翻译强力驱动Possible Duplicate:
Oracle: how to UPSERT (update or insert into a table?)
Hi,
I have a table in which a record has to be modified if it already exists else a new record has to be inserted. Oracle sql doesnt accept IF EXISTS, otherwise I would have done an if - update - else - insert query. I've looked at MERGE but it only works for multiple tables. What do i do?
MERGE doesn't need "multiple tables", but it does need a query as the source. Something like this should work:
MERGE INTO mytable d USING (SELECT 1 id, 'x' name from dual) s ON (d.id = s.id) WHEN MATCHED THEN UPDATE SET d.name = s.name WHEN NOT MATCHED THEN INSERT (id, name) VALUES (s.id, s.name);Alternatively you can do this in PL/SQL:
BEGIN INSERT INTO mytable (id, name) VALUES (1, 'x'); EXCEPTION WHEN DUP_VAL_ON_INDEX THEN UPDATE mytable SET name = 'x' WHERE id = 1; END;merge into MY_TABLE tgt using (select [expressions] from dual ) src on (src.key_condition = tgt.key_condition) when matched then update tgt set tgt.column1 = src.column1 [,...] when not matched then insert into tgt ([list of columns]) values (src.column1 [,...]);The way I always do it (assuming the data is never to be deleted, only inserted) is to
insert, if this fails with a unique constraint violation then you know the row is there,updateUnfortunately many frameworks such as Hibernate treat all database errors (e.g. unique constraint violation) as unrecoverable conditions, so it isn't always easy. (In Hibernate the solution is to open a new session/transaction just to execute this one insert command.)
You can't just do a select count(*) .. where .. as even if that returns zero, and therefore you choose to do an insert, between the time you do the select and the insert someone else might have inserted the row and therefore your insert will fail.
You could use the SQL%ROWCOUNT Oracle variable:
UPDATE table1 SET field2 = value2, field3 = value3 WHERE field1 = value1; IF (SQL%ROWCOUNT = 0) THEN INSERT INTO table (field1, field2, field3) VALUES (value1, value2, value3); END IF; It would be easier just to determine if your primary key (i.e. field1) has a value and then perform an insert or update accordingly. That is, if you use said values as parameters for a stored procedure.
HC-way :)
DECLARE rt_mytable mytable%ROWTYPE; CURSOR update_mytable_cursor(p_rt_mytable IN mytable%ROWTYPE) IS SELECT * FROM mytable WHERE ID = p_rt_mytable.ID FOR UPDATE; BEGIN rt_mytable.ID := 1; rt_mytable.NAME := 'x'; INSERT INTO mytable VALUES (rt_mytable); EXCEPTION WHEN DUP_VAL_ON_INDEX THEN > FOR i IN update_mytable_cursor(rt_mytable) LOOP UPDATE mytable SET NAME = p_rt_mytable.NAME WHERE CURRENT OF update_mytable_cursor; END LOOP update_mytable; END;Please refer to this question if you want to use UPSERT/MERGE command in Oracle. Otherwise, just resolve your issue on the client side by doing a count(1) first and then deciding whether to insert or update.