Error passing an array to PL/pgSQL stored procedure - php

Question

I have this procedure:
CREATE OR REPLACE FUNCTION get_saldo_conto(idUtente integer, idConto integer, categories int[], end_date date) RETURNS numeric(8,2) AS $$
DECLARE
row_transazione transazione%ROWTYPE;
saldoIniziale numeric(8,2);
totale numeric(8,2);
BEGIN
saldoIniziale = (SELECT saldo_iniziale FROM conto WHERE id = idConto AND id_utente = idUtente);
totale = 0;
FOR row_transazione IN SELECT *
FROM transazione
LEFT JOIN categoria ON id_categoria = categoria.id
WHERE id_conto = idConto
AND transazione.id_utente = idUtente
AND id_categoria = ANY (categories)
AND data <= end_date
LOOP
IF(row_transazione.tipo = 'entrata') THEN
totale = totale + row_transazione.importo;
ELSE
totale = totale - row_transazione.importo;
END IF;
END LOOP;
RETURN (saldoIniziale + totale) AS saldo_corrente;
END;
$$ LANGUAGE 'plpgsql';
When I call it with for example with
SELECT get_saldo_conto('1','19','{1,2,4,5,6}', '20/01/2015');
gives me an error
ERROR: op ANY/ALL (array) requires array on right side
Am I doing something wrong passing the array?
I tried also passing like '{1,2,4,5,6}'::int[] with no success.
CREATE TABLE transazione(
id SERIAL PRIMARY KEY,
tipo VARCHAR(7) NOT NULL CHECK(tipo IN('spesa', 'entrata')),
importo NUMERIC(8,2) NOT NULL,
descrizione VARCHAR(40),
data DATE DEFAULT CURRENT_TIMESTAMP,
id_conto INTEGER NOT NULL REFERENCES conto(id) ON UPDATE CASCADE ON DELETE CASCADE,
id_utente INTEGER NOT NULL REFERENCES utente(id) ON UPDATE CASCADE ON DELETE CASCADE,
id_categoria INTEGER REFERENCES categoria(id) ON UPDATE CASCADE ON DELETE SET NULL
);

Answer 1

Debug
You defined the row variable
row_transazione transazione%ROWTYPE;
But then you assign SELECT * FROM transazione LEFT JOIN categoriato it, which obviously does not fit the type.
The error message you display, however, does not make sense. The only case of ANY/ALL in your code looks correct. Are you sure you are calling the function you think you are calling? Investigate with:
SELECT n.nspname, p.proname
, pg_get_function_identity_arguments(p.oid) AS params
FROM pg_proc p
JOIN pg_namespace n ON n.oid = p.pronamespace
WHERE p.proname = 'get_saldo_conto';
.. to find all functions with the given name. And
SHOW search_path;
.. to check if the search_path leads to the right one.
Audited function
Your function would work like this:
CREATE OR REPLACE FUNCTION get_saldo_conto(_id_utente integer
, _id_conto integer
, _categories int[]
, _end_date date)
RETURNS numeric(8,2) AS
$func$
DECLARE
row_trans transazione;
saldoIniziale numeric(8,2) := (SELECT saldo_iniziale
FROM conto
WHERE id_utente = _id_utente
AND id = _id_conto);
totale numeric(8,2) := 0;
BEGIN
FOR row_trans IN
SELECT t.*
FROM transazione t
-- LEFT JOIN categoria ON id_categoria = categoria.id -- useless anyway
WHERE t.id_utente = _id_utente
AND t.id_conto = _id_conto
AND t.id_categoria = ANY (_categories)
AND data <= _end_date
LOOP
IF row_trans.tipo = 'entrata' THEN
totale := totale + row_trans.importo;
ELSE
totale := totale - row_trans.importo;
END IF;
END LOOP;
RETURN (saldoIniziale + totale); -- AS saldo_corrente -- no alias here!
END
$func$ LANGUAGE plpgsql;
But that's just to showcase the syntax. The function is expensive nonsense.
Superior simple query
Replace with a simple SELECT:
SELECT COALESCE((
SELECT saldo_iniziale
FROM conto
WHERE id_utente = _id_utente
AND id = _id_conto), 0)
+ COALESCE((
SELECT sum(CASE WHEN tipo = 'entrata' THEN importo ELSE 0 END)
- sum(CASE WHEN tipo = 'spesa' THEN importo ELSE 0 END)
FROM transazione
WHERE id_conto = _id_conto
AND id_utente = _id_utente
AND id_categoria = ANY (_categories)
AND data <= _end_date), 0) AS saldo;
Assuming rows in conto are unique on (id_utente,id).
Depending on implementation details the best query can vary. I chose a variant that is safe against missing rows and NULL values. Either way, a plain query should be much faster than a loop over all rows.
You can wrap this into a function (SQL or plpgsql) if you want.
Aside:
transazione.tipo should rather be an enum type - or even just "char" or boolean. varchar(7) is a waste for tag with two possible values.
And data DATE DEFAULT CURRENT_TIMESTAMP should really be data DATE DEFAULT CURRENT_DATE.