Update from another table when row does not exist - sql

Question

I am trying to run below query:
UPDATE tempd tvl
SET id = case
when tse.id is null or tse.id = 0 then --coalesce(nullif(tse.id,0),5)
5
else
tse.event_id end
FROM details tse
WHERE
tvl.id = tse.id
AND tvl.name = tse.name
AND tvl.add = tse.add;
Sample for table tempd:
name add id
a xx
b yy
c zz
d ss
details table is blank now. So every row in tempd should have 5, But I am getting always 0 in id column.
Also, can we update from multiple tables? Like: if there is a match from table1 then take that, if match from table2 take that, else use default value.

Answer 1

Rows excluded by predicates in the WHERE clause are not updated at all. That's the core problem of your query.
You want to update all rows of table tempd unconditionally. A correlated subquery in the assigned value does the job - with no outer WHERE clause:
UPDATE tempd tvl
SET id = COALESCE((
SELECT tse.event_id
FROM details tse
WHERE tse.id = tvl.id
AND tse.name = tvl.name
AND tse.add = tvl.add
AND tse.id <> 0 -- excludes NULL & 0
-- ORDER BY ???
-- LIMIT 1 -- can there be > 1 match?
), 5);
Can we update from multiple tables? Like: if there is a match from table1 then take that, if match from table2 take that, else use default value.
One way:
UPDATE tempd tvl
SET id = COALESCE(NULLIF(
(
SELECT tse.event_id
FROM details tse
WHERE tse.id = tvl.id
AND tse.name = tvl.name
AND tse.add = tvl.add
-- AND tse.id <> 0
-- this time we need to accept 0 to exit
-- ORDER BY ???
-- LIMIT 1 -- can there be > 1 match?
)
, (
SELECT t2.some_id
FROM tbl2 t2
WHERE ... -- your filters here
), 0), 5);
All parentheses are required.
If there can be more than 1 match from each correlated subquery, LIMIT 1 will fix that. But, typically, you'll want to add ORDER BY before each LIMIT 1 to get a deterministic pick.
See:
Nullify column in update if subquery returns empty
PostgreSQL update join vs SQL Server update join

Answer 2

It doesn't matter what the where clause is or the update you're asking for is. If you update from an empty table you will always get zero rows updated.
An even simpler example:
sophia=> create table empty_table(x integer);
CREATE TABLE
sophia=> create table foo(x integer);
CREATE TABLE
sophia=> insert into foo(x) values (1),(2),(3);
INSERT 0 3
sophia=> update foo set x = 5 from empty_table where true;
UPDATE 0
sophia=>
As to the solution - you can pick from one of the following depending on your exact use-case:
Get the value to update from a subquery (as per Erwin's example)
Do something so that details always has data
Update the from so the source always has something (e.g. join with something else)