Bingo, this COPY file did not have the datetimeval, so I added a few lines of code to convert it from the ctime-type entry that exists in the record. You would think that postgres could have output a more helpful error message, though.
Thanks a lot for the assist.
Susan
Thanks a lot for the assist.
Susan
On Fri, Mar 7, 2014 at 3:18 PM, Rob Sargent <robjsargent@xxxxxxxxx> wrote:
On 03/07/2014 04:06 PM, Adrian Klaver wrote:
On 03/07/2014 02:48 PM, Susan Cassidy wrote:Is the datetimeval always non-null?
I have another problem with a slightly different trigger. It's very
weird, because it is exactly the same as the first trigger, that now
works, except for the table name.
The error is:
ERROR: query string argument of EXECUTE is null
CONTEXT: PL/pgSQL function metric_int_insert_func() line 5 at EXECUTE
statement
The trigger is:
CREATE OR REPLACE FUNCTION metric_int_insert_func()
RETURNS TRIGGER AS $$
DECLARE insert_sql text;
BEGIN
insert_sql:='insert into metric_int_values_' ||
to_char(NEW.datetimeval,'YYYYMM') || ' values ($1.*)';
EXECUTE insert_sql using NEW;
RETURN NULL;
END;
$$
LANGUAGE plpgsql;
DROP TRIGGER IF EXISTS insert_metric_int_insert_trigger on
metric_int_values;
CREATE TRIGGER insert_metric_int_insert_trigger
BEFORE INSERT ON metric_int_values
FOR EACH ROW EXECUTE PROCEDURE metric_int_insert_func();
which is exactly the same as this one that works:
CREATE OR REPLACE FUNCTION metric_double_insert_func()
RETURNS TRIGGER AS $$
DECLARE insert_sql text;
BEGIN
insert_sql:='insert into metric_double_values_' ||
to_char(NEW.datetimeval,'YYYYMM') || ' values ($1.*)';
EXECUTE insert_sql using NEW;
RETURN NULL;
END;
$$
LANGUAGE plpgsql;
DROP TRIGGER IF EXISTS insert_metric_double_insert_trigger on
metric_double_values;
CREATE TRIGGER insert_metric_double_insert_trigger
BEFORE INSERT ON metric_double_values
FOR EACH ROW EXECUTE PROCEDURE metric_double_insert_func();
I can't seem to figure it out. I've retyped some of the lines, in case
there is a weird character somewhere, but they got there with a vi yank
and put, so that's not likely.
Anyone have any ideas?
Try dropping the function and then creating it, instead of just the create and replace. I have seen issues in the past with a stale copy of a function causing a problem.
Thanks,
Susan
--
Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general