NetNews Usenet Archive 1992 #30/NN_1992_30.iso/spool/comp/database/theory/650

Internet Message Format | 1992-12-14 | 1.6 KB

Path: sparky!uunet!mcsun!uknet!keele!news
From: [e-mail] (Paul Singleton)
Newsgroups: comp.databases.theory
Subject: Re: date -> day of week conversion
Message-ID: <[e-mail]>
Date: 14 Dec 92 16:03:11 GMT
References: <dtb.724154401@otto>
Sender: [e-mail] (UseNet News Service)
Organization: Keele University, England
Lines: 25
Nntp-Posting-Host: seq1.cc.keele.ac.uk

From article <dtb.724154401@otto>, by dtb@otto (David Bath):

> [e-mail] (Paul Singleton) writes:
...
>>Am I right in thinking that this is *not* the sort of functionality which
>>SQL was designed to express, and that in general it is not feasible to
>>define a relation by algorithm (in the way that Prolog programmers are so
>>accustomed to doing)?

> Rumor has it that one SQL based DB, Empress, running on UNIX and VMS,
> permits you to define your own functions that can be embedded in the
> sql, letting you define that function and then
>     select surname, givname
>     where    day_of_week(birthdate) = 'Saturday'
>     from     emp;

OK so I can define a function for use in a join condition, but I really
want to define a virtual table, some of whose columns are functions of
other columns.  I can see that SQL syntax doesn't accommodate this concept,
although I claim it is a reasonable thing to want to do.
----
  __   __    Paul Singleton (Mr)           JANET: [e-mail]
 |__) (__    Computer Science Dept.        other: [e-mail]
 |  .  __).  Keele University, Newcastle,    tel: [phone] x7355
             Staffs ST5 5BG, ENGLAND         fax: [phone]