stanfordnlp/CoreNLP · error · RuntimeException

addFeature was called with a features object that is…

Error message

addFeature was called with a features object that is neither a counter nor a collection!

What it means

The private static helper addFeature expects the features accumulator to be either a Counter<F> (setCount with weight) or a Collection<F> (add). Any other object type is a programming invariant violation, thrown as RuntimeException. This is an internal-type contract inside ColumnDataClassifier's feature extraction.

Solutions

1. Ensure the features accumulator is created as a ClassicCounter<F> (for weighted features) or ArrayList<F>/Collection (for binary features)
2. If you modified the code, restore the standard Counter/Collection container types in the feature extraction methods
3. Cast or wrap custom containers into a Counter/Collection before calling addFeature

Example fix

// before
Map<String,Double> features = new HashMap<>();
addFeature(features, feat, value);
// after
Counter<String> features = new ClassicCounter<>();
addFeature(features, feat, value);

if (!(features instanceof Counter) && !(features instanceof Collection)) throw new IllegalArgumentException("features must be Counter or Collection");

static boolean isFeatureContainer(Object o) { return o instanceof Counter<?> || o instanceof Collection<?>; }

try { extractFeatures(...); } catch (RuntimeException e) { if (e.getMessage().contains("neither a counter nor a collection")) { log.error("Feature accumulator misconfigured", e); } else throw e; }

When it happens

Trigger: Internal feature-extraction code paths in ColumnDataClassifier construction that pass a features object created as neither Counter nor Collection (e.g. after refactoring the accumulator type); not normally triggerable from public API misuse.

Common situations: Subclassing/modifying ColumnDataClassifier and passing a Map or List-of-tuples as features; version drift where custom column classifiers return a non-standard container from their feature factory.

Understand the failure class

Background: Type mismatch errors: IllegalArgumentException, TypeError and type guards across 150 open-source libraries — this error's family across 150 libraries.

Related errors

• Line format error at line
• Error: Line has too few tab-separated columns
• Dataset could not be loaded
• Not enough columns for format
• Unrecognized format specification in

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/d160bd85ca5fbed2. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/classify/ColumnDataClassifier.java:951

      double sqrt = Math.sqrt(value);
      addFeature(featuresC, "Sqrt", sqrt);
    } else {
      addFeature(featuresC, Flags.realValuedFeaturePrefix, value);
    }
  }

  /**
   * This method takes care of adding features to the collection-ish object features via
   * instanceof checks.  Features must be a type of collection or a counter, and value is used
   * iff it is a counter
   */
  private static <F> void addFeature(Object features, F newFeature, double value) {
    if (features instanceof Counter<?>) {
      ErasureUtils.<Counter<F>>uncheckedCast(features).setCount(newFeature, value);
    } else if(features instanceof Collection<?>) {
      ErasureUtils.<Collection<F>>uncheckedCast(features).add(newFeature);
    } else {
      throw new RuntimeException("addFeature was called with a features object that is neither a counter nor a collection!");
    }
  }

  /**
   * Extracts all the features from a certain input column.
   *
   * @param cWord The String to extract data from
   * @param flags Flags specifying which features to extract
   * @param featuresC Some kind of Collection or Counter to put features into
   * @param goldAns The goldAnswer for this whole datum or emptyString if none.
   *                    This is used only for filling in the binned lengths histogram counters
   */
  private void makeDatum(String cWord, Flags flags, Object featuresC, String goldAns) {

    //logger.info("Making features for " + cWord + " flags " + flags);
    if (flags == null) {
      // no features for this column
      return;

View on GitHub (pinned to 1b7edd19c4)