Andreas,
The point of this is to maintain the column position. I don't think that
an alter of a column type should move the column position. It may be
that programmers should not rely on this, but it happens, and in very
large projects. If we can avoid unexpected side-affects like moving the
columns position, then I think we should.
Dave
On Wed, 2003-11-19 at 12:30, Andreas Pflug wrote:
> Christopher Kings-Lynne wrote:
>
> >
> >>> Will adding the logical attribute number break all of the external
> >>> tools? pg_dump, etc are all dependent on attnum now?
> >>>
> >>> Would it be possible to keep the meaning of attnum the same externally
> >>> and add another column internally to represent the physical number?
> >>
> >>
> >>
> >> Interesting idea. It would require a lot of code renaming in the
> >> backend, but it could be done.
> >
> >
> > Given that the interfaces comprise pgadmin, phppgadmin, jdbc and odbc
> > and all the main developers for those lists read all these posts, I
> > think the massive amount of effort to maintain the external interface
> > isn't worth it.
> >
> > I can vouch that it would take me exactly 2 minutes to add support for
> > attlognums in phpPgAdmin.
>
> Lesson 1 in "SQL for beginners" says "Don't use SELECT * if you rely on
> the order of columns". This discussion is about fixing a problem that
> only inexperienced programmers have. It's like an extra set of wheels on
> top of your car, just in case you drive wrong way...
>
> What's happening if you simply delete a column? Ordering 1,2,3,5? Insert
> another column of the same name, as a previously deleted, will it get
> the old position number? And so on. IMHO, way too much effort for
> working around situations that should be avoided anyway.
>
> Regards,
> Andreas
>
>
>