static void s_not_left(void)
{
sgt_printf()->flags.width -= ft_strlen(sgt_printf()->work_buffer)
+ sgt_printf()->flags.prec;
s_if_flags();
if (sgt_printf()->flags.pad == ' ')
{
fill_character(' ');
sgt_printf()->flags.width = 0;
}
s_flags();
s_alt();
sgt_printf()->flags.width += sgt_printf()->flags.prec;
fill_character('0');
add_string(sgt_printf()->work_buffer);
}
void printf_number_left(t_printf *t, unsigned int len)
{
int temp;
s_flags(t);
s_alt(t);
t->flags.width -= len + t->flags.prec;
if (t->flags.prec > 0)
{
temp = t->flags.width;
t->flags.width = t->flags.prec;
if (t->flags.alt && t->flags.base == 8 && t->flags.number.word != 0)
{
t->flags.width -= 1;
temp += 1;
}
printf_fill_character(t, '0');
t->flags.width = temp;
}
printf_add_string(t, t->work_buffer, len);
printf_fill_character(t, ' ');
}
//' Do the GRAS algorithm for updating matrizes
//'
//' This function calculates an updated matrix X, based on a matrix A,
//' which meets the given row and columns totals. Note: A can contain
//' negative elements.
//' @param A the "base" matrix
//' @param u column vector with the row totals
//' @param v row vector with the column totals
//' @param epsilon the error tolerance level, default is 1e-10.
//' @param maxiter maximum number of iterations, default is 10000.
//' @param verbose should some information of the iterations be
//' displayed? Default is FALSE
//' @return the updated matrix X
//' @author Oliver Reiter
//' @references Junius T. and J. Oosterhaven (2003), The solution of
//' updating or regionalizing a matrix with both positive and
//' negative entries, Economic Systems Research, 15, pp. 87-96.
//'
//' Lenzen M., R. Wood and B. Gallego (2007), Some comments on the
//' GRAS method, Economic Systems Research, 19, pp. 461-465.
//'
//' Temurshoev, U., R.E. Miller and M.C. Bouwmeester (2013), A note
//' on the GRAS method, Economic Systems Research, 25, pp. 361-367.
//' @keywords gras, matrix updating
//' @examples
//' ## example from the papers
//' A <- matrix(c(7, 3, 5, -3, 2, 9, 8, 0, -2, 0, 2, 0),
//' ncol = 4, nrow = 3, byrow = TRUE)
//' u <- c(15, 26, -1)
//' v <- c(9, 16, 17, -2)
//' doGRAS(A, u, v)
//' @export
// [[Rcpp::export]]
arma::mat doGRAS(arma::mat A,
arma::colvec u,
arma::rowvec v,
double epsilon = 1e-10,
int maxiter = 1000,
bool verbose = false) {
int nrowsA = A.n_rows;
int ncolsA = A.n_cols;
int nrowsu = u.n_rows;
int ncolsv = v.n_cols;
if(nrowsA != nrowsu) {
// throw std::range_error("Unequal number of rows of A and u!");
::Rf_error("Unequal number of rows of A and u!");
}
if(ncolsA != ncolsv) {
// throw std::range_error("Unequal number of cols of A and v!");
::Rf_error("Unequal number of cols of A and v!");
}
// separate A into a positive (P) and a negative (N) matrix
arma::mat P = arma::mat(A);
P.transform( [](double val) { return val >= 0? val : 0; });
arma::mat N = arma::mat(A);
N.transform( [](double val) { return val < 0? -val : 0; });
// initialize the multiplicators
arma::rowvec s = arma::rowvec(ncolsA).ones();
arma::rowvec s_old, s_alt, pj_r, nj_r = arma::rowvec(ncolsA);
arma::colvec r = arma::colvec(nrowsA).ones();
arma::colvec r_alt, pi_s, ni_s = arma::colvec(nrowsA);
double error = 1;
int iter = 1;
int errorpos = -1;
do {
s_old = s;
// row multiplicator s
pj_r = r.t() * P;
nj_r = sinvc(r).t() * N;
s = sinvr(2 * pj_r) % (v + arma::sqrt(pow(v, 2) + 4 * pj_r % nj_r));
s_alt = -1 * sinvr(v) % nj_r;
s(find(s == 0)) = s_alt(find(s == 0));
// Rcpp::Rcout << "s: " << s; // << std::endl;
// column multiplicator r
pi_s = P * s.t();
ni_s = N * sinvr(s).t();
r = sinvc(2 * pi_s) % (u + arma::sqrt(pow(u, 2) + 4 * pi_s % ni_s));
r_alt = -1 * sinvc(u) % ni_s;
r(find(r == 0)) = r_alt(find(r == 0));
// Rcpp::Rcout << "r': " << r.t(); // << std::endl;
error = abs(s_old - s).max();
if(verbose) {
errorpos = abs(s_old - s).index_max();
Rcpp::Rcout << "iter: " << iter << " -> error: " << error
<< " at " << errorpos << std::endl;
}
iter++;
} while((error > epsilon) & (iter <= maxiter));
if(iter >= maxiter) {
Rcpp::Rcout << "Warning: Reached maxiter!" << std::endl;
}
if(verbose) {
Rcpp::Rcout << " => Iterations needed: " << iter;
Rcpp::Rcout << " => Resulting error: " << error << " at " << errorpos
<< std::endl;
}
// calculate the updated matrizes:
// the multiplicators are both applied to all columns and rows
P.each_row() %= s;
P.each_col() %= r;
N.each_row() %= sinvr(s);
N.each_col() %= sinvc(r);
return P - N;
}
