Sven Willenberger <sven@xxxxxxx> writes:
> Tom Lane wrote:
>> I think we changed this in 8.0. Before 8.0, trying to assign to an
>> element of a NULL array yielded a NULL result array, but I think now
>> we let you produce a one-element array that way.
>>
> Using a 8.0 testbox I find that the arrays still need to be initialized:
[ looks at code... ] Hmm, we changed the main executor but missed plpgsql.
I think this is a bug, since plpgsql is now inconsistent with what
happens in a SQL UPDATE command:
regression=# create table foo (f1 int[]);
CREATE TABLE
regression=# insert into foo values(null);
INSERT 155342 1
regression=# update foo set f1[1] = 33;
UPDATE 1
regression=# select * from foo;
f1
------
{33}
(1 row)
regression=#
I'll see about making plpgsql behave similarly for 8.0.2.
regards, tom lane
---------------------------(end of broadcast)---------------------------
TIP 1: subscribe and unsubscribe commands go to majordomo@xxxxxxxxxxxxxx
