sparse_jac_pattern_xam.cpp#

C++: Jacobian Sparsity Patterns: Example and Test#

# include <cstdio>
# include <cppad/py/cppad_py.hpp>

bool sparse_jac_pattern_xam(void) {
   using cppad_py::a_double;
   using cppad_py::vec_int;
   using cppad_py::vec_double;
   using cppad_py::vec_a_double;
   using cppad_py::d_fun;
   using cppad_py::sparse_rc;
   //
   // initialize return variable
   bool ok = true;
   //------------------------------------------------------------------------
   // number of dependent and independent variables
   int n = 3;
   //
   // create the independent variables ax
   vec_double x(n);
   for(int i = 0; i < n ; i++) {
      x[i] = i + 2.0;
   }
   vec_a_double ax = cppad_py::independent(x);
   //
   // create dependent variables ay with ay[i] = ax[j]
   // where i = mod(j + 1, n)
   vec_a_double ay(n);
   for(int j = 0; j < n ; j++) {
      int i = j+1;
      if( i >= n  ) {
         i = i - n;
      }
      a_double ay_i = ax[j];
      ay[i] = ay_i;
   }
   //
   // define af corresponding to f(x)
   d_fun f(ax, ay);
   //
   // sparsity pattern for identity matrix
   sparse_rc pat_in;
   pat_in.resize(n, n, n);
   for(int k = 0; k < n; k++) {
      pat_in.put(k, k, k);
   }
   //
   // loop over forward and reverse mode
   for(int mode = 0; mode < 2; mode++) {
      sparse_rc pat_out;
      if( mode == 0  ) {
         f.for_jac_sparsity(pat_in, pat_out);
      }
      if( mode == 1  ) {
         f.rev_jac_sparsity(pat_in, pat_out);
      }
      //
      // check that result is sparsity pattern for Jacobian
      ok = ok && n == pat_out.nnz();
      vec_int col_major = pat_out.col_major();
      vec_int row = pat_out.row();
      vec_int col = pat_out.col();
      for(int k = 0; k < n; k++) {
         int ell = col_major[k];
         int r = row[ell];
         int c = col[ell];
         int i = c+1;
         if( i >=  n  ) {
            i = i - n;
         }
         ok = ok && c == k;
         ok = ok && r == i;
      }
   }
   //
   return( ok );
}