]> www.ginac.de Git - ginac.git/blobdiff - ginac/matrix.cpp
fixed a bug where quo() would call vector::reserve() with a negative argument
[ginac.git] / ginac / matrix.cpp
index 28c2f1d25a853faf7750220d1a82867d9f425775..0bedb6bb1a1e506fc0017ea3243d66c64ed3797a 100644 (file)
@@ -20,6 +20,7 @@
  *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 
+#include <iostream>
 #include <algorithm>
 #include <map>
 #include <stdexcept>
@@ -35,7 +36,6 @@
 #include "print.h"
 #include "archive.h"
 #include "utils.h"
-#include "debugmsg.h"
 
 namespace GiNaC {
 
@@ -48,8 +48,7 @@ GINAC_IMPLEMENT_REGISTERED_CLASS(matrix, basic)
 /** Default ctor.  Initializes to 1 x 1-dimensional zero-matrix. */
 matrix::matrix() : inherited(TINFO_matrix), row(1), col(1)
 {
-       debugmsg("matrix default ctor",LOGLEVEL_CONSTRUCT);
-       m.push_back(_ex0());
+       m.push_back(_ex0);
 }
 
 void matrix::copy(const matrix & other)
@@ -75,18 +74,14 @@ DEFAULT_DESTROY(matrix)
 matrix::matrix(unsigned r, unsigned c)
   : inherited(TINFO_matrix), row(r), col(c)
 {
-       debugmsg("matrix ctor from unsigned,unsigned",LOGLEVEL_CONSTRUCT);
-       m.resize(r*c, _ex0());
+       m.resize(r*c, _ex0);
 }
 
 // protected
 
 /** Ctor from representation, for internal use only. */
 matrix::matrix(unsigned r, unsigned c, const exvector & m2)
-  : inherited(TINFO_matrix), row(r), col(c), m(m2)
-{
-       debugmsg("matrix ctor from unsigned,unsigned,exvector",LOGLEVEL_CONSTRUCT);
-}
+  : inherited(TINFO_matrix), row(r), col(c), m(m2) {}
 
 /** Construct matrix from (flat) list of elements. If the list has fewer
  *  elements than the matrix, the remaining matrix elements are set to zero.
@@ -95,8 +90,7 @@ matrix::matrix(unsigned r, unsigned c, const exvector & m2)
 matrix::matrix(unsigned r, unsigned c, const lst & l)
   : inherited(TINFO_matrix), row(r), col(c)
 {
-       debugmsg("matrix ctor from unsigned,unsigned,lst",LOGLEVEL_CONSTRUCT);
-       m.resize(r*c, _ex0());
+       m.resize(r*c, _ex0);
 
        for (unsigned i=0; i<l.nops(); i++) {
                unsigned x = i % c;
@@ -113,7 +107,6 @@ matrix::matrix(unsigned r, unsigned c, const lst & l)
 
 matrix::matrix(const archive_node &n, const lst &sym_lst) : inherited(n, sym_lst)
 {
-       debugmsg("matrix ctor from archive_node", LOGLEVEL_CONSTRUCT);
        if (!(n.find_unsigned("row", row)) || !(n.find_unsigned("col", col)))
                throw (std::runtime_error("unknown matrix dimensions in archive"));
        m.reserve(row * col);
@@ -141,21 +134,22 @@ void matrix::archive(archive_node &n) const
 DEFAULT_UNARCHIVE(matrix)
 
 //////////
-// functions overriding virtual functions from bases classes
+// functions overriding virtual functions from base classes
 //////////
 
 // public
 
 void matrix::print(const print_context & c, unsigned level) const
 {
-       debugmsg("matrix print", LOGLEVEL_PRINT);
-
-       if (is_of_type(c, print_tree)) {
+       if (is_a<print_tree>(c)) {
 
                inherited::print(c, level);
 
        } else {
 
+               if (is_a<print_python_repr>(c))
+                       c.s << class_name() << '(';
+
                c.s << "[";
                for (unsigned y=0; y<row-1; ++y) {
                        c.s << "[";
@@ -174,6 +168,9 @@ void matrix::print(const print_context & c, unsigned level) const
                m[row*col-1].print(c);
                c.s << "]]";
 
+               if (is_a<print_python_repr>(c))
+                       c.s << ')';
+
        }
 }
 
@@ -198,21 +195,9 @@ ex & matrix::let_op(int i)
        return m[i];
 }
 
-/** expands the elements of a matrix entry by entry. */
-ex matrix::expand(unsigned options) const
-{
-       exvector tmp(row*col);
-       for (unsigned i=0; i<row*col; ++i)
-               tmp[i] = m[i].expand(options);
-       
-       return matrix(row, col, tmp);
-}
-
 /** Evaluate matrix entry by entry. */
 ex matrix::eval(int level) const
 {
-       debugmsg("matrix eval",LOGLEVEL_MEMBER_FUNCTION);
-       
        // check if we have to do anything at all
        if ((level==1)&&(flags & status_flags::evaluated))
                return *this;
@@ -232,30 +217,6 @@ ex matrix::eval(int level) const
                                                                                           status_flags::evaluated );
 }
 
-/** Evaluate matrix numerically entry by entry. */
-ex matrix::evalf(int level) const
-{
-       debugmsg("matrix evalf",LOGLEVEL_MEMBER_FUNCTION);
-               
-       // check if we have to do anything at all
-       if (level==1)
-               return *this;
-       
-       // emergency break
-       if (level == -max_recursion_level) {
-               throw (std::runtime_error("matrix::evalf(): recursion limit exceeded"));
-       }
-       
-       // evalf() entry by entry
-       exvector m2(row*col);
-       --level;
-       for (unsigned r=0; r<row; ++r)
-               for (unsigned c=0; c<col; ++c)
-                       m2[r*col+c] = m[r*col+c].evalf(level);
-       
-       return matrix(row, col, m2);
-}
-
 ex matrix::subs(const lst & ls, const lst & lr, bool no_pattern) const
 {
        exvector m2(row * col);
@@ -263,15 +224,15 @@ ex matrix::subs(const lst & ls, const lst & lr, bool no_pattern) const
                for (unsigned c=0; c<col; ++c)
                        m2[r*col+c] = m[r*col+c].subs(ls, lr, no_pattern);
 
-       return ex(matrix(row, col, m2)).bp->basic::subs(ls, lr, no_pattern);
+       return matrix(row, col, m2).basic::subs(ls, lr, no_pattern);
 }
 
 // protected
 
 int matrix::compare_same_type(const basic & other) const
 {
-       GINAC_ASSERT(is_exactly_of_type(other, matrix));
-       const matrix & o = static_cast<matrix &>(const_cast<basic &>(other));
+       GINAC_ASSERT(is_exactly_a<matrix>(other));
+       const matrix &o = static_cast<const matrix &>(other);
        
        // compare number of rows
        if (row != o.rows())
@@ -293,11 +254,21 @@ int matrix::compare_same_type(const basic & other) const
        return 0;
 }
 
+bool matrix::match_same_type(const basic & other) const
+{
+       GINAC_ASSERT(is_exactly_a<matrix>(other));
+       const matrix & o = static_cast<const matrix &>(other);
+       
+       // The number of rows and columns must be the same. This is necessary to
+       // prevent a 2x3 matrix from matching a 3x2 one.
+       return row == o.rows() && col == o.cols();
+}
+
 /** Automatic symbolic evaluation of an indexed matrix. */
 ex matrix::eval_indexed(const basic & i) const
 {
-       GINAC_ASSERT(is_of_type(i, indexed));
-       GINAC_ASSERT(is_ex_of_type(i.op(0), matrix));
+       GINAC_ASSERT(is_a<indexed>(i));
+       GINAC_ASSERT(is_a<matrix>(i.op(0)));
 
        bool all_indices_unsigned = static_cast<const indexed &>(i).all_index_values_are(info_flags::nonnegint);
 
@@ -308,7 +279,7 @@ ex matrix::eval_indexed(const basic & i) const
                if (row != 1 && col != 1)
                        throw (std::runtime_error("matrix::eval_indexed(): vector must have exactly 1 index"));
 
-               const idx & i1 = ex_to_idx(i.op(1));
+               const idx & i1 = ex_to<idx>(i.op(1));
 
                if (col == 1) {
 
@@ -318,7 +289,7 @@ ex matrix::eval_indexed(const basic & i) const
 
                        // Index numeric -> return vector element
                        if (all_indices_unsigned) {
-                               unsigned n1 = ex_to_numeric(i1.get_value()).to_int();
+                               unsigned n1 = ex_to<numeric>(i1.get_value()).to_int();
                                if (n1 >= row)
                                        throw (std::runtime_error("matrix::eval_indexed(): value of index exceeds number of vector elements"));
                                return (*this)(n1, 0);
@@ -332,7 +303,7 @@ ex matrix::eval_indexed(const basic & i) const
 
                        // Index numeric -> return vector element
                        if (all_indices_unsigned) {
-                               unsigned n1 = ex_to_numeric(i1.get_value()).to_int();
+                               unsigned n1 = ex_to<numeric>(i1.get_value()).to_int();
                                if (n1 >= col)
                                        throw (std::runtime_error("matrix::eval_indexed(): value of index exceeds number of vector elements"));
                                return (*this)(0, n1);
@@ -342,8 +313,8 @@ ex matrix::eval_indexed(const basic & i) const
        } else if (i.nops() == 3) {
 
                // Two indices
-               const idx & i1 = ex_to_idx(i.op(1));
-               const idx & i2 = ex_to_idx(i.op(2));
+               const idx & i1 = ex_to<idx>(i.op(1));
+               const idx & i2 = ex_to<idx>(i.op(2));
 
                if (!i1.get_dim().is_equal(row))
                        throw (std::runtime_error("matrix::eval_indexed(): dimension of first index must match number of rows"));
@@ -356,7 +327,7 @@ ex matrix::eval_indexed(const basic & i) const
 
                // Both indices numeric -> return matrix element
                if (all_indices_unsigned) {
-                       unsigned n1 = ex_to_numeric(i1.get_value()).to_int(), n2 = ex_to_numeric(i2.get_value()).to_int();
+                       unsigned n1 = ex_to<numeric>(i1.get_value()).to_int(), n2 = ex_to<numeric>(i2.get_value()).to_int();
                        if (n1 >= row)
                                throw (std::runtime_error("matrix::eval_indexed(): value of first index exceeds number of rows"));
                        if (n2 >= col)
@@ -373,17 +344,17 @@ ex matrix::eval_indexed(const basic & i) const
 /** Sum of two indexed matrices. */
 ex matrix::add_indexed(const ex & self, const ex & other) const
 {
-       GINAC_ASSERT(is_ex_of_type(self, indexed));
-       GINAC_ASSERT(is_ex_of_type(self.op(0), matrix));
-       GINAC_ASSERT(is_ex_of_type(other, indexed));
+       GINAC_ASSERT(is_a<indexed>(self));
+       GINAC_ASSERT(is_a<matrix>(self.op(0)));
+       GINAC_ASSERT(is_a<indexed>(other));
        GINAC_ASSERT(self.nops() == 2 || self.nops() == 3);
 
        // Only add two matrices
        if (is_ex_of_type(other.op(0), matrix)) {
                GINAC_ASSERT(other.nops() == 2 || other.nops() == 3);
 
-               const matrix &self_matrix = ex_to_matrix(self.op(0));
-               const matrix &other_matrix = ex_to_matrix(other.op(0));
+               const matrix &self_matrix = ex_to<matrix>(self.op(0));
+               const matrix &other_matrix = ex_to<matrix>(other.op(0));
 
                if (self.nops() == 2 && other.nops() == 2) { // vector + vector
 
@@ -409,11 +380,11 @@ ex matrix::add_indexed(const ex & self, const ex & other) const
 /** Product of an indexed matrix with a number. */
 ex matrix::scalar_mul_indexed(const ex & self, const numeric & other) const
 {
-       GINAC_ASSERT(is_ex_of_type(self, indexed));
-       GINAC_ASSERT(is_ex_of_type(self.op(0), matrix));
+       GINAC_ASSERT(is_a<indexed>(self));
+       GINAC_ASSERT(is_a<matrix>(self.op(0)));
        GINAC_ASSERT(self.nops() == 2 || self.nops() == 3);
 
-       const matrix &self_matrix = ex_to_matrix(self.op(0));
+       const matrix &self_matrix = ex_to<matrix>(self.op(0));
 
        if (self.nops() == 2)
                return indexed(self_matrix.mul(other), self.op(1));
@@ -424,10 +395,10 @@ ex matrix::scalar_mul_indexed(const ex & self, const numeric & other) const
 /** Contraction of an indexed matrix with something else. */
 bool matrix::contract_with(exvector::iterator self, exvector::iterator other, exvector & v) const
 {
-       GINAC_ASSERT(is_ex_of_type(*self, indexed));
-       GINAC_ASSERT(is_ex_of_type(*other, indexed));
+       GINAC_ASSERT(is_a<indexed>(*self));
+       GINAC_ASSERT(is_a<indexed>(*other));
        GINAC_ASSERT(self->nops() == 2 || self->nops() == 3);
-       GINAC_ASSERT(is_ex_of_type(self->op(0), matrix));
+       GINAC_ASSERT(is_a<matrix>(self->op(0)));
 
        // Only contract with other matrices
        if (!is_ex_of_type(other->op(0), matrix))
@@ -435,14 +406,12 @@ bool matrix::contract_with(exvector::iterator self, exvector::iterator other, ex
 
        GINAC_ASSERT(other->nops() == 2 || other->nops() == 3);
 
-       const matrix &self_matrix = ex_to_matrix(self->op(0));
-       const matrix &other_matrix = ex_to_matrix(other->op(0));
+       const matrix &self_matrix = ex_to<matrix>(self->op(0));
+       const matrix &other_matrix = ex_to<matrix>(other->op(0));
 
        if (self->nops() == 2) {
-               unsigned self_dim = (self_matrix.col == 1) ? self_matrix.row : self_matrix.col;
 
                if (other->nops() == 2) { // vector * vector (scalar product)
-                       unsigned other_dim = (other_matrix.col == 1) ? other_matrix.row : other_matrix.col;
 
                        if (self_matrix.col == 1) {
                                if (other_matrix.col == 1) {
@@ -461,7 +430,7 @@ bool matrix::contract_with(exvector::iterator self, exvector::iterator other, ex
                                        *self = self_matrix.mul(other_matrix.transpose())(0, 0);
                                }
                        }
-                       *other = _ex1();
+                       *other = _ex1;
                        return true;
 
                } else { // vector * matrix
@@ -472,7 +441,7 @@ bool matrix::contract_with(exvector::iterator self, exvector::iterator other, ex
                                        *self = indexed(self_matrix.mul(other_matrix), other->op(2));
                                else
                                        *self = indexed(self_matrix.transpose().mul(other_matrix), other->op(2));
-                               *other = _ex1();
+                               *other = _ex1;
                                return true;
                        }
 
@@ -482,7 +451,7 @@ bool matrix::contract_with(exvector::iterator self, exvector::iterator other, ex
                                        *self = indexed(other_matrix.mul(self_matrix), other->op(1));
                                else
                                        *self = indexed(other_matrix.mul(self_matrix.transpose()), other->op(1));
-                               *other = _ex1();
+                               *other = _ex1;
                                return true;
                        }
                }
@@ -492,28 +461,28 @@ bool matrix::contract_with(exvector::iterator self, exvector::iterator other, ex
                // A_ij * B_jk = (A*B)_ik
                if (is_dummy_pair(self->op(2), other->op(1))) {
                        *self = indexed(self_matrix.mul(other_matrix), self->op(1), other->op(2));
-                       *other = _ex1();
+                       *other = _ex1;
                        return true;
                }
 
                // A_ij * B_kj = (A*Btrans)_ik
                if (is_dummy_pair(self->op(2), other->op(2))) {
                        *self = indexed(self_matrix.mul(other_matrix.transpose()), self->op(1), other->op(1));
-                       *other = _ex1();
+                       *other = _ex1;
                        return true;
                }
 
                // A_ji * B_jk = (Atrans*B)_ik
                if (is_dummy_pair(self->op(1), other->op(1))) {
                        *self = indexed(self_matrix.transpose().mul(other_matrix), self->op(2), other->op(2));
-                       *other = _ex1();
+                       *other = _ex1;
                        return true;
                }
 
                // A_ji * B_kj = (B*A)_ki
                if (is_dummy_pair(self->op(1), other->op(2))) {
                        *self = indexed(other_matrix.mul(self_matrix), other->op(1), self->op(2));
-                       *other = _ex1();
+                       *other = _ex1;
                        return true;
                }
        }
@@ -537,10 +506,10 @@ matrix matrix::add(const matrix & other) const
                throw std::logic_error("matrix::add(): incompatible matrices");
        
        exvector sum(this->m);
-       exvector::iterator i;
-       exvector::const_iterator ci;
-       for (i=sum.begin(), ci=other.m.begin(); i!=sum.end(); ++i, ++ci)
-               (*i) += (*ci);
+       exvector::iterator i = sum.begin(), end = sum.end();
+       exvector::const_iterator ci = other.m.begin();
+       while (i != end)
+               *i++ += *ci++;
        
        return matrix(row,col,sum);
 }
@@ -555,10 +524,10 @@ matrix matrix::sub(const matrix & other) const
                throw std::logic_error("matrix::sub(): incompatible matrices");
        
        exvector dif(this->m);
-       exvector::iterator i;
-       exvector::const_iterator ci;
-       for (i=dif.begin(), ci=other.m.begin(); i!=dif.end(); ++i, ++ci)
-               (*i) -= (*ci);
+       exvector::iterator i = dif.begin(), end = dif.end();
+       exvector::const_iterator ci = other.m.begin();
+       while (i != end)
+               *i++ -= *ci++;
        
        return matrix(row,col,dif);
 }
@@ -625,38 +594,38 @@ matrix matrix::pow(const ex & expn) const
                // Integer cases are computed by successive multiplication, using the
                // obvious shortcut of storing temporaries, like A^4 == (A*A)*(A*A).
                if (expn.info(info_flags::integer)) {
-                       numeric k;
-                       matrix prod(row,col);
+                       numeric b = ex_to<numeric>(expn);
+                       matrix A(row,col);
                        if (expn.info(info_flags::negative)) {
-                               k = -ex_to_numeric(expn);
-                               prod = this->inverse();
+                               b *= -1;
+                               A = this->inverse();
                        } else {
-                               k = ex_to_numeric(expn);
-                               prod = *this;
+                               A = *this;
                        }
-                       matrix result(row,col);
+                       matrix C(row,col);
                        for (unsigned r=0; r<row; ++r)
-                               result.set(r,r,_ex1());
-                       numeric b(1);
-                       // this loop computes the representation of k in base 2 and multiplies
-                       // the factors whenever needed:
-                       while (b.compare(k)<=0) {
-                               b *= numeric(2);
-                               numeric r(mod(k,b));
-                               if (!r.is_zero()) {
-                                       k -= r;
-                                       result = result.mul(prod);
+                               C(r,r) = _ex1;
+                       // This loop computes the representation of b in base 2 from right
+                       // to left and multiplies the factors whenever needed.  Note
+                       // that this is not entirely optimal but close to optimal and
+                       // "better" algorithms are much harder to implement.  (See Knuth,
+                       // TAoCP2, section "Evaluation of Powers" for a good discussion.)
+                       while (b!=1) {
+                               if (b.is_odd()) {
+                                       C = C.mul(A);
+                                       b -= 1;
                                }
-                               prod = prod.mul(prod);
+                               b *= _num1_2;  // b /= 2, still integer.
+                               A = A.mul(A);
                        }
-                       return result;
+                       return A.mul(C);
                }
        }
        throw (std::runtime_error("matrix::pow(): don't know how to handle exponent"));
 }
 
 
-/** operator() to access elements.
+/** operator() to access elements for reading.
  *
  *  @param ro row of element
  *  @param co column of element
@@ -670,19 +639,18 @@ const ex & matrix::operator() (unsigned ro, unsigned co) const
 }
 
 
-/** Set individual elements manually.
+/** operator() to access elements for writing.
  *
+ *  @param ro row of element
+ *  @param co column of element
  *  @exception range_error (index out of range) */
-matrix & matrix::set(unsigned ro, unsigned co, ex value)
+ex & matrix::operator() (unsigned ro, unsigned co)
 {
        if (ro>=row || co>=col)
-               throw (std::range_error("matrix::set(): index out of range"));
-       if (value.return_type() != return_types::commutative)
-               throw std::runtime_error("matrix::set(): non-commutative argument");
-    
+               throw (std::range_error("matrix::operator(): index out of range"));
+
        ensure_if_modifiable();
-       m[ro*col+co] = value;
-       return *this;
+       return m[ro*col+co];
 }
 
 
@@ -723,9 +691,10 @@ ex matrix::determinant(unsigned algo) const
        bool numeric_flag = true;
        bool normal_flag = false;
        unsigned sparse_count = 0;  // counts non-zero elements
-       for (exvector::const_iterator r=m.begin(); r!=m.end(); ++r) {
+       exvector::const_iterator r = m.begin(), rend = m.end();
+       while (r != rend) {
                lst srl;  // symbol replacement list
-               ex rtest = (*r).to_rational(srl);
+               ex rtest = r->to_rational(srl);
                if (!rtest.is_zero())
                        ++sparse_count;
                if (!rtest.info(info_flags::numeric))
@@ -733,6 +702,7 @@ ex matrix::determinant(unsigned algo) const
                if (!rtest.info(info_flags::crational_polynomial) &&
                         rtest.info(info_flags::rational_function))
                        normal_flag = true;
+               ++r;
        }
        
        // Here is the heuristics in case this routine has to decide:
@@ -785,7 +755,7 @@ ex matrix::determinant(unsigned algo) const
                        int sign;
                        sign = tmp.division_free_elimination(true);
                        if (sign==0)
-                               return _ex0();
+                               return _ex0;
                        ex det = tmp.m[row*col-1];
                        // factor out accumulated bogus slag
                        for (unsigned d=0; d<row-2; ++d)
@@ -797,10 +767,13 @@ ex matrix::determinant(unsigned algo) const
                default: {
                        // This is the minor expansion scheme.  We always develop such
                        // that the smallest minors (i.e, the trivial 1x1 ones) are on the
-                       // rightmost column.  For this to be efficient it turns out that
-                       // the emptiest columns (i.e. the ones with most zeros) should be
-                       // the ones on the right hand side.  Therefore we presort the
-                       // columns of the matrix:
+                       // rightmost column.  For this to be efficient, empirical tests
+                       // have shown that the emptiest columns (i.e. the ones with most
+                       // zeros) should be the ones on the right hand side -- although
+                       // this might seem counter-intuitive (and in contradiction to some
+                       // literature like the FORM manual).  Please go ahead and test it
+                       // if you don't believe me!  Therefore we presort the columns of
+                       // the matrix:
                        typedef std::pair<unsigned,unsigned> uintpair;
                        std::vector<uintpair> c_zeros;  // number of zeros in column
                        for (unsigned c=0; c<col; ++c) {
@@ -810,15 +783,15 @@ ex matrix::determinant(unsigned algo) const
                                                ++acc;
                                c_zeros.push_back(uintpair(acc,c));
                        }
-                       sort(c_zeros.begin(),c_zeros.end());
+                       std::sort(c_zeros.begin(),c_zeros.end());
                        std::vector<unsigned> pre_sort;
-                       for (std::vector<uintpair>::iterator i=c_zeros.begin(); i!=c_zeros.end(); ++i)
+                       for (std::vector<uintpair>::const_iterator i=c_zeros.begin(); i!=c_zeros.end(); ++i)
                                pre_sort.push_back(i->second);
                        std::vector<unsigned> pre_sort_test(pre_sort); // permutation_sign() modifies the vector so we make a copy here
                        int sign = permutation_sign(pre_sort_test.begin(), pre_sort_test.end());
                        exvector result(row*col);  // represents sorted matrix
                        unsigned c = 0;
-                       for (std::vector<unsigned>::iterator i=pre_sort.begin();
+                       for (std::vector<unsigned>::const_iterator i=pre_sort.begin();
                                 i!=pre_sort.end();
                                 ++i,++c) {
                                for (unsigned r=0; r<row; ++r)
@@ -874,10 +847,11 @@ ex matrix::charpoly(const symbol & lambda) const
                throw (std::logic_error("matrix::charpoly(): matrix not square"));
        
        bool numeric_flag = true;
-       for (exvector::const_iterator r=m.begin(); r!=m.end(); ++r) {
-               if (!(*r).info(info_flags::numeric)) {
+       exvector::const_iterator r = m.begin(), rend = m.end();
+       while (r!=rend && numeric_flag==true) {
+               if (!r->info(info_flags::numeric))
                        numeric_flag = false;
-               }
+               ++r;
        }
        
        // The pure numeric case is traditionally rather common.  Hence, it is
@@ -924,7 +898,7 @@ matrix matrix::inverse(void) const
        // First populate the identity matrix supposed to become the right hand side.
        matrix identity(row,col);
        for (unsigned i=0; i<row; ++i)
-               identity.set(i,i,_ex1());
+               identity(i,i) = _ex1;
        
        // Populate a dummy matrix of variables, just because of compatibility with
        // matrix::solve() which wants this (for compatibility with under-determined
@@ -932,7 +906,7 @@ matrix matrix::inverse(void) const
        matrix vars(row,col);
        for (unsigned r=0; r<row; ++r)
                for (unsigned c=0; c<col; ++c)
-                       vars.set(r,c,symbol());
+                       vars(r,c) = symbol();
        
        matrix sol(row,col);
        try {
@@ -984,9 +958,11 @@ matrix matrix::solve(const matrix & vars,
        
        // Gather some statistical information about the augmented matrix:
        bool numeric_flag = true;
-       for (exvector::const_iterator r=aug.m.begin(); r!=aug.m.end(); ++r) {
-               if (!(*r).info(info_flags::numeric))
+       exvector::const_iterator r = aug.m.begin(), rend = aug.m.end();
+       while (r!=rend && numeric_flag==true) {
+               if (!r->info(info_flags::numeric))
                        numeric_flag = false;
+               ++r;
        }
        
        // Here is the heuristics in case this routine has to decide:
@@ -1032,19 +1008,18 @@ matrix matrix::solve(const matrix & vars,
                                // assign solutions for vars between fnz+1 and
                                // last_assigned_sol-1: free parameters
                                for (unsigned c=fnz; c<last_assigned_sol-1; ++c)
-                                       sol.set(c,co,vars.m[c*p+co]);
+                                       sol(c,co) = vars.m[c*p+co];
                                ex e = aug.m[r*(n+p)+n+co];
                                for (unsigned c=fnz; c<n; ++c)
                                        e -= aug.m[r*(n+p)+c]*sol.m[c*p+co];
-                               sol.set(fnz-1,co,
-                                               (e/(aug.m[r*(n+p)+(fnz-1)])).normal());
+                               sol(fnz-1,co) = (e/(aug.m[r*(n+p)+(fnz-1)])).normal();
                                last_assigned_sol = fnz;
                        }
                }
                // assign solutions for vars between 1 and
                // last_assigned_sol-1: free parameters
                for (unsigned ro=0; ro<last_assigned_sol-1; ++ro)
-                       sol.set(ro,co,vars(ro,co));
+                       sol(ro,co) = vars(ro,co);
        }
        
        return sol;
@@ -1088,9 +1063,9 @@ ex matrix::determinant_minor(void) const
        //     for (unsigned r=0; r<minorM.rows(); ++r) {
        //         for (unsigned c=0; c<minorM.cols(); ++c) {
        //             if (r<r1)
-       //                 minorM.set(r,c,m[r*col+c+1]);
+       //                 minorM(r,c) = m[r*col+c+1];
        //             else
-       //                 minorM.set(r,c,m[(r+1)*col+c+1]);
+       //                 minorM(r,c) = m[(r+1)*col+c+1];
        //         }
        //     }
        //     // recurse down and care for sign:
@@ -1134,7 +1109,7 @@ ex matrix::determinant_minor(void) const
                        Pkey.push_back(i);
                unsigned fc = 0;  // controls logic for our strange flipper counter
                do {
-                       det = _ex0();
+                       det = _ex0;
                        for (unsigned r=0; r<n-c; ++r) {
                                // maybe there is nothing to do?
                                if (m[Pkey[r]*n+c].is_zero())
@@ -1214,12 +1189,12 @@ int matrix::gauss_elimination(const bool det)
                                }
                                // fill up left hand side with zeros
                                for (unsigned c=0; c<=r1; ++c)
-                                       this->m[r2*n+c] = _ex0();
+                                       this->m[r2*n+c] = _ex0;
                        }
                        if (det) {
                                // save space by deleting no longer needed elements
                                for (unsigned c=r0+1; c<n; ++c)
-                                       this->m[r0*n+c] = _ex0();
+                                       this->m[r0*n+c] = _ex0;
                        }
                        ++r0;
                }
@@ -1261,12 +1236,12 @@ int matrix::division_free_elimination(const bool det)
                                        this->m[r2*n+c] = (this->m[r0*n+r1]*this->m[r2*n+c] - this->m[r2*n+r1]*this->m[r0*n+c]).expand();
                                // fill up left hand side with zeros
                                for (unsigned c=0; c<=r1; ++c)
-                                       this->m[r2*n+c] = _ex0();
+                                       this->m[r2*n+c] = _ex0;
                        }
                        if (det) {
                                // save space by deleting no longer needed elements
                                for (unsigned c=r0+1; c<n; ++c)
-                                       this->m[r0*n+c] = _ex0();
+                                       this->m[r0*n+c] = _ex0;
                        }
                        ++r0;
                }
@@ -1334,13 +1309,13 @@ int matrix::fraction_free_elimination(const bool det)
        matrix tmp_n(*this);
        matrix tmp_d(m,n);  // for denominators, if needed
        lst srl;  // symbol replacement list
-       exvector::iterator it = this->m.begin();
-       exvector::iterator tmp_n_it = tmp_n.m.begin();
-       exvector::iterator tmp_d_it = tmp_d.m.begin();
-       for (; it!= this->m.end(); ++it, ++tmp_n_it, ++tmp_d_it) {
-               (*tmp_n_it) = (*it).normal().to_rational(srl);
-               (*tmp_d_it) = (*tmp_n_it).denom();
-               (*tmp_n_it) = (*tmp_n_it).numer();
+       exvector::const_iterator cit = this->m.begin(), citend = this->m.end();
+       exvector::iterator tmp_n_it = tmp_n.m.begin(), tmp_d_it = tmp_d.m.begin();
+       while (cit != citend) {
+               ex nd = cit->normal().to_rational(srl).numer_denom();
+               ++cit;
+               *tmp_n_it++ = nd.op(0);
+               *tmp_d_it++ = nd.op(1);
        }
        
        unsigned r0 = 0;
@@ -1374,7 +1349,7 @@ int matrix::fraction_free_elimination(const bool det)
                                }
                                // fill up left hand side with zeros
                                for (unsigned c=0; c<=r1; ++c)
-                                       tmp_n.m[r2*n+c] = _ex0();
+                                       tmp_n.m[r2*n+c] = _ex0;
                        }
                        if ((r1<n-1)&&(r0<m-1)) {
                                // compute next iteration's divisor
@@ -1383,8 +1358,8 @@ int matrix::fraction_free_elimination(const bool det)
                                if (det) {
                                        // save space by deleting no longer needed elements
                                        for (unsigned c=0; c<n; ++c) {
-                                               tmp_n.m[r0*n+c] = _ex0();
-                                               tmp_d.m[r0*n+c] = _ex1();
+                                               tmp_n.m[r0*n+c] = _ex0;
+                                               tmp_d.m[r0*n+c] = _ex1;
                                        }
                                }
                        }
@@ -1392,11 +1367,11 @@ int matrix::fraction_free_elimination(const bool det)
                }
        }
        // repopulate *this matrix:
-       it = this->m.begin();
+       exvector::iterator it = this->m.begin(), itend = this->m.end();
        tmp_n_it = tmp_n.m.begin();
        tmp_d_it = tmp_d.m.begin();
-       for (; it!= this->m.end(); ++it, ++tmp_n_it, ++tmp_d_it)
-               (*it) = ((*tmp_n_it)/(*tmp_d_it)).subs(srl);
+       while (it != itend)
+               *it++ = ((*tmp_n_it++)/(*tmp_d_it++)).subs(srl);
        
        return sign;
 }
@@ -1424,12 +1399,12 @@ int matrix::pivot(unsigned ro, unsigned co, bool symbolic)
                        ++k;
        } else {
                // search largest element in column co beginning at row ro
-               GINAC_ASSERT(is_ex_of_type(this->m[k*col+co],numeric));
+               GINAC_ASSERT(is_a<numeric>(this->m[k*col+co]));
                unsigned kmax = k+1;
-               numeric mmax = abs(ex_to_numeric(m[kmax*col+co]));
+               numeric mmax = abs(ex_to<numeric>(m[kmax*col+co]));
                while (kmax<row) {
-                       GINAC_ASSERT(is_ex_of_type(this->m[kmax*col+co],numeric));
-                       numeric tmp = ex_to_numeric(this->m[kmax*col+co]);
+                       GINAC_ASSERT(is_a<numeric>(this->m[kmax*col+co]));
+                       numeric tmp = ex_to<numeric>(this->m[kmax*col+co]);
                        if (abs(tmp) > mmax) {
                                mmax = tmp;
                                k = kmax;
@@ -1467,9 +1442,9 @@ ex lst_to_matrix(const lst & l)
        for (i=0; i<rows; i++)
                for (j=0; j<cols; j++)
                        if (l.op(i).nops() > j)
-                               m.set(i, j, l.op(i).op(j));
+                               m(i, j) = l.op(i).op(j);
                        else
-                               m.set(i, j, ex(0));
+                               m(i, j) = _ex0;
        return m;
 }
 
@@ -1480,7 +1455,7 @@ ex diag_matrix(const lst & l)
        matrix &m = *new matrix(dim, dim);
        m.setflag(status_flags::dynallocated);
        for (unsigned i=0; i<dim; i++)
-               m.set(i, i, l.op(i));
+               m(i, i) = l.op(i);
 
        return m;
 }