I have a delete statement and I want to call a procedure for each of the rows affected by it. Currently I only seem to be able to access a single row of the set that was deleted.
That is if the delete statement below affects multiple rows then I only see a single passed to my analytics procedure
create or replace function platform.update_foo() returns trigger language plpgsql as $$
declare
result record;
begin
with affected_rows as (
delete from private.foo
where environment_id=new.environment_id
and name=new.name
returning *
)
select private.analytics(affected_rows, 'cancelled')
from affected_rows into result;
return new;
end;
$$;
My analytics function has a signature like so
analytics(f private.foo, t varchar default null)
returns void as $$
begin
-- ...
Related
I'm trying to create a trigger with a condition. Based on the geom length I want the attribute "nom" (= name) to be written in upper case or lower case.
here's what I have:
CREATE OR REPLACE FUNCTION public.test_upper_lower()
RETURNS trigger AS
$BODY$
BEGIN
NEW.dummy:= (ST_Length(new.geom));
if (SELECT dummy FROM ligne_ligne)>100
then NEW.nom:= LOWER(nom) FROM ligne_ligne;
else NEW.nom:= UPPER(nom) FROM ligne_ligne;
end if;
RETURN NEW;
END;
$BODY$
LANGUAGE plpgsql;
DROP trigger IF EXISTS test_upper_lower on public.ligne_ligne;
CREATE trigger test_upper_lower BEFORE INSERT OR UPDATE on public.ligne_ligne
FOR EACH ROW
EXECUTE PROCEDURE public.test_upper_lower();
With this I have a "more than one row returned by a subquery" error
Based on other questions on this forum I tried it using case instead of if and using when in the trigger itself not the function but neither are working
Any ideas ?
Thanks
You don't need (or can actually) use SELECT statements to access data from the inserted row.
The part SELECT dummy FROM ligne_ligne returns all rows from that table - not just from the one relevant to the trigger.
As you just want to check the value you just calculated, simply use new.dummy at that point:
CREATE OR REPLACE FUNCTION public.test_upper_lower()
RETURNS trigger AS
$BODY$
BEGIN
NEW.dummy:= ST_Length(new.geom);
if new.dummy > 100 then --<< no SELECT necessary
NEW.nom:= LOWER(new.nom); --<< no "FROM", just access the value
else
NEW.nom:= UPPER(new.nom);
end if;
RETURN NEW;
END;
$BODY$
LANGUAGE plpgsql;
I am using trigger in Postgres database to call function and send newly inserted row to NodeJs application
CREATE OR REPLACE FUNCTION triggerFunction() RETURNS trigger AS $$
DECLARE
BEGIN
PERFORM pg_notify('tableName', row_to_json(NEW)::text );
RETURN NEW;
END;
$$ LANGUAGE plpgsql;
This returns the whole row in json format. However I need to change one of the column name while this row is returned.
Unfortunately AS keywork doesnt work in the row to json with NEW.COLUMN_NAME AS NEW_COLUMN. How can we achieve the solution for this?
CREATE OR REPLACE FUNCTION triggerFunction() RETURNS trigger AS $$
DECLARE
ret json;
BEGIN
select row_to_json(x) into ret from
(select NEW.abc as def, NEW.jkl, NEW.col3) x;
PERFORM pg_notify('tableName', ret::text );
RETURN NEW;
END;
$$ LANGUAGE plpgsql;
This thread is a part challenge of this thread to which I am searching a better solution for one part by BEFORE TRIGGER.
I just want to launch a trigger to convert to correct brackets.
I am thinking whether I should return from the trigger NULL or something else in before trigger.
Code
CREATE OR REPLACE FUNCTION insbef_events_function()
RETURNS TRIGGER AS
$func$
DECLARE
m int[];
BEGIN
FOREACH m SLICE 1 IN ARRAY TG_ARGV[0]::int[]
LOOP
INSERT INTO events (measurement_id, event_index_start, event_index_end)
SELECT NEW.measurement_id, m[1], m[2]; -- Postgres array starts with 1 !
END LOOP;
-- do something with _result ...
RETURN NULL; -- result ignored since this is an BEFORE trigger TODO right?
END
$func$ LANGUAGE plpgsql;
which I use the by the function
CREATE OR REPLACE FUNCTION f_create_my_trigger_events(_arg1 int, _arg2 text, _arg3 text)
RETURNS void AS
$func$
BEGIN
EXECUTE format($$
DROP TRIGGER IF EXISTS insbef_ids ON events
CREATE TRIGGER insbef_ids
BEFORE INSERT ON events
FOR EACH ROW EXECUTE PROCEDURE insbef_events_function(%1$L)$$
, translate(_arg2, '[]', '{}'), translate(_arg3, '[]', '{}')
);
END
$func$ LANGUAGE plpgsql;
I am unsure about this line: RETURN NULL; -- result ignored since this is anBEFOREtrigger TODO right?, since I think this is the case in AFTER trigger but not in before trigger.
I just want to launch a trigger to convert correct brackets.
Test command is sudo -u postgres psql detector -c "SELECT f_create_my_trigger_events(1,'[112]','[113]');" getting the following error because of misunderstanding of the returning -thing, I think.
LINE 3: CREATE TRIGGER insbef_ids
^
QUERY:
DROP TRIGGER IF EXISTS insbef_ids ON events
CREATE TRIGGER insbef_ids
BEFORE INSERT ON events
FOR EACH ROW EXECUTE PROCEDURE insbef_events_function('{112}')
CONTEXT: PL/pgSQL function f_create_my_trigger_events(integer,text,text) line 4 at EXECUTE statement
How can you manage BEFORE triggers in PostgreSQL 9.4?
First of all, you need to pass the row variable in a BEFORE trigger. Passing NULL cancels the operation for the row:
CREATE OR REPLACE FUNCTION insbef_events_function()
RETURNS TRIGGER AS
$func$
DECLARE
m int[];
BEGIN
FOREACH m SLICE 1 IN ARRAY TG_ARGV[0]::int[]
LOOP
INSERT INTO events (measurement_id, event_index_start, event_index_end)
SELECT NEW.measurement_id, m[1], m[2]; -- Postgres array subscripts start with 1
END LOOP;
-- do something with _result ...
RETURN NEW; -- NULL would cancel operation in BEFORE trigger!
END
$func$ LANGUAGE plpgsql;
I demonstrated the use of RETRUN NULL in an AFTER trigger in my previous answer. You can't do the same for a BEFORE trigger. The manual:
Row-level triggers fired BEFORE can return null to signal the trigger
manager to skip the rest of the operation for this row (i.e.,
subsequent triggers are not fired, and the INSERT/UPDATE/DELETE does
not occur for this row). If a nonnull value is returned then the
operation proceeds with that row value.
There is more. Read the manual.
But since you are passing two 1-dimensional arrays instead of one 2-dimensional array now, you need to adapt your trigger logic:
CREATE OR REPLACE FUNCTION insbef_events_function()
LANGUAGE plpgsql RETURNS TRIGGER AS
$func$
DECLARE
a1 int[] := TG_ARGV[1]::int[];
a2 int[] := TG_ARGV[2]::int[];
BEGIN
FOR i in array_lower(a1, 1) .. array_upper(a1, 1)
LOOP
INSERT INTO events (measurement_id, event_index_start, event_index_end)
SELECT NEW.measurement_id -- or TG_ARGV[0]::int instead?
, a1[i], a2[i];
END LOOP;
RETURN NEW; -- NULL would cancel operation in BEFORE trigger!
END
$func$;
It's your responsibility that both arrays have the same number of elements.
The function changing the trigger could look like this now:
CREATE OR REPLACE FUNCTION f_create_my_trigger_events(_arg1 int, _arg2 text, _arg3 text)
LANGUAGE plpgsql RETURNS void AS
$func$
BEGIN
EXECUTE format(
$$DROP TRIGGER IF EXISTS insbef_ids ON measurements; -- on measurements ..
CREATE TRIGGER insbef_ids
BEFORE INSERT ON measurements -- .. according to previous posts!!
FOR EACH ROW EXECUTE PROCEDURE insbef_events_function(%s, %L, %L)$$
, _arg1
, translate(_arg2, '[]', '{}')
, translate(_arg3, '[]', '{}')
);
END
$func$;
You need to understand basics of SQL, PL/pgSQL, trigger functions and array handling before using this advanced automated design.
Is it possible to have a stored procedure behave exactly like a regular select query when no records are found, or is this a driver issue.
For example, with go, a query that returns no rows will return an sql.ErrNoRows error. However, this will not:
create table emptytable(id int);
create function selectany() returns emptytable as $$
DECLARE
_out emptytable;
BEGIN
SELECT * INTO emptytable FROM emptytable limit 1;
RETURN _out;
END;
$$ LANGUAGE PLPGSQL;
I have tried SELECT INTO STRICT, and while that raises a "query returned no rows" error, it is not the same as a non-stored procedure query. Neither is raising NO_DATA_FOUND.
If I understand your requirements correctly:
Return one or no row from a function and allow to do more with the returned row (if any).
Test table:
CREATE TABLE emptytable(id int, txt text); -- multiple columns
To return one or no complete table row:
CREATE OR REPLACE FUNCTION selectany_all()
RETURNS SETOF emptytable AS
$func$
DECLARE
_out emptytable;
BEGIN
FOR _out IN
SELECT * FROM emptytable LIMIT 1
LOOP
-- do something with _out before returning
RAISE NOTICE 'before: %', _out;
RETURN NEXT _out;
-- or do something with _out after returning row
RAISE NOTICE 'after: %', _out;
END LOOP;
END
$func$ LANGUAGE plpgsql;
For a more flexible approach: return arbitrary columns:
CREATE OR REPLACE FUNCTION selectany_any()
RETURNS TABLE (id int, txt text) AS
$func$
BEGIN
FOR id, txt IN
SELECT e.id, e.txt FROM emptytable e LIMIT 1
LOOP
-- do something with id and text before returning
RAISE NOTICE 'before: %, %', id, txt;
RETURN NEXT;
-- or do something with id and text after returning row
RAISE NOTICE 'after: %, %', id, txt;
END LOOP;
END
$func$ LANGUAGE plpgsql;
Note, the LOOP is never entered if there is no row. Accordingly you will get no NOTICE from my test code.
Both functions work for n rows returned as well, LIMIT 1 is just for this particular request.
Closely related, wtih more explanation:
Return multiple fields as a record in PostgreSQL with PL/pgSQL
2.5 options:
1a) If you just need to return a query, you can use SETOF and RETURN QUERY
1b) or just use language SQL as #ClodoaldoNeto, which returns a query natively using sql's SELECT stmt
2) If you need to process the result in the procedure, you must use SETOF and RETURN NEXT, ensuring you check IF FOUND THEN RETURN; (note lack of NEXT, which if given will act as a single blank row is returned)
Ideally, I'd like to not use SETOF for procedures known to return exactly none or 1 rows, but it seems SETOF is required to get a procedure to query like an sql statement from the app and have drivers recognize NO ROWS RETURNED
Examples below:
create table emptytable(id int);
create function selectany() returns setof emptytable as $$
DECLARE
_out emptytable;
BEGIN
SELECT * INTO _out FROM emptytable limit 1;
IF FOUND THEN
RETURN _out;
END IF;
RETURN;
END;
$$ LANGUAGE PLPGSQL;
create function selectany_rq() returns setof emptytable as $$
BEGIN
RETURN QUERY SELECT * INTO _out FROM emptytable limit 1;
END;
$$ LANGUAGE PLPGSQL;
As suggested in the comments do return setof emptytable
create function selectany()
returns setof emptytable as $$
select *
from emptytable
limit 1
;
$$ language sql;
Plain sql can do that
I created a function which writes information about table deletions.
And another function which simply adds a trigger call after delete.
But I would like to store the whole row as string into my table.
According to Postgresql Documentation it should work by adding "OLD.*" into a text based column. But it fails telling me that I try to put too many columns into this table.
OLD is from type RECORD. And i want to have it in my text field like "value1,value2,value3" or it could be "colname:value,colname2:value". I dont care, I just want to see the row which has been deleted.
Another approach can be to log all delete queries from pg_stat_activity. But I don't know how to do that. Simply accessing pg_stat_activity every second would cause too much traffic I guess.
My table is simple:
create table delete_history (date timestamp, tablename varchar(100), data text);
This is my function:
CREATE or REPLACE FUNCTION ondelete() RETURNS TRIGGER AS $$
BEGIN
INSERT INTO delete_history VALUES (CURRENT_TIMESTAMP, TG_TABLE_NAME, OLD.*);
RETURN OLD;
END;
$$ LANGUAGE plpgsql;
This is my trigger:
CREATE OR REPLACE FUNCTION history_create_triggers() RETURNS void
AS $$
DECLARE
r RECORD;
BEGIN
FOR r IN SELECT table_name FROM information_schema.tables WHERE table_schema = 'public' AND table_type='BASE TABLE' LOOP
EXECUTE 'CREATE TRIGGER log_history AFTER DELETE ON public.' || r.table_name || ' FOR EACH ROW EXECUTE PROCEDURE ondelete();';
END LOOP;
END;
$$ LANGUAGE plpgsql;
You can convert type record into text:
CREATE or REPLACE FUNCTION ondelete() RETURNS TRIGGER AS $$
BEGIN
INSERT INTO delete_history VALUES (CURRENT_TIMESTAMP, TG_TABLE_NAME, OLD::text);
RETURN OLD;
END;
$$ LANGUAGE plpgsql;
sql fiddle demo
another approach could be converting your row into JSON with row_to_json function (if you have version 9.2):
CREATE or REPLACE FUNCTION ondelete() RETURNS TRIGGER AS $$
BEGIN
INSERT INTO delete_history VALUES (CURRENT_TIMESTAMP, TG_TABLE_NAME, row_to_json(OLD));
RETURN OLD;
END;
$$ LANGUAGE plpgsql;
sql fiddle demo
Another approach can be convert your data to hstore
CREATE or REPLACE FUNCTION ondelete() RETURNS TRIGGER AS $$
BEGIN
INSERT INTO delete_history VALUES (CURRENT_TIMESTAMP, TG_TABLE_NAME, hstore(OLD));
RETURN OLD;
END;
$$ LANGUAGE plpgsql;
I can't test it now - sqlfiddle is not allowing to use hstore.