Short answer: "depends on the implementation" and
"depends on the
query".
Long answer:
* ASK should finish as soon as the first result is ready.
This
doesn't mean it's always quick; just that it will be between
~0% and
100% of the execution time of the equivalent SELECT query.
* SELECT is the baseline. Its execution time is proportional
to the
complexity of the query, the number of results obtained, and
other
factors.
* DESCRIBE is entirely implementation-dependent, but if a
node has a
particularly large set of relevant statements, it can be
very
expensive indeed.
* Bindings obtained from SELECT are used to evaluated
DESCRIBE and
CONSTRUCT (in most cases). This means that most DESCRIBE or
CONSTRUCT
queries will take longer than the SELECT query, modulo
serializing
the selected results. This is *not* true if you apply
ordering
conditions to the SELECT query; they're irrelevant for
DESCRIBE and
CONSTRUCT, so CONSTRUCT could actually come out quicker.
As you can see, it's not as simple as you think, and the
ratios vary
significantly depending on which language features you use.
Why are you trying to find this out?
-R
On 8 Jan 2008, at 9:04 PM, Ahmad Iqbal wrote:
>
> hi all,
>
> I am wondering, if anybody in this group find out the
relative
> estimated processing time of SPARQL four contstructs;
SELECT, ASK,
> DESCRIBE and CONSTRUCT. Although, I know that
processing time
> varies depending on the complexity of the query (i.e.
graph
> patterns) and size of the dataset but we can always
compare
> relative period between the two. Let's say one query
with SELECT
> construct consumes x msec then the same or similar
query with ASK,
> DESCRIBE and CONSTRUCT may consume 0.6x, 1.3x and 1.5x
msecs
> respectively.
>
> Thanks in advance.
>
> --
> Ahmad
>
>
|
