Class Fory

python/pyfory/_fory.py:82–591

High-performance cross-language serialization framework. Fory provides blazingly-fast serialization for Python objects with support for both Python native mode and xlang mode. Xlang mode handles registered schema objects and cross-language reference metadata; Python native mode han

class Fory:
    """
    High-performance cross-language serialization framework.

    Fory provides blazingly-fast serialization for Python objects with support for
    both Python native mode and xlang mode. Xlang mode handles registered schema
    objects and cross-language reference metadata; Python native mode handles the
    broader Python object graph surface, including circular Python objects.

    In Python native mode (xlang=False), Fory can serialize all Python objects
    including dataclasses, classes with custom serialization methods, and local
    functions/classes, making it a drop-in replacement for pickle.

    In xlang mode, the default, Fory serializes objects in a format that can be
    deserialized by other Fory-supported languages (Java, Go, Rust, C++, etc.).

    Examples:
        >>> import pyfory
        >>> from dataclasses import dataclass
        >>>
        >>> @dataclass
        >>> class Person:
        ...     name: str
        ...     age: pyfory.Int32
        >>>
        >>> fory = pyfory.Fory(xlang=True)
        >>> fory.register(Person, name="example.Person")
        >>> data = fory.serialize(Person("Alice", 30))
        >>> person = fory.deserialize(data)

    See Also:
        ThreadSafeFory: Thread-safe wrapper for concurrent usage
    """

    __slots__ = (
        "config",
        "xlang",
        "compatible",
        "track_ref",
        "type_resolver",
        "write_context",
        "read_context",
        "strict",
        "buffer",
        "max_depth",
        "field_nullable",
        "policy",
    )

    def __init__(
        self,
        xlang: bool = True,
        ref: bool = False,
        strict: bool = True,
        compatible: Optional[bool] = None,
        max_depth: int = 50,
        max_type_fields: int = 512,
        max_type_meta_bytes: int = 4096,