Comments on: Non-execution plans

By: Jonathan Lewis

Jonathan Lewis — Sat, 02 May 2009 06:48:20 +0000

Coskan,
Could you expand you comment about “not working” – what symptoms do you see ?

By: coskan

coskan — Fri, 01 May 2009 11:01:50 +0000

On 11G tablespace online offline trick is not working I think they changed the way it works

By: Conditional SQL « Oracle Scratchpad

Conditional SQL « Oracle Scratchpad — Thu, 25 Jan 2007 01:53:08 +0000

[...] just another example of a “Non-execution plan”. At run-time, when Oracle hits the filter at line 2 it decides whether or not to run the sub-plan [...]

By: Partition Pruning Prevented by Functions on Predicates « The Oracle Sponge

Partition Pruning Prevented by Functions on Predicates « The Oracle Sponge — Wed, 22 Nov 2006 18:07:02 +0000

[...] What I particularly like about this demonstration is the way that it shows both regular check constraints and partition definitions acting together — as Jonathan Lewis demonstrated here the partitioning scheme acts effectively as a series of check constraints on the individual partitions that allow table access to be avoided when it can be deduced that a predicate does not resolve to any partitions at all. [...]

By: David Aldridge

David Aldridge — Tue, 31 Oct 2006 16:57:18 +0000

This seems to me to be a reasonable argument in support of not specifying “values less than (maxvalue)” unless it is absolutely required. I get a general vibe that specifying MAXVALUE is quite widespread in rolling date partitioning and it has always given me an uncomfortable feeling because of the need to split the top partition instead of just adding a new one — with the syntax being more verbose and more operations on the data dictionary being required it tends to rub me the wrong way to see it done. Now here’s a real stick to beat people with.

“Offlining” a tablespace is a nice trick to demonstrate that a segment wasn’t accessed. I used it with Oracle support once to demonstrate that fast refresh of a materlialized view can be based entirely on the MV log without any reference to the master table itself. It’s funny how absolutely convinced they were that the master table was needed until I sent them that demonstration :)

I think that it’s also worth noting that check constraints can operate in the same way as the partition range — a constraint of “id < 400″ allows the segment to not be accessed when a predicate of “id = 500″ is applied to a table scan. Demonstrable by the same offlining of a TS, of course.

By: Mihajlo Tekic

Mihajlo Tekic — Mon, 30 Oct 2006 21:17:45 +0000

Very interesting.

I just reproduce it on 8.1.7.4

–Mihajlo

By: Jonathan Lewis

Jonathan Lewis — Mon, 30 Oct 2006 19:21:09 +0000

Paul
It’s general techology that has been around in one form or another since 7.2.3 – a requirement of partition elimination in partition views. It’s simply a case of recognising new variants of the concept.

Try the query:
select * from pt_range where id > 600;
You find a start/stop key of invalid – just another example of the same thing.

By: Paul Vallee

Paul Vallee — Mon, 30 Oct 2006 19:02:18 +0000

Jonathan -

Fascinating result! How did you first come across this?

Paul