Sure. Sorry to transgress. In this case, since the very same query
(structurally) had very different performance characteristics
depending on which string was used to match a literal, I jumped to a
conclusion. In particular, I'm assuming that the nature of OPTIONAL is
that the 'required' part constrains the second part, I didn't think
that I was inventing something like that.

I would point out that JIRA is a convenient way to pass files around;
would a JIRA not marked 'bug' have been more friendly?


On Thu, Feb 3, 2011 at 4:47 PM, Andy Seaborne
<[email protected]> wrote:
> Hi,
>
> Could we have things like slow queries on the user list to start with,
> please?  If there's an underlying issues, then move to JIRA but until there
> is an addressable issue, JIRA does not work for me for discussing query
> design.  It's quite possible in SPARQL to write very expensive queries,
>
> SELECT * { ?s ?p ?o . ?x ?y ?z }
>
> and ARQ/TDB's design means such queries will execute and finish eventually.
>  But its N^2 results on a database of N.  And N tends to be a decently large
> number. In fact, it wouldn't even use very much memory for that query,
> because of internal streaming.
>
>        Andy
>
> On 03/02/11 15:52, Benson Margulies (JIRA) wrote:
>>
>>     [
>> https://issues.apache.org/jira/browse/JENA-40?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12990142#comment-12990142
>> ]
>>
>> Benson Margulies commented on JENA-40:
>> --------------------------------------
>>
>> Changing the OPTIONAL's to UNION and adding the necessary constraint to
>> both of them also makes it run in reasonable time/
>>
>>
>>> Query that runs more or less forever
>>> ------------------------------------
>>>
>>>                 Key: JENA-40
>>>                 URL: https://issues.apache.org/jira/browse/JENA-40
>>>             Project: Jena
>>>          Issue Type: Bug
>>>          Components: TDB
>>>            Reporter: Benson Margulies
>>>         Attachments: r.rq, yk.txt
>>>
>>>
>>> Using the data set you have from me, the following query, which looks
>>> very much like a series of other queries that run quite rapidly, shows no
>>> sign of finishing.
>>
>

Reply via email to