PY-MODE0004 · manual_model_attribute_projection_count

Count structures that manually repeat fields from one model instance.

This is a deterministic rule for python. Read its implementation.

Definition

Inspect dictionary literals, sequences of key and value pairs, and keyword calls. Group direct attribute reads by their root object when each output key matches the attribute name after hyphen normalization. Report a structure that repeats at least minimum_attributes distinct fields. This is evidence that a Pydantic model, dataclass, or similar typed value already owns the schema and should provide the projection.

Evidence

Each finding records the source range, root object, distinct projected attribute count, and every repeated attribute. The rule does not require static proof of the root type because the matching keys and configured count form the conservative structural signal. The value is the number of structures repeating enough fields of one model instance.

Exceptions

Different output names, computed values involving several attributes, unpacking, positional constructor calls, and projections below the threshold are ignored. Explicit projection can remain when the target schema intentionally differs, excludes secrets, or requires a stable compatibility boundary. Prefer model_dump include or exclude controls, a typed conversion model, or one named serializer over a second handwritten field list.

Examples

Bad

A tuple manually lists `("id", definition.id)`, `("summary", definition.summary)`, and many
more matching fields before rendering them.

Good

definition.model_dump(mode="json", exclude_defaults=True) reads the model-owned schema. A four-field API payload whose names and transformations deliberately differ remains explicit.

References

• Cites “Pydantic documentation”, model serialization. Open reference
• Cites “The Python Standard Library”, dataclasses asdict. Open reference
• Cites “Refactoring”, Data Class and Extract Class