4 template<typename VarType, typename WeightType>
5 template<typename BinType>
6 void Histogram<VarType, WeightType>::ComputationImpl<BinType>::clip(VarType min, VarType max) {
9 throw Elements::Exception("Clipping with min > max can not be done");
11 auto min_bin = m_binning.getBinIndex(min);
12 if (min_bin > m_clip_left && min_bin <= m_clip_right)
13 m_clip_left = min_bin;
15 auto max_bin = m_binning.getBinIndex(max);
16 if (max_bin >= m_clip_left && max_bin < m_clip_right)
17 m_clip_right = max_bin;
21 template<typename VarType, typename WeightType>
22 template<typename BinType>
23 std::tuple<VarType, VarType, VarType> Histogram<VarType, WeightType>::ComputationImpl<BinType>::getStats() const {
24 VarType total = 0, total_count = 0;
27 // Find the mean and standard deviation in one go
28 for (auto i = m_clip_left; i <= m_clip_right; ++i) {
29 auto center = m_binning.getBin(i);
30 total += (*m_counts)[i] * center;
31 total_count += (*m_counts)[i];
32 sigma += (*m_counts)[i] * center * center;
35 VarType mean = total / total_count;
36 sigma = sigma / total_count - mean * mean;
38 sigma = std::sqrt(sigma);
43 WeightType low_sum = 0., high_sum = 0.;
44 auto low_i = m_clip_left, high_i = m_clip_right;
45 while (low_i <= high_i) {
46 if (low_sum < high_sum) {
47 low_sum += (*m_counts)[low_i++];
50 high_sum += (*m_counts)[high_i--];
54 assert(low_sum + high_sum == total_count);
58 auto edges = m_binning.getBinEdges(high_i + 1);
59 auto bin_width = (edges.second - edges.first);
60 auto max_counts = std::max((*m_counts)[low_i], (*m_counts)[high_i]);
61 median = edges.first + bin_width * (high_sum - low_sum) / (2.0 * max_counts);
64 median = m_binning.getBin(0);
67 return std::make_tuple(mean, median, sigma);
71 * This class as a constexpr static member "value" which will be 'true' iff BinType has a method
72 * computeBins that can receive two instances of IterType as parameters
76 template<typename BinType, typename IterType>
79 template <typename U, typename = decltype(std::declval<U>().computeBins(std::declval<IterType>(), std::declval<IterType>()))> struct SFINAE;
80 template<typename U> static char Test(SFINAE<U>*);
81 template<typename U> static int Test(...);
82 static constexpr bool value = sizeof(Test<BinType>(0)) == sizeof(char);
86 * This method is called if BinType has computeBins
88 template<typename BinType, typename IterType>
89 inline void computeBinsForwarder(BinType& binning, IterType begin, IterType end, std::true_type) {
90 binning.computeBins(begin, end);
94 * This method is called if BinType does not have computeBins
96 template<typename BinType, typename IterType>
97 inline void computeBinsForwarder(BinType&, IterType, IterType, std::false_type) {
100 template<typename VarType, typename WeightType>
101 template<typename BinType>
102 template<typename IterType, typename WeightIterType>
103 void Histogram<VarType, WeightType>::ComputationImpl<BinType>::computeBins(IterType begin, IterType end, WeightIterType wbegin) {
104 // This trick should allow the compiler to know the actual binning type, so if they
105 // override methods with *final*, we can skip indirections via vtable
106 computeBinsForwarder(m_binning, begin, end, std::integral_constant<bool, HasComputeBins<BinType, IterType>::value>());
109 m_clip_right = m_binning.getBinCount() - 1;
110 m_counts->resize(m_binning.getBinCount());
112 ssize_t nbins = m_counts->size();
115 for (; i != end; ++i, ++wi) {
116 auto bin = m_binning.getBinIndex(*i);
117 if (bin >= 0 && bin < nbins) {
118 (*m_counts)[bin] += *wi;
123 } // end of namespace Histogram
124 } // end of namespace Euclid