You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@jena.apache.org by Carsten Keßler <ca...@uni-muenster.de> on 2012/10/31 23:14:00 UTC
Fuseki: Named graphs disappear after restart
Hi,
I'm using the latest Fuseki snapshot (0.2.6) with TDB and OWL-Mini
inference (the full configuration file:
https://gist.github.com/3990276). When I upload triples through the
graph store endpoint, this works fine and the triples end up in the
named graph specified in the request. However, once I shut down and
start Fuseki again, the named graphs are gone. The triples I have
inserted are still there, though – sitting in the union graph.
Any hints on what might be going wrong here appreciated!
Thanks
Carsten
---
http://carsten.io
Re: Fuseki: Named graphs disappear after restart
Posted by Andy Seaborne <an...@apache.org>.
On 01/11/12 20:50, Carsten Keßler wrote:
> Thanks for the hint, Andy. So this basically means I cannot have an
> updatable on-disk graph with reasoning enabled? Or is it possible to
> persist at least the triples I upload on disk and only keep the
> inferred ones in memory? I've been reading the docs and googling
> around for a whole afternoon and couldn't find any example of how to
> set this up.
>
> Best, Carsten
You can have inference on an existing graph - it's the fact you added
new (named) graphs, not altered an existing graph, that caused your
disappearing data.
(I hope someone with more inference experience can give more details)
Andy
Re: Fuseki: Named graphs disappear after restart
Posted by Carsten Keßler <ca...@gmail.com>.
Thanks for the hint, Andy. So this basically means I cannot have an updatable on-disk graph with reasoning enabled? Or is it possible to persist at least the triples I upload on disk and only keep the inferred ones in memory? I've been reading the docs and googling around for a whole afternoon and couldn't find any example of how to set this up.
Best,
Carsten
---
http://carsten.io
Am 01.11.2012 um 19:52 schrieb Andy Seaborne <an...@apache.org>:
> On 31/10/12 22:14, Carsten Keßler wrote:
>> Hi,
>>
>> I'm using the latest Fuseki snapshot (0.2.6) with TDB and OWL-Mini
>> inference (the full configuration file:
>> https://gist.github.com/3990276). When I upload triples through the
>> graph store endpoint, this works fine and the triples end up in the
>> named graph specified in the request. However, once I shut down and
>> start Fuseki again, the named graphs are gone. The triples I have
>> inserted are still there, though – sitting in the union graph.
>>
>> Any hints on what might be going wrong here appreciated!
>>
>> Thanks
>> Carsten
>>
>> ---
>> http://carsten.io
>
> The problem is that some of your setup is in-memory and some on disk.
>
>
> <#tdb_dataset_readwrite> rdf:type ja:RDFDataset ;
> ja:defaultGraph <#infModel> .
>
> That is an in-memory dataset with a default graph that is the infModel. By "in-memory" I mean the data structure for the dataset, the named graph map, is in-memory. One graph is persistent - the default graph indirectly
>
> ja:baseModel <#tdbGraph>;
>
> so the base graph is persistent but the inferences are held in memory.
>
> If you add named graphs, they end up in-memory. That's the default place ja:RDFDataset will create graphs.
>
> fuseki:dataset needs to be a tdb:DatasetTDB for the whole dataset to be persistent. Hooking up the (graph-based, memory-based) reasoner isn't going to work for new graphs created in a dataset.
>
> Andy
Re: Fuseki: Named graphs disappear after restart
Posted by Andy Seaborne <an...@apache.org>.
On 31/10/12 22:14, Carsten Keßler wrote:
> Hi,
>
> I'm using the latest Fuseki snapshot (0.2.6) with TDB and OWL-Mini
> inference (the full configuration file:
> https://gist.github.com/3990276). When I upload triples through the
> graph store endpoint, this works fine and the triples end up in the
> named graph specified in the request. However, once I shut down and
> start Fuseki again, the named graphs are gone. The triples I have
> inserted are still there, though – sitting in the union graph.
>
> Any hints on what might be going wrong here appreciated!
>
> Thanks
> Carsten
>
> ---
> http://carsten.io
>
The problem is that some of your setup is in-memory and some on disk.
<#tdb_dataset_readwrite> rdf:type ja:RDFDataset ;
ja:defaultGraph <#infModel> .
That is an in-memory dataset with a default graph that is the infModel.
By "in-memory" I mean the data structure for the dataset, the named
graph map, is in-memory. One graph is persistent - the default graph
indirectly
ja:baseModel <#tdbGraph>;
so the base graph is persistent but the inferences are held in memory.
If you add named graphs, they end up in-memory. That's the default
place ja:RDFDataset will create graphs.
fuseki:dataset needs to be a tdb:DatasetTDB for the whole dataset to be
persistent. Hooking up the (graph-based, memory-based) reasoner isn't
going to work for new graphs created in a dataset.
Andy