Yep, that definitely threw me the first time I encountered it.
Paul B. Anderson wrote:
> It seems that the suggestion to fill intermediate positions with NULLs
> would be preferable to the current behavior.
>
> I know of no requirement to populate arrays in sequence in any other
> language so I think other programmers would be surprised too by the
> current behavior.
>
> Paul
>
>
> Tom Lane wrote:
>> [ expanding this thread, as it now needs wider discussion ]
>>
>> "Paul B. Anderson" <paul(dot)a(at)pnlassociates(dot)com> writes:
>>
>>> Actually, I was not filling all of the arrays in sequential order. I
>>> added code to initialize them in order and the function seems to be
>>> working now. Is that a known problem?
>>>
>>
>> Well, it's a documented behavior: section 8.10.4 saith
>>
>> A stored array value can be enlarged by assigning to an element
>> adjacent to those already present, or by assigning to a slice
>> that is adjacent to or overlaps the data already present.
>>
>> Up to 8.2 we didn't have a lot of choice about this, because without any
>> ability to have nulls embedded in arrays, there wasn't any sane thing to
>> do with the intermediate positions if you assigned to an element not
>> adjacent to the existing range. As of 8.2 we could allow assignment to
>> arbitrary positions by filling the intermediate positions with nulls.
>> The code hasn't actually been changed to allow that, but it's something
>> we could consider doing now.
>>
>> Comments?
>>
>> regards, tom lane
>>
>> ---------------------------(end of broadcast)---------------------------
>> TIP 9: In versions below 8.0, the planner will ignore your desire to
>> choose an index scan if your joining column's datatypes do not
>> match
>>
>> .
>>
>>
--
erik jones <erik(at)myemma(dot)com>
software development
emma(r)