Hello,

I'm using a build of Fuseki from the trunk (revision 1404554) and I'm having 
the following problem.

After running this:

insert data {    graph <http://192.168.6.37/graph/test> {       <x:s> <x:p> 
"FOO"    }}

I verify that the following select query works (I have the option 
tdb:unionDefaultGraph set to true) :
select *  where {  ?s ?p "FOO"}


However, the following Update script does not result in any data added into 
<http://192.168.6.37/graph/new>
insert {    graph <http://192.168.6.37/graph/new> {          ?s ?p "NEW"    }   
      }
where {         ?s ?p "FOO" }


If, however, I use a USING (or wrap the query pattern in a GRAPH) this works 
properly, adding the data to <http://192.168.6.37/graph/new>:

insert {    graph <http://192.168.6.37/graph/new> {          ?s ?p "NEW"    }   
      }
using <http://192.168.6.37/graph/test> where {         ?s ?p "FOO" }


Since I have tdb:unionDefaultGraph set to true, I would expect both INSERTs to 
behave the same way. In my application (which has a much more complicated query 
than the above minimal example), I want the WHERE clause to query against a 
union of all named graphs but I still want to be able to insert into one named 
graph (I obviously can't insert into the read-only union). Is the behavior I'm 
seeing a bug?

As a workaround to the above problem, I could use USING (although it would be 
tedious to always list all named graphs). However, when I specify the USINGs as 
above in my application, Fuseki doesn't complete, it just keeps on going (I 
stopped it after a half hour).  This is true despite the fact that if I take 
the WHERE clause from the INSERT query and put it in a select query (using FROM 
instead of USING), it returns rather quickly. Unfortunately, it is hard for me 
to give you a complete minimal example of this problem since my data and query 
are proprietary, but it doesn't make sense to me that adding a USING to the 
INSERT script should make the query take such a long time. I only have 1.5 
million triples total in my triple store, and the named graphs I specify in the 
USING contain about 150k triples, so I would expect it to run faster with 
USING! If anyone has any ideas as to the solution to this problem also, please 
let me know.


I tested the first problem in the Fuseki 0.2.4 release, and had the same 
problem. However, I didn't have the second problem in 0.2.4 (adding the USING 
doesn't make the query take too long in that version), so something must have 
changed in later versions to cause this problem.


Thank you for your help!

-Elli



>>
>> Thank you,
>> Elli
>>
>

Reply via email to