Histogram.h

/*
 
MIT License
 
Copyright (c) 2017 FMI Open Development / Markus Peura, first.last@fmi.fi
 
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.
 
*/
/*
Part of Rack development has been done in the BALTRAD projects part-financed
by the European Union (European Regional Development Fund and European
Neighbourhood Partnership Instrument, Baltic Sea Region Programme 2007-2013)
*/
#ifndef HISTOGRAM_H_
#define HISTOGRAM_H_
 
//
#include <drain/Log.h>
#include <drain/TypeUtils.h>
#include <typeinfo>
#include <cmath>
 
#include <vector>
#include <iostream>
#include <stdexcept>
#include <map>
 
#include "ValueScaling.h"
 
 
 
namespace drain
{
 
// // // using namespace std;
 
//std::cerr;
 
 
class Histogram : protected std::vector<unsigned long> {
public:
 
    typedef unsigned long count_t;
    typedef std::vector<count_t> vect_t;
 
    Histogram(size_t size=256);
 
    Histogram(const Histogram & histogram);
 
    virtual ~Histogram(){};
 
    ValueScaling scaling;
 
    void setSize(size_t s);
 
    static
    std::size_t recommendSizeByType(const std::type_info & type, std::size_t defaultValue = 256);
 
    inline
    int getSize() const { return size(); };
 
    inline
    int autoSize(const std::type_info & type){
        if (empty()){
            resize(recommendSizeByType(type, 256));
        }
        return size();
    };
 
    inline
    void clearBins(){
        std::fill(begin(), end(), 0);
    }
 
 
    template <class T>
    void compute(const T & src, const std::type_info & type = typeid(double), const UniTuple<double,2>  & scaling = {1.0, 0.0});
 
    //void clearBins();
 
 
    void setSampleCount(long int n){
        //sampleCount = n;
        sampleCount = n;
        sampleCountMedian = static_cast<size_t>(weight * static_cast<double>(n));
    }
 
    size_t getSampleCount() const {
        return sampleCount;
    }
 
 
    void setRange(double dataMin, double dataMax);
 
    inline
    void setRange(const Range<double> & range){
        setRange(range.min, range.max);
    }
 
    inline
    void deriveScaling(const ValueScaling & s, const Type & type){
        setRange(s.fwd(0.0), s.fwd(1 << (8*drain::Type::call<drain::sizeGetter>(type))));
    }
 
    inline
    void setScale(const ValueScaling & scaling){
        drain::Logger mout(__FILE__, __FUNCTION__);
        mout.discouraged("use setRange instead (scaling=", scaling, ")");
        mout.advice("range perhaps: [", this->scaling.getPhysicalRange());
        this->scaling.assignSequence(scaling);
        //int s = 1 << (8*drain::Type::call<drain::sizeGetter>(type));
    }
 
    /*
    inline
    void setScale(const ValueScaling & scaling){
        //scaling.setRange(0.0, bins-1.0, dataMin, dataMax);
    };
    */
 
    inline
    int getInMin() const { return 0; };
 
    inline
    int getUpperBoundIn() const { return size(); };
 
 
    inline
    int getOutMin() const { return scaling.fwd(0); };
 
    inline
    int getUpperBoundOut() const { return scaling.fwd(size()); };
 
 
 
 
    inline
    void setMedianPosition(double pos){
        weight = pos;
        sampleCountMedian = static_cast<size_t>(weight * static_cast<double>(sampleCount));
    };
 
    /*
    template <class T>
    inline
    bool withinLimits(T i) const {
        return  ((i >= inMin) && (i <= inMax));  // check if ok inMax is too much
    }
    */
 
    template <class T>
    inline
    void increment(T i){
        //++(*this)[((i-inMin)*bins)/inSpan];
        ++(*this)[scaling.inv(i)];
        ++sampleCount;  // TODO: slow?
    }
 
    template <class T>
    inline
    void increment(T i, int count){
        (*this)[scaling.inv(i)] += count;
        //(*this)[((i-inMin)*bins)/inSpan] += count;
        sampleCount += count;
    }
 
    template <class T>
    inline
    void incrementRaw(T i){
        ++(*this)[i];
        ++sampleCount;  // TODO: slow?
    }
 
    template <class T>
    inline
    void incrementRaw(T i, int count){
        (*this)[i] += count;
        sampleCount += count;
    }
 
    template <class T>
    inline
    void decrement(T i){
        --(*this)[scaling.inv(i)]; // [((i-inMin)*bins)/inSpan];
        --sampleCount;
    }
 
    template <class T>
    inline
    void decrement(T i, int count){
        (*this)[scaling.inv(i)] -= count; //[((i-inMin)*bins)/inSpan] -= count;
        sampleCount -= count;
    }
 
    template <class T>
    inline
    void decrementRaw(T i){
        --(*this)[i];
        --sampleCount;
    }
 
    template <class T>
    inline
    void decrementRaw(T i, int count){
        (*this)[i] -= count;
        sampleCount -= count;
    }
 
 
    template <class T>
    inline
    T scaleOut(size_type i) const {
        //return static_cast<T>(outMin + (i*outSpan)/bins);
        return static_cast<T>(scaling.fwd(i));
    }
 
 
 
 
 
    //  @param p applies to weighted median; for standard median, p = 0.5.
    template <class T>
    inline
    T getMax() const {
        for (size_type i = size()-1; i > 0; --i)
            if ((*this)[i] > 0)
                return scaleOut<T>(i);
        // cerr << "Warning: Histogram empty.\n";
        return  getUpperBoundOut(); //static_cast<T>(outMax);
    }
 
 
    template <class T>
    inline
    T getMin() const {
        for (size_type i = 0; i < size(); ++i)
            if ((*this)[i] > 0)
                return scaleOut<T>(i);
        //cerr << "Warning: Histogram empty.\n";
        return  getUpperBoundOut(); //static_cast<T>(outMax);
    }
 
 
    template <class T>
    //inline
    T getSum() const {
        double sum = 0;
        //std::cout << __FUNCTION__ << ':' << sum << std::endl;
        for (size_type i = 0; i < size(); i++){
            sum += ((*this)[i] * scaleOut<T>(i)); // count * nominal (should be scaled to middle)
            //if ((*this)[i])
            //  std::cout << i << '\t' << (*this)[i] << '\t' << scaleOut<T>(i) << '\t' << sum << '\n';
        }
        return sum;
    }
 
 
    template <class T>
    inline
    T getMean() const {
        //return scaleOut(getSum()/sampleCount);
        if (sampleCount > 0)
          return getSum<T>()/sampleCount;
        else
          return 0;
    }
 
 
    template <class T>
    inline
    T getMedian() const {
        size_type sum = 0;
        //const
        //size_type limit = static_cast<size_t>(weight*sampleCount);
 
        for (size_type i = 0; i < size(); ++i){
            sum += (*this)[i];
            if (sum >= sampleCountMedian){ // limit){
                //if (sum >= limit){
                return scaleOut<T>(i);
            }
        }
        //return (245*sum)/sampleCount;
 
        //return ::rand(); //static_cast<T>(outMax);
        return getUpperBoundOut(); //static_cast<T>(outMax);
    }
 
 
    template <class T>
    inline
    T getWeightedMedian(float p) const {
        size_type _sum;
        if ((p < 0.0) || (p>1.0)){
            throw std::runtime_error("Histogram<T>::getMedian: median point <0 or >1.0 .");
        }
        const size_t limit =  sampleCountMedian; //   static_cast<size_t>(p * sampleCount);
        _sum = 0;
        for (size_type i = 0; i < size(); i++){
            _sum += (*this)[i];
            if (_sum >= limit){
                return static_cast<T>(i);
            }
        }
        return static_cast<T>(size());
    }
 
 
    //  NOTE: could be pipelined.
    template <class T>
    inline
    T getVariance() const {
        int n;
        long sum;
        long sum2;
        T f;
        T sumT;
 
        sum = 0;
        sum2 = 0;
        for (size_type i = 0; i < size(); i++){
            f = scaleOut<T>(i);
            n = (*this)[i];
            sum  += n * f;
            sum2 += n * (f*f);
        }
        sumT = static_cast<T>(sum)/sampleCount;
        return  static_cast<T>(sum2)/sampleCount - sumT*sumT;
    }
 
 
    template <class T>
    inline
    T getStdDeviation() const {
        return  static_cast<T>(sqrt(getVariance<double>()));
    }
 
 
 
 
 
 
    inline
    const vect_t getVector() const {
        return *this;
    };
 
    std::ostream & toStream(std::ostream & ostr) const;
 
    void dump(std::ostream & ostr = std::cout);
 
    std::string delimiter;
 
    inline
    double getValue(){
        return (this->*statisticPtr)();
    }
 
 
    inline
    double getValue(char c){
        return (this->*getStatisticPtr(c))();
    }
 
 
    inline
    void setValueFunc(char c){
        statisticPtr = getStatisticPtr(c);
    }
 
 
protected:
 
    void initialize();
 
    typedef double (Histogram::*stat_ptr_t)() const;
 
    double (Histogram::*statisticPtr)() const;
 
    stat_ptr_t getStatisticPtr(char c);
 
 
    //double (Histogram::*)getPtr const;
 
private:
    // size_t bins;
 
    //size_type sampleCount;
 
    size_type sampleCount = 0;
 
    // Half of sampleCount.
    size_type sampleCountMedian = 0;
 
    // weight for weighted median;
    float weight = 0.5;
 
    static inline
    bool isSmallInt(const std::type_info & type){
        return (type == typeid(unsigned char)) || (type == typeid(unsigned short int));
    }
 
 
    static inline
    short getBitShift(unsigned int value){
        short i = 0;
        while ((value = (value>>1)) > 0){
            ++i;
        }
        return i;
    }
 
 
};
 
 
template <class T>
void Histogram::compute(const T & dst, const std::type_info & type, const UniTuple<double,2>  & scaling){
 
    drain::Logger mout(__FILE__, __FUNCTION__);
 
    const int size = autoSize(type);
 
    if (size <= 0){
        mout.error(size, " bins, something went wrong");
        return;
    }
 
    if (size > 0x10000){
        mout.error(size, " bins exceeds Histogram size limit: ", 0x10000);
        return;
    }
 
 
    initialize();
 
    ValueScaling inputScaling(scaling);
 
    const bool SAME_SCALING = (inputScaling == this->scaling);
 
    mout.attention("scalings: ", this->scaling, ", ", inputScaling, " same? ", SAME_SCALING);
    if (inputScaling == this->scaling){
 
    }
 
    if (isSmallInt(type)){
        const unsigned short histBits = getBitShift(size);
        if (size == (1<<histBits)){ //
            const signed short dataBits = (8*drain::Type::call<drain::sizeGetter>(type));
            const short int bitShift = dataBits - histBits;
 
            mout.note("Small int (", dataBits, "b) data, histogram[", size,"] ", histBits, " b, computing with bit shift ", bitShift);
            // setRange(inputScaling.fwd(0), inputScaling.fwd(size -1));
            mout.info("Physical range of the histogram: ", this->scaling.getPhysicalRange());
            // mout.warn("Histogram size ", size, "=2^N, using bitShift=", bitShift, ", computing raw values.");
            for (typename T::const_iterator it = dst.begin(); it != dst.end(); ++it){
                this->incrementRaw(static_cast<unsigned short int>(*it) >> bitShift);
            }
        }
        else {
            mout.advice("Consider histogram size of 2^N, e.g.", (1<<histBits));
        }
        return;
    }
 
    //mout.warn("Data range: ", range, ", mapped (physical) Range: ", rangeOut);
    //mout.warn("Size ", s, " != 2^N (slow mode)");
    //mout.advice("Consider histogram size of 2^N, e.g.", (1<<histBits));
    mout.error("Skipping...");
 
 
}
 
inline
std::ostream & operator<<(std::ostream &ostr,const Histogram &h){
    return h.toStream(ostr);
}
 
}
 
//std::ostream & operator<<(std::ostream &ostr,const drain::Histogram &h);
 
//template <class T>
 
 
#endif /*HISTOGRAM_H_*/
 
// Drain