Re: More pain and sufferring with Tropashko's materialized path...

From: Mikito Harakiri <mikharakiri_at_ywho.com>
Date: Thu, 18 Sep 2003 10:31:32 -0700
Message-ID: <s6mab.9$t01.116_at_news.oracle.com>


"--CELKO--" <joe.celko_at_northface.edu> wrote in message news:a264e7ea.0309180921.51b4de35_at_posting.google.com...
> >> but further research leads me to believe this isn't such a good
> idea anyway, because apparently, with MS-SQL at least, applying a UDF
> per row tends to screw up query optimisation on large trees (something
> to do with serialization). <<
>
> The UDF causes SQL Server to revert to row at a time processing rather
> than set oriented processing. It is a big hit, as you found out.

This again looks like non-issue to me. Yes, there cetainly must be a context switch every time user-defined fiunction is invoked (I'm thinking oracle here), but this is an expense that most database systems live with. It is not the user-defined function invocation itself that is a problem, but the way your query and function are written. This is what optimization is all about. If you don't wish to spent any time researching what your optimization bottleneck is, you probably would suffer with any method. Received on Thu Sep 18 2003 - 19:31:32 CEST

Original text of this message