pg.ddx.io  pgsql-sql@postgresql.org mailing list archive  
help / color / mirror / Atom feed
update by one transaction
4+ messages / 3 participants
[nested] [flat]

* update by one transaction
@ 1998-06-11 08:16  Lendvary Gyorgy <gyurika@prolan.hu>
  0 siblings, 1 reply; 4+ messages in thread

From: Lendvary Gyorgy @ 1998-06-11 08:16 UTC (permalink / raw)
  To: pgsql-sql

Hello,

I have tried to write my problem, but I think nobody couldn,t understand
it (because of my bad English, I think).

I have a table for instance boci (row_number int, col2 float, col3
float). The boci table has about 10000 rows. 
I want to update col3 column of the table in every row with different
values. Desired values are in an array has 10000 elements. I want to be
the value of the col3 column in the first row equal to the first element
of the array and so on. 
I've done it with a 'for' cycle, but it is very slow. I want to do
theese updates with one transaction. 

Here is a little piece of my program:

for (i=0; i<10000; i++)
{
    sprintf(buff, "UPDATE boci SET col3 = %f WHERE row_number=%d",
array[i], i);
    PQexec(conn, buff);
}
PQexec(conn, "COMMIT");

This program is very, very slow. Is there any way making this program
much faster (for instance with CURSOR or 'block write' or something
else)? Please write me a little program that describes your ideas!

Thanks for your help in advance!



^ permalink  raw  reply  [nested|flat] 4+ messages in thread

* Re: [SQL] update by one transaction
@ 1998-06-11 08:54  Aleksey Dashevsky <postgres@luckynet.co.il>
  parent: Lendvary Gyorgy <gyurika@prolan.hu>
  0 siblings, 0 replies; 4+ messages in thread

From: Aleksey Dashevsky @ 1998-06-11 08:54 UTC (permalink / raw)
  To: Lendvary Gyorgy <gyurika@prolan.hu>; +Cc: pgsql-sql



On Thu, 11 Jun 1998, Lendvary Gyorgy wrote:

> Hello,
> 
> I have tried to write my problem, but I think nobody couldn,t understand
> it (because of my bad English, I think).
> 
> I have a table for instance boci (row_number int, col2 float, col3
> float). The boci table has about 10000 rows. 
> I want to update col3 column of the table in every row with different
> values. Desired values are in an array has 10000 elements. I want to be
> the value of the col3 column in the first row equal to the first element
> of the array and so on. 
> I've done it with a 'for' cycle, but it is very slow. I want to do
> theese updates with one transaction. 
> 
> Here is a little piece of my program:
> 
> for (i=0; i<10000; i++)
> {
>     sprintf(buff, "UPDATE boci SET col3 = %f WHERE row_number=%d",
> array[i], i);
>     PQexec(conn, buff);
> }
> PQexec(conn, "COMMIT");
> 
> This program is very, very slow. Is there any way making this program
> much faster (for instance with CURSOR or 'block write' or something
> else)? Please write me a little program that describes your ideas!
> 
> Thanks for your help in advance!
> 
It seems you simply forgot to start transaction before the loop. So you
need to prepend this code with begin statement:

PQexec(conn, "BEGIN");

Hope, this will help!

Aleksey.




^ permalink  raw  reply  [nested|flat] 4+ messages in thread

* update by one transaction
@ 1998-06-11 13:26  Lendvary Gyorgy <gyurika@prolan.hu>
  0 siblings, 1 reply; 4+ messages in thread

From: Lendvary Gyorgy @ 1998-06-11 13:26 UTC (permalink / raw)
  To: pgsql-sql

BEGIN isn't the solution, just I forgot to write it. So I repeat my
problem again:
 
I have a table for instance boci (row_number int, col2 float, col3
float). The boci table has about 10000 rows. 
I want to update col3 column of the table in every row with different
values. Desired values are in an array has 10000 elements. I want to be
the value of the col3 column in the first row equal to the first element
of the array and so on. 
I've done it with a 'for' cycle, but it is very slow. I want to do
theese updates with one transaction. 

Here is a little piece of my program:

PQexec(conn, "BEGIN");
for (i=0; i<10000; i++)
{
    sprintf(buff, "UPDATE boci SET col3 = %f WHERE row_number=%d",
array[i], i);
    PQexec(conn, buff);
}
PQexec(conn, "END");

This program is very, very slow. Is there any way making this program
much faster (for instance with CURSOR or 'block write' or something
else)? Please write me a little program that describes your ideas!

Thanks for your help in advance!

Please help me!!!



^ permalink  raw  reply  [nested|flat] 4+ messages in thread

* Re: [SQL] update by one transaction
@ 1998-06-11 17:06  dave madden <dhm@paradigm.webvision.com>
  parent: Lendvary Gyorgy <gyurika@prolan.hu>
  0 siblings, 0 replies; 4+ messages in thread

From: dave madden @ 1998-06-11 17:06 UTC (permalink / raw)
  To: gyurika@prolan.hu; +Cc: pgsql-sql

 =>From: Lendvary Gyorgy <gyurika@prolan.hu>
 =>...
 =>PQexec(conn, "BEGIN");
 =>for (i=0; i<10000; i++)
 =>{
 =>    sprintf(buff, "UPDATE boci SET col3 = %f WHERE row_number=%d",
 =>array[i], i);
 =>    PQexec(conn, buff);
 =>}
 =>PQexec(conn, "END");
 =>
 =>This program is very, very slow.

The first thing I'd check is that you have an index on
boci(row_number).  If you do, or if adding one doesn't help, then
perhaps you could split boci into two tables:

		boci1( row_number int, col2 float )
		boci2( row_number int, col3 float )

Then you could use the COPY command to load data from your array.  To
select things, just join boci1 and boci2 on row_number (or create a
view, if PostgreSQL supports 'em).

d.



^ permalink  raw  reply  [nested|flat] 4+ messages in thread


end of thread, other threads:[~1998-06-11 17:06 UTC | newest]

Thread overview: 4+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
1998-06-11 08:16 update by one transaction Lendvary Gyorgy <gyurika@prolan.hu>
1998-06-11 08:54 ` Aleksey Dashevsky <postgres@luckynet.co.il>
1998-06-11 13:26 update by one transaction Lendvary Gyorgy <gyurika@prolan.hu>
1998-06-11 17:06 ` dave madden <dhm@paradigm.webvision.com>

This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox