After more digging I came across the GraphMaker interface in the DatasetGraphFactory. Since what I want to do is reason over a TDB graph what I am planning on doing is
1) creating a SoftDatasetGraphMap (DatasetGraph) that will discard the datasets when garbage collection needs the space. 2) create a GraphMaker that wraps a Dataset (TDB in my case) and will pull grapsh from the underlying dataset. 3) Given that I have a reasoner figure out how to apply the reasoner to a specifc graph in the TDB dataset and return the resulting InfModel. That model will be placed in the SoftDatasetGraphMap. 4) Use the SoftDatasetGraphMap to construct the Dataset that Fuseki will serve. I am hoping that this strategy will allow me to create new named graphs in TDB and when they are served by Fuseki have them be reasoned over. In addition, when the graph is no longer being used (and memory is at a premium) the graph can be discarded as the reasoner can be recreated from the underlying TDB. Thoughts? Claude On Mon, Jul 15, 2019 at 11:57 AM Andy Seaborne <a...@apache.org> wrote: > Hi Claude, > > On 13/07/2019 22:26, Claude Warren wrote: > > Greetings, > > > > I know that Fuseki serves datasets. > > > > I know that the RDFConnecton will use Fuseki as its storage. > > > > I know that the RDFConnection will create named graphs if they are in the > > dataset in the connection.loadDataset( dataset ) call. > > > > I know that if RDFConnection is talking to Fuseki it will create the > named > > graphs in the dataset on Fuseki. > > > > I have a rulesReasoner and a schema and know how to make and InfModel > from > > it and a schema. > > > > What I don't know is how to get Fuseki to create the new named graphs as > > InfModels when created via the RDFConnection. > > > > Is this possible? > > An InfModel within a dataset needs specific setup in the configuration > file and that isn't accessible via RDFConnection. > > The setup with inference uses the general linked-graph dataset > implementation so there is nothing to stop its configuration while > running. It is specific to that kind of dataset. > > RDFConnection isn't the right place - it is more like a server > administration function. Being able to upload a new, replacement > dataset configuration could be done (and avoid a server restart). Only > (?? I think - it's been a long time since Ii looked) new databases can > be created by configuration upload. > > > > > Additionally is it possible to have the DeductionsModel maintained in a > TDB > > datastore not in-memory as the documentation states? > > Don't know - sorry. > > > I figure I can put the schema in the dataset with the other named graphs > so > > that it is not in memory either. > > > > Any help would be appreciated. > > > > Claude > > > > > > > -- I like: Like Like - The likeliest place on the web <http://like-like.xenei.com> LinkedIn: http://www.linkedin.com/in/claudewarren
