pathwaycom/pathway · error · ValueError

The value column must have one of the following types: {allo

Error message

The value column must have one of the following types: {allowed_value_types}

What it means

Raised by MessageQueueOutputFormat.build for raw/plaintext output when the value column's dtype is not in allowed_value_types (default (BYTES, STR, ANY)). Like keys, raw payloads are serialized as bytes, so only byte-like or string columns are accepted by default; the error names the allowed set. Connectors can pass a different allowed_value_types tuple.

Source

Thrown at python/pathway/io/_utils.py:593

                )
            else:
                column_names = list(table._columns.keys())
                if len(column_names) != 1:
                    raise ValueError(
                        f"'{format}' format without explicit 'value' specification "
                        "can only be used with single-column tables"
                    )
                value = table[column_names[0]]
                value_field_index = cls.add_column_reference_to_extract(
                    value, columns_to_extract, extracted_field_indices
                )

            table = table.select(*columns_to_extract)
            if (
                allowed_value_types is not None
                and table[value._name]._column.dtype not in allowed_value_types
            ):
                raise ValueError(
                    f"The value column must have one of the following types: {allowed_value_types}"
                )

            data_format = api.DataFormat(
                format_type="single_column",
                key_field_names=[],
                value_fields=_format_output_value_fields(table),
                value_field_index=value_field_index,
                schema_registry_settings=maybe_schema_registry_settings(
                    schema_registry_settings
                ),
                subject=subject,
            )
        else:
            raise ValueError(f"Unsupported format: {format}")

        return cls(
            table=table,

View on GitHub (pinned to fa2f74a464)

Solutions

Cast the value column to string/bytes: value on pw.this.count.astype(str).
Or pick a column that is already BYTES/STR as the payload.
If the target system genuinely supports the dtype, pass an extended allowed_value_types tuple.

Example fix

# before
pw.io.kafka.write(t, ..., format='raw')  # single int column 'count'

# after
t = t.select(payload=pw.this.count.astype(str))
pw.io.kafka.write(t, ..., format='raw')

Defensive patterns

Strategy: type-guard

Validate before calling

value_dtype = table[value_name]._column.dtype
assert allowed_value_types is None or value_dtype in allowed_value_types, f"value dtype {value_dtype} not in {allowed_value_types}"

Type guard

def value_dtype_allowed(dtype, allowed=(dt.BYTES, dt.STR, dt.ANY)) -> bool:
    return dtype in allowed

Try / catch

try:
    pw.io.kafka.write(t, ..., format='raw')
except ValueError as e:
    if 'value column' in str(e):
        t = t.with_columns(payload=pw.this[value_name].astype(str))
        pw.io.kafka.write(t.select(pw.this.payload), ..., format='raw')
    else:
        raise

Prevention

Serialize non-string payloads to str/bytes before raw sinks.
Assert payload dtype early in the pipeline, not at sink construction.

When it happens

Trigger: pw.io.kafka.write(t, ..., format='raw') where the single/value column is int, float, or a complex type; schema evolution that retypes the payload column from bytes to something else.

Common situations: Tables whose payload column is numeric (counts, ids) because no serialization step was added; connector-specific serializers that accept a narrower/wider type set.

Related errors

AI-assisted analysis of pathwaycom/pathway@fa2f74a464 (2026-08-15). Data as JSON: /api/errors/08ec84275c17e319. Report an issue: GitHub.