How to validate before update without having the full graph?

[vscaemh]

I have SHACL constraints that requires certain relation between entities. I want to ensure that my graph still adhere to these constraints after adding new knowledge. But some data is in a triple store, and some data is "local", meaning that it has not been added yet.
How do I make a sort of "union" graph between the new data and A SUBSET of the existing data and check my constraints without fetching the entire dataset from my triple store?

Here is an attempt of a better formulation
I have a triple store T.
I have some shapes S that requires the class X to be owned by the class Y.

An instance of Y, Y_inst already exists in T.

Now I want to add an instance of X, X_inst to my triple store.

But it fails the validation before update, because Y_inst is unknown, as it is only in the triple store.

How can I go about this?

It seems to be a relevant research topic e.g. https://ceur-ws.org/Vol-3954/paper1130.pdf
but I guess there is some good patterns that I just don't know about?

[vscaemh]

I guess it could be done with a staging graph? Still concerned that it might take very long, when my graph becomes big.
But perhaps I should split my graph into much smaller graphs and then relate across graphs?

[lisat-dstg]

This is not really an rdflib problem. It's a problem in your data.

You said you want to enforce that X is owned by Y. What you probably mean is that you want any instance of X to be owned by an instance of Y. In this case you'll design a shape constraint that says ex:owner property must exist on instances of X and the range of ex:owner is restricted to Y. So far so good. Then you'll have data like :x1 a ex:X ; ex:ownedBy y1 .. This will be invalid because SHACL doesn't know the type of y1 so you'll have to declare y1 in your data too. :y1 a ex:Y .. You don't have to say anything else about y1, just its type. The triple store will handle all the rest.

[vscaemh]

Hi @lisat-dstg.
Thank you for answering, even though it does not fit the scope, highly appreciated!

It is not part of rdflib, I was just hoping to get some resources on how to get started, and rdflib seemed like a place where people might know how to use rdf in a practical fashion.

My question also concerns the process.
For example:
If you add shapes to apache jena fuseki and then upload data that does not conform to the shapes, then apache jena fuseki happy inserts it. Afterwards you can validate it and be told that it violates the shapes.
I would really rather have that it rejects the data if it does not conform to the shapes.

Otherwise you end up with an invalid knowledge base.

But I can see that e.g. RDF4J checks against shapes before inserting data into the graph.

So maybe it is just a vendor thing?