public double getQuantile()

in src/main/java/org/apache/datasketches/tdigest/TDigestDouble.java [237:284]
41 lines of code
17 McCabe index (conditional complexity)

  public double getQuantile(final double rank) {
    if (isEmpty()) { throw new SketchesStateException(QuantilesAPI.EMPTY_MSG); }
    if (Double.isNaN(rank)) { throw new SketchesArgumentException("Operation is undefined for Nan"); }
    if (rank < 0 || rank > 1) { throw new SketchesArgumentException("Normalized rank must be within [0, 1]"); }

    compress(); // side effect

    if (numCentroids_ == 1) { return centroidMeans_[0]; }

    // at least 2 centroids
    final double weight = rank * centroidsWeight_;
    if (weight < 1) { return minValue_; }
    if (weight > centroidsWeight_ - 1.0) { return maxValue_; }
    final double firstWeight = centroidWeights_[0];
    if (firstWeight > 1 && weight < firstWeight / 2.0) {
      return minValue_ + (weight - 1.0) / (firstWeight / 2.0 - 1.0) * (centroidMeans_[0] - minValue_);
    }
    final double lastWeight = centroidWeights_[numCentroids_ - 1];
    if (lastWeight > 1 && centroidsWeight_ - weight <= lastWeight / 2.0) {
      return maxValue_ + (centroidsWeight_ - weight - 1.0) / (lastWeight / 2.0 - 1.0) * (maxValue_ - centroidMeans_[numCentroids_ - 1]);
    }

    // interpolate between extremes
    double weightSoFar = firstWeight / 2.0;
    for (int i = 0; i < numCentroids_ - 1; i++) {
      final double dw = (centroidWeights_[i] + centroidWeights_[i + 1]) / 2.0;
      if (weightSoFar + dw > weight) {
        // the target weight is between centroids i and i+1
        double leftWeight = 0;
        if (centroidWeights_[i] == 1) {
          if (weight - weightSoFar < 0.5) { return centroidMeans_[i]; }
          leftWeight = 0.5;
        }
        double rightWeight = 0;
        if (centroidWeights_[i + 1] == 1) {
          if (weightSoFar + dw - weight <= 0.5) { return centroidMeans_[i + 1]; }
          rightWeight = 0.5;
        }
        final double w1 = weight - weightSoFar - leftWeight;
        final double w2 = weightSoFar + dw - weight - rightWeight;
        return weightedAverage(centroidMeans_[i], w1, centroidMeans_[i + 1], w2);
      }
      weightSoFar += dw;
    }
    final double w1 = weight - centroidsWeight_ - centroidWeights_[numCentroids_ - 1] / 2.0;
    final double w2 = centroidWeights_[numCentroids_ - 1] / 2.0 - w1;
    return weightedAverage(centroidWeights_[numCentroids_ - 1], w1, maxValue_, w2);
  }