Question

I can’t get information of that simple question over the PostgreSQL documentation, over the Web or even here on StackOverflow… I must not understand something essential here.

I am making a simple SELECT/UPDATE transaction in PostgreSQL:

START TRANSACTION;
SELECT "column" FROM "table" WHERE "criterion" = 'value' AND "activated" = true;
UPDATE "table" SET "activated" = false WHERE "criterion" = 'value';
COMMIT

Basically, I need to get the value of a column when its activated state is true and then deactivate it.
PostgreSQL tells me that there was a 1 row result that has been cancelled

The same happens if I do the following (basically the same transaction without the UPDATE statement):

START TRANSACTION;
SELECT "column" FROM "table" WHERE "criterion" = 'value' AND "activated" = true;
COMMIT

What don’t I understand about transactions? Can’t any SELECT output get out of a transaction block?

Answer 1

This will return all “column”‘s values from the updated rows:

UPDATE "table" SET "activated" = false WHERE "criterion" = 'value' AND "activated" = true
returning "column";

There is no relation to the transaction.

returning will return the values as if a select was issued:

insert into foo (ticket, row, archived) values (3,7,true) returning *;
 ticket | row | archived 
--------+-----+----------
      3 |   7 | t

---

update foo
set archived = true
where not archived
returning *;
 ticket | row | archived 
--------+-----+----------
      2 |   5 | t