#include "UnbinnedLogLikelihoodEstimator.h"
#include "core/Model.h"
#include "fit/data/Data.h"

#include <cmath>
#include <iostream>

UnbinnedLogLikelihoodEstimator::UnbinnedLogLikelihoodEstimator() :
		last_integral(0.0), skip_integral(false), integral_skip_counter(0), integral_skip_amount(
				5) {
	// TODO Auto-generated constructor stub
}

UnbinnedLogLikelihoodEstimator::~UnbinnedLogLikelihoodEstimator() {
	// TODO Auto-generated destructor stub
}

double UnbinnedLogLikelihoodEstimator::eval() {
	//calculate loglikelihood
	// poisson sum_i(y_i * ln (f(x_i)) - f(x_i))

	double loglikelihood = 0.0;

	std::vector<DataPointProxy> &data_points = data->getData();
	// loop over data
	for (unsigned int i = 0; i < data_points.size(); i++) {
		shared_ptr<DataStructs::unbinned_data_point> data_point;
		if (data_points[i].isPointUsed()) {
			data_point = data_points[i].getUnbinnedDataPoint();
			double model_value = fit_model->evaluate(data_point->x);
			// if model is zero at this point should be removed otherwise log(0)!!!
			if (model_value <= 0.0)
				continue;

			loglikelihood += log(model_value);
		}
	}

	// normalization integral
	std::cout << "calculating normalization integral..." << std::endl;
	std::vector<DataStructs::DimensionRange> int_ranges;
	int_ranges.push_back(estimator_options.getFitRangeX());
	int_ranges.push_back(estimator_options.getFitRangeY());

	double integral = fit_model->Integral(int_ranges, 1e-3);
	std::cout << "calculated!" << std::endl;
	std::cout << data_points.size() << std::endl;
	/*if (!skip_integral) {
		if (integral > 0.0) {
			if (fabs(integral - last_integral) / integral < 1e-2) {
				skip_integral = false;
				integral_skip_counter = integral_skip_amount;
			}
		}
		last_integral = integral;
		std::cout << "calculated!" << std::endl;
	} else {
		if (integral_skip_counter < 0)
			skip_integral = false;
		else
			--integral_skip_counter;
	}*/

	return -(loglikelihood-data_points.size()*integral);
}