apache/iceberg · error · UnsupportedOperationException

Not a long column

Error message

Not a long column

What it means

TripleIterator.nextLong() is a default method that throws UnsupportedOperationException; only iterators for INT64 columns override it. Hitting this message means the long accessor was called on a column that is not physically INT64 — the iterator type dispatch does not match the file's column type.

Source

Thrown at parquet/src/main/java/org/apache/iceberg/parquet/TripleIterator.java:81

   * @return the next value as an un-boxed int
   * @throws java.util.NoSuchElementException if there are no more elements
   * @throws UnsupportedOperationException if the underlying data values are not ints
   */
  default int nextInteger() {
    throw new UnsupportedOperationException("Not an integer column");
  }

  /**
   * Returns the next value as an un-boxed long.
   *
   * <p>This method has the same behavior as {@link #next()} and will advance this iterator.
   *
   * @return the next value as an un-boxed long
   * @throws java.util.NoSuchElementException if there are no more elements
   * @throws UnsupportedOperationException if the underlying data values are not longs
   */
  default long nextLong() {
    throw new UnsupportedOperationException("Not a long column");
  }

  /**
   * Returns the next value as an un-boxed float.
   *
   * <p>This method has the same behavior as {@link #next()} and will advance this iterator.
   *
   * @return the next value as an un-boxed float
   * @throws java.util.NoSuchElementException if there are no more elements
   * @throws UnsupportedOperationException if the underlying data values are not floats
   */
  default float nextFloat() {
    throw new UnsupportedOperationException("Not a float column");
  }

  /**
   * Returns the next value as an un-boxed double.
   *

View on GitHub (pinned to 86d9c8fc54)

Solutions

1. Check the physical type in the Parquet footer — nextLong() requires INT64; for INT32 data use nextInteger() and widen
2. Fix the materializer/reader dispatch so long columns bind to the INT64 reader
3. Align the Iceberg schema with the file's physical schema (int vs long promotion must match file content)
4. Rewrite non-conforming data files to the table's current schema
5. In custom iterator code, verify desc.getPrimitiveType() before calling typed accessors

Example fix

// before: assumes INT64
long ts = iterator.nextLong();

// after: narrow by physical type
long ts = desc.getPrimitiveType().getPrimitiveTypeName() == PrimitiveTypeName.INT64
    ? iterator.nextLong()
    : iterator.nextInteger();

if (desc.getPrimitiveType().getPrimitiveTypeName() != PrimitiveTypeName.INT64) {
  throw new IllegalArgumentException("Column " + desc + " is not INT64");
}

boolean isInt64Column(ColumnDescriptor desc) {
  return desc.getPrimitiveType().getPrimitiveTypeName() == PrimitiveTypeName.INT64;
}

try { v = iterator.nextLong(); } catch (UnsupportedOperationException e) { throw new IllegalStateException("Expected INT64 column but got: " + desc.getPrimitiveType(), e); }

When it happens

Trigger: Calling nextLong() on a TripleIterator whose primitive type is not INT64 — commonly when an Iceberg long/timestamp column is actually stored as INT32 (date/time as int) or the custom materializer routes the column to the wrong reader.

Common situations: Timestamp columns stored as INT96/millis-as-int in files written by other engines; Iceberg schema upgraded a column from int to long while files still hold INT32 values; bespoke ParquetValueReaders with incorrect type mapping.

Understand the failure class

Background: Type mismatch errors: IllegalArgumentException, TypeError and type guards across 150 open-source libraries — this error's family across 150 libraries.

Related errors

• Not a boolean column
• Not an integer column
• Not a float column
• Not a double column
• Not a binary column

AI-assisted analysis of apache/iceberg@86d9c8fc54 (2026-09-12). Data as JSON: /api/errors/50a6ca482b525da9. Report an issue: GitHub.