PY-PYDA0003 · imperative_model_input_validation

Find model factories that manually reproduce Pydantic field validation.

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

Definition

Inspect classes derived from Pydantic or configured house model bases. Report a non-validator method only when it checks raw factory input with isinstance, raises a validation-shaped exception, and constructs the enclosing model. Field annotations, nested model types, constrained Annotated aliases, Field, ConfigDict(extra="forbid"), and Pydantic field or model validators should own this work. A thin factory that only calls model_validate is accepted.

Evidence

Each finding identifies the model and factory method. Pydantic then retains nested field paths, aggregates independent failures, and raises one structured ValidationError to the caller.

Exceptions

Boundary code may validate data that is not model input. A model validator may inspect raw input when field declarations cannot express an invariant. Validator code should raise ValueError or PydanticCustomError, not construct ValidationError directly.

Examples

Bad

A `from_table` method checks that `judgments` is a list, checks every item is a dictionary,
rejects unknown keys, and then calls `cls(...)`.

Good

judgments: list[BackendConfiguration] validates the nested collection automatically and ConfigDict(extra="forbid") rejects unknown keys. A model validator handles only a genuine cross-field invariant and raises ValueError when it fails.

References

• Cites “Pydantic documentation”, models, nested models, extra data, and model_validate. Open reference
• Cites “Pydantic documentation”, field and model validators. Open reference
• Cites “Pydantic documentation”, error handling. Open reference