Serialization

How rows are (de)serialized to and from the on-disk JSON format. TinyStore never evals, execs, or pickles data — only json.loads into validated Pydantic models.

serialization

Pydantic <-> JSON row conversion with stable formatting.

TinyStore persists models as rows in JSON table files. Each row is a plain dict produced by model_dump(mode='json') so that datetime/date/UUID/Enum and nested JSON-compatible values are encoded as JSON scalars automatically by Pydantic. Relationship fields are never persisted.

dumps

dumps(obj: Any) -> str

Serialize with stable, human-readable formatting.

model_to_row

model_to_row(
    instance: BaseModel,
    *,
    exclude: frozenset[str] = frozenset(),
) -> dict[str, Any]

Convert a model instance to a persistable row dict.

exclude is the set of relationship field names to drop. The internal optimistic-concurrency version is attached under __version.

row_to_model

row_to_model(
    model_cls: type[T],
    row: dict[str, Any],
    *,
    exclude: frozenset[str] = frozenset(),
) -> T

Convert a stored row dict back into a validated model instance.

Internal fields (__version) are stripped before validation. Relationship fields are excluded (left as their default None).