Re: [rdf4j-dev] potential iterator leak in repository handling?

[Matthew Nguyen <nguyenm9@xxxxxxx>]

I mean where does/should the data for models come from if not backed by a db? Are they calculated on the fly? As noted in a prior email (long back), we're using a version of halyard so maybe its implementation violated something?

Now that I think about it, if models pulls from repository and, iirc, default rdf4j repository is file based this could be the problem bc halyard serializes repo info to the db. 

Could this be the issue?

On Wednesday, April 5, 2023 at 11:50:03 PM EDT, Jeen Broekstra <jeen@xxxxxxxxxxxx> wrote:

On Wed, 5 Apr 2023, at 23:19, Matthew Nguyen via rdf4j-dev wrote:

It may very well be that I don't understand how models initializes itself into a queryable state (it must use the repository interface no?).  Is the intent to load all models into a memorystore prior to use? If so, is that something rdf4j should do prior to using them since it has an opinion on the approach? 

What exactly do you mean with "a queryable state"? RDF4J Models are just Java collections (like a HashSet or an ArrayList) - they are in-memory data structures, not backed by a database. You can 'query' them by getting elements from them or iterating over them (same as you can do with an ArrayList), but they don't support anything like SPARQL execution. 

Generally speaking, for a Model, there is no danger of a memory leak, because the entire thing is in memory anyway.

Jeen

_______________________________________________
rdf4j-dev mailing list
rdf4j-dev@xxxxxxxxxxx
To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/rdf4j-dev