How do I rewrite a query from SQL Server to PostgeSQL? - sql-server

Question

My T-SQL code is:
CREATE PROCEDURE [dbo].[test_mssql]
(#TaskId int = NULL,
#RootTaskId int = NULL,
#SessionUserId int,
#XmlParam xml = NULL,
#OnlyOneRow bit = 0,
#DataType varchar (20) = 'GetChilds')
AS
BEGIN
CREATE TABLE #t (PID int, Id int, UN varchar(500))
INSERT INTO #t (PID, Id, UN)
SELECT osu.PID, osu.Id, u.FN
FROM Users u WITH (NOLOCK)
JOIN uosu WITH (NOLOCK) ON u.UID = uosu.UID
JOIN osu WITH (NOLOCK) ON uosu.OSUID = osu.Id
WHERE u.IsFired_2 = 0
AND uosu.IsPrimary = 1
BEGIN
SELECT PID, Id, UN
FROM #t
END
END
I tried to rewrite it to plpgsql:
CREATE OR REPLACE FUNCTION test_plpgsql(
IN TaskId integer = null,
IN RootTaskId integer = null,
IN SessionUserId integer = null,
IN XmlParam xml = null,
IN OnlyOneRow boolean = false,
IN DataType varchar(20) = 'GetChilds'
) RETURNS TABLE (PID integer, Id integer, UN varchar(500))
AS $$
BEGIN
CREATE TEMPORARY TABLE t (PID integer, Id integer, UN varchar(500));
INSERT INTO t (PID, Id, UN)
SELECT osu.PID, osu.Id, u.FN
FROM Users u
JOIN uosu ON u.UID = uosu.UID
JOIN osu ON uosu.OSUID = osu.Id
WHERE u.IsFired_2 = 0
AND uosu.IsPrimary = 1;
RETURN QUERY SELECT PID, Id, UN FROM t;
END;
$$ LANGUAGE plpgsql;
But I have an error:
POSITION: 178; SQL: ;
---> Npgsql.PostgresException: 42601: syntax error at or near "$1"
Is the error in the plpgsql script I wrote, or in the Npgsql data provider?

Answer 1

Documentation is your friend :-). More in this case, because PL/pgSQL is very different than T-SQL. Good start is forgot all what you know about T-SQL. On second hand - stored procedures in Postgres (PL/pgSQL) are very similar to Oracle (PL/SQL). PL/pgSQL is almost simplified version of PL/SQL (although some part are very different still).
You should to use RETURN QUERY command directly:
CREATE OR REPLACE FUNCTION test_plpgsql(
IN TaskId integer = null,
IN RootTaskId integer = null,
IN SessionUserId integer = null,
IN XmlParam xml = null,
IN OnlyOneRow boolean = false,
IN DataType varchar(20) = 'GetChilds'
) RETURNS TABLE (PID integer, Id integer, UN varchar(500))
AS $$
BEGIN
RETURN QUERY
SELECT osu.PID, osu.Id, u.FN
FROM Users u
JOIN uosu ON u.UID = uosu.UID
JOIN osu ON uosu.OSUID = osu.Id
WHERE u.IsFired_2 = 0
AND uosu.IsPrimary = 1;
END;
$$ LANGUAGE plpgsql;
Maybe you use some very old version of PostgreSQL. In these versions, the PL/pgSQL identifier was preferred always and everywhere. Clause `RETURNS TABLE(PID integer, ..) defines an structure of output relation (columns of output relation), but it defines local variables too (one for any output column).
The statement CREATE TEMPORARY TABLE t (PID integer, Id integer, UN varchar(500)); uses identifier PID. Next statement INSERT INTO t (PID, Id, UN) uses identifier PID too. Old versions (before PostgreSQL 9.0 (2010)) replaces identifies to arguments symbols (like $1) everywhere (without knowledge of context). Modern versions of Postgres (last 12 years) use different mechanism, and there should not be this problem. I was able to create function from your example without any problems.