Extract first line from all subpartitions in all tables in Oracle database - sql

Question

I want to get the first value for all subpartitions of all tables in my Oracle database.
From my resultset, I want to update another table with the values returned and insert a comment('No value' for example) for subpartitions whic are empty.
I have written a piece of code to begin with one table.
Can you please tell me if the algorithm I am using is right and why there is no output?
DECLARE
BEGIN
for i in (select table_name, subpartition_name
from dba_tab_subpartitions
where table_name ='my_table'
order by 1)
loop
execute immediate ' insert into bkp_part (partition_name) values('||i.subpartition_name||')';
commit;
execute immediate 'select *
from '|| i.table_name||' subpartition('||i.subpartition_name||')
where rownum < 2';
end loop;
end;
END;
/

Answer 1

There is no output because you aren't creating any. Your dynamic query isn't going to be executed because you don't select the columns into anything; if you really want the entire row you'd need a %rowtype variable to select into, but I'm not sure if you really want the whole row or just one value from it. And having selected into something, you then need to use that variable - putting it into a table from what you've said, or displaying for debug purpoises via dbms_output.
Your insert statement doesn't need to be dynamic; you can just do:
insert into bkp_part (partition_name) values (i.subpartition_name);
To get the first row from each subpartition, you need to establish what you mean by 'first'. In this example I'm using a hash subpartition, so for the sake of argument I'll decide that the 'first' row is the lowest value in that hash bucket, and to keep it simple I'm only interested in that single column.
I've set up a dummy table with:
create table my_table (id number, part_date date)
partition by range (part_date)
subpartition by hash (id)
subpartition template (
subpartition subpart_1,
subpartition subpart_2,
subpartition subpart_3)
(
partition part_1 values less than (date '2015-01-01')
);
insert into my_table values (1, date '2014-12-29');
insert into my_table values (2, date '2014-12-30');
insert into my_table values (4, date '2014-12-31');
Looking at how the data is distributed, I have nothing in the first subpartition, IDs 1 and 4 in the second, and ID 2 in the third.
That means I can do:
set serveroutput on
declare
l_id my_table.id%type;
begin
for i in (
select table_name, subpartition_name
from user_tab_subpartitions
where table_name = 'MY_TABLE'
order by table_name, subpartition_position
) loop
-- skipping because I don't have that table
-- insert into bkp_part (partition_name) values (i.subpartition_name);
execute immediate 'select min(id) from ' || i.table_name
|| ' subpartition(' || i.subpartition_name || ')'
into l_id;
-- just for debugging/demo
dbms_output.put_line('Subpartition ' || i.subpartition_name
|| ' minimum ID is ' || l_id);
-- do something else with the value; update or insert...
-- update bkp_part set min_id = l_id
-- where partition_name = i.subpartition_name;
end loop;
end;
/
anonymous block completed
Subpartition PART_1_SUBPART_1 minimum ID is
Subpartition PART_1_SUBPART_2 minimum ID is 1
Subpartition PART_1_SUBPART_3 minimum ID is 2
You can adapt that to get the columns you're interested in and do something more useful with them (like inserting/updating your bkp_part table)