• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 // Copyright (C) 2008 Daniel Gomez Ferro <dgomezferro@gmail.com>
6 // Copyright (C) 2013 Désiré Nuentsa-Wakam <desire.nuentsa_wakam@inria.fr>
7 //
8 // This Source Code Form is subject to the terms of the Mozilla
9 // Public License v. 2.0. If a copy of the MPL was not distributed
10 // with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
11 
12 #include "sparse.h"
13 
sparse_basic(const SparseMatrixType & ref)14 template<typename SparseMatrixType> void sparse_basic(const SparseMatrixType& ref)
15 {
16   typedef typename SparseMatrixType::Index Index;
17   typedef Matrix<Index,2,1> Vector2;
18 
19   const Index rows = ref.rows();
20   const Index cols = ref.cols();
21   typedef typename SparseMatrixType::Scalar Scalar;
22   enum { Flags = SparseMatrixType::Flags };
23 
24   double density = (std::max)(8./(rows*cols), 0.01);
25   typedef Matrix<Scalar,Dynamic,Dynamic> DenseMatrix;
26   typedef Matrix<Scalar,Dynamic,1> DenseVector;
27   Scalar eps = 1e-6;
28 
29   Scalar s1 = internal::random<Scalar>();
30   {
31     SparseMatrixType m(rows, cols);
32     DenseMatrix refMat = DenseMatrix::Zero(rows, cols);
33     DenseVector vec1 = DenseVector::Random(rows);
34 
35     std::vector<Vector2> zeroCoords;
36     std::vector<Vector2> nonzeroCoords;
37     initSparse<Scalar>(density, refMat, m, 0, &zeroCoords, &nonzeroCoords);
38 
39     if (zeroCoords.size()==0 || nonzeroCoords.size()==0)
40       return;
41 
42     // test coeff and coeffRef
43     for (int i=0; i<(int)zeroCoords.size(); ++i)
44     {
45       VERIFY_IS_MUCH_SMALLER_THAN( m.coeff(zeroCoords[i].x(),zeroCoords[i].y()), eps );
46       if(internal::is_same<SparseMatrixType,SparseMatrix<Scalar,Flags> >::value)
47         VERIFY_RAISES_ASSERT( m.coeffRef(zeroCoords[0].x(),zeroCoords[0].y()) = 5 );
48     }
49     VERIFY_IS_APPROX(m, refMat);
50 
51     m.coeffRef(nonzeroCoords[0].x(), nonzeroCoords[0].y()) = Scalar(5);
52     refMat.coeffRef(nonzeroCoords[0].x(), nonzeroCoords[0].y()) = Scalar(5);
53 
54     VERIFY_IS_APPROX(m, refMat);
55       /*
56       // test InnerIterators and Block expressions
57       for (int t=0; t<10; ++t)
58       {
59         int j = internal::random<int>(0,cols-1);
60         int i = internal::random<int>(0,rows-1);
61         int w = internal::random<int>(1,cols-j-1);
62         int h = internal::random<int>(1,rows-i-1);
63 
64     //     VERIFY_IS_APPROX(m.block(i,j,h,w), refMat.block(i,j,h,w));
65         for(int c=0; c<w; c++)
66         {
67           VERIFY_IS_APPROX(m.block(i,j,h,w).col(c), refMat.block(i,j,h,w).col(c));
68           for(int r=0; r<h; r++)
69           {
70     //         VERIFY_IS_APPROX(m.block(i,j,h,w).col(c).coeff(r), refMat.block(i,j,h,w).col(c).coeff(r));
71           }
72         }
73     //     for(int r=0; r<h; r++)
74     //     {
75     //       VERIFY_IS_APPROX(m.block(i,j,h,w).row(r), refMat.block(i,j,h,w).row(r));
76     //       for(int c=0; c<w; c++)
77     //       {
78     //         VERIFY_IS_APPROX(m.block(i,j,h,w).row(r).coeff(c), refMat.block(i,j,h,w).row(r).coeff(c));
79     //       }
80     //     }
81       }
82 
83       for(int c=0; c<cols; c++)
84       {
85         VERIFY_IS_APPROX(m.col(c) + m.col(c), (m + m).col(c));
86         VERIFY_IS_APPROX(m.col(c) + m.col(c), refMat.col(c) + refMat.col(c));
87       }
88 
89       for(int r=0; r<rows; r++)
90       {
91         VERIFY_IS_APPROX(m.row(r) + m.row(r), (m + m).row(r));
92         VERIFY_IS_APPROX(m.row(r) + m.row(r), refMat.row(r) + refMat.row(r));
93       }
94       */
95 
96       // test assertion
97       VERIFY_RAISES_ASSERT( m.coeffRef(-1,1) = 0 );
98       VERIFY_RAISES_ASSERT( m.coeffRef(0,m.cols()) = 0 );
99     }
100 
101     // test insert (inner random)
102     {
103       DenseMatrix m1(rows,cols);
104       m1.setZero();
105       SparseMatrixType m2(rows,cols);
106       if(internal::random<int>()%2)
107         m2.reserve(VectorXi::Constant(m2.outerSize(), 2));
108       for (Index j=0; j<cols; ++j)
109       {
110         for (Index k=0; k<rows/2; ++k)
111         {
112           Index i = internal::random<Index>(0,rows-1);
113           if (m1.coeff(i,j)==Scalar(0))
114             m2.insert(i,j) = m1(i,j) = internal::random<Scalar>();
115         }
116       }
117       m2.finalize();
118       VERIFY_IS_APPROX(m2,m1);
119     }
120 
121     // test insert (fully random)
122     {
123       DenseMatrix m1(rows,cols);
124       m1.setZero();
125       SparseMatrixType m2(rows,cols);
126       if(internal::random<int>()%2)
127         m2.reserve(VectorXi::Constant(m2.outerSize(), 2));
128       for (int k=0; k<rows*cols; ++k)
129       {
130         Index i = internal::random<Index>(0,rows-1);
131         Index j = internal::random<Index>(0,cols-1);
132         if ((m1.coeff(i,j)==Scalar(0)) && (internal::random<int>()%2))
133           m2.insert(i,j) = m1(i,j) = internal::random<Scalar>();
134         else
135         {
136           Scalar v = internal::random<Scalar>();
137           m2.coeffRef(i,j) += v;
138           m1(i,j) += v;
139         }
140       }
141       VERIFY_IS_APPROX(m2,m1);
142     }
143 
144     // test insert (un-compressed)
145     for(int mode=0;mode<4;++mode)
146     {
147       DenseMatrix m1(rows,cols);
148       m1.setZero();
149       SparseMatrixType m2(rows,cols);
150       VectorXi r(VectorXi::Constant(m2.outerSize(), ((mode%2)==0) ? m2.innerSize() : std::max<int>(1,m2.innerSize()/8)));
151       m2.reserve(r);
152       for (int k=0; k<rows*cols; ++k)
153       {
154         Index i = internal::random<Index>(0,rows-1);
155         Index j = internal::random<Index>(0,cols-1);
156         if (m1.coeff(i,j)==Scalar(0))
157           m2.insert(i,j) = m1(i,j) = internal::random<Scalar>();
158         if(mode==3)
159           m2.reserve(r);
160       }
161       if(internal::random<int>()%2)
162         m2.makeCompressed();
163       VERIFY_IS_APPROX(m2,m1);
164     }
165 
166   // test innerVector()
167   {
168     DenseMatrix refMat2 = DenseMatrix::Zero(rows, rows);
169     SparseMatrixType m2(rows, rows);
170     initSparse<Scalar>(density, refMat2, m2);
171     Index j0 = internal::random<Index>(0,rows-1);
172     Index j1 = internal::random<Index>(0,rows-1);
173     if(SparseMatrixType::IsRowMajor)
174       VERIFY_IS_APPROX(m2.innerVector(j0), refMat2.row(j0));
175     else
176       VERIFY_IS_APPROX(m2.innerVector(j0), refMat2.col(j0));
177 
178     if(SparseMatrixType::IsRowMajor)
179       VERIFY_IS_APPROX(m2.innerVector(j0)+m2.innerVector(j1), refMat2.row(j0)+refMat2.row(j1));
180     else
181       VERIFY_IS_APPROX(m2.innerVector(j0)+m2.innerVector(j1), refMat2.col(j0)+refMat2.col(j1));
182 
183     SparseMatrixType m3(rows,rows);
184     m3.reserve(VectorXi::Constant(rows,rows/2));
185     for(Index j=0; j<rows; ++j)
186       for(Index k=0; k<j; ++k)
187         m3.insertByOuterInner(j,k) = k+1;
188     for(Index j=0; j<rows; ++j)
189     {
190       VERIFY(j==numext::real(m3.innerVector(j).nonZeros()));
191       if(j>0)
192         VERIFY(j==numext::real(m3.innerVector(j).lastCoeff()));
193     }
194     m3.makeCompressed();
195     for(Index j=0; j<rows; ++j)
196     {
197       VERIFY(j==numext::real(m3.innerVector(j).nonZeros()));
198       if(j>0)
199         VERIFY(j==numext::real(m3.innerVector(j).lastCoeff()));
200     }
201 
202     //m2.innerVector(j0) = 2*m2.innerVector(j1);
203     //refMat2.col(j0) = 2*refMat2.col(j1);
204     //VERIFY_IS_APPROX(m2, refMat2);
205   }
206 
207   // test innerVectors()
208   {
209     DenseMatrix refMat2 = DenseMatrix::Zero(rows, rows);
210     SparseMatrixType m2(rows, rows);
211     initSparse<Scalar>(density, refMat2, m2);
212     if(internal::random<float>(0,1)>0.5) m2.makeCompressed();
213 
214     Index j0 = internal::random<Index>(0,rows-2);
215     Index j1 = internal::random<Index>(0,rows-2);
216     Index n0 = internal::random<Index>(1,rows-(std::max)(j0,j1));
217     if(SparseMatrixType::IsRowMajor)
218       VERIFY_IS_APPROX(m2.innerVectors(j0,n0), refMat2.block(j0,0,n0,cols));
219     else
220       VERIFY_IS_APPROX(m2.innerVectors(j0,n0), refMat2.block(0,j0,rows,n0));
221     if(SparseMatrixType::IsRowMajor)
222       VERIFY_IS_APPROX(m2.innerVectors(j0,n0)+m2.innerVectors(j1,n0),
223                        refMat2.middleRows(j0,n0)+refMat2.middleRows(j1,n0));
224     else
225       VERIFY_IS_APPROX(m2.innerVectors(j0,n0)+m2.innerVectors(j1,n0),
226                       refMat2.block(0,j0,rows,n0)+refMat2.block(0,j1,rows,n0));
227 
228     VERIFY_IS_APPROX(m2, refMat2);
229 
230     m2.innerVectors(j0,n0) = m2.innerVectors(j0,n0) + m2.innerVectors(j1,n0);
231     if(SparseMatrixType::IsRowMajor)
232       refMat2.middleRows(j0,n0) = (refMat2.middleRows(j0,n0) + refMat2.middleRows(j1,n0)).eval();
233     else
234       refMat2.middleCols(j0,n0) = (refMat2.middleCols(j0,n0) + refMat2.middleCols(j1,n0)).eval();
235 
236     VERIFY_IS_APPROX(m2, refMat2);
237 
238   }
239 
240   // test basic computations
241   {
242     DenseMatrix refM1 = DenseMatrix::Zero(rows, rows);
243     DenseMatrix refM2 = DenseMatrix::Zero(rows, rows);
244     DenseMatrix refM3 = DenseMatrix::Zero(rows, rows);
245     DenseMatrix refM4 = DenseMatrix::Zero(rows, rows);
246     SparseMatrixType m1(rows, rows);
247     SparseMatrixType m2(rows, rows);
248     SparseMatrixType m3(rows, rows);
249     SparseMatrixType m4(rows, rows);
250     initSparse<Scalar>(density, refM1, m1);
251     initSparse<Scalar>(density, refM2, m2);
252     initSparse<Scalar>(density, refM3, m3);
253     initSparse<Scalar>(density, refM4, m4);
254 
255     VERIFY_IS_APPROX(m1+m2, refM1+refM2);
256     VERIFY_IS_APPROX(m1+m2+m3, refM1+refM2+refM3);
257     VERIFY_IS_APPROX(m3.cwiseProduct(m1+m2), refM3.cwiseProduct(refM1+refM2));
258     VERIFY_IS_APPROX(m1*s1-m2, refM1*s1-refM2);
259 
260     VERIFY_IS_APPROX(m1*=s1, refM1*=s1);
261     VERIFY_IS_APPROX(m1/=s1, refM1/=s1);
262 
263     VERIFY_IS_APPROX(m1+=m2, refM1+=refM2);
264     VERIFY_IS_APPROX(m1-=m2, refM1-=refM2);
265 
266     if(SparseMatrixType::IsRowMajor)
267       VERIFY_IS_APPROX(m1.innerVector(0).dot(refM2.row(0)), refM1.row(0).dot(refM2.row(0)));
268     else
269       VERIFY_IS_APPROX(m1.innerVector(0).dot(refM2.row(0)), refM1.col(0).dot(refM2.row(0)));
270 
271     VERIFY_IS_APPROX(m1.conjugate(), refM1.conjugate());
272     VERIFY_IS_APPROX(m1.real(), refM1.real());
273 
274     refM4.setRandom();
275     // sparse cwise* dense
276     VERIFY_IS_APPROX(m3.cwiseProduct(refM4), refM3.cwiseProduct(refM4));
277 //     VERIFY_IS_APPROX(m3.cwise()/refM4, refM3.cwise()/refM4);
278 
279     // test aliasing
280     VERIFY_IS_APPROX((m1 = -m1), (refM1 = -refM1));
281     VERIFY_IS_APPROX((m1 = m1.transpose()), (refM1 = refM1.transpose().eval()));
282     VERIFY_IS_APPROX((m1 = -m1.transpose()), (refM1 = -refM1.transpose().eval()));
283     VERIFY_IS_APPROX((m1 += -m1), (refM1 += -refM1));
284   }
285 
286   // test transpose
287   {
288     DenseMatrix refMat2 = DenseMatrix::Zero(rows, rows);
289     SparseMatrixType m2(rows, rows);
290     initSparse<Scalar>(density, refMat2, m2);
291     VERIFY_IS_APPROX(m2.transpose().eval(), refMat2.transpose().eval());
292     VERIFY_IS_APPROX(m2.transpose(), refMat2.transpose());
293 
294     VERIFY_IS_APPROX(SparseMatrixType(m2.adjoint()), refMat2.adjoint());
295   }
296 
297 
298 
299   // test generic blocks
300   {
301     DenseMatrix refMat2 = DenseMatrix::Zero(rows, rows);
302     SparseMatrixType m2(rows, rows);
303     initSparse<Scalar>(density, refMat2, m2);
304     Index j0 = internal::random<Index>(0,rows-2);
305     Index j1 = internal::random<Index>(0,rows-2);
306     Index n0 = internal::random<Index>(1,rows-(std::max)(j0,j1));
307     if(SparseMatrixType::IsRowMajor)
308       VERIFY_IS_APPROX(m2.block(j0,0,n0,cols), refMat2.block(j0,0,n0,cols));
309     else
310       VERIFY_IS_APPROX(m2.block(0,j0,rows,n0), refMat2.block(0,j0,rows,n0));
311 
312     if(SparseMatrixType::IsRowMajor)
313       VERIFY_IS_APPROX(m2.block(j0,0,n0,cols)+m2.block(j1,0,n0,cols),
314                       refMat2.block(j0,0,n0,cols)+refMat2.block(j1,0,n0,cols));
315     else
316       VERIFY_IS_APPROX(m2.block(0,j0,rows,n0)+m2.block(0,j1,rows,n0),
317                       refMat2.block(0,j0,rows,n0)+refMat2.block(0,j1,rows,n0));
318 
319     Index i = internal::random<Index>(0,m2.outerSize()-1);
320     if(SparseMatrixType::IsRowMajor) {
321       m2.innerVector(i) = m2.innerVector(i) * s1;
322       refMat2.row(i) = refMat2.row(i) * s1;
323       VERIFY_IS_APPROX(m2,refMat2);
324     } else {
325       m2.innerVector(i) = m2.innerVector(i) * s1;
326       refMat2.col(i) = refMat2.col(i) * s1;
327       VERIFY_IS_APPROX(m2,refMat2);
328     }
329   }
330 
331   // test prune
332   {
333     SparseMatrixType m2(rows, rows);
334     DenseMatrix refM2(rows, rows);
335     refM2.setZero();
336     int countFalseNonZero = 0;
337     int countTrueNonZero = 0;
338     for (Index j=0; j<m2.outerSize(); ++j)
339     {
340       m2.startVec(j);
341       for (Index i=0; i<m2.innerSize(); ++i)
342       {
343         float x = internal::random<float>(0,1);
344         if (x<0.1)
345         {
346           // do nothing
347         }
348         else if (x<0.5)
349         {
350           countFalseNonZero++;
351           m2.insertBackByOuterInner(j,i) = Scalar(0);
352         }
353         else
354         {
355           countTrueNonZero++;
356           m2.insertBackByOuterInner(j,i) = Scalar(1);
357           if(SparseMatrixType::IsRowMajor)
358             refM2(j,i) = Scalar(1);
359           else
360             refM2(i,j) = Scalar(1);
361         }
362       }
363     }
364     m2.finalize();
365     VERIFY(countFalseNonZero+countTrueNonZero == m2.nonZeros());
366     VERIFY_IS_APPROX(m2, refM2);
367     m2.prune(Scalar(1));
368     VERIFY(countTrueNonZero==m2.nonZeros());
369     VERIFY_IS_APPROX(m2, refM2);
370   }
371 
372   // test setFromTriplets
373   {
374     typedef Triplet<Scalar,Index> TripletType;
375     std::vector<TripletType> triplets;
376     int ntriplets = rows*cols;
377     triplets.reserve(ntriplets);
378     DenseMatrix refMat(rows,cols);
379     refMat.setZero();
380     for(int i=0;i<ntriplets;++i)
381     {
382       Index r = internal::random<Index>(0,rows-1);
383       Index c = internal::random<Index>(0,cols-1);
384       Scalar v = internal::random<Scalar>();
385       triplets.push_back(TripletType(r,c,v));
386       refMat(r,c) += v;
387     }
388     SparseMatrixType m(rows,cols);
389     m.setFromTriplets(triplets.begin(), triplets.end());
390     VERIFY_IS_APPROX(m, refMat);
391   }
392 
393   // test triangularView
394   {
395     DenseMatrix refMat2(rows, rows), refMat3(rows, rows);
396     SparseMatrixType m2(rows, rows), m3(rows, rows);
397     initSparse<Scalar>(density, refMat2, m2);
398     refMat3 = refMat2.template triangularView<Lower>();
399     m3 = m2.template triangularView<Lower>();
400     VERIFY_IS_APPROX(m3, refMat3);
401 
402     refMat3 = refMat2.template triangularView<Upper>();
403     m3 = m2.template triangularView<Upper>();
404     VERIFY_IS_APPROX(m3, refMat3);
405 
406     refMat3 = refMat2.template triangularView<UnitUpper>();
407     m3 = m2.template triangularView<UnitUpper>();
408     VERIFY_IS_APPROX(m3, refMat3);
409 
410     refMat3 = refMat2.template triangularView<UnitLower>();
411     m3 = m2.template triangularView<UnitLower>();
412     VERIFY_IS_APPROX(m3, refMat3);
413 
414     refMat3 = refMat2.template triangularView<StrictlyUpper>();
415     m3 = m2.template triangularView<StrictlyUpper>();
416     VERIFY_IS_APPROX(m3, refMat3);
417 
418     refMat3 = refMat2.template triangularView<StrictlyLower>();
419     m3 = m2.template triangularView<StrictlyLower>();
420     VERIFY_IS_APPROX(m3, refMat3);
421   }
422 
423   // test selfadjointView
424   if(!SparseMatrixType::IsRowMajor)
425   {
426     DenseMatrix refMat2(rows, rows), refMat3(rows, rows);
427     SparseMatrixType m2(rows, rows), m3(rows, rows);
428     initSparse<Scalar>(density, refMat2, m2);
429     refMat3 = refMat2.template selfadjointView<Lower>();
430     m3 = m2.template selfadjointView<Lower>();
431     VERIFY_IS_APPROX(m3, refMat3);
432   }
433 
434   // test sparseView
435   {
436     DenseMatrix refMat2 = DenseMatrix::Zero(rows, rows);
437     SparseMatrixType m2(rows, rows);
438     initSparse<Scalar>(density, refMat2, m2);
439     VERIFY_IS_APPROX(m2.eval(), refMat2.sparseView().eval());
440   }
441 
442   // test diagonal
443   {
444     DenseMatrix refMat2 = DenseMatrix::Zero(rows, rows);
445     SparseMatrixType m2(rows, rows);
446     initSparse<Scalar>(density, refMat2, m2);
447     VERIFY_IS_APPROX(m2.diagonal(), refMat2.diagonal().eval());
448   }
449 
450   // test conservative resize
451   {
452       std::vector< std::pair<Index,Index> > inc;
453       inc.push_back(std::pair<Index,Index>(-3,-2));
454       inc.push_back(std::pair<Index,Index>(0,0));
455       inc.push_back(std::pair<Index,Index>(3,2));
456       inc.push_back(std::pair<Index,Index>(3,0));
457       inc.push_back(std::pair<Index,Index>(0,3));
458 
459       for(size_t i = 0; i< inc.size(); i++) {
460         Index incRows = inc[i].first;
461         Index incCols = inc[i].second;
462         SparseMatrixType m1(rows, cols);
463         DenseMatrix refMat1 = DenseMatrix::Zero(rows, cols);
464         initSparse<Scalar>(density, refMat1, m1);
465 
466         m1.conservativeResize(rows+incRows, cols+incCols);
467         refMat1.conservativeResize(rows+incRows, cols+incCols);
468         if (incRows > 0) refMat1.bottomRows(incRows).setZero();
469         if (incCols > 0) refMat1.rightCols(incCols).setZero();
470 
471         VERIFY_IS_APPROX(m1, refMat1);
472 
473         // Insert new values
474         if (incRows > 0)
475           m1.insert(m1.rows()-1, 0) = refMat1(refMat1.rows()-1, 0) = 1;
476         if (incCols > 0)
477           m1.insert(0, m1.cols()-1) = refMat1(0, refMat1.cols()-1) = 1;
478 
479         VERIFY_IS_APPROX(m1, refMat1);
480 
481 
482       }
483   }
484 
485   // test Identity matrix
486   {
487     DenseMatrix refMat1 = DenseMatrix::Identity(rows, rows);
488     SparseMatrixType m1(rows, rows);
489     m1.setIdentity();
490     VERIFY_IS_APPROX(m1, refMat1);
491   }
492 }
493 
test_sparse_basic()494 void test_sparse_basic()
495 {
496   for(int i = 0; i < g_repeat; i++) {
497     int s = Eigen::internal::random<int>(1,50);
498     EIGEN_UNUSED_VARIABLE(s);
499     CALL_SUBTEST_1(( sparse_basic(SparseMatrix<double>(8, 8)) ));
500     CALL_SUBTEST_2(( sparse_basic(SparseMatrix<std::complex<double>, ColMajor>(s, s)) ));
501     CALL_SUBTEST_2(( sparse_basic(SparseMatrix<std::complex<double>, RowMajor>(s, s)) ));
502     CALL_SUBTEST_1(( sparse_basic(SparseMatrix<double>(s, s)) ));
503     CALL_SUBTEST_1(( sparse_basic(SparseMatrix<double,ColMajor,long int>(s, s)) ));
504     CALL_SUBTEST_1(( sparse_basic(SparseMatrix<double,RowMajor,long int>(s, s)) ));
505 
506     CALL_SUBTEST_1(( sparse_basic(SparseMatrix<double,ColMajor,short int>(short(s), short(s))) ));
507     CALL_SUBTEST_1(( sparse_basic(SparseMatrix<double,RowMajor,short int>(short(s), short(s))) ));
508   }
509 }
510