[maharanay22]

Summary

Writing Pydantic model instances with a plain Enum field failed. pydantic_to_schema maps the field to the type of the enum's values (a dictionary of strings for string values), but the rows come from model_dump(), which keeps the enum members. Arrow can only convert members that also subclass str or int, so lance.write_dataset(models, uri) and LanceDataset.from_pydantic_model raised ArrowTypeError: Expected bytes, got a 'Color' object for a model like this.

class Color(enum.Enum):
    RED = "red"
    BLUE = "blue"

class Item(pydantic.BaseModel):
    color: Color

lance.write_dataset([Item(color=Color.BLUE)], "/tmp/items.lance")

model_to_dict now replaces each enum member with its value, including inside nested models and lists, which matches the schema the model maps to.

Testing

Added test_pydantic_enum_fields_store_enum_values to test_dataset.py for both write_dataset and from_pydantic_model, with a string enum, an integer enum and an enum inside a nested model list. It fails on main and passes with the fix. The existing Pydantic tests in test_dataset.py and test_pydantic.py still pass, and ruff format --check and ruff check are clean on the touched files.

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[lance-gatekeeper]

✅ Gate recommendation: approve.

The scalar-list fast path fixes the previous vector-ingestion regression. Enum values still round-trip through both write entry points, including nested models and lists, and vector ingestion is back near its previous throughput.