apache/druid · error · ISE

Invalid input [%s] of type [%s] for [%s] aggregator [%s]

Error message

Invalid input [%s] of type [%s] for [%s] aggregator [%s]

What it means

HllSketchBuildAggregatorFactory.validateInputs throws IllegalStateException when the input column's capabilities report ValueType.COMPLEX, i.e. the field already contains serialized HLL sketches. Building a new sketch from an already-sketch column is invalid; use the merge-type aggregator (HllSketchMergeFactory) for that. It is invoked while factorizing vectorized/SQL aggregators.

Source

Thrown at extensions-core/datasketches/src/main/java/org/apache/druid/query/aggregation/datasketches/hll/HllSketchBuildAggregatorFactory.java:155

  @Override
  public AggregatorFactory withName(String newName)
  {
    return new HllSketchBuildAggregatorFactory(
        newName,
        getFieldName(),
        getLgK(),
        getTgtHllType(),
        getStringEncoding(),
        isShouldFinalize(),
        isRound()
    );
  }

  private void validateInputs(@Nullable ColumnCapabilities capabilities)
  {
    if (capabilities != null) {
      if (capabilities.is(ValueType.COMPLEX)) {
        throw new ISE(
            "Invalid input [%s] of type [%s] for [%s] aggregator [%s]",
            getFieldName(),
            capabilities.asTypeString(),
            HllSketchModule.BUILD_TYPE_NAME,
            getName()
        );
      }
    }
  }

  private HllSketchUpdater formulateSketchUpdater(ColumnSelectorFactory columnSelectorFactory)
  {
    final ColumnCapabilities capabilities = columnSelectorFactory.getColumnCapabilities(getFieldName());
    validateInputs(capabilities);

    HllSketchUpdater updater = null;
    if (capabilities != null &&
        StringEncoding.UTF8.equals(getStringEncoding()) && ValueType.STRING.equals(capabilities.getType())) {

View on GitHub (pinned to 9b90983fd2)

Solutions

1. Switch the aggregator type from HLLSketchBuild to HLLSketchMerge when the input column already holds sketches.
2. Point the build aggregator at the raw string/numeric column that should be sketched.
3. Fix ingestion spec so the source column stored is raw data, not a serialized sketch, if raw building is intended.
4. In SQL use APPROX_COUNT_DISTINCT_DS_HLL on raw values and DS_HLL(...) on sketch columns appropriately.

Example fix

// before
{"type": "HLLSketchBuild", "name": "s", "fieldName": "sketch_col"}
// after
{"type": "HLLSketchMerge", "name": "s", "fieldName": "sketch_col"}

ColumnCapabilities caps = columnCapabilities(fieldName); if (caps != null && caps.is(ValueType.COMPLEX) && factory.getType().equals("HLLSketchBuild")) { throw new IllegalArgumentException("use HLLSketchMerge for sketch columns"); }

boolean isComplexColumn = caps != null && caps.is(ValueType.COMPLEX);

try { factory.factorize(...) } catch (ISE e) { /* switch to HLLSketchMerge */ }

When it happens

Trigger: Configuring an HLLSketchBuild aggregator whose fieldName points to a COMPLEX column (e.g. an HLLSketch-typed column produced by ingestion or another sketch), during query factorizeVector or formulateSketchUpdater.

Common situations: Mistakenly using 'HLLSketchBuild' on a column ingested as sketch (should be 'HLLSketchMerge'); copying a query spec that had sketch output columns and feeding them back as build inputs; schema drift after switching ingestion to store sketches.

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

• Unsupported type %s
• Not implemented
• Object is not of a type[%s] that can be deserialized to sket
• Expected a number or an instance of DDSketch, but received [
• Expected a number or an instance of DDSketch, but received [

AI-assisted analysis of apache/druid@9b90983fd2 (2026-09-07). Data as JSON: /api/errors/118f66a6cf4d19a5. Report an issue: GitHub.