Re: using sql*loader from pl/sql

"Howard J. Rogers" <hjr_at_dizwell.com> wrote in message news:<419e716a$0$27991$afc38c87_at_news.optusnet.com.au>...
> Prem K Mehrotra wrote:
> > DA Morgan <damorgan_at_x.washington.edu> wrote in message news:<1100759368.620747_at_yasure>...
> >
> >>Howard J. Rogers wrote:
> >>
> >>
> >>>And where do you think this index will reside? In another text file? Or
> >>>inside the database.... and if inside the database, what is the point of
> >>>trying to keep the data itself OUT of the database?!
> >>>
> >>>Whatever...
> >>>
> >>>HJR
> >>
> >>Not to mention the concept of having a rowid that points to a particular
> >>block in a particular data file would have no meaning.
> >
> >
> >
> > Oracle has to figure out someway. How does Oracle do a query today
> > from external tables.
>
> It uses the SQL Loader API to read a text file, and parse it into
> something that looks like a table. No rowids involved.
>
> > I think it will be nice to have some indexes,
>
> Yes, we know you think that. What Daniel and I have tried to point out,
> however, is first the practical difficulties. And, rather more
> important, the logical silliness of the idea. If you want something
> inside the database, bring it inside the database. If you want to keep
> it outside the database, keep it outside the database. But what you want
> is some in, some out.
>
> If what you are really saying is: 'I'd like some way of speeding up
> access to the external data', then say that. It's quite understandable,
> and a perfectly logical request to make. But don't couch that in
> physical/design terms ("I want indexes") that make no logical sense.
>
> You can actually speed up the process of querying an external table by
> parallelizing the select, with a degree of parallelization that is
> independent of the number of text files being accessed.

Good idea, I had doverlooked that.
>
> > One can ofcourse use
> > regular tables,
>
> Precisely.
>
> > but then again I have not found many uses of one time
> > external table access.
>
> In data cleansing exercises, very common.
>
> HJR
Thanks,

Prem Received on Sat Nov 20 2004 - 07:11:59 CST