At 05:43 PM 12/4/2007, Ann W. Harrison wrote:

>Doug Chamberlin wrote:
>
> >
> > Ah, so the problem arises when un-quoted and quoted usage is both used
> over
> > time. Why would anyone do that or why would a tool do that?
>
>Why is a hard question to answer. If you get your table list by
>parsing the output of SHOW TABLES, there's not a lot of place for
>the database to put flags saying "this is pretty printed for the
>user, but is not really a delimited identifier." You're reduced
>to the rules that Martijn described ... and that means you quote
>anything you find with lower case letters.
>
>Why would any tool parse the output of SHOW TABLES ... well, maybe
>because that's as close to a universal metadata interface as exists.
>And the Information Schema (which should be that interface) doesn't
>offer any help on delimited vs. preserved either.

Thanks for clarifying, Ann!

I would think the SHOW TABLES output would, itself, quote the identifiers
that were originally created as quoted identifiers.

> >
> > I always thought if the database was created using un-quoted style then
> you
> > needed to continue to use un-quoted style. Similarly you need to be
> > consistent when using quoted style. Anything else would be problematic. So
> > changing Firebird to preserve case would not exacerbate the problem at all.
>
>You can mix regular and delimited identifiers in a statement. It's not
>a database property at all.

Of course, in one statement you can mix them if dealing with objects that
were created in a mixed way. But mixing usage for the same table name or
field name in one statement? I sure would not do that!