pathwaycom/pathway · error · TypeError

Incompatible types in for a binary operator. The types are:

Error message

Incompatible types in for a binary operator.
The types are: {left} and {right}. You might try casting the expressions to Any type to circumvent this, but this is most probably an error.

What it means

For binary operators (==, <, +, etc.) on pointer columns, Pathway computes the least common ancestor of the two pointer dtypes; when the two pointer types are unrelated (no common supertype other than mismatched generics), types_lca(raising=True) raises and the mapping layer converts it into this TypeError suggesting an explicit cast. The message text is slightly garbled ('Incompatible types in for a binary operator') but always concerns a binary operation between two incompatible Pointer dtypes.

Source

Thrown at python/pathway/internals/operator_mapping.py:204

tuple_handling_operators = {
    operator.eq,
    operator.ne,
    operator.le,
    operator.lt,
    operator.ge,
    operator.gt,
}


def get_binary_operators_mapping(op, left, right):
    if isinstance(left, dt.Array) and isinstance(right, dt.Array):
        left, right = dt.coerce_arrays_pair(left, right)
    if isinstance(left, dt.Pointer) and isinstance(right, dt.Pointer):
        try:
            dt.types_lca(left, right, raising=True)
        except TypeError:
            raise TypeError(
                "Incompatible types in for a binary operator.\n"
                + f"The types are: {left} and {right}. "
                + "You might try casting the expressions to Any type to circumvent this,"
                + " but this is most probably an error."
            )
    return _binary_operators_mapping.get(
        (op, dt.normalize_dtype(left), dt.normalize_dtype(right))
    )


def get_binary_expression(
    left, right, op, left_dtype: dt.DType, right_dtype: dt.DType, default=None
):
    op_engine = _binary_operators_to_engine.get(op)
    left_dtype_engine = left_dtype.to_engine()
    right_dtype_engine = right_dtype.to_engine()
    if op_engine is None:
        return default

View on GitHub (pinned to fa2f74a464)

Solutions

1. Cast one side to a common type as the message suggests: pw.this.id.astype(pw.Pointer) or .astype(Any) on one operand.
2. Align id types at the source: build both tables with the same with_id_type(...) schema so their pointers share a type.
3. Re-check whether you actually want pointer equality — use table.join(...) / ix() for cross-table references instead of == on unrelated pointers.

Example fix

# before
joined = a.filter(a.ref_id == b_row.id)  # incompatible pointer types

# after
joined = a.filter(a.ref_id.astype(pw.Pointer) == b_row.id.astype(pw.Pointer))

from pathway.internals import dtype as dt

def pointers_compatible(l, r) -> bool:
    if isinstance(l, dt.Pointer) and isinstance(r, dt.Pointer):
        try:
            dt.types_lca(l, r, raising=True)
            return True
        except TypeError:
            return False
    return True

When it happens

Trigger: Comparing or joining-related operations between columns of different pointer types, e.g. table_a.some_id == table_b.other_id where one id is Pointer and the other is a different specialized pointer class; also comparisons between id columns of tables built with different with_id_type() schemas.

Common situations: Comparing ids from two differently-typed sources after with_id_type() was applied to one schema, or mixing pw.this.id with a pointer produced by a join/index_with on a differently typed table.

Related errors

• Pathway does not support using binary operator {expression._
• Pathway does not support using reducer {self} on column of t
• Pathway does not support using reducer {self.name} on column
• Pathway does not support using binary operator {operator.__n
• {role} {col._name!r} must be of type str, got {col._column.d

AI-assisted analysis of pathwaycom/pathway@fa2f74a464 (2026-08-15). Data as JSON: /api/errors/27911608f70400e2. Report an issue: GitHub.