PostgreSQL triggers and exceptions - postgresql

Question

I'm trying to get my first ever trigger and function to work, but how I throw exceptions and return data right way?
PostgreSQL 8.4.1
CREATE TABLE "SHIFTS" (
id integer NOT NULL, -- SERIAL
added timestamp without time zone DEFAULT now() NOT NULL,
starts timestamp without time zone NOT NULL,
ends timestamp without time zone NOT NULL,
employee_id integer,
modified timestamp without time zone,
status integer DEFAULT 1 NOT NULL,
billid integer,
CONSTRAINT "SHIFTS_check" CHECK ((starts < ends))
);
-- Check if given shift time overlaps with existing data
CREATE OR REPLACE FUNCTION
shift_overlaps (integer, timestamp, timestamp)
RETURNS
boolean AS $$
DECLARE
_employeeid ALIAS FOR $1;
_start ALIAS FOR $2;
_end ALIAS FOR $3;
BEGIN
SELECT
COUNT(id) AS c
FROM
"SHIFTS"
WHERE
employee_id = _employeeid AND
status = 1 AND
(
(starts BETWEEN _start AND _end)
OR
(ends BETWEEN _start AND _end)
)
;
-- Return boolean
RETURN (c > 0);
END;
$$
LANGUAGE
plpgsql
;
CREATE OR REPLACE FUNCTION
check_shift()
RETURNS trigger AS '
BEGIN
-- Bill ID is set, do not allow update
IF tg_op = "UPDATE" THEN
IF old.billid IS NOT NULL THEN
RAISE EXCEPTION "Shift is locked"
END IF;
END IF;
-- Check for overlap
IF tg_op = "INSERT" THEN
IF new.employee_id IS NOT NULL THEN
IF shift_overlaps(new.employee_id, new.starts, new.ends) THEN
RAISE EXCEPTION "Given time overlaps with shifts"
END IF;
END IF;
END IF;
-- Check for overlap
IF tg_op = "UPDATE" THEN
IF (new.employee_id IS NOT NULL) AND (new.status = 1) THEN
IF shift_overlaps(new.employee_id, new.starts, new.ends) THEN
RAISE EXCEPTION "Given time overlaps with shifts"
END IF;
END IF;
END IF;
RETURN new;
END
'
LANGUAGE
plpgsql
;
-- Shift checker trigger
CREATE TRIGGER
check_shifts
BEFORE
INSERT OR UPDATE
ON
"SHIFTS"
FOR EACH ROW EXECUTE PROCEDURE
check_shift()
;
shift_overlaps():
SQL error: ERROR: query has no destination for result data
check_shift():
SQL error: ERROR: unrecognized exception condition "Shift is locked"

Answer 1

You've got an error here:
SELECT
COUNT(id) AS c
FROM
"SHIFTS"
WHERE
employee_id = _employeeid AND
status = 1 AND
(
(starts BETWEEN _start AND _end)
OR
(ends BETWEEN _start AND _end)
)
;
Such a select in a plpgsql procedure has to be SELECT INTO... like this:
DECLARE
c INTEGER;
BEGIN
SELECT
COUNT(id)
INTO c
FROM
"SHIFTS"
WHERE
employee_id = _employeeid AND
status = 1 AND
(
(starts BETWEEN _start AND _end)
OR
(ends BETWEEN _start AND _end)
)
;
RETURN (c > 0);
END;
And here you've got to have the semicolon at the end of the line:
enter code here`RAISE EXCEPTION "Shift is locked";

Answer 2

Not sure what you're trying to find out. You're managing to raise your own exceptions, so that's good. I would expect that any error handling would be in the code that evokes this method.
If you want to do something inside the procedure, you need an EXCEPTION section:
[ <> ]
[ DECLARE
declarations ]
BEGIN
statements
EXCEPTION
WHEN condition [ OR condition ... ] THEN
handler_statements
[ WHEN condition [ OR condition ... ] THEN
handler_statements
... ]
END;
But generally I would expect you'd handle it in the calling code.

Answer 3

You have to use SELECT INTO to get a value returned by a query
DECLARE
[...]
c boolean;
SELECT
COUNT(id) INTO c
FROM
"SHIFTS"
WHERE
[...]