stanfordnlp/CoreNLP · error · IllegalArgumentException

Cosine is not between -1 and 1: " + cosValue

Error message

Cosine is not between -1 and 1: " + cosValue

What it means

SloppyMath.acos(double cosValue) uses a precomputed lookup table and requires the input cosine to be within [-1, 1]. Floating-point rounding (or a genuinely invalid input) outside that range throws IllegalArgumentException, as documented in its @throws clause.

Solutions

1. Clamp before calling: cosValue = Math.max(-1.0, Math.min(1.0, cosValue))
2. Ensure vectors are properly normalized (divide by norm, guarding norm == 0)
3. Use Math.acos if you don't need this method's approximation

Example fix

// before
double angle = SloppyMath.acos(dot / (normA * normB)); // 1.0000000002
// after
double cos = Math.max(-1.0, Math.min(1.0, dot / (normA * normB)));
double angle = SloppyMath.acos(cos);

if (cosValue < -1.0 || cosValue > 1.0) cosValue = Math.max(-1.0, Math.min(1.0, cosValue));

static double clampUnit(double v) { return Math.max(-1.0, Math.min(1.0, v)); }

try {
  double angle = SloppyMath.acos(cos);
} catch (IllegalArgumentException e) {
  double angle = Math.acos(Math.max(-1.0, Math.min(1.0, cos)));
}

When it happens

Trigger: Calling acos with a cosine computed as 1.0000000001 from rounding (e.g. dot product of normalized vectors), or with unnormalized vectors yielding values like 3.2.

Common situations: Vector similarity code where normalization was skipped or divided by zero norm; accumulated floating-point drift slightly exceeding 1.0.

Understand the failure class

Background: "value must be between 0 and 1" / "out of range" / "must not be negative" errors: fixing range-validation failures across open-source libraries — this error's family across 42 libraries.

Related errors

• vectorName + " element " + i + " is " + vector[i]
• Math.exp(-lambda) +" "+ Math.pow(lambda, x) + ' ' +…
• neg log lik smaller than 0: " + s
• Counters.dotProduct infinite or NaN value for key:
• Gradient is numerically zero, stopped on machine epsilon.

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/220fb54fe91578b2. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/math/SloppyMath.java:644

    }
    else {
      double den = 1.0 + Math.exp(-x);
      return 1.0 / den;
    }
  }


  private static float[] acosCache; // = null;

  /**
   * Compute acos very quickly by directly looking up the value.
   * @param cosValue The cosine of the angle to fine.
   * @return The angle corresponding to the cosine value.
   * @throws IllegalArgumentException if cosValue is not between -1 and 1
   */
  public static double acos(double cosValue) {
    if (cosValue < -1.0 || cosValue > 1.0) {
      throw new IllegalArgumentException("Cosine is not between -1 and 1: " + cosValue);
    }
    int numSamples = 10000;
    if (acosCache == null) {
      acosCache = new float[numSamples + 1];
      for (int i = 0; i <= numSamples; ++i) {
        double x = 2.0 / ((double) numSamples) * ((double) i) - 1.0;
        acosCache[i] = (float) Math.acos(x);
      }
    }

    int i = ((int) (((cosValue + 1.0) / 2.0) * ((double) numSamples)));
    return acosCache[i];
  }


  public static double poisson(int x, double lambda) {
    if (x<0 || lambda<=0.0) throw new RuntimeException("Bad arguments: " + x + " and " + lambda);
    double p = (Math.exp(-lambda) * Math.pow(lambda, x)) / factorial(x);

View on GitHub (pinned to 1b7edd19c4)