Semantics generic parent

[liamhuber]

Some refactoring, namely getting rid of the HasParent interface (which was only used in a single place, Semantic, which can simply have that as part of its interface) and breaking SemanticParent out of the Semantic inheritance tree. These made it easier to then make Semantic generic on its parent type, so that type narrowing in Node(..., Semantic["Composite"], ...) gives us helpful type narrowing.

I think mypy --strict will still complain about the typevar definitions,

ParentType = TypeVar("ParentType", bound="SemanticParent")
ChildType = TypeVar("ChildType", bound="Semantic")

This is because the bounds are generic and we leave the resolution implicitly as typing.Any. At present I'm not concerned about this because (a) let's get mypy happy before worrying about mypy --strict, and (b) because of the circular way that Semantic(Generic[ParentType]) and SemanticParent(Generic[ChildType]), I think we are anyhow actually achieving

ParentType = TypeVar("ParentType", bound="SemanticParent[Semantic[SemanticParent[...]]")
ChildType = TypeVar("ChildType", bound="Semantic[SemanticParent[Semantic[...]]")

even though we obviously can't actually write that out.

TODO: Take a swing at delaying the specification of Semantic's generic type in Node, such that we don't need explicit handling of the parent-most behaviour of Workflow, but can do something like Workflow[Node[None]], while StaticNode[Node[Composite]] continues to behave in the current way. (It might be necessary to delay this until #360 is resolved and we can directly say Workflow[Semantic[None]] and Node[Semantic[Composite]] (or similar).

[github-actions]

👈 Launch a binder notebook on branch pyiron/pyiron_workflow/semantics_generic_parent

[codacy-production]

Coverage summary from Codacy

See diff coverage on Codacy

Coverage variation	Diff coverage
✅ -0.01% (target: -1.00%)	✅ 96.30%

Coverage variation details

	Coverable lines	Covered lines	Coverage
Common ancestor commit (f0deac6)	3408	3118	91.49%
Head commit (72b87c4)	3406 (-2)	3116 (-2)	91.49% (-0.01%)

Coverage variation is the difference between the coverage for the head and common ancestor commits of the pull request branch: <coverage of head commit> - <coverage of common ancestor commit>

Diff coverage details

	Coverable lines	Covered lines	Diff coverage
Pull request (#544)	27	26	96.30%

Diff coverage is the percentage of lines that are covered by tests out of the coverable lines that the pull request added or modified: <covered lines added or modified>/<coverable lines added or modified> * 100%

See your quality gate settings    Change summary preferences

_{Codacy stopped sending the deprecated coverage status on June 5th, 2024. Learn more}

[liamhuber]

TODO: Take a swing at delaying the specification of Semantic's generic type in Node, such that we don't need explicit handling of the parent-most behaviour of Workflow, but can do something like Workflow[Node[None]], while StaticNode[Node[Composite]] continues to behave in the current way. (It might be necessary to delay this until #360 is resolved and we can directly say Workflow[Semantic[None]] and Node[Semantic[Composite]] (or similar).

None is anyhow outside the bounds of the typevar for semantic parents, so indeed let's postpone this change and hope it's even easier later.