From ronz@ravensfield.com Tue May 22 17:35:37 2001 Return-path: Received: from carp.ravensfield.com ([209.41.227.126]) by candle.pha.pa.us (8.10.1/8.10.1) with ESMTP id f4MLZaQ17913 for ; Tue, 22 May 2001 17:35:37 -0400 (EDT) Received: from coho.ravensfield.com (coho [209.41.227.117]) by carp.ravensfield.com (Postfix) with SMTP id 5C2A9800D; Tue, 22 May 2001 16:46:38 -0500 (EST) Content-Type: text/plain; charset="iso-8859-1" From: Andrew Rawnsley Organization: Ravensfield Geographic To: Bruce Momjian Subject: Re: [GENERAL] Queries across multiple databases (was: SELECT from a table in another database). Date: Tue, 22 May 2001 17:37:25 -0400 X-Mailer: KMail [version 1.2] cc: Tom Lane References: <200105220437.f4M4bUA00539@candle.pha.pa.us> In-Reply-To: <200105220437.f4M4bUA00539@candle.pha.pa.us> MIME-Version: 1.0 Message-ID: <01052217372504.01367@coho.ravensfield.com> Content-Transfer-Encoding: 8bit Status: ORr On Tuesday 22 May 2001 12:37am, Bruce Momjian wrote: > Can you send me a little sample of SCHEMA use? Pardon if this is more long-winded or tangental than you are looking for... What may beconfusing many people (not excluding myself from time to time) is that cross-schema queries may have nothing to do with cross-database queries, which is an entirely different kettle of trout.... SCHEMAs as used by at least by Oracle and Sybase are nothing more than users/object owners (I have no experience with DB2 or Informix, or anything more exotic than that). Just off the top of my head, what would satisfy most people would be to be able to refer to objects as OWNER.OBJECT, with owner being 'within' the database (i.e. DATABASE.OWNER.OBJECT, which is how Sybase does it. Oracle has no 'database' parallel like that). Whether you do it Oracle-fashion and use the term SCHEMA for owner pretty universally or Sybase fashion and just pay lip service to the word doesn't really matter (unless there is a standards compliance issue). As to creating schemas...In Oracle you have to execute the CREATE SCHEMA AUTHORIZATION while logged in as that user before you can add objects under that user's ownership. While it seems trivial, if you have a situation where you do not want to grant a user session rights, you have to grant them session rights, log in as them, execute CREATE SCHEMA..., then revoke the session rights. Bah. A table created by user X in schema Y is also owned by user Y, and its user Y that has to have many of the object rights to create that table. In Sybase, its essentially the same except the only real use for the CREATE SCHEMA command is for compliance and to group some DDL commands together. Other than that, Sybase always refers to schemas as owners. You don't have to execute CREATE SCHEMA... to create objects - you just need the rights. I've never used it at least - the only thing I see in it is eliminating the need to type 'go' after every DDL command. As for examples from Oracle space - Here is a foreign key reference with delete triggers from a table in schema/user PROJECT to tables in schemas/users SERVICES and WEBCAL: CREATE TABLE PROJECT.tasks_users (    event_id INTEGER REFERENCES WEBCAL.tasks(event_id) ON DELETE CASCADE,    user_id VARCHAR2(25) REFERENCES SERVICES.users(user_id) ON DELETE CASCADE,    confirmed CHAR(1),    PRIMARY KEY (event_id,user_id) ); A join between tables in would be SELECT   A.SAMPLE_ID,                  A.CONCENTRATION,                  A.CASNO,                  B.PARAMETER,                  C.DESCRIPTION AS STYPE          FROM HAI.RESULTS A, SAMPLETRACK.PARAMETERS B, SAMPLETRACK.SAMPLE_TYPE C                  WHERE A.CASNO = B.CASNO AND A.SAMPLE_TYPE = B.SAMPLE_TYPE In both Oracle and Sybase, all the objects are in the same 'database' (instance in Oracle), as I assume they would be in Postgres. There is I assume a name space issue - one should be able to create a FOO.BAR and a BAR.BAR in the same database. > I may be adding it to > 7.2 inside the same code that maps temp table names to real tables. > Excellent! I see light at the end of the tunnel (I will say the Postgres maintainers are among the most solidly competent around - one never has any real doubts about the system's progress). -- Regards, Andrew Rawnsley Ravensfield Digital Resource Group, Ltd. (740) 587-0114 www.ravensfield.com From reedstrm@rice.edu Wed May 23 10:59:42 2001 Return-path: Received: from ece.rice.edu (ece.rice.edu [128.42.4.34]) by candle.pha.pa.us (8.10.1/8.10.1) with ESMTP id f4NExgQ05774 for ; Wed, 23 May 2001 10:59:42 -0400 (EDT) Received: from wallace.ece.rice.edu (wallace.ece.rice.edu [128.42.12.154]) by ece.rice.edu (Postfix) with ESMTP id A419F68A0E for ; Wed, 23 May 2001 09:59:36 -0500 (CDT) Received: from reedstrm by wallace.ece.rice.edu with local (Exim 3.22 #1 (Debian)) id 152a41-0006E5-00 for ; Wed, 23 May 2001 09:56:41 -0500 Date: Wed, 23 May 2001 09:56:41 -0500 From: "Ross J. Reedstrom" To: Bruce Momjian Subject: Re: [HACKERS] Re: [GENERAL] Re: [GENERAL] Queries across multiple databases ?(was: SELECT from a table in another database). Message-ID: <20010523095641.D23741@rice.edu> References: <004c01c0e214$ce6c4800$1001a8c0@archonet.com> <200105221131.f4MBVIc28574@candle.pha.pa.us> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.3.17i In-Reply-To: <200105221131.f4MBVIc28574@candle.pha.pa.us>; from pgman@candle.pha.pa.us on Tue, May 22, 2001 at 07:31:18AM -0400 Status: ORr Bruce - Around the first of the year, I started playing around with a schema implementation. As you may recall, my first crack at changing file storage names about a year ago was motivated by the need to avoid collisions once schema were available. Anyway, now that Vadim has removed a lot of the internal dependence on relname for keeping track of relations, using the new relfinenode many places relname used to be used, it seems to me that adding a parallel schemaname to all the data structures that use relname isn't as cumbersome as Peter might think. I hadn't tackled the relcache yet: perhaps concatenating the schemaname and relname to use as the hash key is the way to go for that. Unfortunately, all that code is now 4 months old, and on my machine at home. I has started with the rangetable entries, because hacked the parser to allow 'SELECT * FROM schemaname.tablename' was easier than 'select schemaname.tablename.fieldname FROM', since the dot function calling convention isn't allowed in the range list, while it is in the target list. I seem to recall tripping up on the query plan printer, of all things, before other, paying work pushed it aside. I'll see if I can update that code to the current tree, and send you something, if you'd like. Ross On Tue, May 22, 2001 at 07:31:18AM -0400, Bruce Momjian wrote: > > > I'm not sure whether it is quite the way to do it, but I'd have a better > > > time with things if I could span databases in a single request. Are > > > there theoretical problems with spanning databases in a single query? Is > > > it a feature of bad database design & implementation? > > > > I think the developers are planning full schema support for the relatively > > near future (possibly even 7.2, but check the archives and see what's been > > said). Although it looks easy to access a table from another database, > > things can rapidly become more complicated as you start having to deal with > > transactions, triggers, rules, constraints... > > Schema is on my radar screen for 7.2. I am waiting to do some research > in what needs to be done, but my initial idea is to use the system cache > to do namespace mapping, just like is done now for temp tables. > > -- > Bruce Momjian | http://candle.pha.pa.us > pgman@candle.pha.pa.us | (610) 853-3000 > + If your life is a hard drive, | 830 Blythe Avenue > + Christ can be your backup. | Drexel Hill, Pennsylvania 19026 > > ---------------------------(end of broadcast)--------------------------- > TIP 2: you can get off all lists at once with the unregister command > (send "unregister YourEmailAddressHere" to majordomo@postgresql.org)