What causes "readLongs is not supported" in apache/iceberg?

Disable vectorized reads (read.arrow.enabled=false) to use the row-at-a-time Parquet reader.

apache/iceberg · error · UnsupportedOperationException

readLongs is not supported

Error message

readLongs is not supported

What it means

readLongs(total, vec, rowId) is the batch variant for long columns on VectorizedValuesReader; its default body throws UnsupportedOperationException. Real readers override it to bulk-decode longs into a BigIntVector. nextVals() dispatching here means the active reader class does not implement the batch long path.

Source

Thrown at arrow/src/main/java/org/apache/iceberg/arrow/vectorized/parquet/VectorizedValuesReader.java:90

  }

  /**
   * Read binary data of some length
   *
   * @param len The number of bytes to read
   */
  default Binary readBinary(int len) {
    throw new UnsupportedOperationException("readBinary is not supported");
  }

  /** Read `total` integers into `vec` starting at `vec[rowId]` */
  default void readIntegers(int total, FieldVector vec, int rowId) {
    throw new UnsupportedOperationException("readIntegers is not supported");
  }

  /** Read `total` longs into `vec` starting at `vec[rowId]` */
  default void readLongs(int total, FieldVector vec, int rowId) {
    throw new UnsupportedOperationException("readLongs is not supported");
  }

  /** Read `total` floats into `vec` starting at `vec[rowId]` */
  default void readFloats(int total, FieldVector vec, int rowId) {
    throw new UnsupportedOperationException("readFloats is not supported");
  }

  /** Read `total` doubles into `vec` starting at `vec[rowId]` */
  default void readDoubles(int total, FieldVector vec, int rowId) {
    throw new UnsupportedOperationException("readDoubles is not supported");
  }

  /**
   * Initialize the reader from a page. See {@link ValuesReader#initFromPage(int,
   * ByteBufferInputStream)}.
   */
  void initFromPage(int valueCount, ByteBufferInputStream in) throws IOException;
}

View on GitHub (pinned to 86d9c8fc54)

Solutions

Disable vectorized reads (read.arrow.enabled=false) to use the row-at-a-time Parquet reader.
Upgrade Iceberg to get batch long support for this type/encoding.
Override readLongs() in the concrete reader to fill vec[rowId..rowId+total).

Example fix

// before
// default readLongs throws
// after
@Override
public void readLongs(int total, FieldVector vec, int rowId) {
  for (int i = 0; i < total; i++) ((BigIntVector) vec).setSafe(rowId + i, readLong());
}

Defensive patterns

Strategy: try-catch

Validate before calling

// confirm long column before batch path
if (!columnType.equals(Types.LongType.get())) props.put("read.arrow.enabled", "false");

Try / catch

try {
  taskReader.readBatch();
} catch (UnsupportedOperationException e) {
  return fallbackRowReader();
}

Prevention

Ensure the selected reader class overrides readLongs before enabling Arrow batch scans.
Avoid mixing custom readers from different Iceberg versions.
Add a unit test scanning long columns with vectorization enabled.

When it happens

Trigger: nextVals(total) attempts bulk fill of an Arrow long vector for a Parquet INT64 column, but the active VectorizedValuesReader subclass does not override readLongs().

Common situations: Vectorized scans of bigint/timestamp columns where the reader factory selected a fallback reader; encodings whose dedicated reader only implements batch integers; custom reader implementations.

Understand the failure class

Background: UnsupportedOperationException and "is not supported" errors: when a library deliberately refuses a call — this error's family across 30 libraries.

Related errors

AI-assisted analysis of apache/iceberg@86d9c8fc54 (2026-09-12). Data as JSON: /api/errors/c567667b9523b258. Report an issue: GitHub.