Andre Holzner <Andre.Holzner@cern.ch> writes:
> Am I doing something wrong or is this a 'feature' ?
What's biting you is that the array slice operator uses the provided
lower bounds in the resultant array. For example:
regression=# select pcha from zdec_bhab; pcha
------------------------------------{{11,12,13},{21,22,23},{31,32,33}}
(1 row)
regression=# select array_dims(pcha) from zdec_bhab;array_dims
------------[1:3][1:3]
(1 row)
regression=# select pcha[2:2][1:3] from zdec_bhab; pcha
--------------{{21,22,23}}
(1 row)
regression=# select array_dims(pcha[2:2][1:3]) from zdec_bhab;array_dims
------------[2:2][1:3]
(1 row)
So your function receives an array with first index starting at 2,
which it's not expecting; its attempt to fetch element [1][1] is out
of bounds and produces a NULL.
Offhand this behavior seems like a misfeature: perhaps it'd be more
sensible for the extracted slice to always have index lower bounds
set to 1. But I'd like to see some discussion before changing it
(and I don't plan to touch it before 7.2 release, in any case ;-)).
Comments anyone?
regards, tom lane