[Mulgara-general] iTQL question regarding the relationshipbetween query size and query speed

[Paul Gearon]

On Thu, Apr 17, 2008 at 9:18 AM, David Moll <DMoll at myaperio.com> wrote:
>
>
>  > Any chance you can go to 1.2? It's definitely better.
>
>  I pulled down Rev 793 (1.2.1) and set it up.  We're also using the
>  1.6_02 JVM - is there any reason to use a 1.5 JVM over the 1.6?

Nope.

Not that it's a big deal, but which OS are you on?

>  That's just for the unit test that we set up to measure performance.  In
>  our production server, <Asset:num> is actually:
>
>  <http://datamanagement.viewpointusa.com/2007/02/23/aperio#Asset_9fa48c4b
>  -f496-4f94-bd17-a87f76f28302>
>
>  It's just that the test wasn't set up to generate a GUID for each Asset.
>  I was not aware that the full URI was treated differently than the
>  abbreviated version when no aliasing was defined.

Aliasing is the same as an XML namespace. If you don't use it, then a
URI of the form:
  <Asset:9fa48c4b-f496-4f94-bd17-a87f76f28302>
Is NOT a valid URL, but is instead a URN. However, the moment you
define 'Asset' as a namespace, then the query engine knows to expand
it whenever it sees it.

> I also assumed that
>  the 22-rdf-syntax-ns was abbreviated to rdfs, but I usually just throw
>  in shorthand because it's faster than copy/pasting URIs from various
>  places.

There are two separate namespaces for RDF and RDF-Schema. These are:
rdf: http://www.w3.org/1999/02/22-rdf-syntax-ns#
rdfs: http://www.w3.org/2000/01/rdf-schema#

I don't always understand why certain elements are in one and not the
other, but it's a standard, so I follow it. In this case, the "type"
predicate is in rdf and not rdfs.

>  We're testing against 1.1 and 1.2 now, the test was taking longer after
>  switching to 1.2 last night, but I had not set up the logging that
>  Andrae suggested.  I'm setting up tests right now to confirm prior
>  behavior and verify what I was seeing yesterday as I am not convinced I
>  didn't change too many things and compromised the validity of the
>  testing environment.

Moving from 1.1 to 1.2 certainly speeds up loading data. I can't
recall anything that would slow down queries. If you find yourself in
a position where you can compare version directly (without changing
other parameters) and the queries HAVE slowed down, then please let us
know and I'll try doing a comparison to see where the time went.

Now that you mention it, it might be a good idea to set up a standard
box that we do a standard set of queries on with each release. That
way we have a means of comparing the speed of one release to the next.
Mind you... that's not always easy to maintain when operating systems
need regular upgrading. I have an iMac that was doing a "dist" build
in just over a minute 18 months ago. After several OS and JVM updates
and one upgrade (to Leopard) it now takes 2 and half minutes!  :-(

Paul