Re: Derived class guesses table name even when parent specifies it

On Sat, Jul 16, 2005 at 06:34:11PM +0100, Hamish Lawson wrote:
> It looks like there are two different ends to which class inheritance
> can be put. The purpose you mention (and which I hadn't thought about)
> is modelling table inheritance. On the other hand I was using class
> inheritance to allow me to define additional methods without modifying
> the autogenerated table class. I think my proposal may moderately
> convenience the latter without affecting the former.
> 
> My proposal only changes what happens when a table name is specified
> in some class but not in a derived class: namely, that the derived
> class should use the table name specified in its parent rather than
> guess it. Since your scenario above doesn't specify ant table names,
> it would be unaffected by this change. Moreover it's my guess that the
> pattern I'm considering (A specifying a table name but B and C not
> doing so) is unlikely to be used when modelling table inheritance;
> either no class would specify a table name or all would. Instead the
> pattern in question is most likely to be used when using class
> inheritance for the purpose of API extension; and that, I suggest, is
> when you'd want the semantics that I'm proposing.

That's convincing to me.  I've implemented it (I think) and am
checking it in wnth a test case.  

js

-- 
Jacob Smullyan

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFC4TMuuqamFyFXXLIRAhdoAKC1SlfVLyrX8Ai4wua56WNRVF/oQgCeNOqz
TbjnfSM6d6BFpM4v849HVxQ=
=lUC4
-----END PGP SIGNATURE-----