stanfordnlp/CoreNLP · critical · RuntimeException

Could not open temporary feature index file for writing.

Error message

Could not open temporary feature index file for writing.

What it means

With flags.saveFeatureIndexToDisk enabled, training writes the feature index to a temp file per fold via IOUtils.writeObjectToTempFile. If that write throws IOException (unwritable temp dir, no disk space, security restrictions), it is wrapped in this RuntimeException because training cannot proceed without persisting the index.

Solutions

  1. Check disk space and make java.io.tmpdir writable, or set -Djava.io.tmpdir to a writable directory.
  2. Disable saveFeatureIndexToDisk so the feature index stays in memory if it fits.
  3. Inspect the wrapped IOException (root cause) for the precise IO failure and fix it.

Example fix

// before
java -Djava.io.tmpdir=/nonexistent -cp ... edu.stanford.nlp.ie.crf.CRFClassifier ...
// after
java -Djava.io.tmpdir=/data/tmp -cp ... edu.stanford.nlp.ie.crf.CRFClassifier ...
// or remove saveFeatureIndexToDisk=true from the properties
Defensive patterns

Strategy: try-catch

Validate before calling

File tmpDir = new File(System.getProperty("java.io.tmpdir"));
if (!tmpDir.canWrite() || tmpDir.getUsableSpace() < 100L * 1024 * 1024) {
  throw new IllegalStateException("tmpdir not writable or low on space: " + tmpDir);
}
// or simply don't set saveFeatureIndexToDisk=true when the index fits in memory

Try / catch

try {
  classifier.train(files);
} catch (RuntimeException e) {
  if (String.valueOf(e.getMessage()).contains("temporary feature index file for writing")) {
    props.remove("saveFeatureIndexToDisk"); // keep index in memory
    // rebuild classifier and retry, or fix tmpdir first
  } else throw e;
}

Prevention

When it happens

Trigger: flags.saveFeatureIndexToDisk=true and IOUtils.writeObjectToTempFile throws IOException — e.g. java.io.tmpdir not writable, disk full, or a security manager blocking temp file creation.

Common situations: Read-only or full /tmp on shared servers and containers; restrictive tmpdir; sandboxed CI environments with no temp write access.

Understand the failure class

Background: "failed to write file", "Could not save figure", "Error saving remote file" — file write failed: causes and fixes across languages and libraries — this error's family across 38 libraries.

Related errors

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/b1f03dfddc68483b. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/ie/crf/CRFClassifier.java:1646

        evaluators = new Evaluator[evaluatorList.size()];
        evaluatorList.toArray(evaluators);
      }

      if (flags.numTimesPruneFeatures == i) {
        docs = null; // hopefully saves memory
      }
      // save feature index to disk and read in later
      File featIndexFile = null;

      // CRFLogConditionalObjectiveFunction.featureIndex = featureIndex;
      // int numFeatures = featureIndex.size();
      if (flags.saveFeatureIndexToDisk) {
        try {
          log.info("Writing feature index to temporary file.");
          featIndexFile = IOUtils.writeObjectToTempFile(featureIndex, "featIndex" + i + ".tmp");
          // featureIndex = null;
        } catch (IOException e) {
          throw new RuntimeException("Could not open temporary feature index file for writing.");
        }
      }

      // first index is the number of the document
      // second index is position in the document also the index of the
      // clique/factor table
      // third index is the number of elements in the clique/window these
      // features are for (starting with last element)
      // fourth index is position of the feature in the array that holds them
      // element in data[i][j][k][m] is the index of the mth feature occurring
      // in position k of the jth clique of the ith document
      int[][][][] data = dataAndLabelsAndFeatureVals.first();
      // first index is the number of the document
      // second index is the position in the document
      // element in labels[i][j] is the index of the correct label (if it
      // exists) at position j in document i
      int[][] labels = dataAndLabelsAndFeatureVals.second();
      double[][][][] featureVals = dataAndLabelsAndFeatureVals.third();

View on GitHub (pinned to 1b7edd19c4)