histogram.h

Go to the documentation of this file.

/* This file is part of the Tomographer project, which is distributed under the
 * terms of the MIT license.
 *
 * The MIT License (MIT)
 *
 * Copyright (c) 2015 ETH Zurich, Institute for Theoretical Physics, Philippe Faist
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to deal
 * in the Software without restriction, including without limitation the rights
 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 * copies of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */

#ifndef HISTOGRAM_H
#define HISTOGRAM_H

#include <cmath>
#include <cstdlib>

#include <iostream>
#include <iomanip> // std::setprecision
#include <sstream> // std::stringstream
#include <stdexcept> // std::out_of_range
#include <type_traits> // std::enable_if

#include <boost/math/constants/constants.hpp>

#include <Eigen/Core>

#include <tomographer/tools/fmt.h>
#include <tomographer/qit/util.h>


namespace Tomographer {


template<typename Scalar_, typename CountType_ = unsigned int>
struct UniformBinsHistogram
{
  typedef Scalar_ Scalar;

  typedef CountType_ CountType;

  static constexpr bool HasErrorBars = false;

  struct Params {
    Params(Scalar min_ = 0.0, Scalar max_ = 1.0, std::size_t num_bins_ = 50)
      : min(min_), max(max_), num_bins(num_bins_)
    {
    }
    template<typename OtherParams/*,
             typename std::enable_if<(tomo_internal::is_histogram_params_type<OtherParams>::value), bool>::type
             dummy2 = true*/>
    Params(const OtherParams& other)
      : min(other.min), max(other.max), num_bins(other.num_bins)
    {
    }

    Scalar min;
    Scalar max;
    std::size_t num_bins;

    inline bool is_within_bounds(Scalar value) const
    {
      return std::isfinite(value) && value >= min && value < max;
    }
    inline std::size_t bin_index(Scalar value) const
    {
      if ( !is_within_bounds(value) ) {
        throw std::out_of_range(streamstr("UniformBinsHistogram::Params: Value "<<value
                                          <<" out of range ["<<min<<","<<max<<"["));
      }
      return bin_index_unsafe(value);
    }
    inline std::size_t bin_index_unsafe(Scalar value) const
    {
      return (std::size_t)((value-min) / (max-min) * num_bins);
    }
    inline Scalar bin_lower_value(std::size_t index) const
    {
      eigen_assert(Tools::is_positive(index) && (std::size_t)index < num_bins);
      return min + index * (max-min) / num_bins;
    }
    inline Scalar bin_center_value(std::size_t index) const
    {
      eigen_assert(Tools::is_positive(index) && (std::size_t)index < num_bins);
      return min + (index+boost::math::constants::half<Scalar>()) * (max-min) / num_bins;
    }
    inline Scalar bin_upper_value(std::size_t index) const
    {
      eigen_assert(Tools::is_positive(index) && (std::size_t)index < num_bins);
      return min + (index+1) * (max-min) / num_bins;
    }
    inline Scalar bin_resolution() const
    {
      return (max - min) / num_bins;
    }
  };

  Params params;
  Eigen::Array<CountType, Eigen::Dynamic, 1> bins;
  CountType off_chart;

  UniformBinsHistogram(Params p = Params())
    : params(p), bins(Eigen::Array<CountType,Eigen::Dynamic,1>::Zero(p.num_bins)),
      off_chart(0)
  {
  }

  UniformBinsHistogram(Scalar min_, Scalar max_, std::size_t num_bins)
    : params(min_, max_, num_bins), bins(Eigen::Array<CountType,Eigen::Dynamic,1>::Zero(num_bins)),
      off_chart(0)
  {
  }

  template<typename HistogramType>
  UniformBinsHistogram(const HistogramType & other)
    : params(other.params), bins(other.params.num_bins), off_chart(other.off_chart)
  {
    bins = other.bins.template cast<CountType>();
  }

  inline void reset()
  {
    bins.resize(params.num_bins);
    bins.setZero();
    off_chart = 0;
  }

  template<typename EigenType>
  inline void load(const Eigen::DenseBase<EigenType> & x, CountType off_chart_ = 0)
  {
    eigen_assert(x.cols() == 1);
    eigen_assert((std::size_t)x.rows() == params.num_bins);
    bins = x.derived().template cast<CountType>();
    off_chart = off_chart_;
  }

  template<typename EigenType>
  inline void add(const Eigen::ArrayBase<EigenType> & x, CountType off_chart_ = 0)
  {
    // the argument must be of ArrayBase type (as opposed to load() where we can
    // also accept MatrixBase types) because Eigen doesn't allow operator+=
    // between Arrays and Matrices, but has an operator= .
    eigen_assert(x.cols() == 1);
    eigen_assert((std::size_t)x.rows() == params.num_bins);
    bins += x.derived().template cast<CountType>();
    off_chart += off_chart_;
  }

  template<typename OtherScalar, typename OtherCountType>
  inline void add(const UniformBinsHistogram<OtherScalar,OtherCountType> & x)
  {
    eigen_assert(x.bins.cols() == 1);
    eigen_assert((std::size_t)x.bins.rows() == params.num_bins);
    eigen_assert(std::fabs(x.params.min - params.min) < 1e-8);
    eigen_assert(std::fabs(x.params.max - params.max) < 1e-8);
    bins += x.bins.template cast<CountType>();
    off_chart += x.off_chart;
  }

  inline std::size_t num_bins() const
  {
    return params.num_bins;
  }

  inline CountType count(std::size_t i) const
  {
    return bins(i);
  }

  inline bool is_within_bounds(Scalar value) const
  {
    return params.is_within_bounds(value);
  }
  inline std::size_t bin_index(Scalar value) const
  {
    return params.bin_index(value);
  }
  inline Scalar bin_lower_value(int index) const
  {
    return params.bin_lower_value(index);
  }
  inline Scalar bin_center_value(std::size_t index) const
  {
    return params.bin_center_value(index);
  }
  inline Scalar bin_upper_value(std::size_t index) const
  {
    return params.bin_upper_value(index);
  }

  inline Scalar bin_resolution() const
  {
    return params.bin_resolution();
  }

  inline std::size_t record(Scalar value)
  {
    if ( !is_within_bounds(value) ) {
      ++off_chart;
      return std::numeric_limits<std::size_t>::max();
    }
    // calling bin_index_unsafe because we have already checked that value is in range.
    const std::size_t index = params.bin_index_unsafe(value);
    ++bins( index );
    return index;
  }

  inline std::size_t record(Scalar value, CountType weight)
  {
    if ( !is_within_bounds(value) ) {
      off_chart += weight;
      return std::numeric_limits<std::size_t>::max();
    }
    // calling bin_index_unsafe is safe here because we have already checked that value is
    // in range.
    const std::size_t index = params.bin_index_unsafe(value);
    bins(index) += weight;
    return index;
  }

  inline std::string pretty_print(int max_width = 0) const
  {
    return histogram_pretty_print(*this, max_width);
  }

};



template<typename Scalar_, typename CountType_ = double>
struct UniformBinsHistogramWithErrorBars : public UniformBinsHistogram<Scalar_, CountType_>
{
  typedef Scalar_ Scalar;
  typedef CountType_ CountType;

  typedef UniformBinsHistogram<Scalar_, CountType_> Base_;
  typedef typename Base_::Params Params;
  
  static constexpr bool HasErrorBars = true;

  Eigen::Array<CountType, Eigen::Dynamic, 1> delta;

  // make these accessible without having to use the "Base_::member" syntax all the time
  using Base_::params;
  using Base_::bins;
  using Base_::off_chart;


  UniformBinsHistogramWithErrorBars(Params params = Params())
    : Base_(params), delta(Eigen::Array<CountType, Eigen::Dynamic, 1>::Zero(params.num_bins))
  {
  }

  UniformBinsHistogramWithErrorBars(Scalar min, Scalar max, std::size_t num_bins)
    : Base_(min, max, num_bins), delta(Eigen::Array<CountType, Eigen::Dynamic, 1>::Zero(num_bins))
  {
  }

  inline void reset()
  {
    Base_::reset();
    delta.resize(Base_::num_bins());
    delta.setZero();
  }
  
  inline CountType errorbar(std::size_t i) const
  {
    return delta(i);
  }

  std::string pretty_print(int max_width = 0) const
  {
    return histogram_pretty_print(*this, max_width);
  }
  
};







template<typename HistogramType_, typename RealAvgType = double>
struct AveragedHistogram
  : public UniformBinsHistogramWithErrorBars<typename HistogramType_::Scalar, RealAvgType>
{
  typedef HistogramType_ HistogramType;
  typedef UniformBinsHistogramWithErrorBars<typename HistogramType_::Scalar, RealAvgType> Base_;

  typedef typename Base_::Params Params;
  typedef typename Base_::Scalar Scalar;
  typedef typename Base_::CountType CountType;

  static constexpr bool HasErrorBars = true;

  int num_histograms;


  AveragedHistogram(const Params& params = Params())
    : Base_(params), num_histograms(0)
  {
  }

  inline void reset(const Params& params_)
  {
    Base_::params = params_;
    Base_::reset();
    num_histograms = 0;
  }

  inline void reset()
  {
    Base_::reset();
    num_histograms = 0;
  }

  // ---------------------------------------------------------------------------
  // Implementation in case the added histograms don't have their own error bars
  // ---------------------------------------------------------------------------

  template<bool dummy = true,
           typename std::enable_if<dummy && (!HistogramType::HasErrorBars), bool>::type dummy2 = true>
  inline void add_histogram(const HistogramType& histogram)
  {
    // bins collects the sum of the histograms
    // delta for now collects the sum of squares. delta will be normalized in run_finished().

    eigen_assert((typename HistogramType::CountType)histogram.num_bins() ==
                 (typename HistogramType::CountType)Base_::num_bins());

    for (std::size_t k = 0; k < histogram.num_bins(); ++k) {
      RealAvgType binvalue = histogram.count(k);
      Base_::bins(k) += binvalue;
      Base_::delta(k) += binvalue * binvalue;
    }

    Base_::off_chart += histogram.off_chart;
    ++num_histograms;
  }
  template<bool dummy = true,
           typename std::enable_if<dummy && (!HistogramType::HasErrorBars), bool>::type dummy2 = true>
  inline void finalize()
  {
    Base_::bins /= num_histograms;
    Base_::delta /= num_histograms;
    Base_::off_chart /= num_histograms;

    // delta = sqrt(< X^2 > - < X >^2) / sqrt(Nrepeats-1)
    auto finhist2 = Base_::bins*Base_::bins; // for array, this is c-wise product
    Base_::delta = ( (Base_::delta - finhist2) / (num_histograms-1) ).sqrt();
  }

  // ---------------------------------------------------------------------------
  // Implementation in case the added histograms do have their own error bars
  // ---------------------------------------------------------------------------

  template<bool dummy = true,
           typename std::enable_if<dummy && (HistogramType::HasErrorBars), bool>::type dummy2 = true>
  inline void add_histogram(const HistogramType& histogram)
  {
    // bins collects the sum of the histograms
    // delta for now collects the sum of squares. delta will be normalized in run_finished().

    eigen_assert((typename HistogramType::CountType)histogram.num_bins() == Base_::num_bins());

    for (std::size_t k = 0; k < histogram.num_bins(); ++k) {
      RealAvgType binvalue = histogram.count(k);
      Base_::bins(k) += binvalue;
      RealAvgType bindelta = histogram.errorbar(k);
      Base_::delta(k) += bindelta*bindelta;
    }

    Base_::off_chart += histogram.off_chart;
    ++num_histograms;
  }
  template<bool dummy = true,
           typename std::enable_if<dummy && (HistogramType::HasErrorBars), bool>::type dummy2 = true>
  inline void finalize()
  {
    Base_::bins /= num_histograms;
    Base_::off_chart /= num_histograms;

    Base_::delta = Base_::delta.sqrt();
    Base_::delta /= num_histograms;
  }

};





namespace tomo_internal {
// internal helpers
//
// get labels left of histogram (generic HistogramType interface: no information, just bin #)
template<typename HistogramType>
struct histogram_pretty_print_label
{
  static inline void getLabels(std::vector<std::string> & labels, const HistogramType & hist)
  {
    labels.resize(hist.num_bins());
    for (std::size_t k = 0; k < hist.num_bins(); ++k) {
      labels[k] = std::to_string(k);
    }
  }
};
// get labels left of histogram (for the family of histograms with 'Params':
// UniformBinsHistogram, UniformBinsHistogramWithErrorBars etc.)
//
// common code for several specializations of histogram_pretty_print_label
template<typename HistogramType>
inline void histogram_get_labels_for_hist_params(std::vector<std::string> & labels, const HistogramType& hist)
{
  labels.resize(hist.num_bins());

  const double max_label_val = std::max(hist.bin_center_value(0), hist.bin_center_value(hist.num_bins()-1));
  const int powten = (int)std::floor(std::log10(max_label_val));
  const int relprecision = 4;
  const int precision = (powten > relprecision) ? 0 : (relprecision - powten - 1);

  for (std::size_t k = 0; k < hist.num_bins(); ++k) {
    std::ostringstream ss;
    ss << std::fixed << std::setprecision(precision) << std::right << hist.bin_center_value(k);
    labels[k] = ss.str();
  }
}
// specialize histogram pretty-print labels using the code above
template<typename Scalar_, typename CountType_>
struct histogram_pretty_print_label<UniformBinsHistogram<Scalar_, CountType_> >
{
  typedef UniformBinsHistogram<Scalar_, CountType_> HistogramType;
  static inline void getLabels(std::vector<std::string> & labels, const HistogramType & hist)
  {
    histogram_get_labels_for_hist_params<HistogramType>(labels, hist);
  }
};
// specialize histogram pretty-print labels using the code above
template<typename Scalar_, typename CountType_>
struct histogram_pretty_print_label<UniformBinsHistogramWithErrorBars<Scalar_, CountType_> >
{
  typedef UniformBinsHistogramWithErrorBars<Scalar_, CountType_> HistogramType;
  static inline void getLabels(std::vector<std::string> & labels, const HistogramType & hist)
  {
    histogram_get_labels_for_hist_params<HistogramType>(labels, hist);
  }
};
// specialize histogram pretty-print labels using the code above
template<typename BaseHistogramType_, typename RealAvgType_>
struct histogram_pretty_print_label<AveragedHistogram<BaseHistogramType_, RealAvgType_> >
{
  typedef AveragedHistogram<BaseHistogramType_, RealAvgType_> HistogramType;
  static inline void getLabels(std::vector<std::string> & labels, const HistogramType & hist)
  {
    histogram_get_labels_for_hist_params<HistogramType>(labels, hist);
  }
};

// format bin counts nicely
template<typename HistogramType>
struct histogram_pretty_print_value
{
  const HistogramType & hist;
  histogram_pretty_print_value(const HistogramType & hist_) : hist(hist_) { }

  template<bool dummy = true,
           typename std::enable_if<dummy && !HistogramType::HasErrorBars, bool>::type dummy2 = true>
  static inline void getStrValues(std::vector<std::string> & svalues, const HistogramType & hist)
  {
    svalues.resize(hist.num_bins());

    std::size_t k;
    double max_val = 1.0;
    for (k = 0; k < hist.num_bins(); ++k) {
      if (k == 0 || hist.count(k) > max_val) {
        max_val = hist.count(k);
      }
    }

    const int powten = (int)std::floor(std::log10(max_val));
    const int relprecision = 3;
    const int precision = abs_precision_for(powten, relprecision);
    const int w = (precision > 0) ? (precision+powten+1) : (relprecision+2);

    for (std::size_t k = 0; k < hist.num_bins(); ++k) {
      std::stringstream ss;
      ss << std::setprecision(precision) << std::fixed << std::right << std::setw(w)
         << hist.count(k);
      svalues[k] = ss.str();
    }
  }
  template<bool dummy = true,
           typename std::enable_if<dummy && HistogramType::HasErrorBars, bool>::type dummy2 = true>
  static inline void getStrValues(std::vector<std::string> & svalues, const HistogramType & hist)
  {
    svalues.resize(hist.num_bins());

    std::size_t k;
    double max_val = 1.0;
    for (k = 0; k < hist.num_bins(); ++k) {
      if (k == 0 || hist.count(k) > max_val) {
        max_val = hist.count(k);
      }
    }

    const int powten = (int)std::floor(std::log10(max_val)); // floor of log_{10}(...)
    const int relprecision = 3;
    const int precision = abs_precision_for(powten, relprecision);
    const int w = (precision > 0) ? (precision+powten+1) : (relprecision+2);

    for (k = 0; k < hist.num_bins(); ++k) {
      std::stringstream ss;
      ss << std::setprecision(precision) << std::fixed << std::right << std::setw(w)
         << hist.count(k)
         << " +- "
         << std::setprecision(abs_precision_for(powten-1, relprecision-1)) << std::setw(w)
         << hist.errorbar(k);
      svalues[k] = ss.str();
    }
  }
private:
  static inline int abs_precision_for(const int powten, const int relprecision)
  {
    return (powten > relprecision) ? 0 : (relprecision - powten - 1);
  }
};

//
// access this with public API using histogram_pretty_print().
//
template<typename HistogramType>
struct histogram_pretty_printer
{
  const HistogramType & hist;
  const int max_width;

  const std::string lsep;
  const std::string rsep;

  std::vector<std::string> labels;
  int maxlabelwidth;

  std::vector<std::string> svalues;
  int maxsvaluewidth;

  double max_value;

  int max_bar_width;
  double barscale;

  histogram_pretty_printer(const HistogramType & hist_, const int max_width_)
    : hist(hist_), max_width(max_width_), lsep(" |"), rsep("  ")
  {
    // first pass:
    //  -  determine the maximum value attained in the histogram
    //  -  determine maximum width of formatted label & value fields.

    labels.resize(hist.num_bins());
    svalues.resize(hist.num_bins());

    histogram_pretty_print_label<HistogramType>::getLabels(labels, hist);
    histogram_pretty_print_value<HistogramType>::getStrValues(svalues, hist);

    std::size_t k;
    for (k = 0; k < hist.num_bins(); ++k) {
      double val = maxval(k);

      if (k == 0 || val > max_value) {
        max_value = val;
      }
      if (k == 0 || (int)labels[k].size() > maxlabelwidth) {
        maxlabelwidth = labels[k].size();
      }
      if (k == 0 || (int)svalues[k].size() > maxsvaluewidth) {
        maxsvaluewidth = svalues[k].size();
      }
    }

    max_bar_width = max_width - maxlabelwidth - maxsvaluewidth - lsep.size() - rsep.size();
    if (max_bar_width < 2) {
      max_bar_width = 2;
    }
    barscale = ((max_value > 0) ? (max_value / max_bar_width) : 1.0);
  }

  inline int value_to_bar_length(double val) const
  {
    if (val < 0) {
      val = 0;
    }
    int l = (int)(val/barscale+0.5);
    if (l >= max_bar_width) {
      return max_bar_width-1;
    }
    return l;
  }

  inline void fill_str_len(std::string & s, double valstart, double valend,
                           char c, char clside, char crside) const
  {
    int vs = value_to_bar_length(valstart);
    int ve = value_to_bar_length(valend);
    eigen_assert(vs >= 0);
    eigen_assert(vs < (int)s.size());
    eigen_assert(ve >= 0);
    eigen_assert(ve < (int)s.size());
    for (int j = vs; j < ve; ++j) {
      s[j] = c;
    }
    if (clside && crside && clside != crside && vs == ve) {
      if (ve < (int)s.size()-1) {
        ++ve;
      } else if (vs > 1) {
        --vs;
      }
    }
    if (clside) {
      s[vs] = clside;
    }
    if (crside) {
      s[ve] = crside;
    }
  }

  inline void pretty_print(std::ostream & str) const
  {
    // perform now second pass:
    //   - display the histogram line by line, with the calculated widths.

    std::size_t k;

    for (k = 0; k < hist.num_bins(); ++k) {
      str << std::setw(maxlabelwidth) << labels[k] << lsep
          << make_bar(k) << rsep << std::setw(maxsvaluewidth) << svalues[k] << "\n";
    }
  }

private:
  // maxval(k): how much this bar may extend in length
  template<bool dummy = true,
           typename std::enable_if<dummy && !HistogramType::HasErrorBars, bool>::type dummy2 = true>
  inline double maxval(const std::size_t k) const
  {
    return (double)hist.count(k);
  }
  template<bool dummy = true,
           typename std::enable_if<dummy && HistogramType::HasErrorBars, bool>::type dummy2 = true>
  inline double maxval(const std::size_t k) const
  {
    return (double)(hist.count(k) + hist.errorbar(k));
  }
  // make_bar(k): produce the histogram bar in characters...
  template<bool dummy = true,
           typename std::enable_if<dummy && !HistogramType::HasErrorBars, bool>::type dummy2 = true>
  inline std::string make_bar(std::size_t k) const
  {
    std::string sbar(max_bar_width, ' ');
    fill_str_len(sbar, 0.0, hist.count(k), '*', 0, 0);
    return sbar;
  }
  template<bool dummy = true,
           typename std::enable_if<dummy && HistogramType::HasErrorBars, bool>::type dummy2 = true>
  inline std::string make_bar(std::size_t k) const
  {
    std::string sbar(max_bar_width, ' ');
    const double binval = hist.count(k);
    const double binerr = hist.errorbar(k);
    fill_str_len(sbar, 0.0, binval - binerr, '*', '*', '*');
    fill_str_len(sbar, binval - binerr, binval + binerr, '-', '|', '|');
    return sbar;
  }
};

template<typename HistogramType>
inline std::string histogram_short_bar_fmt(const HistogramType & histogram, const bool log_scale,
                                           const int max_width)
{
  std::string s = Tools::fmts("%.2g|", (double)histogram.bin_lower_value(0));
  std::string send = Tools::fmts("|%.2g", (double)histogram.bin_upper_value(histogram.num_bins()-1));
  if (histogram.off_chart > 0) {
    send += Tools::fmts(" [+%.1g off]", (double)histogram.off_chart);
  }
  
  const int maxbarwidth = max_width - s.size() - send.size();
  const int numdiv = (int)(std::ceil((float)histogram.num_bins() / maxbarwidth) + 0.5f);
  const int barwidth = (int)(std::ceil((float)histogram.num_bins() / numdiv) + 0.5f);

  Eigen::Array<typename HistogramType::CountType,Eigen::Dynamic,1> vec(barwidth);
  Eigen::ArrayXf veclog(barwidth);

  int k;
  float minlogval = 0;
  float maxlogval = 0;
  for (k = 0; k < barwidth; ++k) {
    vec(k) = histogram.bins.segment(numdiv*k, std::min((std::size_t)numdiv,
                                                       (std::size_t)(histogram.bins.size()-numdiv*k))).sum();
    if (vec(k) > 0) {
      if (log_scale) {
        veclog(k) = std::log((float)vec(k));
      } else {
        veclog(k) = (float)vec(k);
      }
      if (k == 0 || minlogval > veclog(k)) {
        minlogval = veclog(k);
      }
      if (k == 0 || maxlogval < veclog(k)) {
        maxlogval = veclog(k) + 1e-6f;
      }
    } else {
      veclog(k) = 0.f;
    }
  }

  // now, prepare string
  const std::string chars = ".-+ox%#";
  for (k = 0; k < barwidth; ++k) {
    if (vec(k) <= 0) {
      s += ' ';
    } else {
      int i = (int)(chars.size() * (veclog(k) - minlogval) / (maxlogval - minlogval));
      if (i < 0) { i = 0; }
      if (i >= (int)chars.size()) { i = chars.size()-1; }
      s += chars[i];
    }
  }

  s += send;
  
  return s;
}

int maybe_default_col_width(int max_width = 0)
{
  if (max_width <= 0) {
    const int offset = max_width;
    // decide of a maximum width to display
    max_width = 100; // default maximum width
    // If the user provided a value for the terminal width, use it. Note that $COLUMNS is
    // not in the environment usually, so you have to set it manually with e.g.
    //    shell> export COLUMNS=$COLUMNS
    const char * cols_s = std::getenv("COLUMNS");
    if (cols_s != NULL) {
      max_width = std::atoi(cols_s);
    }
    max_width += offset; // if we had given, e.g. maxwidth=-4
  }
  return max_width;
}

} // namespace tomo_internal


template<typename HistogramType>
inline void histogram_pretty_print(std::ostream & str, const HistogramType & histogram, int max_width = 0)
{
  eigen_assert(Tools::is_positive(histogram.params.num_bins));

  if (histogram.params.num_bins == 0) {
    str << "<empty histogram: no bins>\n";
    return;
  }

  max_width = tomo_internal::maybe_default_col_width(max_width);
  tomo_internal::histogram_pretty_printer<HistogramType>(histogram, max_width).pretty_print(str);
}

template<typename HistogramType>
inline std::string histogram_pretty_print(const HistogramType & histogram, int max_width = 0)
{
  std::ostringstream ss;
  histogram_pretty_print<HistogramType>(ss, histogram, max_width);
  return ss.str();
}

template<typename HistogramType>
inline int histogram_short_bar(std::ostream & str, const HistogramType & histogram,
                               bool log_scale = true, int max_width = 0)
{
  eigen_assert(Tools::is_positive(histogram.params.num_bins));

  max_width = tomo_internal::maybe_default_col_width(max_width);

  std::string s;
  if (histogram.params.num_bins == 0) {
    s = "<empty histogram: no bins>";
  } else {
    s = tomo_internal::histogram_short_bar_fmt<HistogramType>(histogram, log_scale, max_width);
  }

  str << s;
  return max_width - s.size();
}
template<typename HistogramType>
inline std::string histogram_short_bar(const HistogramType & histogram, bool log_scale = true, int max_width = 0)
{
  eigen_assert(Tools::is_positive(histogram.params.num_bins));

  if (histogram.params.num_bins == 0) {
    return "<empty histogram: no bins>";
  }

  max_width = tomo_internal::maybe_default_col_width(max_width);
  return tomo_internal::histogram_short_bar_fmt<HistogramType>(histogram, log_scale, max_width);
}






} // namespace Tomographer




#endif