apache/iceberg · error · UnsupportedOperationException

Cannot convert unknown primitive type

Error message

Cannot convert unknown primitive type: %s

What it means

Thrown by MessageTypeToType.primitive when converting a Parquet MessageType primitive column into an Iceberg Type and the primitive name matches no known case. Indicates the Parquet schema contains a primitive this Iceberg version cannot map.

Solutions

1. Upgrade Iceberg so the primitive type is mapped
2. Rewrite the data with supported Parquet primitive types
3. Inspect the offending column with parquet-tools and adjust the schema

Example fix

// before
// reading file with newer parquet type using old iceberg
// after
// upgrade dependency
// implementation 'org.apache.iceberg:iceberg-spark-runtime:latest'

MessageType parquetSchema = ParquetFileReader.readFooter(conf, path).getFileMetaData().getSchema();
parquetSchema.getColumns().forEach(c -> {
  if (!KNOWN_PRIMITIVES.contains(c.getPrimitiveTypeName())) throw new IllegalStateException("Unmappable type: " + c.getPrimitiveTypeName());
});

try {
  Type icebergType = new MessageTypeToType(parquetSchema).convert(field);
} catch (UnsupportedOperationException e) {
  if (e.getMessage().startsWith("Cannot convert unknown primitive type")) { /* upgrade iceberg */ }
}

When it happens

Trigger: Converting a Parquet schema (MessageType) to an Iceberg Type where the column's primitive falls through the switch, e.g. a type added in newer Parquet releases.

Common situations: Reading files produced by newer writers than the Iceberg version in use; INT96 or other edge types before mapping existed; incompatible library versions on the classpath.

Understand the failure class

Background: "is not a compatible type" / "cannot merge" errors: when a value's type doesn't match what the library requires — this error's family across 65 libraries.

Related errors

• Avro writer does not support variant types
• Byte stream split encoding is not supported for type " +…
• Can't handle
• Cannot decode dictionary of type
• Not a boolean column

AI-assisted analysis of apache/iceberg@86d9c8fc54 (2026-09-12). Data as JSON: /api/errors/31c2949f0dd91433. Report an issue: GitHub.

Appendix: source

Thrown at parquet/src/main/java/org/apache/iceberg/parquet/MessageTypeToType.java:181

      case BOOLEAN:
        return Types.BooleanType.get();
      case INT32:
        return Types.IntegerType.get();
      case INT64:
        return Types.LongType.get();
      case FLOAT:
        return Types.FloatType.get();
      case DOUBLE:
        return Types.DoubleType.get();
      case FIXED_LEN_BYTE_ARRAY:
        return Types.FixedType.ofLength(primitive.getTypeLength());
      case INT96:
        return Types.TimestampType.withZone();
      case BINARY:
        return Types.BinaryType.get();
    }

    throw new UnsupportedOperationException("Cannot convert unknown primitive type: " + primitive);
  }

  @Override
  public Type variant(GroupType variant) {
    return Types.VariantType.get();
  }

  private static class ParquetLogicalTypeVisitor
      implements LogicalTypeAnnotation.LogicalTypeAnnotationVisitor<Type> {
    private static final ParquetLogicalTypeVisitor INSTANCE = new ParquetLogicalTypeVisitor();

    private static ParquetLogicalTypeVisitor get() {
      return INSTANCE;
    }

    @Override
    public Optional<Type> visit(LogicalTypeAnnotation.StringLogicalTypeAnnotation stringType) {
      return Optional.of(Types.StringType.get());

View on GitHub (pinned to 86d9c8fc54)