Re: The TransRelational Model: Clarifications

"Alfredo Novoa" <alfredo_at_ncs.es> wrote in message news:4199eeb0.8029546_at_news.wanadoo.es...
> On Mon, 15 Nov 2004 14:17:52 -0800, "Mikito Harakiri"
> <mikharakiri_at_iahu.com> wrote:
> >Also, is there such thing as query optimization in TRM?
>
> No, query optimization is a previous step, but there are complex
> operations (like joins), that are very straightforward with the TRM.
>
> In many cases it is rather easy to find the BEST plan for the current
> structures.

More clarification might help. Query optimization can be roughly divided into 2 components:
i. physical optimization
ii. logical optimization
The first one is essentially optimization of a single query block which is normally just a simple restrict-project-join query. It is comparatively easy. The major difficulty there is estimating sizes of intermediate relations. Does TRM eliminate alternative join order access paths, or merely makes their performance close so that the join order cost doesn't really vary that much anymore?

The second part is query transformations. Symbolic manipulation in general, and query rewrite, in particular, is really hard. One can have have multiple subqueries and inner views within a query and the search space explosion of various subquery unnesting and view merging possibilities. Granted, Date doesn't allow scalar subqueries, the task becomes somewhat less difficult. Still optimizing [complex] view merging must be a challenge that have to be addressed. Received on Tue Nov 16 2004 - 18:43:12 CET