pentaho/pentaho-kettle · error · KettleException

FilterRows.Exception.UnexpectedErrorFoundInEvaluationFuction

FilterRows.Exception.UnexpectedErrorFoundInEvaluationFuction

Error message

FilterRows.Exception.UnexpectedErrorFoundInEvaluationFuction

What it means

FilterRows evaluates its Condition against each incoming row to decide the true/false target. keepRow() catches any Exception thrown during that evaluation, logs the row that failed, and rethrows a KettleException with this message - meaning the filter expression itself blew up while processing a specific row.

Solutions

1. Inspect the log line 'Error occurred while filtering rows' + the row dump to identify the offending field/value.
2. Check the filter condition's field types vs the actual incoming row metadata (right-click > Show input fields / preview upstream).
3. Add a 'Value Mapper'/'If field value is null' or 'Data Grid'-style cleanup step upstream to normalize nulls/types before filtering.

Example fix

// before: filter condition compares field 'amount' (Number) to string '10'
// after: upstream add a Select Values step converting 'amount' from String to Number,
// or change the condition value type to Number 10

// Check condition field types match the incoming metadata before running:
for (String field : condition.getUsedFields()) {
  ValueMetaInterface vm = prevRowMeta.searchValueMeta(field);
  if (vm == null) throw new IllegalStateException("Filter field missing: " + field);
}

try {
  trans.execute(null);
} catch (KettleException e) {
  if (e.getMessage().contains("UnexpectedErrorFoundInEvaluationFuction")) {
    log.error("Filter condition failed on a row - check field types/nulls in the logged row dump", e);
  }
  throw e;
}

When it happens

Trigger: keepRow() -> condition.evaluate(rowMeta, row) throws: comparing incompatible field types (string vs number), null values against non-null-safe comparisons, or a condition referencing a field whose data changed shape after metadata was cached.

Common situations: Filtering on a numeric field that sometimes receives non-numeric/null data; comparing a date field against a wrongly formatted string; upstream step changed field types after the filter was configured.

Understand the failure class

Background: "Must be a positive integer", "Invalid value", "Unsupported": the invalid-argument-value error family, when a library rejects the value you pass — this error's family across 35 libraries.

Related errors

• Error adding values to row!
• Error during processing a row
• Failed to add Cell Row:
• Failed to process some row
• SalesforceInput.Exception.CanNotReadFromSalesforce

AI-assisted analysis of pentaho/pentaho-kettle@f3058517a1 (2026-09-13). Data as JSON: /api/errors/6d68732b4e51af64. Report an issue: GitHub.

Appendix: source

Thrown at engine/src/main/java/org/pentaho/di/trans/steps/filterrows/FilterRows.java:59

  private FilterRowsMeta meta;
  private FilterRowsData data;

  public FilterRows( StepMeta stepMeta, StepDataInterface stepDataInterface, int copyNr, TransMeta transMeta,
    Trans trans ) {
    super( stepMeta, stepDataInterface, copyNr, transMeta, trans );
  }

  private synchronized boolean keepRow( RowMetaInterface rowMeta, Object[] row ) throws KettleException {
    try {
      return meta.getCondition().evaluate( rowMeta, row );
    } catch ( Exception e ) {
      String message =
        BaseMessages.getString( PKG, "FilterRows.Exception.UnexpectedErrorFoundInEvaluationFuction" );
      logError( message );
      logError( BaseMessages.getString( PKG, "FilterRows.Log.ErrorOccurredForRow" ) + rowMeta.getString( row ) );
      logError( Const.getStackTracker( e ) );
      throw new KettleException( message, e );
    }
  }

  public boolean processRow( StepMetaInterface smi, StepDataInterface sdi ) throws KettleException {
    meta = (FilterRowsMeta) smi;
    data = (FilterRowsData) sdi;

    boolean keep;

    Object[] r = getRow(); // Get next usable row from input rowset(s)!
    if ( r == null ) { // no more input to be expected...

      setOutputDone();
      return false;
    }

    if ( first ) {
      first = false;

View on GitHub (pinned to f3058517a1)