[hrithiksaini99]

Fixes #16542

NetworkX's set_edge_attributes and set_node_attributes accept either a dict-like mapping (mapping edges/nodes to values) or any scalar value (e.g. int, float, str, etc.) applied to all edges/nodes when name is provided:

>>> import networkx as nx
>>> G = nx.path_graph(3)
>>> nx.set_edge_attributes(G, 5, "some name")
>>> G.edges[0, 1]["some name"]
5

Previously, the stubs only accepted SupportsItems or dict when name: str, causing type checkers to reject calls with scalar values and an attribute name.

This PR:

• Adds overloads for set_edge_attributes and set_node_attributes with values: object when name: str.
• Adds test cases in stubs/networkx/@tests/test_cases/check_tricky_function_params-py312.py.

[srittau]

See previous comments about tests.

[hrithiksaini99]

Removed — the test cases added by this PR are gone, and the Scalar alias is folded into the existing mapping overloads instead of adding a separate object overload. Thanks for the pointer to tests/REGRESSION.md.

[srittau]

I'm not certain what this overload does, especially since it completely overrides the previous one. object also doesn't mean "any scalar value", we need a better annotation and can then fold this overload into the previous one.

[hrithiksaini99]

Reworked. There is no longer a separate object overload: the existing values: SupportsItems[_Node, Unused] overload now reads values: SupportsItems[_Node, Unused] | Scalar, where Scalar is the bool | int | float | complex | str | bytes | np.generic alias already defined in networkx._typing (it was previously only inlined into _ScalarT's bound). That names the scalar set explicitly instead of using object, and the overload no longer shadows the previous one.

[github-actions]

According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉