apache/iceberg · error · DataException

Unable to create table from empty object

Error message

Unable to create table from empty object

What it means

IcebergWriterFactory.autoCreateTable infers the Iceberg schema for the new table from a sample SinkRecord. When the record has no value schema, it infers the type from the raw value; if that inference returns null (value is null or an empty/unsupported object), table creation cannot proceed and a DataException is thrown.

Solutions

1. Pre-create the target Iceberg table with an explicit schema instead of relying on auto-creation.
2. Ensure the topic carries records with real, non-null, non-empty values and a value schema (use Avro/Protobuf with schema registry, or ensure JSON converter emits schemas).
3. Check the producer/converter config so tombstones or empty messages are not the records used for schema inference.
4. Verify table_create/auto-create properties point at the correct topic-to-table mapping with a valid sample record available.

Example fix

// before
// auto-create from a topic whose sampled record has null schema/value
// "iceberg.sink.auto-create-table": "true"

// after
// pre-create the table explicitly
spark.sql("CREATE TABLE catalog.db.topic_table (id long, ts timestamp) USING iceberg");

// Java — check the sample before relying on auto-creation
boolean canAutoCreate = sample.value() != null &&
    (sample.valueSchema() != null || sample.value() instanceof Map && !((Map<?, ?>) sample.value()).isEmpty());

try {
  Table table = factory.autoCreateTable(tableName, sample);
} catch (DataException e) {
  if (String.valueOf(e.getMessage()).contains("Unable to create table from empty object")) {
    LOG.error("Sample record has no schema/empty value — pre-create table {} explicitly", tableName, e);
  }
  throw e;
}

When it happens

Trigger: auto-create-table is enabled and the sampled record's valueSchema() is null while its value() yields no inferable type — e.g. null payloads (tombstones), empty maps/structs, or unsupported value shapes reaching the sampler.

Common situations: Topic's first/most recent record is a tombstone (null value) at connector start; producer sending schema-less formats (raw JSON/bytes) with empty objects; misconfigured value converter producing null values.

Understand the failure class

Background: "must not be empty", "cannot be empty" — required-field validation errors across open-source libraries — this error's family across 41 libraries.

Related errors

• An error occurred closing catalog instance, ignoring...
• An error occurred converting record, topic
• Cannot convert date
• Cannot convert java.util.Date to variant without a…
• Cannot convert map to variant: keys must be non-null…

AI-assisted analysis of apache/iceberg@86d9c8fc54 (2026-09-12). Data as JSON: /api/errors/4fd11076de5245a7. Report an issue: GitHub.

Appendix: source

Thrown at kafka-connect/kafka-connect/src/main/java/org/apache/iceberg/connect/data/IcebergWriterFactory.java:89

    UUID tableUuid = table.uuid();
    if (tableUuid == null) {
      LOG.warn(
          "Table {} does not have a UUID, this may cause issues with commit coordination on table replace",
          identifier);
    }
    TableReference tableReference = TableReference.of(catalog.name(), identifier, tableUuid);

    return new IcebergWriter(table, tableReference, config);
  }

  @VisibleForTesting
  Table autoCreateTable(String tableName, SinkRecord sample) {
    StructType structType;
    if (sample.valueSchema() == null) {
      Type type = SchemaUtils.inferIcebergType(sample.value(), config);
      if (type == null) {
        throw new DataException("Unable to create table from empty object");
      }
      structType = type.asStructType();
    } else {
      structType = SchemaUtils.toIcebergType(sample.valueSchema(), config).asStructType();
    }

    org.apache.iceberg.Schema schema = new org.apache.iceberg.Schema(structType.fields());
    TableIdentifier identifier = TableIdentifier.parse(tableName);

    createNamespaceIfNotExist(catalog, identifier.namespace());

    List<String> partitionBy = config.tableConfig(tableName).partitionBy();
    PartitionSpec spec;
    try {
      spec = SchemaUtils.createPartitionSpec(schema, partitionBy);
    } catch (Exception e) {
      LOG.error(
          "Unable to create partition spec {}, table {} will be unpartitioned",

View on GitHub (pinned to 86d9c8fc54)