> Background: ?this is a :DataSet and it has a :records object property,
> range :Record and a :numberOfRecords computed via a spin rule. :Record has
> properties :rate (xsd:float), :time (xsd:dateTime), and :cumulative
> (xsd:float). Each :Records has :rate and :time values assigned. The :Record
> :index starts at 1 for the earliest time. The first :record (:index =1 ) has
> a :cumulative value 0.
>
> CONSTRUCT {
> ?record2 :cumulative ?cumulative2 }
> WHERE {
> ?this :numberRecords ?N
> ?j tops:for (2 N ) .
> LET ( ?i := ?j -1)
> ?record1 :index ?i .
> ?record1 ^:records ?this.
> ?record1 :time ?t1 .
> ?record1 :cumulative ?cumulative1
> ?record2 :index ?j .
> ?record2 ^:records ?this.
> ?record2 :time ?t2 .
> ?record2 :rate ?rate2 .
> LET (?cumulative2 := (?cumulative1 + (?rate2 *
> datetime:daysBetween(?t1, ?t2)))) .
> }
>
> We are concerned about 2 things 1) the scalability of this kind query up to
> 10,000 records, 2) how to generalize this for arbitrary windows, and 3) the
> index could become a problem. That is why we are investigating other ways to
> do it.
Hi Arthur,
without knowing details, the above solution indeed does not look efficient,
especially because it relies on the chaining of SPIN rules in an "outer" loop.
Also you are required to store intermediate values in records, while in reality
you seem to be more interested in the end result. An alternative that should
work already is to use recursive SPIN functions like I am doing in the SPIN
Result Sets implementation (see last email). This is very geeky stuff but works
- basically the idea is to have something like
function step(n)
... FILTER (n > 0) .
... LET (?result := ?cumulative + step(n-1)
Those approaches require the definition of SPIN functions, and are limited by
potential stack overflows for very large data sets (because of the recursion at
the end adds to the stack). Doing such aggregations will become significantly
easier with the upcoming release and the SPIN Result Sets support (via
SPARQLMotion). If you start using JavaScript you could as well just drop the
use of RDF and SPARQL altogether, as you would work on entirely different data
structures that have no way back (i.e. it is not possible to call SPARQL or RDF
functions from JavaScript, so anything like a daysBetween function would need
to be coded in JS, etc etc).
A side question: your use of inverse property paths (^) is for readability
only? You could of course also write ?this :records ?record2 etc.
Holger
--
You received this message because you are subscribed to the Google Groups
"TopBraid Composer Users" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to
[email protected].
For more options, visit this group at
http://groups.google.com/group/topbraid-composer-users?hl=en.
