stanfordnlp/CoreNLP · error · RuntimeException

Error loading flags.readerAndWriter

Error message

Error loading flags.readerAndWriter: '%s'

What it means

AbstractSequenceClassifier.makeReaderAndWriter loads the DocumentReaderAndWriter class named by flags.readerAndWriter via reflection; any failure (class missing, no default constructor, init error) is wrapped in a RuntimeException naming the flag value. The root cause is the wrapped exception `e`.

Solutions

1. Check the cause (`e`) of the RuntimeException for ClassNotFoundException or InstantiationException details
2. Verify flags.readerAndWriter matches an existing DocumentReaderAndWriter implementation (e.g. edu.stanford.nlp.sequences.ColumnDocumentReaderAndWriter)
3. Ensure the class is on the classpath (add your jar or the module containing it)
4. Confirm the class has a public no-argument constructor

Example fix

// before (in props)
readerAndWriter = edu.stanford.nlp.sequences.ColumnDocumentReaderAndWritter
// after
readerAndWriter = edu.stanford.nlp.sequences.ColumnDocumentReaderAndWriter

try {
  Class.forName(flags.readerAndWriter);
} catch (ClassNotFoundException e) {
  throw new IllegalStateException("readerAndWriter class not on classpath: " + flags.readerAndWriter);
}

try { rw = classifier.makeReaderAndWriter(); }
catch (RuntimeException e) { log.error("Bad flags.readerAndWriter: {}", e.getMessage(), e.getCause()); throw e; }

When it happens

Trigger: Setting flags.readerAndWriter to a class name that does not exist, is not on the classpath, lacks a no-arg constructor, is not a DocumentReaderAndWriter, or whose constructor throws.

Common situations: Typo in the readerAndWriter property of a classifier config/serialized flag file; using a custom reader class not included in the jar; classpath differences between training and inference environments.

Related errors

• Error loading flags.plainTextDocumentReaderAndWriter
• Unknown class
• Couldn't instantiate CategoryMergingGrammarCompactor.
• Loading: failed with
• Could not load tokenizer factory

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/1be871f064c96374. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/ie/AbstractSequenceClassifier.java:241

  public Set<String> getKnownLCWords() {
    return knownLCWords;
  }

  /**
   * Makes a DocumentReaderAndWriter based on the flags the CRFClassifier
   * was constructed with.  Will create an instance of the class specified in
   * the property flags.readerAndWriter and
   * initialize it with the CRFClassifier's flags.
   *
   * @return The appropriate ReaderAndWriter for training/testing this classifier
   */
  public DocumentReaderAndWriter<IN> makeReaderAndWriter() {
    DocumentReaderAndWriter<IN> readerAndWriter;
    try {
      readerAndWriter = ReflectionLoading.loadByReflection(flags.readerAndWriter);
    } catch (Exception e) {
      throw new RuntimeException(String.format("Error loading flags.readerAndWriter: '%s'", flags.readerAndWriter), e);
    }
    readerAndWriter.init(flags);
    return readerAndWriter;
  }

  /**
   * Makes a DocumentReaderAndWriter based on
   * flags.plainTextReaderAndWriter.  Useful for reading in
   * untokenized text documents or reading plain text from the command
   * line.  An example of a way to use this would be to return a
   * edu.stanford.nlp.wordseg.Sighan2005DocumentReaderAndWriter for
   * the Chinese Segmenter.
   */
  public static <INN extends CoreMap> DocumentReaderAndWriter<INN> makePlainTextReaderAndWriter(SeqClassifierFlags flags) {
    String readerClassName = flags.plainTextDocumentReaderAndWriter;
    // We set this default here if needed because there may be models
    // which don't have the reader flag set
    if (readerClassName == null) {

View on GitHub (pinned to 1b7edd19c4)