Message-ID: <4795C3D7.6050209@lelarge.info>
Date: Tue, 22 Jan 2008 11:22:15 +0100
From: Guillaume Lelarge <guillaume@lelarge.info>
User-Agent: Thunderbird 2.0.0.9 (X11/20071112)
MIME-Version: 1.0
To: sad <sad@bankir.ru>
CC: pgsql-sql@postgresql.org
Subject: Re: currval() within one statement
References: <4795984E.7030904@bankir.ru>
	<20080122073637.GB21676@a-kretschmer.de>
	<4795A152.3080400@bankir.ru> <4795B3B6.5040004@lelarge.info>
	<4795C20A.30606@bankir.ru>
In-Reply-To: <4795C20A.30606@bankir.ru>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

sad wrote:
> Guillaume Lelarge wrote:
>> sad wrote:
>>> A. Kretschmer wrote:
>>>
>>>>> is it expected that the currval() changes its value between calls 
>>>>> within one statement ?
>>>
>>>> Conclusion, don't call nextval() within a TRIGGER, and insert either
>>>> nextval() for the column or omit this column.
>>>
>>> I only note that i still want to discuss the titled problem or to be
>>> given an exact pointer to documentation regarding the currval() behavior
>>> in the described situation, that i had.
>>>
>>
>> Well, your situation is weird, to say the least. currval() doesn't 
>> change the value of a sequence. Adding a trigger that calls nextval() 
>> will change the value. But you're not telling us which kind of 
>> trigger... per statement or for each row ? if it's for each row, then 
>> that's quite understandable.
> 
> my fault --  i forgot to say.
> The trigger is "for each row".
> 
> It is understandable, i agree, but it is surprising -- alternative 
> behavior (constant result off currval()) is understandable too.
> 

It isn't surprising. A "for each row" trigger will execute the trigger 
function for each individual row, which will executes nextval each 
time... so each currval will get a different value.


-- 
Guillaume.
  http://www.postgresqlfr.org
  http://dalibo.com