I was afraid that was the answer. I am thinking that I might try to build a new reasoner model implementation that would specify 3 graphs: base truth graph, rules and derived truth graph. The graph would expose base truth, rules, and base+derived truth as endpoints for query and allow updating for base truth or rules. I think the reasoner has the ability to accept new rules it is just the fuseki interface that does not provide access to them.
Not sure when I can get to explore this idea. :( Claude On Sun, Dec 30, 2018 at 5:14 PM ajs6f <aj...@apache.org> wrote: > Claude-- > > I don't think you ever got a response on this. I'm not very expert with > the inference framework, but am I correct in understanding your question to > be: how, using only the assembler tools and Fuseki, can one update the > schema of an inference model? > > If so, I don't think that _is_ possible (I hope I'm wrong and someone > corrects me!), but some options that might be interesting: > > • Building a new dataset on-demand for schema changes. I think you should > be able to use a new assembler doc for each change: > > > https://jena.apache.org/documentation/fuseki2/fuseki-server-protocol.html#assembler-example > > (apparently we need an example there, :sigh:) and if your actual triples > are resident on disk (in TDB, say) you should be able to avoid copying them. > > • Using a Fuseki extension. I haven't looked at this much (it's pretty > new) but Andy has done a large amount of work to offer custom endpoints for > Fuseki. It might be possible to offer an endpoint to load a new schema or > do some other coarser manipulation to get what you need done. > > ajs6f > > > On Dec 6, 2018, at 9:55 AM, Claude Warren <cla...@xenei.com> wrote: > > > > Somehow I missed this response. > > > > If the schema and the data are in the same graph everything is fine, but > I > > have schema and data in separate graphs. > > > > So without putting the graphs together into a single graph is it > possible, > > via fuseki, to update the "schema" rules in an inference graph? > > > > Even if I put the schema and the data together in a union graph and run > the > > inferencer on that the updates still have to go to one or the other graph > > right? So i have to start mixing the schema and the data. > > > > Is there a way around this? If not, does anyone have any idea how hard > it > > would be to implement an inference model that would allow updates of the > > schema data. I figure if that is available I can plug that into Fuseki > > with a couple of custom hooks and do the updates. > > > > Claude > > > > On Fri, Nov 30, 2018 at 1:11 PM Andy Seaborne <a...@apache.org> wrote: > > > >> > >> > >> On 27/11/2018 16:57, Claude Warren wrote: > >>> I have a case using Fuseki. I have 2 named graphs call them "data" and > >>> "schema". Data contains all the data, schema contains all the RDFS > based > >>> triples. > >>> > >>> I can configure Fuseki so that an inference model uses an RDF Reasoner > to > >>> apply the "schmea" rules to the "data", call this graph "inf". In > >> addition > >>> I think fuseki can be configured so that any updates to "inf" are added > >> as > >>> triples to "data". > >>> > >>> My issue is that there are periodic updates to the "schema" (e.g. when > >> new > >>> a RDF class is created). My understanding is that the reasoners do not > >>> like it when you make updates to the graphs they are manipulating > without > >>> going through the reasoner itself. So is there a way to make updates > to > >>> the "schema" graph via fuseki such that the "inf" graph/reasoner will > be > >>> happy? > >> > >> What happens if schema and data are in the same graph and so the updates > >> do go via the inference engine? > >> > >>> > >>> Claude > >>> > >> > > > > > > -- > > I like: Like Like - The likeliest place on the web > > <http://like-like.xenei.com> > > LinkedIn: http://www.linkedin.com/in/claudewarren > > -- I like: Like Like - The likeliest place on the web <http://like-like.xenei.com> LinkedIn: http://www.linkedin.com/in/claudewarren
