pathwaycom/pathway · error · ValueError

The column {column} doesn't belong to the target table {tabl

Error message

The column {column} doesn't belong to the target table {table}

What it means

DataSink.check_sort_by_columns validates a sink's sort_by columns before writing: every ColumnReference in sort_by must belong to the exact Table being written (column._table is the table object). Columns coming from another table (e.g. the pre-join inputs, or an intermediate table) fail this identity check.

Source

Thrown at python/pathway/internals/datasink.py:28

from pathway.internals import api
from pathway.internals.expression import ColumnReference

if TYPE_CHECKING:
    from pathway.internals.table import Table


class DataSink(ABC):
    @property
    def name(self) -> str:
        return type(self).__qualname__.lower().removesuffix("datasink")

    def check_sort_by_columns(self, table: Table):
        sort_by = getattr(self, "sort_by", None)
        if sort_by is None:
            return
        for column in sort_by:
            if column._table != table:
                raise ValueError(
                    f"The column {column} doesn't belong to the target table {table}"
                )


@dataclass(frozen=True)
class GenericDataSink(DataSink):
    datastorage: api.DataStorage
    dataformat: api.DataFormat
    datasink_name: str
    unique_name: str | None
    sort_by: Iterable[ColumnReference] | None = None
    on_pipeline_finished: Callable | None = None

    @property
    def name(self) -> str:
        return self.datasink_name

    @property

View on GitHub (pinned to fa2f74a464)

Solutions

1. Re-derive the columns from the table being written: out = t.select(...); pw.io.csv.write(out, sort_by=[out.x, out.y]).
2. Chain immediately: write the table whose attribute columns you use in sort_by in the same expression scope.
3. Drop sort_by if ordering is not required.

Example fix

# before
sorted_cols = [t.x, t.y]
out = t.select(x=pw.this.x, y=pw.this.y)
pw.io.csv.write(out, path, sort_by=sorted_cols)  # cols belong to t, not out

# after
out = t.select(x=pw.this.x, y=pw.this.y)
pw.io.csv.write(out, path, sort_by=[out.x, out.y])

assert all(c._table is table_being_written for c in (sort_by or [])), 'sort_by columns must come from the exact table written'

def sort_by_columns_valid(table, sort_by) -> bool:
    return all(c._table is table for c in (sort_by or []))

When it happens

Trigger: Passing sort_by=[t_src.col] where t_src is not the table handed to pw.io.json.write / csv.write etc., typically after a select/rename/join so the saved table object is a new instance while sort_by still references the old table's columns.

Common situations: Reusing column references captured before a transformation: cols = [t.x, t.y]; out = t.select(...); pw.io.csv.write(out, ..., sort_by=cols). Join outputs where sort_by references an input table's column.

Related errors

• direction argument of join should be of type asof_join.Direc
• The behavior argument of join should be of type pathway.temp
• The interval argument of a join should be of a type pathway.
• Join received extra kwargs. You probably want to use TableLi
• Expected a ColumnReference, found a string. Did you mean thi

AI-assisted analysis of pathwaycom/pathway@fa2f74a464 (2026-08-15). Data as JSON: /api/errors/4b46dba3951f1a01. Report an issue: GitHub.