Re: Oracle View Machanism - Opinions Please!!!!!!

From: Alastair Newson <anewson_at_cix.compulink.co.uk>
Date: Fri, 13 May 1994 06:36:08 GMT
Message-ID: <CpqAC8.Enq_at_cix.compulink.co.uk>


I can think of 2 dangers with views-on-views.

  1. The optimser of the chosen server may not be able to 'decompose' the query. It and ends up evaluating the whole view, before passing results into the next part of the query. This was a regular problem with Oracle v4 and v5. The v6 optimiser does a lot more to decompose the view into the overal query plan - so the order of access of tables in and outside the view may be mixed up. The work around in earlier versions was to have a parameter table that was part of the view. In that way the results could be restricted before being returned to the rest of the query.
  2. Unless a system is very well documented, and continually supported, it's quite easy for the precise 'meaning' of views to be lost. What may be a valid aggregation at one level (accessible as a view), may not be higher up. If the original view is modified its 'meaning' in the context of where its used may be correct, - but as the basis for other work may not. Views can hide the complexitly of an underlying data-model. This is double edged. It makes it easier to use the view - but the underlying assumptions may disappear into the mists of time. I've seen customers missed of a discounts cheque run because an assumption was made about the view used.

There my 0.02 worth

Alastair Received on Fri May 13 1994 - 08:36:08 CEST

Original text of this message