apache/iceberg · error · IllegalArgumentException

Cannot get value for invalid index

Error message

Cannot get value for invalid index: <index>

What it means

Thrown by a positional accessor (e.g. the delete-vector record reader's get) when asked for a value at an index that the underlying structure does not have. These composite readers only expose a fixed number of positional fields (here 0 and 1 for path and pos of a delete); any other index is invalid and raises IllegalArgumentException.

Solutions

1. Only access indices 0..numFields-1 of the composite reader; verify the index against the struct width
2. Update the switch mapping if new fields (e.g. sequence numbers) were added to the delete schema
3. Check the writerToFieldIndex mapping when Unknown columns are skipped so indices align

Example fix

// before: blindly accessing a third column
Object seq = reader.get(2);
// after: guard the index
if (index >= 0 && index < numFields) { value = reader.get(index); }

if (index < 0 || index >= reader.numFields()) {
  throw new IndexOutOfBoundsException("reader index " + index + " out of range");
}

boolean isValidIndex(int index, int numFields) { return index >= 0 && index < numFields; }

try {
  value = accessor.get(index);
} catch (IllegalArgumentException e) {
  if (e.getMessage().startsWith("Cannot get value for invalid index")) {
    // re-map index via writerToFieldIndex before retrying
  } else throw e;
}

When it happens

Trigger: Calling get(int index) on the reader's internal position-accessor with an index other than 0 or 1, usually from code that assumes a different field layout or passes a writer/field index computed against a different struct type.

Common situations: Positional deletes/vectorized delete readers where the writer-index mapping is off by the number of columns; code changes adding columns without updating the switch; generic code iterating indices beyond the record width.

Related errors

• AlwaysFalse is a placeholder only
• AlwaysTrue is a placeholder only
• Avro writer does not support variant types
• Batch reading is not supported in non-vectorized reader
• Buffer size of is larger than requested size of

AI-assisted analysis of apache/iceberg@86d9c8fc54 (2026-09-12). Data as JSON: /api/errors/acc9a2eeb0f122ff. Report an issue: GitHub.

Appendix: source

Thrown at parquet/src/main/java/org/apache/iceberg/parquet/ParquetValueWriters.java:733

  public static class PositionDeleteStructWriter<R> extends StructWriter<PositionDelete<R>> {
    private final Function<CharSequence, ?> pathTransformFunc;

    public PositionDeleteStructWriter(
        StructWriter<?> replacedWriter, Function<CharSequence, ?> pathTransformFunc) {
      super(Arrays.asList(replacedWriter.writers));
      this.pathTransformFunc = pathTransformFunc;
    }

    @Override
    protected Object get(PositionDelete<R> delete, int index) {
      switch (index) {
        case 0:
          return pathTransformFunc.apply(delete.path());
        case 1:
          return delete.pos();
      }
      throw new IllegalArgumentException("Cannot get value for invalid index: " + index);
    }
  }

  /** Returns a mapping from writer index to field index, skipping Unknown columns. */
  static int[] writerToFieldIndex(Types.StructType struct, int numWriters) {
    if (null == struct) {
      return IntStream.rangeClosed(0, numWriters).toArray();
    }

    List<Types.NestedField> recordFields = struct.fields();

    // value writer index to record field index
    int[] indexes = new int[numWriters];
    int writerIndex = 0;
    for (int pos = 0; pos < recordFields.size(); pos += 1) {
      if (recordFields.get(pos).type().typeId() != org.apache.iceberg.types.Type.TypeID.UNKNOWN) {
        indexes[writerIndex] = pos;
        writerIndex += 1;

View on GitHub (pinned to 86d9c8fc54)