pathwaycom/pathway · error · ValueError

'delimiter' is only meaningful for the 'dsv' format, but {fo

Error message

'delimiter' is only meaningful for the 'dsv' format, but {format!r} was specified. Drop the 'delimiter' argument or use format='dsv'.

What it means

Raised by MessageQueueOutputFormat.build when the output format is not 'dsv' but a non-default 'delimiter' was passed. Delimiter only controls field separation for CSV/DSV output; for 'json', 'raw', or 'plaintext' it has no effect, so passing one is treated as a likely mistake (note the default ',' is allowed).

Source

Thrown at python/pathway/io/_utils.py:453

    @classmethod
    def construct(
        cls,
        table: Table,
        *,
        format: str = "json",
        delimiter: str = ",",
        key: ColumnReference | None = None,
        value: ColumnReference | None = None,
        headers: Iterable[ColumnReference] | None = None,
        topic_name: ColumnReference | None = None,
        schema_registry_settings: SchemaRegistrySettings | None = None,
        subject: str | None = None,
        allowed_key_types: tuple[dt.DType, ...] | None = (dt.BYTES, dt.STR, dt.ANY),
        allowed_value_types: tuple[dt.DType, ...] | None = (dt.BYTES, dt.STR, dt.ANY),
    ) -> MessageQueueOutputFormat:
        if delimiter != "," and format != "dsv":
            raise ValueError(
                f"'delimiter' is only meaningful for the 'dsv' format, but "
                f"{format!r} was specified. Drop the 'delimiter' argument "
                f"or use format='dsv'."
            )
        if subject is not None and schema_registry_settings is None:
            raise ValueError(
                "'subject' was provided without 'schema_registry_settings'. "
                "The 'subject' parameter only has an effect when a schema "
                "registry is configured; either pass 'schema_registry_settings' "
                "or remove 'subject'."
            )
        if schema_registry_settings is not None and subject is None:
            raise ValueError(
                "'schema_registry_settings' was provided without 'subject'. "
                "When a schema registry is configured, 'subject' must also be "
                "set so the formatter knows which subject to encode under."
            )
        if subject is not None and not subject:

View on GitHub (pinned to fa2f74a464)

Solutions

Drop the 'delimiter' argument when format is not 'dsv'.
Or change format to 'dsv' if you actually intended delimited output.

Example fix

# before
pw.io.kafka.write(t, ..., format='json', delimiter=';')

# after
pw.io.kafka.write(t, ..., format='json')

Defensive patterns

Strategy: validation

Validate before calling

if format != 'dsv':
    assert delimiter == ',', "delimiter is only accepted for format='dsv'"

Try / catch

try:
    pw.io.kafka.write(t, ..., format=format, delimiter=delim)
except ValueError as e:
    if "'delimiter' is only meaningful" in str(e):
        pw.io.kafka.write(t, ..., format=format)  # retry without delimiter
    else:
        raise

Prevention

Only set 'delimiter' in dsv-specific config branches.
Treat non-default delimiter as a dsv-only option in your config schema.

When it happens

Trigger: pw.io.kafka.write(table, ..., format='json', delimiter=';'); switching an output from dsv to json while keeping the delimiter argument; reusing one kwargs dict across several writers.

Common situations: Shared config dicts reused for multiple sinks; refactoring a CSV output into JSON output without pruning format-specific options.

Related errors

AI-assisted analysis of pathwaycom/pathway@fa2f74a464 (2026-08-15). Data as JSON: /api/errors/8a991b953990a83b. Report an issue: GitHub.