source: pacpussensors/trunk/Vislab/lib3dv/eigen/test/array.cpp@ 138

Last change on this file since 138 was 136, checked in by ldecherf, 8 years ago

Doc

File size: 11.6 KB
Line 
1// This file is part of Eigen, a lightweight C++ template library
2// for linear algebra.
3//
4// Copyright (C) 2008-2009 Gael Guennebaud <gael.guennebaud@inria.fr>
5//
6// This Source Code Form is subject to the terms of the Mozilla
7// Public License v. 2.0. If a copy of the MPL was not distributed
8// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
9
10#include "main.h"
11
12template<typename ArrayType> void array(const ArrayType& m)
13{
14 typedef typename ArrayType::Index Index;
15 typedef typename ArrayType::Scalar Scalar;
16 typedef Array<Scalar, ArrayType::RowsAtCompileTime, 1> ColVectorType;
17 typedef Array<Scalar, 1, ArrayType::ColsAtCompileTime> RowVectorType;
18
19 Index rows = m.rows();
20 Index cols = m.cols();
21
22 ArrayType m1 = ArrayType::Random(rows, cols),
23 m2 = ArrayType::Random(rows, cols),
24 m3(rows, cols);
25
26 ColVectorType cv1 = ColVectorType::Random(rows);
27 RowVectorType rv1 = RowVectorType::Random(cols);
28
29 Scalar s1 = internal::random<Scalar>(),
30 s2 = internal::random<Scalar>();
31
32 // scalar addition
33 VERIFY_IS_APPROX(m1 + s1, s1 + m1);
34 VERIFY_IS_APPROX(m1 + s1, ArrayType::Constant(rows,cols,s1) + m1);
35 VERIFY_IS_APPROX(s1 - m1, (-m1)+s1 );
36 VERIFY_IS_APPROX(m1 - s1, m1 - ArrayType::Constant(rows,cols,s1));
37 VERIFY_IS_APPROX(s1 - m1, ArrayType::Constant(rows,cols,s1) - m1);
38 VERIFY_IS_APPROX((m1*Scalar(2)) - s2, (m1+m1) - ArrayType::Constant(rows,cols,s2) );
39 m3 = m1;
40 m3 += s2;
41 VERIFY_IS_APPROX(m3, m1 + s2);
42 m3 = m1;
43 m3 -= s1;
44 VERIFY_IS_APPROX(m3, m1 - s1);
45
46 // scalar operators via Maps
47 m3 = m1;
48 ArrayType::Map(m1.data(), m1.rows(), m1.cols()) -= ArrayType::Map(m2.data(), m2.rows(), m2.cols());
49 VERIFY_IS_APPROX(m1, m3 - m2);
50
51 m3 = m1;
52 ArrayType::Map(m1.data(), m1.rows(), m1.cols()) += ArrayType::Map(m2.data(), m2.rows(), m2.cols());
53 VERIFY_IS_APPROX(m1, m3 + m2);
54
55 m3 = m1;
56 ArrayType::Map(m1.data(), m1.rows(), m1.cols()) *= ArrayType::Map(m2.data(), m2.rows(), m2.cols());
57 VERIFY_IS_APPROX(m1, m3 * m2);
58
59 m3 = m1;
60 m2 = ArrayType::Random(rows,cols);
61 m2 = (m2==0).select(1,m2);
62 ArrayType::Map(m1.data(), m1.rows(), m1.cols()) /= ArrayType::Map(m2.data(), m2.rows(), m2.cols());
63 VERIFY_IS_APPROX(m1, m3 / m2);
64
65 // reductions
66 VERIFY_IS_APPROX(m1.abs().colwise().sum().sum(), m1.abs().sum());
67 VERIFY_IS_APPROX(m1.abs().rowwise().sum().sum(), m1.abs().sum());
68 using std::abs;
69 VERIFY_IS_MUCH_SMALLER_THAN(abs(m1.colwise().sum().sum() - m1.sum()), m1.abs().sum());
70 VERIFY_IS_MUCH_SMALLER_THAN(abs(m1.rowwise().sum().sum() - m1.sum()), m1.abs().sum());
71 if (!internal::isMuchSmallerThan(abs(m1.sum() - (m1+m2).sum()), m1.abs().sum(), test_precision<Scalar>()))
72 VERIFY_IS_NOT_APPROX(((m1+m2).rowwise().sum()).sum(), m1.sum());
73 VERIFY_IS_APPROX(m1.colwise().sum(), m1.colwise().redux(internal::scalar_sum_op<Scalar>()));
74
75 // vector-wise ops
76 m3 = m1;
77 VERIFY_IS_APPROX(m3.colwise() += cv1, m1.colwise() + cv1);
78 m3 = m1;
79 VERIFY_IS_APPROX(m3.colwise() -= cv1, m1.colwise() - cv1);
80 m3 = m1;
81 VERIFY_IS_APPROX(m3.rowwise() += rv1, m1.rowwise() + rv1);
82 m3 = m1;
83 VERIFY_IS_APPROX(m3.rowwise() -= rv1, m1.rowwise() - rv1);
84}
85
86template<typename ArrayType> void comparisons(const ArrayType& m)
87{
88 using std::abs;
89 typedef typename ArrayType::Index Index;
90 typedef typename ArrayType::Scalar Scalar;
91 typedef typename NumTraits<Scalar>::Real RealScalar;
92
93 Index rows = m.rows();
94 Index cols = m.cols();
95
96 Index r = internal::random<Index>(0, rows-1),
97 c = internal::random<Index>(0, cols-1);
98
99 ArrayType m1 = ArrayType::Random(rows, cols),
100 m2 = ArrayType::Random(rows, cols),
101 m3(rows, cols);
102
103 VERIFY(((m1 + Scalar(1)) > m1).all());
104 VERIFY(((m1 - Scalar(1)) < m1).all());
105 if (rows*cols>1)
106 {
107 m3 = m1;
108 m3(r,c) += 1;
109 VERIFY(! (m1 < m3).all() );
110 VERIFY(! (m1 > m3).all() );
111 }
112 VERIFY(!(m1 > m2 && m1 < m2).any());
113 VERIFY((m1 <= m2 || m1 >= m2).all());
114
115 // comparisons to scalar
116 VERIFY( (m1 != (m1(r,c)+1) ).any() );
117 VERIFY( (m1 > (m1(r,c)-1) ).any() );
118 VERIFY( (m1 < (m1(r,c)+1) ).any() );
119 VERIFY( (m1 == m1(r,c) ).any() );
120
121 // test Select
122 VERIFY_IS_APPROX( (m1<m2).select(m1,m2), m1.cwiseMin(m2) );
123 VERIFY_IS_APPROX( (m1>m2).select(m1,m2), m1.cwiseMax(m2) );
124 Scalar mid = (m1.cwiseAbs().minCoeff() + m1.cwiseAbs().maxCoeff())/Scalar(2);
125 for (int j=0; j<cols; ++j)
126 for (int i=0; i<rows; ++i)
127 m3(i,j) = abs(m1(i,j))<mid ? 0 : m1(i,j);
128 VERIFY_IS_APPROX( (m1.abs()<ArrayType::Constant(rows,cols,mid))
129 .select(ArrayType::Zero(rows,cols),m1), m3);
130 // shorter versions:
131 VERIFY_IS_APPROX( (m1.abs()<ArrayType::Constant(rows,cols,mid))
132 .select(0,m1), m3);
133 VERIFY_IS_APPROX( (m1.abs()>=ArrayType::Constant(rows,cols,mid))
134 .select(m1,0), m3);
135 // even shorter version:
136 VERIFY_IS_APPROX( (m1.abs()<mid).select(0,m1), m3);
137
138 // count
139 VERIFY(((m1.abs()+1)>RealScalar(0.1)).count() == rows*cols);
140
141 // and/or
142 VERIFY( (m1<RealScalar(0) && m1>RealScalar(0)).count() == 0);
143 VERIFY( (m1<RealScalar(0) || m1>=RealScalar(0)).count() == rows*cols);
144 RealScalar a = m1.abs().mean();
145 VERIFY( (m1<-a || m1>a).count() == (m1.abs()>a).count());
146
147 typedef Array<typename ArrayType::Index, Dynamic, 1> ArrayOfIndices;
148
149 // TODO allows colwise/rowwise for array
150 VERIFY_IS_APPROX(((m1.abs()+1)>RealScalar(0.1)).colwise().count(), ArrayOfIndices::Constant(cols,rows).transpose());
151 VERIFY_IS_APPROX(((m1.abs()+1)>RealScalar(0.1)).rowwise().count(), ArrayOfIndices::Constant(rows, cols));
152}
153
154template<typename ArrayType> void array_real(const ArrayType& m)
155{
156 using std::abs;
157 using std::sqrt;
158 typedef typename ArrayType::Index Index;
159 typedef typename ArrayType::Scalar Scalar;
160 typedef typename NumTraits<Scalar>::Real RealScalar;
161
162 Index rows = m.rows();
163 Index cols = m.cols();
164
165 ArrayType m1 = ArrayType::Random(rows, cols),
166 m2 = ArrayType::Random(rows, cols),
167 m3(rows, cols);
168
169 Scalar s1 = internal::random<Scalar>();
170
171 // these tests are mostly to check possible compilation issues.
172 VERIFY_IS_APPROX(m1.sin(), sin(m1));
173 VERIFY_IS_APPROX(m1.cos(), cos(m1));
174 VERIFY_IS_APPROX(m1.asin(), asin(m1));
175 VERIFY_IS_APPROX(m1.acos(), acos(m1));
176 VERIFY_IS_APPROX(m1.tan(), tan(m1));
177
178 VERIFY_IS_APPROX(cos(m1+RealScalar(3)*m2), cos((m1+RealScalar(3)*m2).eval()));
179
180 VERIFY_IS_APPROX(m1.abs().sqrt(), sqrt(abs(m1)));
181 VERIFY_IS_APPROX(m1.abs(), sqrt(numext::abs2(m1)));
182
183 VERIFY_IS_APPROX(numext::abs2(numext::real(m1)) + numext::abs2(numext::imag(m1)), numext::abs2(m1));
184 VERIFY_IS_APPROX(numext::abs2(real(m1)) + numext::abs2(imag(m1)), numext::abs2(m1));
185 if(!NumTraits<Scalar>::IsComplex)
186 VERIFY_IS_APPROX(numext::real(m1), m1);
187
188 // shift argument of logarithm so that it is not zero
189 Scalar smallNumber = NumTraits<Scalar>::dummy_precision();
190 VERIFY_IS_APPROX((m1.abs() + smallNumber).log() , log(abs(m1) + smallNumber));
191
192 VERIFY_IS_APPROX(m1.exp() * m2.exp(), exp(m1+m2));
193 VERIFY_IS_APPROX(m1.exp(), exp(m1));
194 VERIFY_IS_APPROX(m1.exp() / m2.exp(),(m1-m2).exp());
195
196 VERIFY_IS_APPROX(m1.pow(2), m1.square());
197 VERIFY_IS_APPROX(pow(m1,2), m1.square());
198
199 ArrayType exponents = ArrayType::Constant(rows, cols, RealScalar(2));
200 VERIFY_IS_APPROX(Eigen::pow(m1,exponents), m1.square());
201
202 m3 = m1.abs();
203 VERIFY_IS_APPROX(m3.pow(RealScalar(0.5)), m3.sqrt());
204 VERIFY_IS_APPROX(pow(m3,RealScalar(0.5)), m3.sqrt());
205
206 // scalar by array division
207 const RealScalar tiny = sqrt(std::numeric_limits<RealScalar>::epsilon());
208 s1 += Scalar(tiny);
209 m1 += ArrayType::Constant(rows,cols,Scalar(tiny));
210 VERIFY_IS_APPROX(s1/m1, s1 * m1.inverse());
211
212 // check inplace transpose
213 m3 = m1;
214 m3.transposeInPlace();
215 VERIFY_IS_APPROX(m3,m1.transpose());
216 m3.transposeInPlace();
217 VERIFY_IS_APPROX(m3,m1);
218}
219
220template<typename ArrayType> void array_complex(const ArrayType& m)
221{
222 typedef typename ArrayType::Index Index;
223
224 Index rows = m.rows();
225 Index cols = m.cols();
226
227 ArrayType m1 = ArrayType::Random(rows, cols),
228 m2(rows, cols);
229
230 for (Index i = 0; i < m.rows(); ++i)
231 for (Index j = 0; j < m.cols(); ++j)
232 m2(i,j) = sqrt(m1(i,j));
233
234 VERIFY_IS_APPROX(m1.sqrt(), m2);
235 VERIFY_IS_APPROX(m1.sqrt(), Eigen::sqrt(m1));
236}
237
238template<typename ArrayType> void min_max(const ArrayType& m)
239{
240 typedef typename ArrayType::Index Index;
241 typedef typename ArrayType::Scalar Scalar;
242
243 Index rows = m.rows();
244 Index cols = m.cols();
245
246 ArrayType m1 = ArrayType::Random(rows, cols);
247
248 // min/max with array
249 Scalar maxM1 = m1.maxCoeff();
250 Scalar minM1 = m1.minCoeff();
251
252 VERIFY_IS_APPROX(ArrayType::Constant(rows,cols, minM1), (m1.min)(ArrayType::Constant(rows,cols, minM1)));
253 VERIFY_IS_APPROX(m1, (m1.min)(ArrayType::Constant(rows,cols, maxM1)));
254
255 VERIFY_IS_APPROX(ArrayType::Constant(rows,cols, maxM1), (m1.max)(ArrayType::Constant(rows,cols, maxM1)));
256 VERIFY_IS_APPROX(m1, (m1.max)(ArrayType::Constant(rows,cols, minM1)));
257
258 // min/max with scalar input
259 VERIFY_IS_APPROX(ArrayType::Constant(rows,cols, minM1), (m1.min)( minM1));
260 VERIFY_IS_APPROX(m1, (m1.min)( maxM1));
261
262 VERIFY_IS_APPROX(ArrayType::Constant(rows,cols, maxM1), (m1.max)( maxM1));
263 VERIFY_IS_APPROX(m1, (m1.max)( minM1));
264
265}
266
267void test_array()
268{
269 for(int i = 0; i < g_repeat; i++) {
270 CALL_SUBTEST_1( array(Array<float, 1, 1>()) );
271 CALL_SUBTEST_2( array(Array22f()) );
272 CALL_SUBTEST_3( array(Array44d()) );
273 CALL_SUBTEST_4( array(ArrayXXcf(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
274 CALL_SUBTEST_5( array(ArrayXXf(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
275 CALL_SUBTEST_6( array(ArrayXXi(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
276 }
277 for(int i = 0; i < g_repeat; i++) {
278 CALL_SUBTEST_1( comparisons(Array<float, 1, 1>()) );
279 CALL_SUBTEST_2( comparisons(Array22f()) );
280 CALL_SUBTEST_3( comparisons(Array44d()) );
281 CALL_SUBTEST_5( comparisons(ArrayXXf(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
282 CALL_SUBTEST_6( comparisons(ArrayXXi(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
283 }
284 for(int i = 0; i < g_repeat; i++) {
285 CALL_SUBTEST_1( min_max(Array<float, 1, 1>()) );
286 CALL_SUBTEST_2( min_max(Array22f()) );
287 CALL_SUBTEST_3( min_max(Array44d()) );
288 CALL_SUBTEST_5( min_max(ArrayXXf(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
289 CALL_SUBTEST_6( min_max(ArrayXXi(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
290 }
291 for(int i = 0; i < g_repeat; i++) {
292 CALL_SUBTEST_1( array_real(Array<float, 1, 1>()) );
293 CALL_SUBTEST_2( array_real(Array22f()) );
294 CALL_SUBTEST_3( array_real(Array44d()) );
295 CALL_SUBTEST_5( array_real(ArrayXXf(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
296 }
297 for(int i = 0; i < g_repeat; i++) {
298 CALL_SUBTEST_4( array_complex(ArrayXXcf(internal::random<int>(1,EIGEN_TEST_MAX_SIZE), internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
299 }
300
301 VERIFY((internal::is_same< internal::global_math_functions_filtering_base<int>::type, int >::value));
302 VERIFY((internal::is_same< internal::global_math_functions_filtering_base<float>::type, float >::value));
303 VERIFY((internal::is_same< internal::global_math_functions_filtering_base<Array2i>::type, ArrayBase<Array2i> >::value));
304 typedef CwiseUnaryOp<internal::scalar_sum_op<double>, ArrayXd > Xpr;
305 VERIFY((internal::is_same< internal::global_math_functions_filtering_base<Xpr>::type,
306 ArrayBase<Xpr>
307 >::value));
308}
Note: See TracBrowser for help on using the repository browser.