ALL-DUPL0001 · repeated_class_method_count

Count exact sibling methods that can be pulled into their shared direct base.

This is a deterministic rule for all languages. Read its implementation.

Definition

Parse the configured project sources and compare methods declared directly in classes sharing one meaningful direct base. Normalize each complete method abstract syntax tree by removing its leading docstring and source positions. Names, signatures, decorators, annotations, and executable statements must still match exactly. The value counts definitions beyond the first stable path and line in every qualified group. Module functions, nested functions, and classes without a shared ownership boundary are not compared.

Evidence

Every extra definition receives a finding at its own location. Each finding cites every definition in the group, including the first, and measures affected definitions and files. Every reported group already has one shared direct base, so the finding identifies that concrete pull-up boundary rather than proposing a speculative new abstraction. The value is the number of definitions past the first in every qualified group.

Exceptions

Docstring-only bodies, pass, ellipsis, raise NotImplementedError, and return NotImplemented are placeholders rather than duplicated implementations. Unrelated classes are excluded because an extraction owner cannot be proven deterministically. Provider selection decides whether tests and vendored paths are in scope because thin adapters and protocol doubles often repeat intentionally. Comments and formatting are absent from Python syntax trees. Similar methods with different names, signatures, decorators, annotations, or statements remain separate.

Examples

Bad

`PromptCategory.key` and `JudgmentCriterion.key` both use `@property`, accept only `self`, and
return `normalize_name(self.name)`. Different human docstrings do not hide the repeated
implementation. If both classes directly inherit `KeyedStrEnum`, the finding can suggest
moving the shared contract there.

Good

Two protocol methods whose bodies are ... are ignored. Two unrelated analyzer result methods and two identical test-backend properties are also ignored because neither establishes a production pull-up boundary.

References

• Cites “Refactoring”, Pull Up Method. Open reference
• Cites “The Python Standard Library”, ast, abstract syntax trees. Open reference
• Cites “The Python Standard Library”, ast.get_docstring. Open reference