I have a range partitioned Oracle table which I want to update using partition key.
Is there a way to dynamically create update script which takes into account partition key as follows, WITHOUT having to manually maintain such script
update table where date between 'a' and 'b'
update table where date between 'b' and 'c'.
I think it would not be necessary to specify the range in the where clause to update individual partitions. You may use the partition_extension_clause of update dynamically.
BEGIN
FOR r IN (
SELECT partition_name
FROM user_tab_partitions
WHERE table_name = 'YOUR_TABLE'
) LOOP
EXECUTE IMMEDIATE 'UPDATE YOUR_TABLE (' || r.partition_name || ') SET somecol = somevalue where someother_clause';
COMMIT; --if it's necessary
END LOOP;
END;
/
Related
I have a table called schema.exampletab, which has partitions where the partition name is on the form Pxxxxxx where the x's are partition periods (year and month YYYYMM) and then there is a local PK. I want to drop a partition, but I get the error message ORA-14006: Invalid partition name when attempting to drop it, but when I try to simply query data from a partition it selects and displays the data from the partition in question. Note that I am using a scheduler which uses date as a parameter using YYYYMMDD, so I usually substring.
For instance I can query data by using the following statement:
SELECT *
FROM schema.exampletab
PARTITION (P202110);
This returns 20 rows of data in my table.
I then try to drop the partition using the following statement:
ALTER TABLE schema.exampletab DROP PARTITION (concat(P,substr('20211011',1,6))
UPDATE INDEXES;
This however, leads to the ORA-14006 error message. In order to control that the partition exists I tried to look up that the schema, table, and partition exists in the all_tab_partitions where all partitions in the database I work on are logged.
select partition_name from all_tab_partitions
where table_owner = 'schema' and table_name = 'exampletab' and
substr(partition_name,2,7) = substr('20211022',1,6);
This returns the partition name P202110 in the query result.
I hoped I could use the drop partition with a subquery like this:
ALTER TABLE schema.exampletab DROP PARTITION select partition_name from all_tab_partitions
where table_owner = 'schema' and table_name = 'exampletab' and
substr(partition_name,2,7) = substr('20211022',1,6);
However, this still leads to ORA-14006 error. I have tried to write the partition name like 'P202110' and P202110 instead of the parenteces, but no luck.
How can I write the drop partition statement such that it drops the partition instead of giving the ORA-14006 error?
This is something which I have to do routinely and so it would be nice to know how to properly select and drop partitions, or truncate them, etc. Also, I use a scheduler which runs the sql queries at specified intervals so there I have to specify dates as a parameter, meaning that in my example code the YYYYMM is a parametervalue that gets parsed in and so I need to concactinate the P with this outputted parameter value, because if I type only P202110 it actually drops it.
You have two options, keeping in consideration that you are using sysdate to know which partition you should drop.
Option 1 -> Dynamic SQL and the result must be executed outside the query
select ' alter table '||table_owner||'.'||table_name||' drop partition '||partition_name||' update indexes ; '
from
dba_Tab_partitions
where table_owner = 'your_schema'
and table_name = 'your_table'
and partition_name = 'P'||substr(to_char(sysdate,'yyyymmdd'),1,6) ; ;
This query gives you the command output, but you need to execute it
alter table yourschema.yourtable drop partition P202110 update indexes;
Example in my own environment
SQL> select ' alter table '||table_owner||'.'||table_name||' drop partition '||partition_name||' update indexes ; '
from
dba_Tab_partitions
where table_owner = 'FDM_DATA'
and table_name = 'FDM_DIM_CUSTOMER'
and partition_name = 'P_'||substr(to_char(sysdate,'yyyymmdd'),1,6) ;
'ALTERTABLE'||TABLE_OWNER||'.'||TABLE_NAME||'DROPPARTITION'||PARTITION_NAME||'UP
--------------------------------------------------------------------------------
alter table FDM_DATA.FDM_DIM_CUSTOMER drop partition P_202110 update indexes ;
Option 2 -> PLSQL
A better option is to use PLSQL. A small example when you want to drop only one partition based on the current sysdate. You can extend / modify this code to cover any kind of time frame.
declare
v_owner varchar2(128) := 'YOUR_SCHEMA';
v_table_name varchar2(128) := 'YOUR_TABLE';
v_partition_name varchar2(128);
begin
select partition_name into v_partition_name from all_tab_partitions
where table_owner = v_owner and
table_name = v_table_name and
partition_name = 'P'||substr(to_char(sysdate,'yyyymmdd'),1,6) ;
execute immediate 'alter table '||v_owner||'.'||v_table_name||' drop partition '||v_partition_name||' update indexes' ;
exception
when no_data_found then null; -- if there is no partition, nothing to do and no error is raised
when others then raise;
end;
/
You cannot use "expressions" in partition name
This is wrong:
ALTER TABLE schema.exampletab DROP PARTITION (concat(P,substr('20211011',1,6))
UPDATE INDEXES;
this is correct:
ALTER TABLE schema.exampletab DROP PARTITION P20211011
UPDATE INDEXES;
You may use this code to drop partitions dynamically. Just update C_OWNER ,C_TABLE_NAME , C_PARTITION_TEMPLATE in the header. Please pay attention that the sql command is built dynamically, but when it is ready is has the partition name fully parsed.
declare
C_OWNER varchar2(128) := 'MYOWNER';
C_TABLE_NAME varchar2(128) := 'MYTABLE';
C_PARTITION_TEMPLATE varchar2(128) := '2011';
cursor part_cur is
select *
from all_tab_partitions
where table_owner=C_OWNER and
table_name = C_TABLE_NAME and
partition_name like '%'||C_PARTITION_TEMPLATE||'%'
order by partition_position;
BEGIN
for part_rec in part_cur loop
execute immediate 'ALTER TABLE "'||part_rec.table_owner||'"."'||part_rec.table_name||'"'||
' DROP PARTITION ("'||part_rec.partition_name||'") UPDATE INDEXES';
end loop;
END;
/
How can I drop multiple tables in the same database that starts with same prefix?
Ex:
Query to delete 1 table
drop table project_id.db.test_table_<some_random_string>
But how can I drop all tables that start with the same prefix test_table_ in the same db?
A possible work around would be. (region is set eu)
BEGIN
DECLARE drop_statments ARRAY<string>;
DECLARE len int64 default 1;
SET drop_statments = (SELECT ARRAY_AGG( 'drop table ' || table_schema ||'.' || table_name)
FROM `region-eu.INFORMATION_SCHEMA.TABLES`
WHERE table_schema = 'db' and table_name like 'Table_Prefix%'
);
WHILE ARRAY_LENGTH(drop_statments) >= len DO
EXECUTE IMMEDIATE drop_statments[offset(len-1)];
SET len = len +1;
END WHILE ;
END;
You may use any of below INFORMATION_SCHEMA dataset.
-- Returns metadata for tables in a single dataset.
SELECT * FROM myDataset.INFORMATION_SCHEMA.TABLES;
-- Returns metadata for tables in a region.
SELECT * FROM region-us.INFORMATION_SCHEMA.TABLES;
Similar to narendra# solution, but using FOR..IN:
FOR drop_statement IN
(SELECT CONCAT("drop table ",table_schema,".", table_name, ";" ) AS value
FROM dataset.INFORMATION_SCHEMA.TABLES -- or region.INFORMATION_SCHEMA.TABLES
WHERE table_name LIKE "table_prefix%"
ORDER BY table_name DESC)
DO
EXECUTE IMMEDIATE(drop_statement.value); -- Here the table is dropped
END FOR;
It's also worth to mention, that you can change "table" to "view" also depending on the information returned from the INFORMATION_SCHEMA views.
I have a database with the column "endpointid" in a lot of tables. I am looking for a search function that would find every table containing a specific endpointid in order to write a query to delete that endpoint. I have tried a delete function to delete it from all tables but that is not working properly since a specific endpointid might not be in all tables. I know the following query gives all tables with the column name:
select table_name from all_tab_columns where lower(column_name) like lower('%endpointid%');
How can I extend that query to search for a specific record of endpointid?
Here is an example to delete rows with a specific endpointid value:
CREATE TABLE mytest (
endpointid NUMBER
);
INSERT INTO mytest VALUES ( 1 );
INSERT INTO mytest VALUES ( 2 );
DECLARE
ep NUMBER := 2;
BEGIN
FOR t_rec IN (
SELECT
table_name
FROM
all_tab_columns
WHERE
lower(column_name) LIKE lower('%endpointid%')
) LOOP
EXECUTE IMMEDIATE 'delete from '
|| t_rec.table_name
|| ' where endpointid = :1'
USING ep;
END LOOP;
END;
Note that if these tables have foreign key relationships, this may fail, as it does not take into account the ordering of the table references. If that is needed, then you would need to structure your metatada query to find those relationships.
I have some big tables (30+ columns) with NOT NULL constraints. I would like to change all those constraints to NULL. To do it for a single column I can use
ALTER TABLE <your table> MODIFY <column name> NULL;
Is there a way to do it for all columns in one request ? Or should I copy/paste this line for all columns (><) ?
Is there a way to do it for all columns in one request ?
Yes. By (ab)using EXECUTE IMMEDIATE in PL/SQL. Loop through all the columns by querying USER_TAB_COLUMNS view.
For example,
FOR i IN
( SELECT * FROM user_tab_columns WHERE table_name = '<TABLE_NAME>' AND NULLABLE='N'
)
LOOP
EXECUTE IMMEDIATE 'ALTER TABLE <TABLE_NAME> MODIFY i.COLUMN_NAME NULL';
END LOOP;
In my opinion, by the time you would write the PL/SQL block, you could do it much quickly by using a good text editor. In pure SQL you just need 30 queries for 30 columns.
For a single table you can issue a single alter table command to set the listed columns to allow null, which is a little more efficient than running one at a time, but you still have to list every column.
alter table ...
modify (
col1 null,
col1 null,
col3 null);
If you were applying not null constraints then this would be more worthwhile, as they require a scan of the table to ensure that no nulls are present, and (I think) an exclusive table lock.
You can query user_tab_cols and combine it with a FOR cursor & EXECUTE IMMEDIATE to modify all not null columns - the PL/SQL block for doing like that would look like so:
DECLARE
v_sql_statement VARCHAR2(2000);
BEGIN
FOR table_recs IN (SELECT table_name, column_name
FROM user_tab_cols
WHERE nullable = 'N') LOOP
v_sql_statement :=
'ALTER TABLE ' || table_recs.table_name || ' MODIFY ' || table_recs.column_name || ' NULL';
EXECUTE IMMEDIATE v_sql_statement;
END LOOP;
END;
If you want to do it for all columns in a database instead of the ones in current schema, you can replace user_tab_cols and put in dba_tab_cols; I'd just run the query in the FOR to ensure that the columns being fetched are indeed the correct ones to be modified
I have a problem in that I want my output to be a single table (lets call it Output) with 2 columns: one for the "TableName" and one for the DateTime of the last update (using the scn_to_timestamp(max(ora_rowscn)) command).
I have 100 tables and I want to pull in the last update date/times for all these tables into the Output table.
So I can do this:
insert into Output(TableName)
select table_name
from all_tables;
which will put all the tables I have from my database into the TableName column. But I don't know how to loop through each entry and use the tablename as a variable and pass this into the scn_to_timestamp(ora_rowscn).
I thought I would try something like below:
for counter in Output(TableName) LOOP
insert into Output(UpdateDate)
select scn_to_timestamp(max(ora_rowscn))
from counter;
END LOOP;
Any suggestions?
Thank you
This query is a little bit clumsy as it uses xmlgen to execute dynamic sql in a query, but it might work for you.
select x.*
from all_tables t,
xmltable('/ROWSET/ROW' passing
dbms_xmlgen.getxmltype('select ''' || t.table_name ||
''' tab_name, max(ora_rowscn) as la from ' ||
t.table_name)
COLUMNS tab_name varchar2(30) PATH 'TAB_NAME',
max_scn number PATH 'LA') x
Here is a sqlfiddle demo
You can also use PLSQL and then use execute immediate