databendlabs/databend · error

Unsupported column type for length calculation

Error message

Unsupported column type for length calculation

What it means

The variable-size row encoder's length-calculation visitor computes encoded lengths before writing sort rows. Its default visit_typed_column arm raises unimplemented!("Unsupported column type for length calculation") for types it cannot measure. Sorting on such a column with the variable encoder is unsupported.

Solutions

1. Cast the sort key to a supported type before ordering (e.g. col::string)
2. Restructure the query to sort by a primitive column or expression
3. Check the Databend release notes for sort-encoder support of the type; upgrade if added
4. Implement the missing length-measurement arm in the variable encoder if you maintain the code

Example fix

// before
SELECT * FROM t ORDER BY map_col;
// after
SELECT * FROM t ORDER BY map_col['key']::string;

-- sql: keep variable-width sort keys to supported types (String primarily)
SELECT data_type FROM information_schema.columns
WHERE table_name = 't' AND column_name IN ('sort_col');
-- cast non-string variable keys: ORDER BY sort_col::string

fn variable_encodable(ty: &DataType) -> bool {
    matches!(ty, DataType::String | DataType::Nullable(box inner)
        if matches!(**inner, DataType::String))
        || matches!(ty, DataType::String)
}

When it happens

Trigger: A sort/ORDER BY (or window partition key) includes a column type whose encoded length cannot be computed by the variable encoder — types not routed to a proper visitor arm.

Common situations: ORDER BY on newly added or nested types (arrays, maps, variants) that only partially support variable encoding; optimizer selecting variable encoding for an unsupported type; version skew between data type support and sorter support.

Related errors

• Unsupported column type for encoding
• internal error: entered unreachable code
• {}
• Temp table id used up
• Invalid temp table desc

AI-assisted analysis of databendlabs/databend@288d84d76e (2026-09-11). Data as JSON: /api/errors/9beee89e75213b90. Report an issue: GitHub.

Appendix: source

Thrown at src/query/pipeline/transforms/src/processors/transforms/sorts/core/row_convert/variable.rs:482

        self.visit_variable_length_data(column.iter(), encoded_len)
    }

    fn visit_string(&mut self, column: StringColumn) -> Result<()> {
        self.visit_variable_length_data(column.iter(), |s, is_null| {
            encoded_len(s.as_bytes(), is_null)
        })
    }

    fn visit_variant(&mut self, column: BinaryColumn) -> Result<()> {
        self.visit_variable_length_data(column.iter(), encoded_len)
    }

    fn visit_typed_column<T: ValueType>(
        &mut self,
        _column: T::Column,
        _data_type: &DataType,
    ) -> Result<()> {
        unimplemented!("Unsupported column type for length calculation")
    }
}

struct EncodeVisitor<'a> {
    out: &'a mut BinaryColumnBuilder,
    validity: (bool, Option<&'a Bitmap>),
    field: &'a RowSortField,
}

impl EncodeVisitor<'_> {
    fn visit_const_column(&mut self, scalar: &Scalar, data_type: &DataType) -> Result<()> {
        let is_null = scalar.is_null();
        debug_assert!(data_type.is_nullable_or_null() || !is_null);
        match data_type.remove_nullable() {
            DataType::Null => {}
            DataType::Boolean => {
                let scalar = if is_null {
                    false

View on GitHub (pinned to 288d84d76e)