Geant4-11
RandGeneral.cc
Go to the documentation of this file.
1// -*- C++ -*-
2//
3// -----------------------------------------------------------------------
4// HEP Random
5// --- RandGeneral ---
6// class implementation file
7// -----------------------------------------------------------------------
8
9// =======================================================================
10// S.Magni & G.Pieri - Created: 5th September 1995
11// G.Cosmo - Added constructor using default engine from the
12// static generator. Simplified shoot() and
13// shootArray() (not needed in principle!): 20th Aug 1998
14// M.G.Pia & G.Cosmo - Fixed bug in computation of theIntegralPdf in
15// two constructors: 5th Jan 1999
16// S.Magni & G.Pieri - Added linear interpolation: 24th Mar 1999
17// M. Fischler - General cleanup: 14th May 1999
18// + Eliminated constructor code replication by factoring
19// common code into prepareTable.
20// + Eliminated fire/shoot code replication by factoring
21// out common code into mapRandom.
22// + A couple of methods are moved inline to avoid a
23// speed cost for factoring out mapRandom: fire()
24// and shoot(anEngine).
25// + Inserted checks for negative weight and zero total
26// weight in the bins.
27// + Modified the binary search loop to avoid incorrect
28// behavior when rand is example on a boundary.
29// + Moved the check of InterpolationType up into
30// the constructor. A type other than 0 or 1
31// will give the interpolated distribution (instead of
32// a distribution that always returns 0).
33// + Modified the computation of the returned value
34// to use algeraic simplification to improve speed.
35// Eliminated two of the three divisionns, made
36// use of the fact that nabove-nbelow is always 1, etc.
37// + Inserted a check for rand hitting the boundary of a
38// zero-width bin, to avoid dividing 0/0.
39// M. Fischler - Minor correction in assert 31 July 2001
40// + changed from assert (above = below+1) to ==
41// M Fischler - put and get to/from streams 12/15/04
42// + Modifications to use a vector as theIntegraPdf
43// M Fischler - put/get to/from streams uses pairs of ulongs when
44// + storing doubles avoid problems with precision
45// 4/14/05
46//
47// =======================================================================
48
51#include <cassert>
52#include <iostream>
53#include <string>
54#include <vector>
55
56namespace CLHEP {
57
58std::string RandGeneral::name() const {return "RandGeneral";}
60
61
63// Constructors
65
66RandGeneral::RandGeneral( const double* aProbFunc,
67 int theProbSize,
68 int IntType )
69 : HepRandom(),
70 localEngine(HepRandom::getTheEngine(), do_nothing_deleter()),
71 nBins(theProbSize),
72 InterpolationType(IntType)
73{
74 prepareTable(aProbFunc);
75}
76
78 const double* aProbFunc,
79 int theProbSize,
80 int IntType )
81: HepRandom(),
82 localEngine(&anEngine, do_nothing_deleter()),
83 nBins(theProbSize),
84 InterpolationType(IntType)
85{
86 prepareTable(aProbFunc);
87}
88
90 const double* aProbFunc,
91 int theProbSize,
92 int IntType )
93: HepRandom(),
94 localEngine(anEngine),
95 nBins(theProbSize),
96 InterpolationType(IntType)
97{
98 prepareTable(aProbFunc);
99}
100
101void RandGeneral::prepareTable(const double* aProbFunc) {
102//
103// Private method called only by constructors. Prepares theIntegralPdf.
104//
105 if (nBins < 1) {
106 std::cerr <<
107 "RandGeneral constructed with no bins - will use flat distribution\n";
109 return;
110 }
111
112 theIntegralPdf.resize(nBins+1);
113 theIntegralPdf[0] = 0;
114 int ptn;
115 double weight;
116
117 for ( ptn = 0; ptn<nBins; ++ptn ) {
118 weight = aProbFunc[ptn];
119 if ( weight < 0 ) {
120 // We can't stomach negative bin contents, they invalidate the
121 // search algorithm when the distribution is fired.
122 std::cerr <<
123 "RandGeneral constructed with negative-weight bin " << ptn <<
124 " = " << weight << " \n -- will substitute 0 weight \n";
125 weight = 0;
126 }
127 // std::cout << ptn << " " << weight << " " << theIntegralPdf[ptn] << "\n";
128 theIntegralPdf[ptn+1] = theIntegralPdf[ptn] + weight;
129 }
130
131 if ( theIntegralPdf[nBins] <= 0 ) {
132 std::cerr <<
133 "RandGeneral constructed nothing in bins - will use flat distribution\n";
135 return;
136 }
137
138 for ( ptn = 0; ptn < nBins+1; ++ptn ) {
140 // std::cout << ptn << " " << theIntegralPdf[ptn] << "\n";
141 }
142
143 // And another useful variable is ...
144 oneOverNbins = 1.0 / nBins;
145
146 // One last chore:
147
148 if ( (InterpolationType != 0) && (InterpolationType != 1) ) {
149 std::cerr <<
150 "RandGeneral does not recognize IntType " << InterpolationType
151 << "\n Will use type 0 (continuous linear interpolation \n";
153 }
154
155} // prepareTable()
156
158//
159// Private method called only by prepareTables in case of user error.
160//
161 nBins = 1;
162 theIntegralPdf.resize(2);
163 theIntegralPdf[0] = 0;
164 theIntegralPdf[1] = 1;
165 oneOverNbins = 1.0;
166 return;
167
168} // UseFlatDistribution()
169
171// Destructor
173
175}
176
177
179// mapRandom(rand)
181
182double RandGeneral::mapRandom(double rand) const {
183//
184// Private method to take the random (however it is created) and map it
185// according to the distribution.
186//
187
188 int nbelow = 0; // largest k such that I[k] is known to be <= rand
189 int nabove = nBins; // largest k such that I[k] is known to be > rand
190 int middle;
191
192 while (nabove > nbelow+1) {
193 middle = (nabove + nbelow+1)>>1;
194 if (rand >= theIntegralPdf[middle]) {
195 nbelow = middle;
196 } else {
197 nabove = middle;
198 }
199 } // after this loop, nabove is always nbelow+1 and they straddle rad:
200 assert ( nabove == nbelow+1 );
201 assert ( theIntegralPdf[nbelow] <= rand );
202 assert ( theIntegralPdf[nabove] >= rand );
203 // If a defective engine produces rand=1, that will
204 // still give sensible results so we relax the > rand assertion
205
206 if ( InterpolationType == 1 ) {
207
208 return nbelow * oneOverNbins;
209
210 } else {
211
212 double binMeasure = theIntegralPdf[nabove] - theIntegralPdf[nbelow];
213 // binMeasure is always aProbFunc[nbelow],
214 // but we don't have aProbFunc any more so we subtract.
215
216 if ( binMeasure == 0 ) {
217 // rand lies right in a bin of measure 0. Simply return the center
218 // of the range of that bin. (Any value between k/N and (k+1)/N is
219 // equally good, in this rare case.)
220 return (nbelow + .5) * oneOverNbins;
221 }
222
223 double binFraction = (rand - theIntegralPdf[nbelow]) / binMeasure;
224
225 return (nbelow + binFraction) * oneOverNbins;
226 }
227
228} // mapRandom(rand)
229
231 const int size, double* vect )
232{
233 int i;
234
235 for (i=0; i<size; ++i) {
236 vect[i] = shoot(anEngine);
237 }
238}
239
240void RandGeneral::fireArray( const int size, double* vect )
241{
242 int i;
243
244 for (i=0; i<size; ++i) {
245 vect[i] = fire();
246 }
247}
248
249std::ostream & RandGeneral::put ( std::ostream & os ) const {
250 int pr=os.precision(20);
251 std::vector<unsigned long> t(2);
252 os << " " << name() << "\n";
253 os << "Uvec" << "\n";
254 os << nBins << " " << oneOverNbins << " " << InterpolationType << "\n";
256 os << t[0] << " " << t[1] << "\n";
257 assert (static_cast<int>(theIntegralPdf.size())==nBins+1);
258 for (unsigned int i=0; i<theIntegralPdf.size(); ++i) {
260 os << theIntegralPdf[i] << " " << t[0] << " " << t[1] << "\n";
261 }
262 os.precision(pr);
263 return os;
264}
265
266std::istream & RandGeneral::get ( std::istream & is ) {
267 std::string inName;
268 is >> inName;
269 if (inName != name()) {
270 is.clear(std::ios::badbit | is.rdstate());
271 std::cerr << "Mismatch when expecting to read state of a "
272 << name() << " distribution\n"
273 << "Name found was " << inName
274 << "\nistream is left in the badbit state\n";
275 return is;
276 }
277 if (possibleKeywordInput(is, "Uvec", nBins)) {
278 std::vector<unsigned long> t(2);
280 is >> t[0] >> t[1]; oneOverNbins = DoubConv::longs2double(t);
281 theIntegralPdf.resize(nBins+1);
282 for (unsigned int i=0; i<theIntegralPdf.size(); ++i) {
283 is >> theIntegralPdf[i] >> t[0] >> t[1];
285 }
286 return is;
287 }
288 // is >> nBins encompassed by possibleKeywordInput
290 theIntegralPdf.resize(nBins+1);
291 for (unsigned int i=0; i<theIntegralPdf.size(); ++i) is >> theIntegralPdf[i];
292 return is;
293}
294
295} // namespace CLHEP
static double longs2double(const std::vector< unsigned long > &v)
Definition: DoubConv.cc:110
static std::vector< unsigned long > dto2longs(double d)
Definition: DoubConv.cc:94
void prepareTable(const double *aProbFunc)
Definition: RandGeneral.cc:101
void useFlatDistribution()
Definition: RandGeneral.cc:157
RandGeneral(const double *aProbFunc, int theProbSize, int IntType=0)
Definition: RandGeneral.cc:66
std::istream & get(std::istream &is)
Definition: RandGeneral.cc:266
virtual ~RandGeneral()
Definition: RandGeneral.cc:174
std::ostream & put(std::ostream &os) const
Definition: RandGeneral.cc:249
void fireArray(const int size, double *vect)
Definition: RandGeneral.cc:240
std::string name() const
Definition: RandGeneral.cc:58
HepRandomEngine & engine()
Definition: RandGeneral.cc:59
void shootArray(const int size, double *vect)
std::vector< double > theIntegralPdf
Definition: RandGeneral.h:131
double mapRandom(double rand) const
Definition: RandGeneral.cc:182
std::shared_ptr< HepRandomEngine > localEngine
Definition: RandGeneral.h:130
Definition: DoubConv.h:17
bool possibleKeywordInput(IS &is, const std::string &key, T &t)
Definition: RandomEngine.h:166