apache/iceberg · error · UnsupportedOperationException

Expected value to be date or timestamp:

Error message

Expected value to be date or timestamp: 

What it means

Bind-time type validation in DaysFunction.doBind: the argument to system.days is neither DateType, TimestampType nor TimestampNTZType (the three temporal kinds the day transform accepts). The appended value is the offending type; the failure occurs when Spark binds the function, before any row is evaluated.

Solutions

1. Cast to date/timestamp first: system.days(CAST(col AS DATE)) or system.days(CAST('2024-01-01' AS DATE))
2. Use to_date/to_timestamp to convert strings: system.days(to_date('2024-01-01'))
3. For epoch-second bigint columns, convert: system.days(timestamp_millis(col * 1000))

Example fix

// before
SELECT * FROM t WHERE system.days(ts_col_string) = ...;  -- string
// after
SELECT * FROM t WHERE system.days(CAST(ts_col_string AS TIMESTAMP)) = ...;

if (!(value.dataType() instanceof DateType) && !(value.dataType() instanceof TimestampType) && !(value.dataType() instanceof TimestampNTZType)) {
  throw new IllegalArgumentException("days() requires date or timestamp, got: " + value.dataType());
}

boolean temporal(DataType t) {
  return t instanceof DateType || t instanceof TimestampType || t instanceof TimestampNTZType;
}

try {
  bound = daysFn.doBind(valueType);
} catch (UnsupportedOperationException e) {
  throw new IllegalArgumentException("Cast the expression to DATE or TIMESTAMP", e);
}

When it happens

Trigger: Calling system.days(<expr>) where expr is a string, long, or other non-temporal type, e.g. system.days('2024-01-01') or system.days(id).

Common situations: Passing date strings instead of date literals; applying days() to bigint epoch columns; comparing partition columns with uncast values in SQL predicates.

Understand the failure class

Background: "Must be a positive integer", "Invalid value", "Unsupported": the invalid-argument-value error family, when a library rejects the value you pass — this error's family across 35 libraries.

Related errors

• Expected value to be date or timestamp:
• Expected value to be timestamp:
• Expected column to be date, tinyint, smallint, int, bigint…
• Expected number of buckets to be tinyint, shortint or int
• Expected number of buckets to be tinyint, shortint or int

AI-assisted analysis of apache/iceberg@86d9c8fc54 (2026-09-12). Data as JSON: /api/errors/74a753a22b408701. Report an issue: GitHub.

Appendix: source

Thrown at spark/v4.1/spark/src/main/java/org/apache/iceberg/spark/functions/DaysFunction.java:48

import org.apache.spark.sql.types.TimestampType;

/**
 * A Spark function implementation for the Iceberg day transform.
 *
 * <p>Example usage: {@code SELECT system.days('source_col')}.
 */
public class DaysFunction extends UnaryUnboundFunction {

  @Override
  protected BoundFunction doBind(DataType valueType) {
    if (valueType instanceof DateType) {
      return new DateToDaysFunction();
    } else if (valueType instanceof TimestampType) {
      return new TimestampToDaysFunction();
    } else if (valueType instanceof TimestampNTZType) {
      return new TimestampNtzToDaysFunction();
    } else {
      throw new UnsupportedOperationException(
          "Expected value to be date or timestamp: " + valueType.catalogString());
    }
  }

  @Override
  public String description() {
    return name()
        + "(col) - Call Iceberg's day transform\n"
        + "  col :: source column (must be date or timestamp)";
  }

  @Override
  public String name() {
    return "days";
  }

  protected abstract static class BaseToDaysFunction extends BaseScalarFunction<Integer>
      implements ReducibleFunction<Integer, Integer> {

View on GitHub (pinned to 86d9c8fc54)