Received: from candle.pha.pa.us (root@s5-03.ppp.op.net [209.152.195.67]) by hub.org (8.9.3/8.9.3) with ESMTP id VAA19688; Wed, 2 Feb 2000 21:31:02 -0500 (EST) (envelope-from pgman@candle.pha.pa.us) Received: (from pgman@localhost) by candle.pha.pa.us (8.9.0/8.9.0) id VAA28084; Wed, 2 Feb 2000 21:08:46 -0500 (EST) From: Bruce Momjian Message-Id: <200002030208.VAA28084@candle.pha.pa.us> Subject: Re: [GENERAL] Proposed Changes to PostgreSQL In-Reply-To: <3898DA32.793A98BA@nimrod.itg.telecom.com.au> from Chris Bitmead at "Feb 3, 2000 12:30:26 pm" To: chris@bitmead.com Date: Wed, 2 Feb 2000 21:08:46 -0500 (EST) CC: "pgsql-hackers@postgreSQL.org" , pgsql-sql@postgresql.org, pgsql-general@postgresql.org X-Mailer: ELM [version 2.4ME+ PL66 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit > THESE CHANGES DON'T AFFECT YOU IF YOU DON'T USE INHERITANCE. > > Speak now about these changes or please, forever hold your peace. Of > course you can comment later if I screw up implementation. > > The proposed changes are.... > > 1) An imaginary field in every tuple that tells you the class it came > from. > This is useful when you select from table* and want to know which > relation the object actually came from. It wouldn't be stored on disk, > and like oid it wouldn't be displayed when you do SELECT *. The field > would be called classname. So you could have... > SELECT p.classname, p.name FROM person p; > person | Fred > student | Bill > employee | Jim > person | Chris So the field is created on the fly to show what table it came from. Seems like a good idea, though implementing another usually-invisible column will be tough. However, because it is not really a column like the oid is a column, it should be ok. Of course, internally it is relid. > 2) Changing the sense of the default for getting inherited tuples. > Currently you only get inherited tuples if you specify "tablename*". > This would be changed so that you get all sub-class tuples too by > default unless you specify "ONLY tablename". There are several > rationale for this. Firstly this is what Illustra/Informix have > implemented. Secondly, I believe it is more logical from an OO > perspective as well as giving a more useful default. If a politician > IS a person and I say SELECT * from person, then logically I should > see all the politicians because they are people too (so they claim > :). Thirdly, there are a whole range of SQL statements that should > probably be disallowed without including sub-classes. e.g. an ALTER > TABLE ADD COLUMN that does not include sub-classes is almost certainly > undesirable. It seems ashame to have to resort to non-standard SQL > with the "*" syntax in this case when it is really your only > choice. Basicly, wanting ONLY a classname is a far more unusual > choice, and leaving off the "*" is a common error. Fourthly, it seems > out of character for the SQL language to have this single character > operator. The SQL style is to use wordy descriptions of the operators > meaning. "ONLY" fits well here because it describes its own meaning > perfectly whereas to the unitiated, "*" is harder to guess at. While > this change is an incompatibility I hope for those few people using > inheritance they can accept the need to move forward without > over-burden of backwards compatibility. Sounds fine to me. Just realize you are taking on a long-overdue but big job here. > > 3) The ability to return different types of rows from a SELECT. This > is to allow implementation of ODBMS functionality where a query could > be required to instantiate objects of differing types with differing > attributes. This bothers me. We return relational data, showing the same number of columns and types for every query. I don't think we want to change that, even for OO. How are you going to return that info the the client side? -- Bruce Momjian | http://www.op.net/~candle 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