changeset 8303:b11c31849b44

improve norm computation capabilities
author Jaroslav Hajek <highegg@gmail.com>
date Fri, 31 Oct 2008 08:05:32 +0100
parents f2e050b62199
children eeaee297c0da
files liboctave/CSparse.cc liboctave/CSparse.h liboctave/ChangeLog liboctave/MArray-C.cc liboctave/MArray-d.cc liboctave/MArray-defs.h liboctave/MArray-f.cc liboctave/MArray-fC.cc liboctave/Makefile.in liboctave/dSparse.cc liboctave/dSparse.h liboctave/oct-norm.cc liboctave/oct-norm.h scripts/ChangeLog scripts/linear-algebra/Makefile.in scripts/linear-algebra/__norm__.m src/ChangeLog src/Makefile.in src/data.cc src/xnorm.cc src/xnorm.h
diffstat 21 files changed, 1054 insertions(+), 369 deletions(-) [+]
line wrap: on
line diff
--- a/liboctave/CSparse.cc	Sun Nov 02 06:44:12 2008 +0100
+++ b/liboctave/CSparse.cc	Fri Oct 31 08:05:32 2008 +0100
@@ -520,6 +520,37 @@
   return result;
 }
 
+ComplexRowVector 
+SparseComplexMatrix::row (octave_idx_type i) const
+{
+  octave_idx_type nc = columns ();
+  ComplexRowVector retval (nc, 0);
+
+  for (octave_idx_type j = 0; j < nc; j++)
+    for (octave_idx_type k = cidx (j); k < cidx (j+1); k++)
+      {
+        if (ridx (k) == i)
+          {
+            retval(j) = data (k);
+            break;
+          }
+      }
+
+  return retval;
+}
+
+ComplexColumnVector 
+SparseComplexMatrix::column (octave_idx_type i) const
+{
+  octave_idx_type nr = rows ();
+  ComplexColumnVector retval (nr);
+
+  for (octave_idx_type k = cidx (i); k < cidx (i+1); k++)
+    retval(ridx (k)) = data (k);
+
+  return retval;
+}
+
 // destructive insert/delete/reorder operations
 
 SparseComplexMatrix&
--- a/liboctave/CSparse.h	Sun Nov 02 06:44:12 2008 +0100
+++ b/liboctave/CSparse.h	Fri Oct 31 08:05:32 2008 +0100
@@ -127,6 +127,12 @@
 
   friend SparseComplexMatrix conj (const SparseComplexMatrix& a);
 
+  // extract row or column i.
+
+  ComplexRowVector row (octave_idx_type i) const;
+
+  ComplexColumnVector column (octave_idx_type i) const;
+
 private:
   SparseComplexMatrix dinverse (MatrixType &mattyp, octave_idx_type& info, 
 				double& rcond, const bool force = false, 
--- a/liboctave/ChangeLog	Sun Nov 02 06:44:12 2008 +0100
+++ b/liboctave/ChangeLog	Fri Oct 31 08:05:32 2008 +0100
@@ -1,3 +1,20 @@
+2008-10-31  Jaroslav Hajek  <highegg@gmail.com>
+
+	* oct-norm.h: New header file.
+	* oct-norm.cc: New source.
+	* CSparse.cc (SparseComplexMatrix::row, SparseComplexMatrix::column):
+	New member functions.
+	* CSparse.h (SparseComplexMatrix): Declare them.
+	* dSparse.cc (SparseMatrix::row, SparseMatrix::column):
+	New member functions.
+	* dSparse.h (SparseMatrix): Declare them.
+	* MArray-C.cc (MArray<Complex>::norm), 
+	MArray-d.cc (MArray<double>::norm), 
+	MArray-fC.cc (MArray<FloatComplex>::norm), 
+	MArray-f.cc (MArray<float>::norm): Wrap a call to xnorm. 
+
+	* MArray-defs.h (MARRAY_NORM_BODY): Remove.
+
 2008-11-02  Jaroslav Hajek <highegg@gmail.com>
 
 	* idx-vector.cc (idx_vector::is_complement): Set resulting extent
--- a/liboctave/MArray-C.cc	Sun Nov 02 06:44:12 2008 +0100
+++ b/liboctave/MArray-C.cc	Fri Oct 31 08:05:32 2008 +0100
@@ -28,23 +28,17 @@
 // Instantiate MArrays of Complex values.
 
 #include "oct-cmplx.h"
-#include "f77-fcn.h"
-
-extern "C"
-{
-  F77_RET_T
-  F77_FUNC (xdznrm2, XDZNRM2) (const octave_idx_type&, const Complex*,
-			       const octave_idx_type&, double&);
-}
 
 #include "MArray.h"
 #include "MArray.cc"
+#include "CColVector.h"
+#include "oct-norm.h"
 
 template <>
 OCTAVE_API double
 MArray<Complex>::norm (double p) const
 {
-  MARRAY_NORM_BODY (Complex, double, xdznrm2, XDZNRM2, octave_NaN);
+  return xnorm (ComplexColumnVector (*this), p);
 }
 
 template class OCTAVE_API MArray<Complex>;
--- a/liboctave/MArray-d.cc	Sun Nov 02 06:44:12 2008 +0100
+++ b/liboctave/MArray-d.cc	Fri Oct 31 08:05:32 2008 +0100
@@ -26,23 +26,16 @@
 
 // Instantiate MArrays of double values.
 
-#include "f77-fcn.h"
-
-extern "C"
-{
-  F77_RET_T
-  F77_FUNC (xdnrm2, XDNRM2) (const octave_idx_type&, const double*,
-			     const octave_idx_type&, double&);
-}
-
 #include "MArray.h"
 #include "MArray.cc"
+#include "dColVector.h"
+#include "oct-norm.h"
 
 template <>
 OCTAVE_API double
 MArray<double>::norm (double p) const
 {
-  MARRAY_NORM_BODY (double, double, xdnrm2, XDNRM2, octave_NaN);
+  return xnorm (ColumnVector (*this), p);
 }
 
 template class OCTAVE_API MArray<double>;
--- a/liboctave/MArray-defs.h	Sun Nov 02 06:44:12 2008 +0100
+++ b/liboctave/MArray-defs.h	Fri Oct 31 08:05:32 2008 +0100
@@ -343,90 +343,6 @@
   MDIAGARRAY2_DADA_BINOP_FWD_DEFS \
     (R, T, dynamic_cast<const B<T>&>, R, dynamic_cast<const B<T>&>, R)
 
-#define MARRAY_NORM_BODY(TYPE, RTYPE, blas_norm, BLAS_NORM, NAN_VALUE)	\
- \
-  RTYPE retval = NAN_VALUE; \
- \
-  octave_idx_type len = length (); \
- \
-  if (len > 0) \
-    { \
-      const TYPE *d = data (); \
- \
-      if (p == -1) \
-	{ \
-	  /* Frobenius norm.  */ \
-	  retval = 0; \
- \
-          /* precondition */ \
-          RTYPE inf_norm = 0.; \
-	  for (octave_idx_type i = 0; i < len; i++) \
-	    { \
-              RTYPE d_abs = std::abs (d[i]); \
-              if (d_abs > inf_norm) \
-                inf_norm = d_abs; \
-            } \
-          inf_norm = (inf_norm == octave_Inf || inf_norm == 0. ? 1.0 : \
-		      inf_norm); \
-          RTYPE scale = 1. / inf_norm; \
-\
-	  for (octave_idx_type i = 0; i < len; i++) \
-	    { \
-	      RTYPE d_abs = std::abs (d[i]) * scale; \
-	      retval += d_abs * d_abs; \
-	    } \
- \
-	  retval = ::sqrt (retval) * inf_norm; \
-	} \
-      else if (p == 2) \
-	F77_FCN (blas_norm, BLAS_NORM) (len, d, 1, retval); \
-      else if (xisinf (p)) \
-	{ \
-	  octave_idx_type i = 0; \
- \
-	  while (i < len && xisnan (d[i])) \
-	    i++; \
- \
-	  if (i < len) \
-	    retval = std::abs (d[i]); \
- \
-	  if (p > 0) \
-	    { \
-	      while (i < len) \
-		{ \
-		  RTYPE d_abs = std::abs (d[i++]); \
- \
-		  if (d_abs > retval) \
-		    retval = d_abs; \
-		} \
-	    } \
-	  else \
-	    { \
-	      while (i < len) \
-		{ \
-		  RTYPE d_abs = std::abs (d[i++]); \
- \
-		  if (d_abs < retval) \
-		    retval = d_abs; \
-		} \
-	    } \
-	} \
-      else \
-	{ \
-	  retval = 0; \
- \
-	  for (octave_idx_type i = 0; i < len; i++) \
-	    { \
-	      RTYPE d_abs = std::abs (d[i]); \
-	      retval += pow (d_abs, p); \
-	    } \
- \
-	  retval = pow (retval, 1/p); \
-	} \
-    } \
- \
-  return retval
-
 // Now we have all the definitions we need.
 
 #endif
--- a/liboctave/MArray-f.cc	Sun Nov 02 06:44:12 2008 +0100
+++ b/liboctave/MArray-f.cc	Fri Oct 31 08:05:32 2008 +0100
@@ -26,23 +26,16 @@
 
 // Instantiate MArrays of float values.
 
-#include "f77-fcn.h"
-
-extern "C"
-{
-  F77_RET_T
-  F77_FUNC (xsnrm2, XSNRM2) (const octave_idx_type&, const float*,
-			     const octave_idx_type&, float&);
-}
-
 #include "MArray.h"
 #include "MArray.cc"
+#include "fColVector.h"
+#include "oct-norm.h"
 
 template <>
 OCTAVE_API float
 MArray<float>::norm (float p) const
 {
-  MARRAY_NORM_BODY (float, float, xsnrm2, XSNRM2, octave_Float_NaN);
+  return xnorm (FloatColumnVector (*this));
 }
 
 template class OCTAVE_API MArray<float>;
--- a/liboctave/MArray-fC.cc	Sun Nov 02 06:44:12 2008 +0100
+++ b/liboctave/MArray-fC.cc	Fri Oct 31 08:05:32 2008 +0100
@@ -28,23 +28,17 @@
 // Instantiate MArrays of FloatComplex values.
 
 #include "oct-cmplx.h"
-#include "f77-fcn.h"
-
-extern "C"
-{
-  F77_RET_T
-  F77_FUNC (xscnrm2, XSCNRM2) (const octave_idx_type&, const FloatComplex*,
-			       const octave_idx_type&, float&);
-}
 
 #include "MArray.h"
 #include "MArray.cc"
+#include "fCColVector.h"
+#include "oct-norm.h"
 
 template <>
 OCTAVE_API float
 MArray<FloatComplex>::norm (float p) const
 {
-  MARRAY_NORM_BODY (FloatComplex, float, xscnrm2, XSCNRM2, octave_Float_NaN);
+  return xnorm (FloatComplexColumnVector (*this));
 }
 
 template class OCTAVE_API MArray<FloatComplex>;
--- a/liboctave/Makefile.in	Sun Nov 02 06:44:12 2008 +0100
+++ b/liboctave/Makefile.in	Fri Oct 31 08:05:32 2008 +0100
@@ -88,9 +88,9 @@
 	lo-ieee.h lo-mappers.h lo-math.h lo-specfun.h \
 	lo-sysdep.h lo-utils.h mach-info.h md5.h oct-alloc.h oct-cmplx.h \
 	oct-env.h oct-fftw.h oct-getopt.h oct-group.h oct-inttypes.h \
-	oct-lookup.h oct-md5.h oct-mutex.h oct-passwd.h oct-rand.h oct-rl-edit.h \
-	oct-rl-hist.h oct-shlib.h oct-sort.h oct-spparms.h oct-syscalls.h \
-	oct-sparse.h oct-time.h oct-uname.h \
+	oct-lookup.h oct-md5.h oct-mutex.h oct-norm.h oct-passwd.h \
+	oct-rand.h oct-rl-edit.h oct-rl-hist.h oct-shlib.h oct-sort.h \
+	oct-spparms.h oct-syscalls.h oct-sparse.h oct-time.h oct-uname.h \
 	pathlen.h pathsearch.h prog-args.h \
 	randgamma.h randmtzig.h randpoisson.h regex-match.h \
 	sparse-sort.h statdefs.h str-vec.h \
@@ -148,7 +148,8 @@
 	file-ops.cc file-stat.cc glob-match.cc idx-vector.cc \
 	lo-ieee.cc lo-mappers.cc lo-specfun.cc lo-sysdep.cc \
 	lo-utils.cc mach-info.cc oct-alloc.cc oct-env.cc \
-	oct-fftw.cc oct-group.cc oct-mutex.cc oct-md5.cc oct-passwd.cc oct-rand.cc \
+	oct-fftw.cc oct-group.cc oct-mutex.cc oct-md5.cc \
+	oct-norm.cc oct-passwd.cc oct-rand.cc \
 	oct-shlib.cc oct-spparms.cc oct-syscalls.cc oct-time.cc oct-uname.cc \
 	prog-args.cc regex-match.cc \
 	sparse-sort.cc sparse-util.cc str-vec.cc \
--- a/liboctave/dSparse.cc	Sun Nov 02 06:44:12 2008 +0100
+++ b/liboctave/dSparse.cc	Fri Oct 31 08:05:32 2008 +0100
@@ -515,6 +515,37 @@
   return result;
 }
 
+RowVector 
+SparseMatrix::row (octave_idx_type i) const
+{
+  octave_idx_type nc = columns ();
+  RowVector retval (nc, 0);
+
+  for (octave_idx_type j = 0; j < nc; j++)
+    for (octave_idx_type k = cidx (j); k < cidx (j+1); k++)
+      {
+        if (ridx (k) == i)
+          {
+            retval(j) = data (k);
+            break;
+          }
+      }
+
+  return retval;
+}
+
+ColumnVector 
+SparseMatrix::column (octave_idx_type i) const
+{
+  octave_idx_type nr = rows ();
+  ColumnVector retval (nr);
+
+  for (octave_idx_type k = cidx (i); k < cidx (i+1); k++)
+    retval(ridx (k)) = data (k);
+
+  return retval;
+}
+
 SparseMatrix
 SparseMatrix::concat (const SparseMatrix& rb, const Array<octave_idx_type>& ra_idx)
 {
--- a/liboctave/dSparse.h	Sun Nov 02 06:44:12 2008 +0100
+++ b/liboctave/dSparse.h	Fri Oct 31 08:05:32 2008 +0100
@@ -121,6 +121,12 @@
     }
   SparseMatrix hermitian (void) const { return transpose (); }
 
+  // extract row or column i.
+
+  RowVector row (octave_idx_type i) const;
+
+  ColumnVector column (octave_idx_type i) const;
+
 private:
   SparseMatrix dinverse (MatrixType &mattyp, octave_idx_type& info, 
 			 double& rcond, const bool force = false, 
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/liboctave/oct-norm.cc	Fri Oct 31 08:05:32 2008 +0100
@@ -0,0 +1,562 @@
+/*
+
+Copyright (C) 2008 VZLU Prague, a.s.
+
+This file is part of Octave.
+
+Octave is free software; you can redistribute it and/or modify it
+under the terms of the GNU General Public License as published by the
+Free Software Foundation; either version 3 of the License, or (at your
+option) any later version.
+
+Octave is distributed in the hope that it will be useful, but WITHOUT
+ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+for more details.
+
+You should have received a copy of the GNU General Public License
+along with Octave; see the file COPYING.  If not, see
+<http://www.gnu.org/licenses/>.
+
+*/
+
+// author: Jaroslav Hajek <highegg@gmail.com>
+
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include <cassert>
+#include <cfloat>
+#include <cmath>
+
+#include <iostream>
+
+#include "oct-cmplx.h"
+#include "lo-error.h"
+#include "lo-ieee.h"
+#include "Array.h"
+#include "Array.cc"
+#include "Array-util.h"
+#include "CMatrix.h"
+#include "dMatrix.h"
+#include "fCMatrix.h"
+#include "fMatrix.h"
+#include "CColVector.h"
+#include "dColVector.h"
+#include "CRowVector.h"
+#include "dRowVector.h"
+#include "fCColVector.h"
+#include "fColVector.h"
+#include "fCRowVector.h"
+#include "fRowVector.h"
+#include "CSparse.h"
+#include "dSparse.h"
+#include "dbleSVD.h"
+#include "CmplxSVD.h"
+#include "floatSVD.h"
+#include "fCmplxSVD.h"
+
+// Theory: norm accumulator is an object that has an accum method able to handle
+// both real and complex element, and a cast operator returning the intermediate
+// norm.
+
+// norm accumulator for the p-norm
+template <class R>
+class norm_accumulator_p
+{
+  R p,scl,sum;
+public:
+  norm_accumulator_p () {} // we need this one for Array
+  norm_accumulator_p (R pp) : p(pp), scl(0), sum(1) {}
+
+  template<class U> 
+  void accum (U val)
+    {
+      R t = std::abs (val);
+      if (scl == t) // we need this to handle Infs properly
+        sum += 1;
+      else if (scl < t)
+        {
+          sum *= std::pow (scl/t, p);
+          sum += 1;
+          scl = t;
+        }
+      else if (t != 0)
+        sum += std::pow (t/scl, p);
+    }
+  operator R () { return scl * std::pow (sum, 1/p); }
+};
+
+// norm accumulator for the minus p-pseudonorm
+template <class R>
+class norm_accumulator_mp
+{
+  R p,scl,sum;
+public:
+  norm_accumulator_mp () {} // we need this one for Array
+  norm_accumulator_mp (R pp) : p(pp), scl(0), sum(1) {}
+
+  template<class U> 
+  void accum (U val)
+    {
+      R t = 1 / std::abs (val);
+      if (scl == t)
+        sum += 1;
+      else if (scl < t)
+        {
+          sum *= std::pow (scl/t, p);
+          sum += 1;
+          scl = t;
+        }
+      else if (t != 0)
+        sum += std::pow (t/scl, p);
+    }
+  operator R () { return scl * std::pow (sum, -1/p); }
+};
+
+// norm accumulator for the 2-norm (euclidean)
+template <class R>
+class norm_accumulator_2
+{
+  R scl,sum;
+  static R pow2 (R x) { return x*x; }
+public:
+  norm_accumulator_2 () : scl(0), sum(1) {}
+
+  void accum (R val)
+    {
+      R t = std::abs (val);
+      if (scl == t)
+        sum += 1;
+      else if (scl < t)
+        {
+          sum *= pow2 (scl/t);
+          sum += 1;
+          scl = t;
+        }
+      else if (t != 0)
+        sum += pow2 (t/scl);
+    }
+
+  void accum (std::complex<R> val)
+    {
+      accum (val.real ());
+      accum (val.imag ());
+    }
+
+  operator R () { return scl * std::sqrt (sum); }
+};
+
+// norm accumulator for the 1-norm (city metric)
+template <class R>
+class norm_accumulator_1
+{
+  R sum;
+public:
+  norm_accumulator_1 () : sum (0) {}
+  template<class U> 
+  void accum (U val)
+    {
+      sum += std::abs (val); 
+    }
+  operator R () { return sum; }
+};
+
+// norm accumulator for the inf-norm (max metric)
+template <class R>
+class norm_accumulator_inf
+{
+  R max;
+public:
+  norm_accumulator_inf () : max (0) {}
+  template<class U> 
+  void accum (U val)
+    {
+      max = std::max (max, std::abs (val));
+    }
+  operator R () { return max; }
+};
+
+// norm accumulator for the -inf pseudonorm (min abs value)
+template <class R>
+class norm_accumulator_minf
+{
+  R min;
+public:
+  norm_accumulator_minf () : min (octave_Inf) {}
+  template<class U> 
+  void accum (U val)
+    {
+      min = std::min (min, std::abs (val));
+    }
+  operator R () { return min; }
+};
+
+// norm accumulator for the 0-pseudonorm (hamming distance)
+template <class R>
+class norm_accumulator_0
+{
+  unsigned int num;
+public:
+  norm_accumulator_0 () : num (0) {}
+  template<class U> 
+  void accum (U val)
+    {
+      if (val != static_cast<U> (0)) ++num;
+    }
+  operator R () { return num; }
+};
+
+
+// OK, we're armed :) Now let's go for the fun
+
+template <class T, class R, class ACC>
+inline void vector_norm (const Array<T>& v, R& res, ACC acc)
+{
+  for (octave_idx_type i = 0; i < v.numel (); i++)
+    acc.accum (v(i));
+
+  res = acc;
+}
+
+// dense versions
+template <class T, class R, class ACC>
+void column_norms (const MArray2<T>& m, MArray<R>& res, ACC acc)
+{
+  res.resize (m.columns ());
+  for (octave_idx_type j = 0; j < m.columns (); j++)
+    {
+      ACC accj = acc;
+      for (octave_idx_type i = 0; i < m.rows (); i++)
+        accj.accum (m(i, j));
+
+      res.xelem (j) = accj;
+    }
+}
+
+template <class T, class R, class ACC>
+void row_norms (const MArray2<T>& m, MArray<R>& res, ACC acc)
+{
+  res.resize (m.rows ());
+  Array<ACC> acci (m.rows (), acc); 
+  for (octave_idx_type j = 0; j < m.columns (); j++)
+    {
+      for (octave_idx_type i = 0; i < m.rows (); i++)
+        acci.xelem (i).accum (m(i, j));
+    }
+
+  for (octave_idx_type i = 0; i < m.rows (); i++)
+    res.xelem (i) = acci(i);
+}
+
+// sparse versions
+template <class T, class R, class ACC>
+void column_norms (const MSparse<T>& m, MArray<R>& res, ACC acc)
+{
+  res.resize (m.columns ());
+  for (octave_idx_type j = 0; j < m.columns (); j++)
+    {
+      ACC accj = acc;
+      for (octave_idx_type k = m.cidx (j); k < m.cidx (j+1); k++)
+        accj.accum (m.data (k));
+
+      res.xelem (j) = accj;
+    }
+}
+
+template <class T, class R, class ACC>
+void row_norms (const MSparse<T>& m, MArray<R>& res, ACC acc)
+{
+  res.resize (m.rows ());
+  Array<ACC> acci (m.rows (), acc); 
+  for (octave_idx_type j = 0; j < m.columns (); j++)
+    {
+      for (octave_idx_type k = m.cidx (j); k < m.cidx (j+1); k++)
+        acci(m.ridx (k)).accum (m.data (k));
+    }
+
+  for (octave_idx_type i = 0; i < m.rows (); i++)
+    res.xelem (i) = acci(i);
+}
+
+// now the dispatchers
+#define DEFINE_DISPATCHER(FUNC_NAME, ARG_TYPE, RES_TYPE) \
+template <class T, class R> \
+RES_TYPE FUNC_NAME (const ARG_TYPE& v, R p) \
+{ \
+  RES_TYPE res; \
+  if (p == 2) \
+    FUNC_NAME (v, res, norm_accumulator_2<R> ()); \
+  else if (p == 1) \
+    FUNC_NAME (v, res, norm_accumulator_1<R> ()); \
+  else if (lo_ieee_isinf (p)) \
+    { \
+      if (p > 0) \
+        FUNC_NAME (v, res, norm_accumulator_inf<R> ()); \
+      else \
+        FUNC_NAME (v, res, norm_accumulator_minf<R> ()); \
+    } \
+  else if (p == 0) \
+    FUNC_NAME (v, res, norm_accumulator_0<R> ()); \
+  else if (p > 0) \
+    FUNC_NAME (v, res, norm_accumulator_p<R> (p)); \
+  else \
+    FUNC_NAME (v, res, norm_accumulator_mp<R> (p)); \
+  return res; \
+}
+
+DEFINE_DISPATCHER (vector_norm, MArray<T>, R)
+DEFINE_DISPATCHER (vector_norm, MArray2<T>, R)
+DEFINE_DISPATCHER (column_norms, MArray2<T>, MArray<R>)
+DEFINE_DISPATCHER (row_norms, MArray2<T>, MArray<R>)
+DEFINE_DISPATCHER (column_norms, MSparse<T>, MArray<R>)
+DEFINE_DISPATCHER (row_norms, MSparse<T>, MArray<R>)
+
+// The approximate subproblem in Higham's method. Find lambda and mu such that
+// norm ([lambda, mu], p) == 1 and norm (y*lambda + col*mu, p) is maximized.
+// Real version. As in Higham's paper.
+template <class ColVectorT, class R>
+static void 
+higham_subp (const ColVectorT& y, const ColVectorT& col, 
+             octave_idx_type nsamp, R p, R& lambda, R& mu)
+{
+  R nrm = 0;
+  for (octave_idx_type i = 0; i < nsamp; i++)
+    {
+      R fi = i*M_PI/nsamp, lambda1 = cos (fi), mu1 = sin (fi);
+      R lmnr = std::pow (std::pow (std::abs (lambda1), p) + 
+                         std::pow (std::abs (mu1), p), 1/p);
+      lambda1 /= lmnr; mu1 /= lmnr;
+      R nrm1 = vector_norm (lambda1 * y + mu1 * col, p);
+      if (nrm1 > nrm)
+        {
+          lambda = lambda1;
+          mu = mu1;
+          nrm = nrm1;
+        }
+    }
+}
+
+// Complex version. Higham's paper does not deal with complex case, so we use a simple
+// extension. First, guess the magnitudes as in real version, then try to rotate lambda
+// to improve further.
+template <class ColVectorT, class R>
+static void 
+higham_subp (const ColVectorT& y, const ColVectorT& col, 
+             octave_idx_type nsamp, R p, 
+             std::complex<R>& lambda, std::complex<R>& mu)
+{
+  typedef std::complex<R> CR;
+  R nrm = 0;
+  lambda = 1.0;
+  CR lamcu = lambda / std::abs (lambda);
+  // Probe magnitudes
+  for (octave_idx_type i = 0; i < nsamp; i++)
+    {
+      R fi = i*M_PI/nsamp, lambda1 = cos (fi), mu1 = sin (fi);
+      R lmnr = std::pow (std::pow (std::abs (lambda1), p) + 
+                         std::pow (std::abs (mu1), p), 1/p);
+      lambda1 /= lmnr; mu1 /= lmnr;
+      R nrm1 = vector_norm (lambda1 * lamcu * y + mu1 * col, p);
+      if (nrm1 > nrm)
+        {
+          lambda = lambda1 * lamcu;
+          mu = mu1;
+          nrm = nrm1;
+        }
+    }
+  R lama = std::abs (lambda);
+  // Probe orientation
+  for (octave_idx_type i = 0; i < nsamp; i++)
+    {
+      R fi = i*M_PI/nsamp;
+      lamcu = CR (cos (fi), sin (fi));
+      R nrm1 = vector_norm (lama * lamcu * y + mu * col, p);
+      if (nrm1 > nrm)
+        {
+          lambda = lama * lamcu;
+          nrm = nrm1;
+        }
+    }
+}
+
+// the p-dual element (should work for both real and complex)
+template <class T, class R>
+inline T elem_dual_p (T x, R p)
+{
+  return signum (x) * std::pow (std::abs (x), p-1);
+}
+
+// the VectorT is used for vectors, but actually it has to be
+// a Matrix type to allow all the operations. For instance SparseMatrix
+// does not support multiplication with column/row vectors.
+// the dual vector
+template <class VectorT, class R>
+VectorT dual_p (const VectorT& x, R p, R q)
+{
+  VectorT res (x.dims ());
+  for (octave_idx_type i = 0; i < x.numel (); i++)
+    res.xelem (i) = elem_dual_p (x(i), p);
+  return res / vector_norm (res, q);
+}
+
+// Higham's hybrid method
+template <class MatrixT, class VectorT, class R>
+R higham (const MatrixT& m, R p, R tol, int maxiter,
+          VectorT& x)
+{
+  x.resize (m.columns (), 1);
+  // the OSE part
+  VectorT y(m.rows (), 1, 0), z(m.rows (), 1);
+  typedef typename VectorT::element_type RR;
+  RR lambda = 0, mu = 0;
+  for (octave_idx_type k = 0; k < m.columns (); k++)
+    {
+      VectorT col (m.column (k));
+      if (k > 0)
+        higham_subp (y, col, 4*k, p, lambda, mu);
+      for (octave_idx_type i = 0; i < k; i++)
+        x(i) *= lambda;
+      x(k) = mu;
+      y = lambda * y + mu * col;
+    }
+  
+  // the PM part
+  x = x / vector_norm (x, p);
+  R q = p/(p-1);
+
+  R gamma = 0, gamma1;
+  int iter = 0;
+  while (iter < maxiter)
+    {
+      y = m*x;
+      gamma1 = gamma;
+      gamma = vector_norm (y, p);
+      z = dual_p (y, p, q);
+      z = z.hermitian ();
+      z = z * m;
+
+      if (iter > 0 && (vector_norm (z, q) <= gamma
+                       || (gamma - gamma1) <= tol*gamma))
+        break;
+
+      z = z.hermitian ();
+      x = dual_p (z, q, p);
+      iter ++;
+    }
+
+  return gamma;
+}
+
+// TODO: is there a better way?
+inline float get_eps (float) { return FLT_EPSILON; }
+inline double get_eps (double) { return DBL_EPSILON; }
+// derive column vector and SVD types 
+
+static const char *p_less1_gripe = "xnorm: p must be at least 1";
+
+// Static constant to control the maximum number of iterations. 100 seems to be a good value.
+// Eventually, we can provide a means to change this constant from Octave.
+static int max_norm_iter = 100;
+
+// version with SVD for dense matrices
+template <class MatrixT, class VectorT, class SVDT, class R>
+R matrix_norm (const MatrixT& m, R p, VectorT, SVDT)
+{
+  R res = 0;
+  if (p == 2)
+    {
+      SVDT svd (m, SVD::sigma_only);
+      res = svd.singular_values () (0,0);
+    }
+  else if (p == 1)
+    res = xcolnorms (m, 1).max ();
+  else if (lo_ieee_isinf (p))
+    res = xrownorms (m, 1).max ();
+  else if (p > 1)
+    {
+      VectorT x;
+      res = higham (m, p, std::sqrt (get_eps (p)), max_norm_iter, x);
+    }
+  else
+    (*current_liboctave_error_handler) (p_less1_gripe); 
+
+  return res;
+}
+
+// SVD-free version for sparse matrices
+template <class MatrixT, class VectorT, class R>
+R matrix_norm (const MatrixT& m, R p, VectorT)
+{
+  R res = 0;
+  if (p == 1)
+    res = xcolnorms (m, 1).max ();
+  else if (lo_ieee_isinf (p))
+    res = xrownorms (m, 1).max ();
+  else if (p > 1)
+    {
+      VectorT x;
+      res = higham (m, p, std::sqrt (get_eps (p)), max_norm_iter, x);
+    }
+  else
+    (*current_liboctave_error_handler) (p_less1_gripe); 
+
+  return res;
+}
+
+// and finally, here's what we've promised in the header file
+
+#define DEFINE_XNORM_FUNCS(PREFIX, RTYPE) \
+  RTYPE xnorm (const PREFIX##ColumnVector& x, RTYPE p) \
+  { return vector_norm (x, p); } \
+  RTYPE xnorm (const PREFIX##RowVector& x, RTYPE p) \
+  { return vector_norm (x, p); } \
+  RTYPE xnorm (const PREFIX##Matrix& x, RTYPE p) \
+  { return matrix_norm (x, p, PREFIX##Matrix (), PREFIX##SVD ()); } \
+  RTYPE xfrobnorm (const PREFIX##Matrix& x) \
+  { return vector_norm (x, static_cast<RTYPE> (2)); }
+
+DEFINE_XNORM_FUNCS(, double)
+DEFINE_XNORM_FUNCS(Complex, double)
+DEFINE_XNORM_FUNCS(Float, float)
+DEFINE_XNORM_FUNCS(FloatComplex, float)
+
+// this is needed to avoid copying the sparse matrix for xfrobnorm
+template <class T, class R>
+inline void array_norm_2 (const T* v, octave_idx_type n, R& res)
+{
+  norm_accumulator_2<R> acc;
+  for (octave_idx_type i = 0; i < n; i++)
+    acc.accum (v[i]);
+
+  res = acc;
+}
+
+#define DEFINE_XNORM_SPARSE_FUNCS(PREFIX, RTYPE) \
+  RTYPE xnorm (const Sparse##PREFIX##Matrix& x, RTYPE p) \
+  { return matrix_norm (x, p, PREFIX##Matrix ()); } \
+  RTYPE xfrobnorm (const Sparse##PREFIX##Matrix& x) \
+  { \
+    RTYPE res; \
+    array_norm_2 (x.data (), x.nnz (), res); \
+    return res; \
+  }
+
+DEFINE_XNORM_SPARSE_FUNCS(, double)
+DEFINE_XNORM_SPARSE_FUNCS(Complex, double)
+
+#define DEFINE_COLROW_NORM_FUNCS(PREFIX, RPREFIX, RTYPE) \
+  extern RPREFIX##RowVector xcolnorms (const PREFIX##Matrix& m, RTYPE p) \
+  { return column_norms (m, p); } \
+  extern RPREFIX##ColumnVector xrownorms (const PREFIX##Matrix& m, RTYPE p) \
+  { return row_norms (m, p); } \
+
+DEFINE_COLROW_NORM_FUNCS(, , double)
+DEFINE_COLROW_NORM_FUNCS(Complex, , double)
+DEFINE_COLROW_NORM_FUNCS(Float, Float, float)
+DEFINE_COLROW_NORM_FUNCS(FloatComplex, Float, float)
+
+DEFINE_COLROW_NORM_FUNCS(Sparse, , double)
+DEFINE_COLROW_NORM_FUNCS(SparseComplex, , double)
+
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/liboctave/oct-norm.h	Fri Oct 31 08:05:32 2008 +0100
@@ -0,0 +1,60 @@
+/*
+
+Copyright (C) 2008 VZLU Prague, a.s.
+
+This file is part of Octave.
+
+Octave is free software; you can redistribute it and/or modify it
+under the terms of the GNU General Public License as published by the
+Free Software Foundation; either version 3 of the License, or (at your
+option) any later version.
+
+Octave is distributed in the hope that it will be useful, but WITHOUT
+ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+for more details.
+
+You should have received a copy of the GNU General Public License
+along with Octave; see the file COPYING.  If not, see
+<http://www.gnu.org/licenses/>.
+
+*/
+
+// author: Jaroslav Hajek <highegg@gmail.com>
+
+#if !defined (octave_xnorm_h)
+#define octave_xnorm_h 1
+
+#include "oct-cmplx.h"
+
+#define DECLARE_XNORM_FUNCS(PREFIX, RTYPE) \
+  class PREFIX##Matrix; \
+  class PREFIX##ColumnVector; \
+  class PREFIX##RowVector; \
+  \
+  extern RTYPE xnorm (const PREFIX##ColumnVector&, RTYPE p = 2); \
+  extern RTYPE xnorm (const PREFIX##RowVector&, RTYPE p = 2); \
+  extern RTYPE xnorm (const PREFIX##Matrix&, RTYPE p = 2); \
+  extern RTYPE xfrobnorm (const PREFIX##Matrix&); 
+
+DECLARE_XNORM_FUNCS(, double)
+DECLARE_XNORM_FUNCS(Complex, double)
+DECLARE_XNORM_FUNCS(Float, float)
+DECLARE_XNORM_FUNCS(FloatComplex, float)
+
+DECLARE_XNORM_FUNCS(Sparse, double)
+DECLARE_XNORM_FUNCS(SparseComplex, double)
+
+#define DECLARE_COLROW_NORM_FUNCS(PREFIX, RPREFIX, RTYPE) \
+  extern RPREFIX##RowVector xcolnorms (const PREFIX##Matrix&, RTYPE p = 2); \
+  extern RPREFIX##ColumnVector xrownorms (const PREFIX##Matrix&, RTYPE p = 2); \
+
+DECLARE_COLROW_NORM_FUNCS(, , double)
+DECLARE_COLROW_NORM_FUNCS(Complex, , double)
+DECLARE_COLROW_NORM_FUNCS(Float, Float, float)
+DECLARE_COLROW_NORM_FUNCS(FloatComplex, Float, float)
+
+DECLARE_COLROW_NORM_FUNCS(Sparse, , double)
+DECLARE_COLROW_NORM_FUNCS(SparseComplex, , double)
+
+#endif
--- a/scripts/ChangeLog	Sun Nov 02 06:44:12 2008 +0100
+++ b/scripts/ChangeLog	Fri Oct 31 08:05:32 2008 +0100
@@ -1,3 +1,7 @@
+2008-10-31  Jaroslav Hajek <highegg@gmail.com>
+
+	* linear-algebra/__norm__.m: Remove.
+
 2008-10-31  David Bateman  <dbateman@free.fr>
 
 	* plot/__contour__.m: Exclude infinite values when calculating contour
--- a/scripts/linear-algebra/Makefile.in	Sun Nov 02 06:44:12 2008 +0100
+++ b/scripts/linear-algebra/Makefile.in	Fri Oct 31 08:05:32 2008 +0100
@@ -33,7 +33,7 @@
 INSTALL_PROGRAM = @INSTALL_PROGRAM@
 INSTALL_DATA = @INSTALL_DATA@
 
-SOURCES = __norm__.m commutation_matrix.m cond.m condest.m cross.m \
+SOURCES = commutation_matrix.m cond.m condest.m cross.m \
   dmult.m dot.m duplication_matrix.m housh.m krylov.m krylovb.m logm.m \
   null.m onenormest.m orth.m planerot.m qzhess.m rank.m rref.m subspace.m \
   trace.m vec.m vech.m
--- a/scripts/linear-algebra/__norm__.m	Sun Nov 02 06:44:12 2008 +0100
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,142 +0,0 @@
-## Copyright (C) 1996, 1997, 2007 John W. Eaton
-##
-## This file is part of Octave.
-##
-## Octave is free software; you can redistribute it and/or modify it
-## under the terms of the GNU General Public License as published by
-## the Free Software Foundation; either version 3 of the License, or (at
-## your option) any later version.
-##
-## Octave is distributed in the hope that it will be useful, but
-## WITHOUT ANY WARRANTY; without even the implied warranty of
-## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-## General Public License for more details.
-##
-## You should have received a copy of the GNU General Public License
-## along with Octave; see the file COPYING.  If not, see
-## <http://www.gnu.org/licenses/>.
-
-## Undocumented internal function.
-
-## Author: jwe
-
-function retval = __norm__ (x, p)
-
-  if (nargin < 1 || nargin > 2)
-    print_usage ();
-  endif
-
-  if (isempty (x))
-    retval = [];
-    return;
-  endif
-
-  if (ndims (x) > 2)
-    error ("norm: only valid on 2-D objects")
-  endif
-
-  if (nargin == 1)
-    p = 2;
-  endif
-
-  ## Do we have a vector or matrix as the first argument?
-  if (ndims(x) == 2 && (rows (x) == 1 || columns (x) == 1))
-    if (ischar (p))
-      if (strcmp (p, "fro"))
-        inf_norm = norm (x, "inf");
-        if (inf_norm && finite (inf_norm)) 
-          retval = inf_norm .* sqrt (sum (abs (x ./ inf_norm) .^ 2));
-        else
-          retval = inf_norm;
-        endif
-      elseif (strcmp (p, "inf"))
-        retval = max (abs (x));
-      else
-        error ("norm: unrecognized norm");
-      endif
-    else
-      if (p == Inf)
-        retval = max (abs (x));
-      elseif (p == -Inf)
-        retval = min (abs (x));
-      elseif (p == 1)
-        retval = sum (abs (x));
-      elseif (p == 2)
-        if (iscomplex (x))
-          x = abs (x);
-        endif
-        retval = sqrt (sumsq (x));
-      else
-        retval = sum (abs (x) .^ p) ^ (1/p);
-      endif
-    endif
-  else
-    if (ischar (p))
-      if (strcmp (p, "fro"))
-        inf_norm = norm (x, "inf");
-        if (inf_norm && finite (inf_norm))
-          retval = inf_norm .* sqrt (sum (sum (abs (x ./ inf_norm) .^ 2)));
-        else
-          retval = inf_norm;
-        endif
-      elseif (strcmp (p, "inf"))
-        retval = max (sum (abs (x), 2));
-      else
-        error ("norm: unrecognized vector norm");
-      endif
-    else
-      if (p == 1)
-        retval = max (sum (abs (x)));
-      elseif (p == 2)
-        s = svd (x);
-        retval = s (1);
-      elseif (p == Inf)
-        retval = max (sum (abs (x), 2));
-      else
-        error ("norm: unrecognized matrix norm");
-      endif
-    endif
-  endif
-
-endfunction
-
-%!test
-%! assert (__norm__ (magic (3)), 15, -2*eps);
-%! assert (__norm__ (magic (3) * i), 15, -2*eps);
-
-%!test
-%! assert (__norm__ (zeros (5), "fro"), 0);
-%! assert (__norm__ (ones (5), "fro"), 5);
-%! assert (__norm__ (zeros (5,1), "fro"), 0);
-%! assert (__norm__ (2*ones (5,3), "fro"), sqrt (60));
-
-%!test
-%! assert (__norm__ (zeros (5), "inf"), 0);
-%! assert (__norm__ (ones (5), "inf"), 5);
-%! assert (__norm__ (2*ones (5,1), "inf"), 2);
-%! assert (__norm__ (2*ones (5,3), "inf"), 6);
-
-%!test
-%! assert (__norm__ (zeros (5), 1), 0);
-%! assert (__norm__ (ones (5), 1), 5);
-%! assert (__norm__ (2*ones (1,5), 1), 10);
-%! assert (__norm__ (2*ones (3,5), 1), 6);
-
-
-%!test
-%! assert (__norm__ (1e304 * ones (5, 3), "fro"), 1e304 * sqrt (15));
-%! assert (__norm__ (1e-320 * ones (5, 3), "fro"), 1e-320 * sqrt (15));
-%! assert (x = __norm__ ([1, 2; 3, Inf], "fro"), Inf);
-%! assert (x = __norm__ ([1, 2, 3, Inf], "fro"), Inf);
-%! assert (x = __norm__ ([1, -Inf; 3, 4], "fro"), Inf);
-%! assert (x = __norm__ ([1, 2; 3, NaN], "fro"), NaN);
-
-%!test
-%! assert (__norm__ (1e304 * ones (5, 3), "inf"), 3e304);
-%! assert (__norm__ (1e-320 * ones (5, 3), "inf"), 3e-320);
-%! assert (x = __norm__ ([1, 2; 3, Inf], "inf"), Inf);
-%! assert (x = __norm__ ([1, 2, 3, Inf], "inf"), Inf);
-%! assert (x = __norm__ ([1, -Inf; 3, 4], "inf"), Inf);
-%! assert (x = __norm__ ([1, 2; 3, NaN], "inf"), 3);
-
-
--- a/src/ChangeLog	Sun Nov 02 06:44:12 2008 +0100
+++ b/src/ChangeLog	Fri Oct 31 08:05:32 2008 +0100
@@ -1,3 +1,12 @@
+2008-10-31  Jaroslav Hajek <highegg@gmail.com>
+
+	* xnorm.cc: New source.
+	* xnorm.h: New header file.
+	* Makefile.in: Include xnorm.cc in the build process.
+	* data.cc (Fnorm): Call xnorm, xcolnorms, xrownorms or xfrobnorm
+	to do the actual work.
+
+
 2008-10-30  John W. Eaton  <jwe@octave.org>
 
 	* oct-map.cc (Octave_map::index): Copy key_list.
--- a/src/Makefile.in	Sun Nov 02 06:44:12 2008 +0100
+++ b/src/Makefile.in	Fri Oct 31 08:05:32 2008 +0100
@@ -128,7 +128,7 @@
 	pager.h parse.h pr-output.h procstream.h sighandlers.h \
 	siglist.h sparse-xdiv.h sparse-xpow.h symtab.h sysdep.h \
 	token.h toplev.h unwind-prot.h utils.h variables.h \
-	version.h xdiv.h xpow.h \
+	version.h xdiv.h xnorm.h xpow.h \
 	$(OV_INCLUDES) \
 	$(PT_INCLUDES) \
 	$(OV_SPARSE_INCLUDES)
@@ -207,7 +207,7 @@
 	parse.y pr-output.cc procstream.cc sighandlers.cc \
 	siglist.c sparse-xdiv.cc sparse-xpow.cc strfns.cc \
 	syscalls.cc symtab.cc sysdep.cc token.cc toplev.cc \
-	unwind-prot.cc utils.cc variables.cc xdiv.cc xpow.cc \
+	unwind-prot.cc utils.cc variables.cc xdiv.cc xnorm.cc xpow.cc \
 	$(OV_SRC) \
 	$(PT_SRC)
 
--- a/src/data.cc	Sun Nov 02 06:44:12 2008 +0100
+++ b/src/data.cc	Fri Oct 31 08:05:32 2008 +0100
@@ -62,6 +62,7 @@
 #include "utils.h"
 #include "variables.h"
 #include "pager.h"
+#include "xnorm.h"
 
 #if ! defined (HAVE_HYPOTF) && defined (HAVE__HYPOTF)
 #define hypotf _hypotf
@@ -4574,11 +4575,11 @@
 
 DEFUN (norm, args, ,
   "-*- texinfo -*-\n\
-@deftypefn {Function File} {} norm (@var{a}, @var{p})\n\
+@deftypefn {Function File} {} norm (@var{a}, @var{p}, @var{opt})\n\
 Compute the p-norm of the matrix @var{a}.  If the second argument is\n\
 missing, @code{p = 2} is assumed.\n\
 \n\
-If @var{a} is a matrix:\n\
+If @var{a} is a matrix (or sparse matrix):\n\
 \n\
 @table @asis\n\
 @item @var{p} = @code{1}\n\
@@ -4594,6 +4595,10 @@
 @item @var{p} = @code{\"fro\"}\n\
 @cindex Frobenius norm\n\
 Frobenius norm of @var{a}, @code{sqrt (sum (diag (@var{a}' * @var{a})))}.\n\
+\n\
+@item other @var{p}, @code{@var{p} > 1}\n\
+@cindex general p-norm \n\
+maximum @code{norm (A*x, p)} such that @code{norm (x, p) == 1}\n\
 @end table\n\
 \n\
 If @var{a} is a vector or a scalar:\n\
@@ -4608,20 +4613,27 @@
 @item @var{p} = @code{\"fro\"}\n\
 Frobenius norm of @var{a}, @code{sqrt (sumsq (abs (a)))}.\n\
 \n\
-@item other\n\
+@item @var{p} = 0\n\
+Hamming norm - the number of nonzero elements.\n\
+\n\
+@item other @var{p}, @code{@var{p} > 1}\n\
 p-norm of @var{a}, @code{(sum (abs (@var{a}) .^ @var{p})) ^ (1/@var{p})}.\n\
+\n\
+@item other @var{p} @code{@var{p} < 1}\n\
+the p-pseudonorm defined as above.\n\
 @end table\n\
+\n\
+If @code{\"rows\"} is given as @var{opt}, the norms of all rows of the matrix @var{a} are\n\
+returned as a column vector. Similarly, if @code{\"columns\"} or @code{\"cols\"} is passed\n\
+column norms are computed.\n\
 @seealso{cond, svd}\n\
 @end deftypefn")
 {
-  // Currently only handles vector norms for full double/complex
-  // vectors internally.  Other cases are handled by __norm__.m.
-
   octave_value_list retval;
 
   int nargin = args.length ();
 
-  if (nargin == 1 || nargin == 2)
+  if (nargin >= 1 && nargin <= 3)
     {
       octave_value x_arg = args(0);
 
@@ -4634,86 +4646,48 @@
 	}
       else if (x_arg.ndims () == 2)
 	{
-	  if ((x_arg.rows () == 1 || x_arg.columns () == 1)
-	      && ! (x_arg.is_sparse_type () || x_arg.is_integer_type ()))
-	    {
-	      double p_val = 2;
-
-	      if (nargin == 2)
-		{
-		  octave_value p_arg = args(1);
-
-		  if (p_arg.is_string ())
-		    {
-		      std::string p = args(1).string_value ();
-
-		      if (p == "inf")
-			p_val = octave_Inf;
-		      else if (p == "fro")
-			p_val = -1;
-		      else
-			error ("norm: unrecognized norm `%s'", p.c_str ());
-		    }
-		  else
-		    {
-		      p_val = p_arg.double_value ();
-
-		      if (error_state)
-			error ("norm: unrecognized norm value");
-		    }
-		}
-
-	      if (x_arg.is_single_type ())
-		{
-		  if (! error_state)
-		    {
-		      if (x_arg.is_real_type ())
-			{
-			  MArray<float> x (x_arg.float_array_value ());
-
-			  if (! error_state)
-			    retval(0) = x.norm (static_cast<float>(p_val));
-			  else
-			    error ("norm: expecting real vector");
-			}
-		      else
-			{
-			  MArray<FloatComplex> x (x_arg.float_complex_array_value ());
-
-			  if (! error_state)
-			    retval(0) = x.norm (static_cast<float>(p_val));
-			  else
-			    error ("norm: expecting complex vector");
-			}
-		    }
-		}
-	      else
-		{
-		  if (! error_state)
-		    {
-		      if (x_arg.is_real_type ())
-			{
-			  MArray<double> x (x_arg.array_value ());
-
-			  if (! error_state)
-			    retval(0) = x.norm (p_val);
-			  else
-			    error ("norm: expecting real vector");
-			}
-		      else
-			{
-			  MArray<Complex> x (x_arg.complex_array_value ());
-
-			  if (! error_state)
-			    retval(0) = x.norm (p_val);
-			  else
-			    error ("norm: expecting complex vector");
-			}
-		    }
-		}
-	    }
-	  else
-	    retval = feval ("__norm__", args);
+          enum { sfmatrix, sfcols, sfrows, sffrob, sfinf } strflag = sfmatrix;
+          if (nargin > 1 && args(nargin-1).is_string ())
+            {
+              std::string str = args(nargin-1).string_value ();
+              if (str == "cols" || str == "columns")
+                strflag = sfcols;
+              else if (str == "rows")
+                strflag = sfrows;
+              else if (str == "fro")
+                strflag = sffrob;
+              else if (str == "inf")
+                strflag = sfinf;
+              else
+                error ("norm: unrecognized option: %s", str.c_str ());
+              // we've handled the last parameter, so act as if it was removed
+              nargin --;
+            }
+          else if (nargin > 1 && ! args(1).is_scalar_type ())
+            gripe_wrong_type_arg ("norm", args(1), true);
+
+          if (! error_state)
+            {
+              octave_value p_arg = (nargin > 1) ? args(1) : octave_value (2);
+              switch (strflag)
+                {
+                case sfmatrix:
+                  retval(0) = xnorm (x_arg, p_arg);
+                  break;
+                case sfcols:
+                  retval(0) = xcolnorms (x_arg, p_arg);
+                  break;
+                case sfrows:
+                  retval(0) = xrownorms (x_arg, p_arg);
+                  break;
+                case sffrob:
+                  retval(0) = xfrobnorm (x_arg);
+                  break;
+                case sfinf:
+                  retval(0) = xnorm (x_arg, octave_Inf);
+                  break;
+                }
+            }
 	}
       else
 	error ("norm: only valid for 2-D objects");
@@ -4721,17 +4695,6 @@
   else
     print_usage ();
 
-  // Should not return a sparse type
-  if (retval(0).is_sparse_type ())
-    {
-      if (retval(0).type_name () == "sparse matrix") 
-	retval(0) = retval(0).matrix_value ();
-      else if (retval(0).type_name () == "sparse complex matrix")
-	retval(0) = retval(0).complex_matrix_value ();
-      else if (retval(0).type_name () == "sparse bool matrix")
-	retval(0) = retval(0).bool_matrix_value ();
-    }
-
   return retval;
 }
 
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/src/xnorm.cc	Fri Oct 31 08:05:32 2008 +0100
@@ -0,0 +1,212 @@
+/*
+
+Copyright (C) 2008 VZLU Prague, a.s.
+
+This file is part of Octave.
+
+Octave is free software; you can redistribute it and/or modify it
+under the terms of the GNU General Public License as published by the
+Free Software Foundation; either version 3 of the License, or (at your
+option) any later version.
+
+Octave is distributed in the hope that it will be useful, but WITHOUT
+ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+for more details.
+
+You should have received a copy of the GNU General Public License
+along with Octave; see the file COPYING.  If not, see
+<http://www.gnu.org/licenses/>.
+
+*/
+
+// author: Jaroslav Hajek <highegg@gmail.com>
+
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include <cassert>
+#include <cfloat>
+#include <cmath>
+
+#include "oct-norm.h"
+
+#include "error.h"
+#include "xnorm.h"
+#include "ov.h"
+#include "gripes.h"
+
+octave_value xnorm (const octave_value& x, const octave_value& p)
+{
+  octave_value retval;
+
+  bool isvector = (x.columns () == 1 || x.rows () == 1);
+  bool iscomplex = x.is_complex_type ();
+  bool issparse = x.is_sparse_type ();
+  bool isfloat = x.is_single_type ();
+
+  if (isfloat || x.is_double_type ())
+    {
+      if (isvector)
+        {
+          if (isfloat & iscomplex)
+            retval = xnorm (x.float_complex_column_vector_value (),
+                            p.float_value ());
+          else if (isfloat)
+            retval = xnorm (x.float_column_vector_value (),
+                            p.float_value ());
+          else if (iscomplex)
+            retval = xnorm (x.complex_column_vector_value (),
+                            p.double_value ());
+          else 
+            retval = xnorm (x.column_vector_value (),
+                            p.double_value ());
+        }
+      else if (issparse)
+        {
+          if (iscomplex)
+            retval = xnorm (x.sparse_complex_matrix_value (),
+                            p.double_value ());
+          else 
+            retval = xnorm (x.sparse_matrix_value (),
+                            p.double_value ());
+        }
+      else
+        {
+          if (isfloat & iscomplex)
+            retval = xnorm (x.float_complex_matrix_value (),
+                            p.float_value ());
+          else if (isfloat)
+            retval = xnorm (x.float_matrix_value (),
+                            p.float_value ());
+          else if (iscomplex)
+            retval = xnorm (x.complex_matrix_value (),
+                            p.double_value ());
+          else 
+            retval = xnorm (x.matrix_value (),
+                            p.double_value ());
+        }
+    }
+  else
+    gripe_wrong_type_arg ("xnorm", x, true);
+
+  return retval;
+}
+
+octave_value xcolnorms (const octave_value& x, const octave_value& p)
+{
+  octave_value retval;
+
+  bool iscomplex = x.is_complex_type ();
+  bool issparse = x.is_sparse_type ();
+  bool isfloat = x.is_single_type ();
+
+  if (isfloat || x.is_double_type ())
+    {
+      if (issparse)
+        {
+          if (iscomplex)
+            retval = xcolnorms (x.sparse_complex_matrix_value (),
+                                p.double_value ());
+          else 
+            retval = xcolnorms (x.sparse_matrix_value (),
+                                p.double_value ());
+        }
+      else
+        {
+          if (isfloat & iscomplex)
+            retval = xcolnorms (x.float_complex_matrix_value (),
+                                p.float_value ());
+          else if (isfloat)
+            retval = xcolnorms (x.float_matrix_value (),
+                                p.float_value ());
+          else if (iscomplex)
+            retval = xcolnorms (x.complex_matrix_value (),
+                                p.double_value ());
+          else 
+            retval = xcolnorms (x.matrix_value (),
+                                p.double_value ());
+        }
+    }
+  else
+    gripe_wrong_type_arg ("xcolnorms", x, true);
+
+  return retval;
+}
+
+octave_value xrownorms (const octave_value& x, const octave_value& p)
+{
+  octave_value retval;
+
+  bool iscomplex = x.is_complex_type ();
+  bool issparse = x.is_sparse_type ();
+  bool isfloat = x.is_single_type ();
+
+  if (isfloat || x.is_double_type ())
+    {
+      if (issparse)
+        {
+          if (iscomplex)
+            retval = xrownorms (x.sparse_complex_matrix_value (),
+                                p.double_value ());
+          else 
+            retval = xrownorms (x.sparse_matrix_value (),
+                                p.double_value ());
+        }
+      else
+        {
+          if (isfloat & iscomplex)
+            retval = xrownorms (x.float_complex_matrix_value (),
+                                p.float_value ());
+          else if (isfloat)
+            retval = xrownorms (x.float_matrix_value (),
+                                p.float_value ());
+          else if (iscomplex)
+            retval = xrownorms (x.complex_matrix_value (),
+                                p.double_value ());
+          else 
+            retval = xrownorms (x.matrix_value (),
+                                p.double_value ());
+        }
+    }
+  else
+    gripe_wrong_type_arg ("xrownorms", x, true);
+
+  return retval;
+}
+
+octave_value xfrobnorm (const octave_value& x)
+{
+  octave_value retval;
+
+  bool iscomplex = x.is_complex_type ();
+  bool issparse = x.is_sparse_type ();
+  bool isfloat = x.is_single_type ();
+
+  if (isfloat || x.is_double_type ())
+    {
+      if (issparse)
+        {
+          if (iscomplex)
+            retval = xfrobnorm (x.sparse_complex_matrix_value ());
+          else 
+            retval = xfrobnorm (x.sparse_matrix_value ());
+        }
+      else
+        {
+          if (isfloat & iscomplex)
+            retval = xfrobnorm (x.float_complex_matrix_value ());
+          else if (isfloat)
+            retval = xfrobnorm (x.float_matrix_value ());
+          else if (iscomplex)
+            retval = xfrobnorm (x.complex_matrix_value ());
+          else 
+            retval = xfrobnorm (x.matrix_value ());
+        }
+    }
+  else
+    gripe_wrong_type_arg ("xfrobnorm", x, true);
+
+  return retval;
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/src/xnorm.h	Fri Oct 31 08:05:32 2008 +0100
@@ -0,0 +1,35 @@
+/*
+
+Copyright (C) 2008 VZLU Prague, a.s.
+
+This file is part of Octave.
+
+Octave is free software; you can redistribute it and/or modify it
+under the terms of the GNU General Public License as published by the
+Free Software Foundation; either version 3 of the License, or (at your
+option) any later version.
+
+Octave is distributed in the hope that it will be useful, but WITHOUT
+ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+for more details.
+
+You should have received a copy of the GNU General Public License
+along with Octave; see the file COPYING.  If not, see
+<http://www.gnu.org/licenses/>.
+
+*/
+
+// author: Jaroslav Hajek <highegg@gmail.com>
+
+#if !defined (octave_xnorm_h)
+#define octave_xnorm_h 1
+
+class octave_value;
+
+extern octave_value xnorm (const octave_value& x, const octave_value& p);
+extern octave_value xcolnorms (const octave_value& x, const octave_value& p);
+extern octave_value xrownorms (const octave_value& x, const octave_value& p);
+extern octave_value xfrobnorm (const octave_value& x);
+
+#endif