Contour levels discretization setting ?

[jschueller]

What is the idea?

After the recent Contour changes I wonder if we should have a Contour levels discretization setting.

void Contour::buildDefaultLevels(const UnsignedInteger number)
{
  const UnsignedInteger size = data_.getSize();
  levels_.resize(number);
  if (levelsDiscretization_ == "linear")
  {
    const Scalar xMin = data_.getMin()[0];
    const Scalar xMax = data_.getMax()[0];
    for (UnsignedInteger i = 0; i < number; ++ i)
      levels_[i] = xMin + (i * 1.0 / (number - 1.0)) * (xMax - xMin);
  }
  else if (levelsDiscretization_ == "ranks")
  {
    // Use the empirical quantiles
    const Sample sortedData(data_.sort(0));
    for (UnsignedInteger i = 0; i < number; ++ i)
      levels_[i] = sortedData(static_cast<UnsignedInteger>(size * (i + 0.5) / number), 0);
    levels_.erase(std::unique(levels_.begin(), levels_.end()), levels_.end());
  }
}

Currently if we want a plot with linear discretization we should erase the default ranks discretization and let matplotlib compute its own.

Why is this needed?

No response

Additional Context

No response

[mbaudin47]

@josephmure : What's your point of view on this topic?