1 // This file is part of Eigen, a lightweight C++ template library
2 // for linear algebra.
3 //
4 // Copyright (C) 2008-2011 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 "sparse.h"
11 
sparse_vector(int rows,int cols)12 template<typename Scalar,typename StorageIndex> void sparse_vector(int rows, int cols)
13 {
14   double densityMat = (std::max)(8./(rows*cols), 0.01);
15   double densityVec = (std::max)(8./(rows), 0.1);
16   typedef Matrix<Scalar,Dynamic,Dynamic> DenseMatrix;
17   typedef Matrix<Scalar,Dynamic,1> DenseVector;
18   typedef SparseVector<Scalar,0,StorageIndex> SparseVectorType;
19   typedef SparseMatrix<Scalar,0,StorageIndex> SparseMatrixType;
20   Scalar eps = 1e-6;
21 
22   SparseMatrixType m1(rows,rows);
23   SparseVectorType v1(rows), v2(rows), v3(rows);
24   DenseMatrix refM1 = DenseMatrix::Zero(rows, rows);
25   DenseVector refV1 = DenseVector::Random(rows),
26               refV2 = DenseVector::Random(rows),
27               refV3 = DenseVector::Random(rows);
28 
29   std::vector<int> zerocoords, nonzerocoords;
30   initSparse<Scalar>(densityVec, refV1, v1, &zerocoords, &nonzerocoords);
31   initSparse<Scalar>(densityMat, refM1, m1);
32 
33   initSparse<Scalar>(densityVec, refV2, v2);
34   initSparse<Scalar>(densityVec, refV3, v3);
35 
36   Scalar s1 = internal::random<Scalar>();
37 
38   // test coeff and coeffRef
39   for (unsigned int i=0; i<zerocoords.size(); ++i)
40   {
41     VERIFY_IS_MUCH_SMALLER_THAN( v1.coeff(zerocoords[i]), eps );
42     //VERIFY_RAISES_ASSERT( v1.coeffRef(zerocoords[i]) = 5 );
43   }
44   {
45     VERIFY(int(nonzerocoords.size()) == v1.nonZeros());
46     int j=0;
47     for (typename SparseVectorType::InnerIterator it(v1); it; ++it,++j)
48     {
49       VERIFY(nonzerocoords[j]==it.index());
50       VERIFY(it.value()==v1.coeff(it.index()));
51       VERIFY(it.value()==refV1.coeff(it.index()));
52     }
53   }
54   VERIFY_IS_APPROX(v1, refV1);
55 
56   // test coeffRef with reallocation
57   {
58     SparseVectorType v4(rows);
59     DenseVector v5 = DenseVector::Zero(rows);
60     for(int k=0; k<rows; ++k)
61     {
62       int i = internal::random<int>(0,rows-1);
63       Scalar v = internal::random<Scalar>();
64       v4.coeffRef(i) += v;
65       v5.coeffRef(i) += v;
66     }
67     VERIFY_IS_APPROX(v4,v5);
68   }
69 
70   v1.coeffRef(nonzerocoords[0]) = Scalar(5);
71   refV1.coeffRef(nonzerocoords[0]) = Scalar(5);
72   VERIFY_IS_APPROX(v1, refV1);
73 
74   VERIFY_IS_APPROX(v1+v2, refV1+refV2);
75   VERIFY_IS_APPROX(v1+v2+v3, refV1+refV2+refV3);
76 
77   VERIFY_IS_APPROX(v1*s1-v2, refV1*s1-refV2);
78 
79   VERIFY_IS_APPROX(v1*=s1, refV1*=s1);
80   VERIFY_IS_APPROX(v1/=s1, refV1/=s1);
81 
82   VERIFY_IS_APPROX(v1+=v2, refV1+=refV2);
83   VERIFY_IS_APPROX(v1-=v2, refV1-=refV2);
84 
85   VERIFY_IS_APPROX(v1.dot(v2), refV1.dot(refV2));
86   VERIFY_IS_APPROX(v1.dot(refV2), refV1.dot(refV2));
87 
88   VERIFY_IS_APPROX(m1*v2, refM1*refV2);
89   VERIFY_IS_APPROX(v1.dot(m1*v2), refV1.dot(refM1*refV2));
90   {
91     int i = internal::random<int>(0,rows-1);
92     VERIFY_IS_APPROX(v1.dot(m1.col(i)), refV1.dot(refM1.col(i)));
93   }
94 
95 
96   VERIFY_IS_APPROX(v1.squaredNorm(), refV1.squaredNorm());
97 
98   VERIFY_IS_APPROX(v1.blueNorm(), refV1.blueNorm());
99 
100   // test aliasing
101   VERIFY_IS_APPROX((v1 = -v1), (refV1 = -refV1));
102   VERIFY_IS_APPROX((v1 = v1.transpose()), (refV1 = refV1.transpose().eval()));
103   VERIFY_IS_APPROX((v1 += -v1), (refV1 += -refV1));
104 
105   // sparse matrix to sparse vector
106   SparseMatrixType mv1;
107   VERIFY_IS_APPROX((mv1=v1),v1);
108   VERIFY_IS_APPROX(mv1,(v1=mv1));
109   VERIFY_IS_APPROX(mv1,(v1=mv1.transpose()));
110 
111   // check copy to dense vector with transpose
112   refV3.resize(0);
113   VERIFY_IS_APPROX(refV3 = v1.transpose(),v1.toDense());
114   VERIFY_IS_APPROX(DenseVector(v1),v1.toDense());
115 
116   // test conservative resize
117   {
118     std::vector<StorageIndex> inc;
119     if(rows > 3)
120       inc.push_back(-3);
121     inc.push_back(0);
122     inc.push_back(3);
123     inc.push_back(1);
124     inc.push_back(10);
125 
126     for(std::size_t i = 0; i< inc.size(); i++) {
127       StorageIndex incRows = inc[i];
128       SparseVectorType vec1(rows);
129       DenseVector refVec1 = DenseVector::Zero(rows);
130       initSparse<Scalar>(densityVec, refVec1, vec1);
131 
132       vec1.conservativeResize(rows+incRows);
133       refVec1.conservativeResize(rows+incRows);
134       if (incRows > 0) refVec1.tail(incRows).setZero();
135 
136       VERIFY_IS_APPROX(vec1, refVec1);
137 
138       // Insert new values
139       if (incRows > 0)
140         vec1.insert(vec1.rows()-1) = refVec1(refVec1.rows()-1) = 1;
141 
142       VERIFY_IS_APPROX(vec1, refVec1);
143     }
144   }
145 
146 }
147 
test_sparse_vector()148 void test_sparse_vector()
149 {
150   for(int i = 0; i < g_repeat; i++) {
151     int r = Eigen::internal::random<int>(1,500), c = Eigen::internal::random<int>(1,500);
152     if(Eigen::internal::random<int>(0,4) == 0) {
153       r = c; // check square matrices in 25% of tries
154     }
155     EIGEN_UNUSED_VARIABLE(r+c);
156 
157     CALL_SUBTEST_1(( sparse_vector<double,int>(8, 8) ));
158     CALL_SUBTEST_2(( sparse_vector<std::complex<double>, int>(r, c) ));
159     CALL_SUBTEST_1(( sparse_vector<double,long int>(r, c) ));
160     CALL_SUBTEST_1(( sparse_vector<double,short>(r, c) ));
161   }
162 }
163 
164