Re: Column naming convention good or bad?

From: Paul Singleton <p.singleton_at_keele.ac.uk>
Date: Fri, 17 Nov 2000 15:31:51 +0000
Message-ID: <3A154F67.56076FBC_at_keele.ac.uk>


Jan-Erik Rosinowski wrote:

> >Yes, it's bad. Think about what happens when a column changes type
> >from integer to long. If you prefixed it with "int", you have to
> >change it to "lng".
> >
> >Now your views don't work, because they're looking for "int". Any
> >client software that's based on those views breaks, too.
 

> if you change a type from int to long (i.e. i32->i64) your clientsw
> will break anyway if you don't change it
 

> >On top of all that, it reveals storage-level details--one of the
> >things relational systems are supposed to hide from us.
>
> huh. :-) ever seen users joining varchars with numbers?
>
> types are not a storage detail but a domain classification.

Many programming systems support user-defined types (where "user" means "developer" :-) such that you can code client and server in terms of these types, alter the type definition (within reason/limits), recompile everything, and it all still works.

Consistent use of UDTs (even if they are only aliases for physical types) can reveal the scope of the impact of a proposed change.

Database app development really needs something like this, preferably as an open standard.

AFAIK, the open database standards which are widely implemented don't support anything like this.

So maybe Ali's company should bite the bullet and commit itself to some decent proprietary 4GL database application generator which manages types and changes for them automatically?

(rather than invent a manual system of their own :-)

Paul S. Received on Fri Nov 17 2000 - 16:31:51 CET

Original text of this message