Left join update statement - sql

Question

I couldn't find any easy of doing table update with value from left join.
I want to do something like this:
UPDATE tbl1 p
LEFT JOIN (
select column1 , column2 from tbl2
union
select column1,column2 from tbl3
) c on c.column2=p.column2
SET p.column1 = nvl(c.column1, 'dummy');

Answer 1

UPDATE p SET p.column1 = nvl(c.column1, 'dummy')
from tbl1 p
LEFT JOIN (
select column1 , column2 from tbl2
union
select column1,column2 from tbl3
) c on c.column2=p.column2

Answer 2

I would recommend expressing this as:
UPDATE tbl1 p
SET p.column1 = COALESCE((SELECT column1 from tbl2 t2 WHERE t2.column2 = p.column2),
(SELECT column1 from tbl3 t3 WHERE t3.column2 = p.column2),
'dummy'
);
Each subquery can make use of indexes, so this should have better performance as well.

Answer 3

Try and see whether below code gives performance with CTE
Update tbl1 p
set column1 = nvl((with CTE as
(select column1 , column2 from tbl2
union
select column1,column2 from tbl3) select column1 from CTE where CTE.column2= p.column2 ),'dummy');

Answer 4

Use merge:
merge into tbl1 tgt
using (select min(column1) column1, column2
from (select column1, column2 from tbl2 union all
select column1, column2 from tbl3)
group by column2) src
on (tgt.column2 = src.column2)
when matched then update set tgt.column1 = src.column1
Let's say you have these tables:
create table tbl1(column1, column2) as (
select 0, 'A' from dual union all
select 0, 'B' from dual union all
select 0, 'C' from dual union all
select 0, 'D' from dual );
create table tbl2(column1, column2) as (
select 1, 'A' from dual union all
select 41, 'D' from dual );
create table tbl3(column1, column2) as (
select 2, 'B' from dual union all
select 42, 'D' from dual );
We have value for key A in tbl2, value for B in tbl3, value for C is absent in both source tables, and we have two problematic values for key D. You have to decide what to do in last case, use any aggregation function like min(), avg() or listagg() for strings. If such situation is not possible then you can simplify my statement, replace source subquery with simple union.
You could also use update, but in this case you have to add where clause and check existence of keys to avoid nullifying values and this makes code longer.
Result of merge:
COLUMN1 COLUMN2
---------- -------
1 A
2 B
0 C
41 D

Answer 5

Merge statement could have been the perfect solution if it was possible to update the same field in WHEN MATCHED and in WHEN NOT MATCHED. Anyway i fixed it like that:
The table is getting filled from external table using procedure.
Setting the field to 'dummy' for all records in the procedure inserting statement and use only MERGE WHEN MATCHED after works as i expected
PS: i couldnt make it work as #Ondřej Crha said.