Update graph operator overloading for subclasses

[shreyasnagare]

This PR makes the set operations for rdflib.Graph subclasses (like g1 = g1 + g2) more intuitive and similar to the results of in-place set operations (like g1 += g2).

Example:

If you were to do a union of a ConjunctiveGraph (a subclass of Graph) instance with itself or any other ConjunctiveGraph instance using something like g = g + g, the instance would lose all the methods specific to the ConjunctiveGraph.

Steps to reproduce the issue:

from rdflib import Graph

class MyGraph(Graph):
    def perform_reasoning(self):
        pass

g = MyGraph()
"perform_reasoning" in dir(g)
# True

g += g
"perform_reasoning" in dir(g)
# True

g = g + g
"perform_reasoning" in dir(g)
# False

Proposed Changes

• Use type(self)() instead of Graph() as the initial return value for set operations.
• This would allow Graph to pass down similar behavior to its derived classes (unless overridden) without breaking any existing functionality.

[nicholascar]

Looks sensible to me.

The only test failure is an address assignment, so not a real one.

[nicholascar]

@shreyasnagare thinks for this PR - looks good to me. Can you please add in your "steps to reproduce" as an actual test though? So test_graph_operator.py or similar? Thanks.

[nicholascar]

@shreyasnagare thanks for that rapid turn-around! Just waiting for another review from one of my co-maintainers and then we'll likely merge.

Please do feel free to make more good updates like this!

[gtfierro]

Great to see this change, and thanks @shreyasnagare for finding and addressing it. I'd like to incorporate these changes in the brickschema package --- is there a timeline for the next release of RDFlib?

[nicholascar]

@gtfierro 6.0.0 is only a couple of days away! We merged in the JSON-LD plug-in, sorted out namespace issues and merged in heaps of smaller PRs like this. Even updated docco. Just waiting for some final CI/CD steps and we will have the package into PyPI. Again, only a few days!

[gtfierro]

Great news! Congrats on the release