PostgreSQL Error: set-valued function called in context that connot accept a set - postgresql

Question

I'm trying to convert a T-SQL procedure to PL/PGSQL procedure,
When i run my function i have this following error: "set-valued function called in context that cannot accept a set".
without the "RETURN QUERY" i will have an error "query has no destination for result data" that why i added it
Find below the error message and my function
create or replace function public.UP_GetCumulPerformancesParPortefeuille(strMatricule VARCHAR(20), strDevise varchar(3), dateDebut DATE) returns setof record
language plpgsql
as $$
BEGIN
DECLARE NO_PTF_V INT;
DT_CRS_V DATE;
PC_PRF_V FLOAT;
ResultF FLOAT = 0;
PreviousResult FLOAT = 0;
PreviousCPA INT = 0;
Performances public.performancestb;
PerfCumul public.perfcumultb;
curseur CURSOR FOR
SELECT NO_PTF, DT_CRS, SUM(PC_PRF * MT_DEM)
FROM (
SELECT D.NO_PTF, P.DT_CRS, P.PC_PRF
, D.MT_DEM/100 as MT_DEM
FROM public.TB_Demande D
INNER JOIN
public.performancestb P ON D.ID_CPA = P.ID_CPA
AND D.MC_UTL = strMatricule
) Q
WHERE DT_CRS >= dateDebut
GROUP BY NO_PTF, DT_CRS
ORDER BY NO_PTF, DT_CRS;
-- Chargement de l'historique des performances
--
BEGIN
INSERT INTO public.performancestb (ID_CPA, DT_CRS, PC_PRF)
SELECT ID_CPA, DT_CRS, PC_PRF FROM public.UF_GetHistoriquePerformances(strDevise);
OPEN curseur;
FETCH NEXT FROM curseur
INTO NO_PTF_V, DT_CRS_V, PC_PRF_V;
WHILE(found) loop
IF PreviousCPA = 0 OR PreviousCPA <> NO_PTF_V then
PreviousResult := 0;
PreviousCPA := NO_PTF_V;
end if;
ResultF := PreviousResult + PC_PRF_V * (PreviousResult + 100);
PreviousResult := ResultF;
INSERT INTO public.perfcumultb (NO_PTF, DT_CRS, MT_PRF)
VALUES (NO_PTF_V, DT_CRS_V, ResultF);
FETCH NEXT FROM curseur
INTO NO_PTF_V, DT_CRS_V, PC_PRF_V;
end loop;
CLOSE curseur;
Return query(SELECT NO_PTF, DT_CRS, MT_PRF, MT_PRF + 100 as MT_PRF_BSE_100
, CASE WHEN DT_CRS = FIRST_VALUE(DT_CRS) OVER (PARTITION BY NO_PTF ORDER BY NO_PTF,DT_CRS)
THEN 0
ELSE ((100 + MT_PRF) / (100 + LAG(MT_PRF, 1) OVER (PARTITION BY NO_PTF ORDER BY NO_PTF, DT_CRS))) - 1
END
* 100 AS MT_VOL
, CASE WHEN MT_PRF = FIRST_VALUE(MT_PRF) OVER (PARTITION BY NO_PTF ORDER BY NO_PTF, DT_CRS)
AND MT_PRF < 0
THEN MT_PRF
ELSE ((MT_PRF + 100) / MAX(MT_PRF + 100) OVER (PARTITION BY NO_PTF ORDER BY NO_PTF, DT_CRS
ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) - 1 )
* 100
END AS MT_MAX_DDO
FROM public.perfcumultb);
end;
END;
$$;

Answer 1

I finally found the solution by creating an intermediate table as return type instead of using setof record as return type.
Find below the solution
CREATE OR REPLACE FUNCTION public.up_getcumulperformancesparportefeuille(strmatricule character varying, strdevise character varying, datedebut date) returns setof resultfinal
LANGUAGE plpgsql
AS $function$
BEGIN
DECLARE NO_PTF_V INT;
DT_CRS_V DATE;
PC_PRF_V FLOAT;
ResultF FLOAT = 0;
PreviousResult FLOAT = 0;
PreviousCPA INT = 0;
Performances public.performancestb;
PerfCumul public.perfcumultb;
curseur CURSOR FOR
SELECT NO_PTF, DT_CRS, SUM(PC_PRF * MT_DEM)
FROM (
SELECT D.NO_PTF, P.DT_CRS, P.PC_PRF
, D.MT_DEM/100 as MT_DEM
FROM public.TB_Demande D
INNER JOIN
public.performancestb P ON D.ID_CPA = P.ID_CPA
AND D.MC_UTL = strMatricule
) Q
WHERE DT_CRS >= dateDebut
GROUP BY NO_PTF, DT_CRS
ORDER BY NO_PTF, DT_CRS;
-- Chargement de l'historique des performances
BEGIN
INSERT INTO public.performancestb (ID_CPA, DT_CRS, PC_PRF)
SELECT ID_CPA, DT_CRS, PC_PRF FROM public.UF_GetHistoriquePerformances(strDevise);
OPEN curseur;
FETCH NEXT FROM curseur
INTO NO_PTF_V, DT_CRS_V, PC_PRF_V;
WHILE(found) loop
IF PreviousCPA = 0 OR PreviousCPA <> NO_PTF_V then
PreviousResult := 0;
PreviousCPA := NO_PTF_V;
end if;
ResultF := PreviousResult + PC_PRF_V * (PreviousResult + 100);
PreviousResult := ResultF;
INSERT INTO public.perfcumultb (NO_PTF, DT_CRS, MT_PRF)
VALUES (NO_PTF_V, DT_CRS_V, ResultF);
FETCH NEXT FROM curseur
INTO NO_PTF_V, DT_CRS_V, PC_PRF_V;
end loop;
CLOSE curseur;
return query(SELECT NO_PTF, DT_CRS, MT_PRF, MT_PRF + 100 as MT_PRF_BSE_100
, CASE WHEN DT_CRS = FIRST_VALUE(DT_CRS) OVER (PARTITION BY NO_PTF ORDER BY NO_PTF,DT_CRS)
THEN 0
ELSE ((100 + MT_PRF) / (100 + LAG(MT_PRF, 1) OVER (PARTITION BY NO_PTF ORDER BY NO_PTF, DT_CRS))) - 1
END
* 100 AS MT_VOL
, CASE WHEN MT_PRF = FIRST_VALUE(MT_PRF) OVER (PARTITION BY NO_PTF ORDER BY NO_PTF, DT_CRS)
AND MT_PRF < 0
THEN MT_PRF
ELSE ((MT_PRF + 100) / MAX(MT_PRF + 100) OVER (PARTITION BY NO_PTF ORDER BY NO_PTF, DT_CRS
ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) - 1 )
* 100
END AS MT_MAX_DDO
FROM public.perfcumultb);
end;
END;
$function$;

Answer 2

General advise - when you come from MS SQL world, forget all and start with reading documentation - it has only 50 pages related to PLpgSQL. Almost everything is different in Postgres.
Postgres doesn't supports unbound queries - every query should to have some target - or you can use PERFORM statement, that throw result of query. If you want to push query's result to output, you should to use RETURN QUERY statement.
Syntax of RETURN QUERY is RETURN QUERY query. Parenthesis there are useless. But in your example is visible MS SQL pattern that is not used in Postgres. There (MSSQL) you fill temp table, and on the end you read this table. Porting this pattern to Postgres is antipattern - instead filling aux table, you should to use RETURN NEXT and return row immediately. Maybe you need it (I don't know because you use window functions in final query, but there will be significant overhead).
Instead manual fetching from cursor and manipulation with cursor, you can use FOR IN cursor statement.
Your function returns undefined recordset, because you used returns setof record. Then you should to call this function with special syntax - SELECT * FROM fx(args) as (colname typename, ...). Better to use OUT arguments, and then you don't need to specify result structure in query.
CREATE OR REPLACE FUNCTION foo(IN nrows int, OUT a int, OUT b int)
RETURNS SETOF RECORD AS $$
BEGIN
RETURN QUERY SELECT i, i+1 FROM generate_series(1,nrows) g(i);
END;
$$ LANGUAGE plpgsql;
SELECT * FROM foo(3);
┌───┬───┐
│ a │ b │
╞═══╪═══╡
│ 1 │ 2 │
│ 2 │ 3 │
│ 3 │ 4 │
└───┴───┘
(3 rows)
Please, start by reading documentation - T-SQL is really different language and environment, and moving to PLpgSQL is not intuitive.