On 26 February 2010 14:33, Raymond O'Donnell <rod@iol.ie> wrote:
> On 26/02/2010 12:15, Thom Brown wrote:
>> On 26 February 2010 12:02, Anton Maksimenkov <anton200@gmail.com> wrote:
>>> Hi.
>>>
>>> I have a simple function.
>>> CREATE OR REPLACE FUNCTION myf_convert_phone18digits(
>>> in_phone VARCHAR
>>> ) RETURNS BIGINT
>>> -- IMMUTABLE
>>> AS $$
>>> DECLARE
>>> t_extent_len BIGINT;
>>> t_phone_18 BIGINT;
>>> t_multiplier BIGINT;
>>> BEGIN
>>>
>>> IF in_phone IS NULL OR in_phone = '' THEN
>>> RAISE EXCEPTION 'in_phone[%] IS NULL OR =''''!', in_phone;
>>> END IF;
>>>
>>> t_extent_len := 18 - length(in_phone);
>>> t_multiplier := 10::BIGINT ^ t_extent_len::BIGINT; --<<< ERROR HERE
>>> t_phone_18 := (in_phone)::BIGINT;
>>> t_phone_18 := t_phone_18 * t_multiplier;
>>>
>>> RETURN t_phone_18;
>>>
>>> END;
>>> $$ LANGUAGE plpgsql;
>>>
>>>
>>> When I try it in pgAdmin, first time it show me error:
>>> --
>>> ERROR: value out of range: underflow
>>> CONTEXT: PL/pgSQL function "myf_convert_phone18digits" line 12 at assignment
>>> --
>>> SECOND (and consequences) time it works and just show result 771506000000000000
>>>
>>> I get same problem when I try to call query from web (php).
>>> I can't do "second time" here, so web always fail with "ERROR: value
>>> out of range: underflow"
>>>
>>> What's the problem? What can I do with it?
>>> --
>>> antonvm
>>>
>>
>> If t_extent_len is 19 or more, it will fail as it will exceed the
>> maximum range of bigint.
>
> I'm curious as to why it worked the second time, when invoked twice in a
> row from pgAdmin (assuming that it was called with the same argument -
> the OP didn't say, but I'd imagine that was the case).
>
I'm also curious to know why there's an underflow error instead of
overflow. And in fact, even if a 19 digit phone number were passed
in, it would only result in a 10^-1 calculation, resulting in 0.1.
I think we need to know what value is being passed in to generate this error.
Thom