pathwaycom/pathway · error · RuntimeError

output schema validation error, received {output.as_typehint

Error message

output schema validation error, received {output.as_typehints()} vs expected {cls.output_schema.typehints()}

What it means

When a @pw.row_transformer class declares an explicit output= schema, Pathway validates that the schema generated from the transformer's output attributes is a subschema of that declared output. If an output attribute's dtype or name does not fit the declared output schema, this RuntimeError is raised (note: received/expected wording — the transformer's attributes must be assignable to `output`).

Source

Thrown at python/pathway/internals/row_transformer.py:173

        """Pseudo-random hash of its argument. Produces pointer types. Applied value-wise."""
        return ref_scalar(*args, optional=optional)

    def __init_subclass__(
        cls,
        input=Any,
        output=Any,
    ):
        cls._attributes = {
            attr.name: attr for attr in attrs_of_type(cls, AbstractAttribute)
        }
        cls.input_schema = input
        cls.output_schema = schema_from_types(
            **{attr.output_name: attr.dtype for attr in cls._output_attributes.values()}
        )
        if output is not Any and not schema.is_subschema(cls.output_schema, output):
            print(output)
            print(cls.output_schema)
            raise RuntimeError(
                f"output schema validation error, received {output.as_typehints()} vs expected {cls.output_schema.typehints()}"  # noqa
            )
        for attr in cls._attributes.values():
            attr.class_arg = cls


class AbstractAttribute(ABC):
    is_method = False
    is_output = False
    _dtype: dt.DType | None = None
    class_arg: ClassArgMeta  # lateinit by parent ClassArg

    def __init__(self, **params) -> None:
        super().__init__()
        self.params = params
        self.name = self.params.get("name", None)
        if "dtype" in self.params:
            self._dtype = dt.wrap(self.params["dtype"])

View on GitHub (pinned to fa2f74a464)

Solutions

Align the declared output schema with the transformer's output attributes: same column names, and attribute dtypes must be subtypes of the declared field types.
Reorder/fix type annotations on output attributes in the ClassArg so each matches the corresponding field of output=.
As a last resort omit output= and let the schema be derived from the attributes (only if downstream does not require the exact type).

Example fix

# before
@pw.row_transformer(input=In, output=Out)  # Out.value: float
class T:
    class output(pw.ClassArg):
        value: int = pw.input_method()  # mismatch

# after
@pw.row_transformer(input=In, output=Out2)  # Out2.value: int
class T:
    class output(pw.ClassArg):
        value: int = pw.input_method()

Defensive patterns

Strategy: validation

Validate before calling

import pathway as pw

def output_matches(derived: type[pw.Schema], declared: type[pw.Schema]) -> bool:
    return pw.Schema.is_subschema(derived, declared) if hasattr(pw.Schema, 'is_subschema') else declared.is_subschema_of(derived) if False else __import__('pathway.internals.schema', fromlist=['schema']).schema.is_subschema(derived, declared)

Prevention

Unit-test each row transformer against its declared output schema right after defining it.
Regenerate the declared output schema from the transformer attributes when you change either side.

When it happens

Trigger: Declaring @pw.row_transformer(input=InputSchema, output=OutputSchema) where an output attribute's dtype (e.g. int attribute vs float field in OutputSchema, or a missing/extra column) makes schema.is_subschema(cls.output_schema, output) false.

Common situations: Editing the transformer's output attributes without updating the declared output schema (or vice versa); version upgrades where attribute dtype inference changed (e.g. int vs float).

Related errors

AI-assisted analysis of pathwaycom/pathway@fa2f74a464 (2026-08-15). Data as JSON: /api/errors/734abc48a494c9d6. Report an issue: GitHub.