Can u find (number of rows) + (SELECT * FROM emp) in 1 query?

Question

We have a requirement like this: We need to pass a SQL statement as a SYS_REFCURSOR OUT variable from a SP. Problem is, if the SQL returned 0 rows we have to send 1 row with all NULLS to the SYS_REFCURSOR.
Is there any way to find out the number of rows returned from a SQL without having to put a SELECT COUNT(*) again using the "same" SQL?

Answer 1

Peter:
That only works if the cursor actually returns rows. Consider:
SQL> create table t as
  2  select rownum id, to_char(to_date(rownum, 'J'), 'Jsp') descr
  3  from user_objects
  4  where rownum <= 5;
Table created.
SQL> insert into t select * from t;
5 rows created.
SQL> commit;
Commit complete.
SQL> create function f (p_id in number) return sys_refcursor as
  2     l_cur sys_refcursor;
  3  begin
  4     open l_cur for
  5        select id, descr, count(*) over() cnt
  6        from t
  7        where id = p_id;
  8     return l_cur;
  9  end;
10  /
Function created.
SQL> var cur refcursor
SQL> exec :cur := f(1);
PL/SQL procedure successfully completed.
SQL> print cur
        ID DESCR             CNT
         1 One                 2
         1 One                 2
SQL> exec :cur := f(42);
PL/SQL procedure successfully completed.
SQL> print cur
no rows selectedSince the OP's requirement (or ""solution" coming from someone who misunderstood the concept of cursors") is that "if the SQL returned 0 rows we have to send 1 row with all NULLS to the SYS_REFCURSOR" the analytic count doesn't help.
@OP
The only reliable (for certain definitions of reliable) way would be to consume the first row of the cursor to see if it did in fact return any rows.
However, if the cursor was initally empty, we then need to generate a new cursor with all NULLS to satisfy the request. Not a big deal, as the new cursor would reflect the state of the database at the time of the initial query, even if a qualifying row was inserted and committed between the first query and the second, but it doesn't work the other way around.
If I consume the first row and find a record, then I need to re-do the query to get all of the rows. but what happens if another process changes/deletes the qualifying row in between and commits? The second query will have no rows and you will return an empty cursor to the caller.
Also, counting the rows (which the analytic count will need to do) could have a significant impact on performance since all the qualifying rows need to be read before the first row can be returned.
John