]>
Commit | Line | Data |
---|---|---|
03896fc4 | 1 | // |
2 | // | |
3 | // Nikolai Amelin, Ludmila Malinina, Timur Pocheptsov (C) JINR/Dubna | |
4 | // amelin@sunhe.jinr.ru, malinina@sunhe.jinr.ru, pocheptsov@sunhe.jinr.ru | |
5 | // November. 2, 2005 | |
6 | // | |
7 | // | |
8 | // This class is taken from the GEANT4 tool kit and changed!!!!! | |
b1c2e580 | 9 | |
03896fc4 | 10 | #include <TError.h> |
b1c2e580 | 11 | #include "RandArrayFunction.h" |
b1c2e580 | 12 | |
13 | ||
786056a2 | 14 | RandArrayFunction::RandArrayFunction(const Double_t *aProbFunc, Int_t theProbSize, Int_t intType): |
15 | fIntegralPdf(), | |
16 | fNBins(theProbSize), | |
17 | fOneOverNbins(0), | |
18 | fInterpolationType(intType) | |
b1c2e580 | 19 | { |
20 | PrepareTable(aProbFunc); | |
21 | } | |
22 | ||
786056a2 | 23 | RandArrayFunction::RandArrayFunction(Int_t theProbSize, Int_t intType): |
24 | fIntegralPdf(), | |
25 | fNBins(theProbSize), | |
26 | fOneOverNbins(0), | |
27 | fInterpolationType(intType) | |
b1c2e580 | 28 | {} |
29 | ||
30 | void RandArrayFunction::PrepareTable(const Double_t* aProbFunc) { | |
31 | //Prepares fIntegralPdf. | |
32 | if(fNBins < 1) { | |
33 | Error("RandArrayFunction::PrepareTable", | |
34 | "RandArrayFunction constructed with no bins - will use flat distribution."); | |
35 | UseFlatDistribution(); | |
36 | return; | |
37 | } | |
38 | ||
39 | fIntegralPdf.resize(fNBins + 1); | |
40 | fIntegralPdf[0] = 0; | |
41 | Int_t ptn; | |
42 | for (ptn = 0; ptn < fNBins; ++ptn ) { | |
43 | Double_t weight = aProbFunc[ptn]; | |
44 | if (weight < 0.) { | |
45 | // We can't stomach negative bin contents, they invalidate the | |
46 | // search algorithm when the distribution is fired. | |
47 | Warning("RandArrayFunction::PrepareTable", | |
48 | "RandArrayFunction constructed with negative-weight bin %d == %f -- will substitute 0 weight", | |
49 | ptn, weight); | |
50 | weight = 0.; | |
51 | } | |
52 | fIntegralPdf[ptn + 1] = fIntegralPdf[ptn] + weight; | |
53 | } | |
54 | ||
55 | if (fIntegralPdf[fNBins] <= 0.) { | |
56 | Warning("RandArrayFunction::PrepareTable", | |
57 | "RandArrayFunction constructed with nothing in bins - will use flat distribution"); | |
58 | UseFlatDistribution(); | |
59 | return; | |
60 | } | |
61 | ||
62 | for (ptn = 0; ptn < fNBins + 1; ++ptn) | |
63 | fIntegralPdf[ptn] /= fIntegralPdf[fNBins]; | |
64 | ||
65 | // And another useful variable is ... | |
66 | fOneOverNbins = 1.0 / fNBins; | |
67 | // One last chore: | |
68 | if (fInterpolationType && fInterpolationType != 1) { | |
69 | Info("RandArrayFunction::PrepareTable", | |
70 | "RandArrayFunction does not recognize fInterpolationType %d \n" | |
71 | "Will use type 0 (continuous linear interpolation)", fInterpolationType); | |
72 | fInterpolationType = 0; | |
73 | } | |
74 | } | |
75 | ||
76 | void RandArrayFunction::UseFlatDistribution() { | |
77 | //Called only by PrepareTable in case of user error. | |
78 | fNBins = 1; | |
79 | fIntegralPdf.resize(2); | |
80 | fIntegralPdf[0] = 0; | |
81 | fIntegralPdf[1] = 1; | |
82 | fOneOverNbins = 1.0; | |
83 | } | |
84 | ||
85 | Double_t RandArrayFunction::MapRandom(Double_t rand) const { | |
86 | // Private method to take the random (however it is created) and map it | |
87 | // according to the distribution. | |
88 | ||
89 | Int_t nBelow = 0; // largest k such that I[k] is known to be <= rand | |
90 | Int_t nAbove = fNBins; // largest k such that I[k] is known to be > rand | |
91 | Int_t middle; | |
92 | ||
93 | while (nAbove > nBelow+1) { | |
94 | middle = (nAbove + nBelow+1)>>1; | |
95 | rand >= fIntegralPdf[middle] ? nBelow = middle : nAbove = middle; | |
96 | }// after this loop, nAbove is always nBelow+1 and they straddle rad: | |
97 | ||
98 | /*assert ( nAbove = nBelow+1 ); | |
99 | assert ( fIntegralPdf[nBelow] <= rand ); | |
100 | assert ( fIntegralPdf[nAbove] >= rand );*/ | |
101 | // If a defective engine produces rand=1, that will | |
102 | // still give sensible results so we relax the > rand assertion | |
103 | ||
104 | if (fInterpolationType == 1) { | |
105 | return nBelow * fOneOverNbins; | |
106 | } | |
107 | else { | |
108 | Double_t binMeasure = fIntegralPdf[nAbove] - fIntegralPdf[nBelow]; | |
109 | // binMeasure is always aProbFunc[nBelow], | |
110 | // but we don't have aProbFunc any more so we subtract. | |
111 | ||
112 | if (!binMeasure) { | |
113 | // rand lies right in a bin of measure 0. Simply return the center | |
114 | // of the range of that bin. (Any value between k/N and (k+1)/N is | |
115 | // equally good, in this rare case.) | |
116 | return (nBelow + .5) * fOneOverNbins; | |
117 | } | |
118 | ||
119 | Double_t binFraction = (rand - fIntegralPdf[nBelow]) / binMeasure; | |
120 | ||
121 | return (nBelow + binFraction) * fOneOverNbins; | |
122 | } | |
123 | } | |
124 | ||
125 | void RandArrayFunction::FireArray(Int_t size, Double_t *vect) const { | |
126 | for (Int_t i = 0; i < size; ++i) | |
127 | vect[i] = Fire(); | |
128 | } |