Re: [HACKERS] [hackers]development suggestion needed

From: Don Baccus <dhogaza(at)pacifier(dot)com>
To: Tatsuo Ishii <t-ishii(at)sra(dot)co(dot)jp>, tgl(at)sss(dot)pgh(dot)pa(dot)us
Cc: xun(at)cs(dot)ucsb(dot)edu, pgsql-hackers(at)postgreSQL(dot)org
Subject: Re: [HACKERS] [hackers]development suggestion needed
Date: 2000-01-14 02:56:39
Message-ID: 3.0.1.32.20000113185639.01072980@mail.pacifier.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

At 11:42 AM 1/14/00 +0900, Tatsuo Ishii wrote:

>This is possible since PostgreSQL was born unless I misunderstand what
>you are saying...
>
>test=> create table "/tmp/t1" (i int);

Clever...

Of course, when I'm porting over thousands of lines of an Oracle
data model and tens of thousands of lines of scripting code that
refers to these tables via queries this is a very inconvenient
way to locate a particular table in a particular place. It involves
changing a lot of code...

Besides being somewhat ... baroque? :)

>BTW, it would be nice to add a "table space" concept to the create
>table statement.

I figured you felt that way!

>
>-- reserve a table space named 'foo' which is physically located under
>-- /pg/myspace. Only PostgreSQL super user can execute this command
>-- to avoid security risks.
>create table space foo as '/pg/myspace';
>
>-- create table t1 under /pg/myspace
>create table t1 (i int) with table space 'foo';

Yes, that's the Oracle-ish style of it. Of course, Oracle allows
all sorts of anal retentive features like allowing a DBA to
restrict the size of the tablespace, etc that I personally
don't care about...

Though I understand why they're important to some.

Oracle tables and indices within a single tablespace all live in
one file (if you're using filesystem rather than raw I/O), so
they also provide features which allow you to specify how big
a chunk to allocate per extent (Oracle pre-allocates to avoid
running out of disk space while you're running except in ways
that you control, and in hopes of getting contiguous chunks of
disk storage because they hope you're using previously empty
disks used only for Oracle).

Features like this don't fit well with each table/index residing
in its own file. Personally I don't have any need for them, either,
but as Postgres gets more popular (as it will as it continues to
improve) it may attract the attention of folks with traditional
DBA requirements like this.

Of course, that would require a new storage manager, one similar
in concept to what would be needed to implement raw I/O.

- Don Baccus, Portland OR <dhogaza(at)pacifier(dot)com>
Nature photos, on-line guides, Pacific Northwest
Rare Bird Alert Service and other goodies at
http://donb.photo.net.

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Bruce Momjian 2000-01-14 03:18:14 Re: [HACKERS] [hackers]development suggestion needed
Previous Message Tatsuo Ishii 2000-01-14 02:42:03 Re: [HACKERS] [hackers]development suggestion needed