pathwaycom/pathway · error · ValueError

Can't use 'id' as a column name

Error message

Can't use 'id' as a column name

What it means

Pathway reserves the column name 'id' for the table's primary key (the auto-generated pointer column), so user data columns cannot be named 'id'. combine_args_kwargs raises this ValueError as soon as any positional smart_name or kwarg key equals 'id', before the expression is ever evaluated.

Source

Thrown at python/pathway/internals/desugaring.py:306

    return {
        name: _desugar_this_arg(substitution, arg) for name, arg in new_kwargs.items()
    }


def combine_args_kwargs(
    args: Iterable[expr.ColumnReference],
    kwargs: Mapping[str, Any],
    exclude_columns: set[str] | None = None,
) -> dict[str, expr.ColumnExpression]:
    all_args = {}

    def add(name, expression):
        if exclude_columns is not None and name in exclude_columns:
            return
        if name in all_args:
            raise ValueError(f"Duplicate expression value given for {name}")
        if name == "id":
            raise ValueError("Can't use 'id' as a column name")
        if not isinstance(expression, expr.ColumnExpression):
            expression = expr.ColumnConstExpression(expression)
        all_args[name] = expression

    for expression in args:
        add(expr.smart_name(expression), expression)
    for name, expression in kwargs.items():
        add(name, expression)

    return all_args


class DesugaringContext:
    _substitution: dict[thisclass.ThisMetaclass, table.Joinable] = {}

    @property
    @abstractmethod
    def _desugaring(self) -> DesugaringTransform:

View on GitHub (pinned to fa2f74a464)

Solutions

1. Rename the column to anything else, e.g. select(row_id=table.id) or use a schema with field name like object_id
2. In a schema definition, declare the key with a different name (e.g. class Schema(pw.Schema, id_column='object_id')) so the key column is not called 'id'
3. When reading a data source with an 'id' column, map it in the connector/schema to a different output name before selecting

Example fix

// before
table.select(id=pw.this.left)
// after
class Out(pw.Schema, id_column='object_id'):
    object_id: int
    left: int
table.select(object_id=pw.this.left)

RESERVED = {"id"}

def validate_column_names(kwargs: dict):
    bad = RESERVED & set(kwargs)
    if bad:
        raise ValueError(f"reserved column names not allowed: {bad}")

When it happens

Trigger: table.select(id=table.key) or table.select(table.id_col) where the referenced column is literally named 'id'; with_columns(id=...); passing a schema class or kwargs containing an 'id' field through the same desugaring path.

Common situations: Porting pandas/SQL code where 'id' is a natural column name; reading external data whose columns include 'id' and then doing select(*cols) or rename operations; joining and trying to re-expose the key as 'id'.

Related errors

• Incorrect S3 path: {s3_path}
• SchemaRegistryHeader.key must be a str, got {type(self.key).
• Duplicate expression value given for {name}
• Iceberg reader requires explicit primary key fields specific
• primary_key column {primary_key._name!r} does not belong to

AI-assisted analysis of pathwaycom/pathway@fa2f74a464 (2026-08-15). Data as JSON: /api/errors/555e26ef54c5f970. Report an issue: GitHub.