Re: optimal insert

From: "George Pavlov" <gpavlov(at)mynewplace(dot)com>
To: "Dirk Jagdmann" <jagdmann(at)gmail(dot)com>
Cc: <pgsql-sql(at)postgresql(dot)org>
Subject: Re: optimal insert
Date: 2006-10-10 21:04:10
Message-ID: 8C5B026B51B6854CBE88121DBF097A864DF146@ehost010-33.exch010.intermedia.net
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-sql

And don't forget that \COPY and especially COPY are usually much faster
(and, IMHO, easier to compose/maintain) than gobs of INSERTs.

> -----Original Message-----
> From: pgsql-sql-owner(at)postgresql(dot)org
> [mailto:pgsql-sql-owner(at)postgresql(dot)org] On Behalf Of Aaron Bono
> Sent: Tuesday, October 10, 2006 1:46 PM
> To: Dirk Jagdmann
> Cc: pgsql-sql(at)postgresql(dot)org
> Subject: Re: [SQL] optimal insert
>
> On 10/8/06, Dirk Jagdmann <jagdmann(at)gmail(dot)com> wrote:
>
> Hello experts,
>
> I have a database that contains three tables:
>
> create table a (
> id serial primary key,
> ... -- some more fields not relevant for my question
> );
> create table b (
> id serial primary key,
> ... -- some more fields not relevant for my question
> );
> create table a_b (
> a int not null references a,
> b int not null references b
> );
>
> Tables a and b have already been filled with lots of
> rows. Now my
> application needs to insert the relationship of a to b
> into table a_b
> and is currently doing it with inserts like the following:
>
> insert into a_b(a,b) values(1,100);
> insert into a_b(a,b) values(1,200);
> insert into a_b(a,b) values(1,54);
> insert into a_b(a,b) values(1,4577);
>
> So for a batch of inserts the value of a stays the
> same, while for by
> arbitrary values are inserted. Now I have wondered if
> PostreSQL offers
> a smarter way to insert those values? A solution can
> contains usage of
> some plpgsql code.
>
>
> It depends on your logic. If you can write a query that
> selects out the a and b records, there is a smart way:
>
> insert into a_b(a, b)
> select 1, b.id from b where b.id in (100, 200, 54, 4577);
>
> This is not really smart because you already have the id
> values - the select may diminish your performance rather than
> help. But if the select is "smarter" then you don't even
> need to know what the b.id <http://b.id> values are:
>
> insert into a_b(a, b)
> select 1, b.id from b where b.somecolumn = 'somevalue';
>
> Hope this helps.
>
>
>
> ==================================================================
> Aaron Bono
> Aranya Software Technologies, Inc.
> http://www.aranya.com
> http://codeelixir.com
>
> ==================================================================
>
>

In response to

Responses

Browse pgsql-sql by date

  From Date Subject
Next Message Dirk Jagdmann 2006-10-10 21:12:28 Re: optimal insert
Previous Message Aaron Bono 2006-10-10 20:45:30 Re: optimal insert