Thread: EXECUTE in trigger functions.

Hi,

Is there a way of using EXECUTE in trigger functions to to do
something like:

CREATE OR REPLACE FUNCTION insert_trigger()
   RETURNS trigger AS
$BODY$
BEGIN
    EXECUTE('INSERT INTO public_partitions.table_'
    || date_part('year',NEW.eventdate)::VarChar
    || lpad(date_part('month',NEW.eventdate)::Varchar,2,'0')
    || lpad(date_part('day',NEW.eventdate)::Varchar,2,'0')
    || ' VALUES (NEW.*)');
    RETURN NULL;
END;
$BODY$
   LANGUAGE 'plpgsql'

This would obviously be very useful for partitioning - however if I
try this I get:

SQL state: 42601
Context: SQL statement "INSERT INTO public_partitions.summary_20080101
VALUES (NEW.*)"
PL/pgSQL function "insert_trigger" line 2 at EXECUTE statement

Thanks,

David.

David Brain wrote:
> Is there a way of using EXECUTE in trigger functions to to do something
> like:
>
> CREATE OR REPLACE FUNCTION insert_trigger()
>   RETURNS trigger AS
> $BODY$
> BEGIN
>     EXECUTE('INSERT INTO public_partitions.table_'
>     || date_part('year',NEW.eventdate)::VarChar
>     || lpad(date_part('month',NEW.eventdate)::Varchar,2,'0')
>     || lpad(date_part('day',NEW.eventdate)::Varchar,2,'0')
>     || ' VALUES (NEW.*)');
>     RETURN NULL;
> END;
> $BODY$
>   LANGUAGE 'plpgsql'
>
> This would obviously be very useful for partitioning - however if I try
> this I get:

Execute does execute a given string of SQL. To my knowledge there is no
way you can pass new.* to that statement. What should work is to use
prepare and bind all fields of new.* separately.

http://www.postgresql.org/docs/current/static/sql-prepare.html


--
Best regards,
Hannes Dorbath