Home -> Usenet -> comp.databases.theory -> Re: Type-free Circles and Ellipses [T]

Re: Type-free Circles and Ellipses [T]

From: Owen Rees <owenrees_at_waitrose.deletethis.com>
Date: Fri, 31 Aug 2001 00:51:27 +0100
Message-ID: <1tbtot027vurc13fvjd698rh95a66tcmhv_at_4ax.com> 







On Thu, 30 Aug 2001 10:05:31 GMT, dmitry_at_elros.cbb-automation.de
(Dmitry A. Kazakov) wrote:



>If you are arguing against the limitations of the subtype relations

>usual to the modern OO languages, I am with you. There is nothing

>fundamental that requires the subtype relation be a tree.




In most of the useful theory that I have seen, the subtype relation is
described as forming a lattice. Of course, for that to work, you have
to be able to think about types that have never been written down, but
which are derived as meets or joins of other types. Of the types you
start with, some are written down explicitly, and others are inferred
from various language constructs, and the types used in them.



In the language we invented when I was working on the ANSA
computational model, expressions and variables had different kinds of
types. The two kinds of types had conformance relations which each
formed semi-lattices, but opposite ways up. The other thing we did was
to ban the use of the word 'subtype' in our reports and discussions;
we did that because the emotive connotations, the confusion with
inheritance and the potential to be confused about which way was 'up'
made it a liability.



Ever since then I have found the type systems in languages that
purport to be OO deeply unsatisfying.


-- 
Owen Rees - opinions expressed here are mine; for the full disclaimer
visit <http://www.users.waitrose.com/~owenrees/index.html#disclaimer>


Received on Fri Aug 31 2001 - 01:51:27 CEST












Original text of this message