Performance Problems with Oracle Table Functions

Good afternoon,

I have got two functions returning the results of an SQL execution:
 - a pipelined function returning a table of records.
 - a non-pipelined function returning a table of objects.
I call those two functions using an Oracle Table function. (s. Code Snippet)

As an alternative I have got the same SQL, hard coded in an Oracle View.

If I compare these alternatives, the execution of the View takes much less time than the other two alternatives. In detail, all operations need more or less the same time except that the table fucntion version appends another sort output operation. Here are some long-op stats taken from Toad:
 - for the View:
     Table scan (114389 Blocks):         21s
     Hash Join (56854 Blocks):             45s
     Sort Output I (48867 Blocks):       71s

 - for the function returning the table of records:
     Table scan (114389 Blocks):         21s
     Hash Join (56854 Blocks):             43s
     Sort Output I (48867 Blocks):       69s
     Sort Output II (117763 Blocks):  132s

I noticed, that this effect does not depend on the overall execution time of the SQL code, but on the overall volume of returned datasets. For example, the above stats result in 4m datasets. The overall execution time is round about 150secs (and 350 secs for version with the table functions). The problem does not occur for an SQL running 400secs but only returning 1k datasets. So I am pretty sure, that the "waste of time" happens due to the outside table-function. Nevertheless I have got no clue, how to handle this problem.

Any ideas around? Is there any kind of way to get around the second sort output?

Best regards and thanks in advance
FINARIS

1:
2:
3:
4:
5:
6:
7:
8:
9:
10:
11:
12:
13:
14:
15:
16:
17:
18:
19:
20:
21:
22:
23:

function get_my_table return my_types.my_table_type PIPELINED is
 
	stmt long := null;
 
	v_rec	my_types.my_record_type;
 
	TYPE t_cur IS REF CURSOR;
	cur t_cur;
 
begin
	stmt := get_my_sql_statement;
	open cur for stmt;
	loop
		fetch cur into v_rec;
		exit when cur%notfound;
		PIPE ROW(v_rec);
	end loop;
	close cur;
	return;
end get_my_table;
 
 
select * from table(my_pack.get_my_table())