Обсуждение: About updates

Hello,

postgresql 8.0.1, in a plpgsql function

To update columns' values in a table (without OID), if I ran:
1. "update table1 set col1 = ..., col2 = ... ... col100 ="

or
2.
"update table1 set col1 = "
...
"update table1 set col100 = "

way 1 only has one disk I/O, right? While way 2 is more time consuming
since there is disk I/O when a update is ran.

Thanks a lot,
Ying

am  10.03.2006, um 10:46:39 -0500 mailte Emi Lu folgendes:
> Hello,
>
> postgresql 8.0.1, in a plpgsql function
>
> To update columns' values in a table (without OID), if I ran:
> 1. "update table1 set col1 = ..., col2 = ... ... col100 ="
>
> or
> 2.
> "update table1 set col1 = "
> ...
> "update table1 set col100 = "
>
> way 1 only has one disk I/O, right? While way 2 is more time consuming
> since there is disk I/O when a update is ran.

Because of MVCC every UPDATE is practical a DELETE + INSERT.
Way1: you have only one DELETE+INSERT, Way2 one hundred, and you have
100 dead rows until the next VACUUM.


HTH, Andreas
--
Andreas Kretschmer    (Kontakt: siehe Header)
Heynitz:  035242/47215,      D1: 0160/7141639
GnuPG-ID 0x3FFF606C http://wwwkeys.de.pgp.net
 ===    Schollglas Unternehmensgruppe    ===

Thanks Andreas. That was a quick response.
So way 1 must be quicker.



>am  10.03.2006, um 10:46:39 -0500 mailte Emi Lu folgendes:
>
>
>>Hello,
>>
>>postgresql 8.0.1, in a plpgsql function
>>
>>To update columns' values in a table (without OID), if I ran:
>>1. "update table1 set col1 = ..., col2 = ... ... col100 ="
>>
>>or
>>2.
>>"update table1 set col1 = "
>>...
>>"update table1 set col100 = "
>>
>>way 1 only has one disk I/O, right? While way 2 is more time consuming
>>since there is disk I/O when a update is ran.
>>
>>
>
>Because of MVCC every UPDATE is practical a DELETE + INSERT.
>Way1: you have only one DELETE+INSERT, Way2 one hundred, and you have
>100 dead rows until the next VACUUM.
>
>
>HTH, Andreas
>
>