk_id
                    开发者_如何学编程    order by t2.date_col desc)
                 where rownum = 1)

But, because the reference to table1 (t1) in the correlated subquery is two-levels deep, it pops up with an Oracle error (invalid id t1). I need to be able to rewrite this, but the one caveat is that only the where clause may be changed (i.e. the initial select and from must remain unchanged). Can it be done?

Here's a different analytic approach:

select t1.pk_id
  from table1 t1
 where 'XXX' = (select distinct first_value(t2.a_col)
                                  over (order by t2.date_col desc)
                  from table2 t2
                  where t2.fk_id = t1.pk_id)

And here's the same idea using a ranking function:

select t1.pk_id
  from table1 t1
 where 'XXX' = (select max(t2.a_col) keep
                          (dense_rank first order by t2.date_col desc)
                  from table2 t2
                  where t2.fk_id = t1.pk_id)

you could use analytics here: join table1 to table2, take the most recent table2 record for each element in table1 and verify that this most recent element has a value of 'XXX':

SELECT *
  FROM (SELECT t1.*, 
               t2.a_col, 
               row_number() over (PARTITION BY t1.pk 
                                  ORDER BY t2.date_col DESC) rnk
           FROM table1 t1
           JOIN table2 t2 ON t2.fk_id = t1.pk_id)
 WHERE rnk = 1
   AND a_col = 'XXX'

Update: Without modifying the top-level SELECT, you could write a query like this:

SELECT t1.pk_id
  FROM table1 t1
 WHERE 'XXX' =
       (SELECT a_col
          FROM (SELECT a_col, 
                       t2_in.fk_id, 
                       row_number() over(PARTITION BY t2_in.fk_id 
                                         ORDER BY t2_in.date_col DESC) rnk
                   FROM table2 t2_in) t2
         WHERE rnk = 1
           AND t2.fk_id = t1.pk_id)

Basically you only join (SEMI-JOIN) the rows from table2 that are the most recent for each fk_id

Try this:

select t1.pk_id   
from table1 t1  
where 'XXX' = 
(select a_col 
 from table2 t2                           
 where t2.fk_id = t1.pk_id                         
 and t2.date_col =
   (select max(t3.date_col)
    from table2 t3
    where t3.fk_id = t2.fk_id)
)

Does this do what you are looking for?

select t1.pk_id
  from table1 t1
 where 'XXX' = (  select a_col
                    from table2 t2
                   where t2.fk_id = t1.pk_id
                    t2.date_col = (select max(date_col) from table2 where fk_id = t1.pk_id)
                )