romanbsd-gsl 1.11.2.2

data/ext/sum.c

@@ -0,0 +1,168 @@
+/*
+  sum.c
+  Ruby/GSL: Ruby extension library for GSL (GNU Scientific Library)
+    (C) Copyright 2001-2006 by Yoshiki Tsunesada
+
+  Ruby/GSL is free software: you can redistribute it and/or modify it
+  under the terms of the GNU General Public License.
+  This library is distributed in the hope that it will be useful, but
+  WITHOUT ANY WARRANTY.
+*/
+#include "rb_gsl_config.h"
+#include "rb_gsl_array.h"
+#include <gsl/gsl_sum.h>
+
+static VALUE rb_gsl_sum_accel(VALUE obj)
+{
+  gsl_sum_levin_u_workspace *w = NULL;
+  double sum, err, sum_plain, *ptr;
+  size_t terms_used, n, stride;  
+  ptr = get_vector_ptr(obj, &stride, &n);
+  w = gsl_sum_levin_u_alloc(n);
+  gsl_sum_levin_u_accel(ptr, n, w, &sum, &err);
+  sum_plain = w->sum_plain;
+  terms_used = w->terms_used;
+  gsl_sum_levin_u_free(w);
+ return rb_ary_new3(4, rb_float_new(sum), rb_float_new(err), 
+		    rb_float_new(sum_plain), INT2FIX(terms_used));
+}
+ 
+static VALUE rb_gsl_utrunc_accel(VALUE obj)
+{
+  gsl_sum_levin_utrunc_workspace *w = NULL;
+  double sum, err, sum_plain, *ptr;
+  size_t terms_used, n, stride;
+  ptr = get_vector_ptr(obj, &stride, &n);
+  w = gsl_sum_levin_utrunc_alloc(n);
+  gsl_sum_levin_utrunc_accel(ptr, n, w, &sum, &err);
+  sum_plain = w->sum_plain;
+  terms_used = w->terms_used;
+  gsl_sum_levin_utrunc_free(w);
+  return rb_ary_new3(4, rb_float_new(sum), rb_float_new(err), 
+		     rb_float_new(sum_plain), INT2FIX(terms_used));
+}
+
+static VALUE rb_gsl_sum_levin_u_new(VALUE klass, VALUE nn)
+{
+  gsl_sum_levin_u_workspace *w = NULL;
+  CHECK_FIXNUM(nn);
+  w = gsl_sum_levin_u_alloc(FIX2INT(nn));
+  return Data_Wrap_Struct(klass, 0, gsl_sum_levin_u_free, w);
+}
+
+static VALUE rb_gsl_sum_levin_utrunc_new(VALUE klass, VALUE nn)
+{
+  gsl_sum_levin_utrunc_workspace *w = NULL;
+  CHECK_FIXNUM(nn);
+  w = gsl_sum_levin_utrunc_alloc(FIX2INT(nn));
+  return Data_Wrap_Struct(klass, 0, gsl_sum_levin_utrunc_free, w);
+}
+
+/* singleton */
+static VALUE rb_gsl_sum_levin_u_accel2(VALUE obj, VALUE vv)
+{
+  gsl_sum_levin_u_workspace *w = NULL;
+  double sum, err, sum_plain;
+  size_t terms_used, n, stride;
+  double *ptr;
+  ptr = get_vector_ptr(vv, &stride, &n);
+  w = gsl_sum_levin_u_alloc(n);
+  gsl_sum_levin_u_accel(ptr, n, w, &sum, &err);
+  sum_plain = w->sum_plain;
+  terms_used = w->terms_used;
+  gsl_sum_levin_u_free(w);
+  return rb_ary_new3(4, rb_float_new(sum), rb_float_new(err), 
+		     rb_float_new(sum_plain), INT2FIX(terms_used));
+}
+
+static VALUE rb_gsl_sum_levin_utrunc_accel2(VALUE obj, VALUE vv)
+{
+  gsl_sum_levin_utrunc_workspace *w = NULL;
+  double sum, err, sum_plain;
+  size_t terms_used, n, stride;
+  double *ptr;
+  ptr = get_vector_ptr(vv, &stride, &n);
+  w = gsl_sum_levin_utrunc_alloc(n);
+  gsl_sum_levin_utrunc_accel(ptr, n, w, &sum, &err);
+  sum_plain = w->sum_plain;
+  terms_used = w->terms_used;
+  gsl_sum_levin_utrunc_free(w);
+  return rb_ary_new3(4, rb_float_new(sum), rb_float_new(err), 
+		     rb_float_new(sum_plain), INT2FIX(terms_used));
+}
+
+static VALUE rb_gsl_sum_levin_u_sum_plain(VALUE obj)
+{
+  gsl_sum_levin_u_workspace *w = NULL;
+  Data_Get_Struct(obj, gsl_sum_levin_u_workspace, w);
+  return rb_float_new(w->sum_plain);
+}
+
+static VALUE rb_gsl_sum_levin_u_terms_used(VALUE obj)
+{
+  gsl_sum_levin_u_workspace *w = NULL;
+  Data_Get_Struct(obj, gsl_sum_levin_u_workspace, w);
+  return INT2FIX(w->terms_used);
+}
+
+static VALUE rb_gsl_sum_levin_utrunc_sum_plain(VALUE obj)
+{
+  gsl_sum_levin_utrunc_workspace *w = NULL;
+  Data_Get_Struct(obj, gsl_sum_levin_utrunc_workspace, w);
+  return rb_float_new(w->sum_plain);
+}
+
+static VALUE rb_gsl_sum_levin_utrunc_terms_used(VALUE obj)
+{
+  gsl_sum_levin_utrunc_workspace *w = NULL;
+  Data_Get_Struct(obj, gsl_sum_levin_utrunc_workspace, w);
+  return INT2FIX(w->terms_used);
+}
+
+void Init_gsl_sum(VALUE module) 
+{
+  VALUE mgsl_sum;
+  VALUE cgsl_sum_levin_u, cgsl_sum_levin_utrunc;
+
+  mgsl_sum = rb_define_module_under(module, "Sum");
+  cgsl_sum_levin_u = rb_define_class_under(mgsl_sum, 
+					   "Levin_u", cGSL_Object);
+  cgsl_sum_levin_utrunc = rb_define_class_under(mgsl_sum, 
+						"Levin_utrunc", cGSL_Object);
+
+  rb_define_singleton_method(cgsl_sum_levin_u, "new", rb_gsl_sum_levin_u_new, 1);
+  rb_define_singleton_method(cgsl_sum_levin_u, "alloc", rb_gsl_sum_levin_u_new, 1);
+  rb_define_singleton_method(cgsl_sum_levin_utrunc, "new", 
+			     rb_gsl_sum_levin_utrunc_new, 1);
+  rb_define_singleton_method(cgsl_sum_levin_utrunc, "alloc", 
+			     rb_gsl_sum_levin_utrunc_new, 1);
+  rb_define_singleton_method(cgsl_sum_levin_u, "accel", 
+			     rb_gsl_sum_levin_u_accel2, 1);
+
+  rb_define_singleton_method(cgsl_sum_levin_utrunc, "accel", 
+			     rb_gsl_sum_levin_utrunc_accel2, 1);
+  rb_define_method(cgsl_sum_levin_u, "accel", rb_gsl_sum_levin_u_accel2, 1);
+  rb_define_method(cgsl_sum_levin_utrunc, "accel", 
+		   rb_gsl_sum_levin_utrunc_accel2, 1);
+
+  rb_define_method(cgsl_sum_levin_u, "sum_plain", rb_gsl_sum_levin_u_sum_plain, 0);
+  rb_define_method(cgsl_sum_levin_u, "terms_used", 
+		   rb_gsl_sum_levin_u_terms_used, 0);
+  rb_define_method(cgsl_sum_levin_utrunc, "sum_plain", 
+		   rb_gsl_sum_levin_utrunc_sum_plain, 0);
+  rb_define_method(cgsl_sum_levin_utrunc, "terms_used", 
+		   rb_gsl_sum_levin_utrunc_terms_used, 0);
+  /***/
+
+  rb_define_method(cgsl_vector, "accel_sum", rb_gsl_sum_accel, 0);
+  rb_define_alias(cgsl_vector, "accel", "accel_sum");
+  rb_define_alias(cgsl_vector, "sum_accel", "accel_sum");
+  rb_define_method(cgsl_vector, "utrunc_accel", rb_gsl_utrunc_accel, 0);
+
+#ifdef HAVE_NARRAY_H
+  rb_define_method(cNArray, "accel_sum", rb_gsl_sum_accel, 0);
+  rb_define_alias(cNArray, "accel", "accel_sum");
+  rb_define_alias(cNArray, "sum_accel", "accel_sum");
+  rb_define_method(cNArray, "utrunc_accel", rb_gsl_utrunc_accel, 0);
+#endif
+}

data/ext/tamu_anova.c

@@ -0,0 +1,56 @@
+#include "rb_gsl.h"
+
+#ifdef HAVE_TAMU_ANOVA_TAMU_ANOVA_H
+VALUE rb_tamu_anova_alloc(int argc, VALUE *argv, VALUE klass)
+{
+  gsl_vector *data;
+  gsl_vector_long *factor;
+  long I, J;
+  struct tamu_anova_table *table;
+  switch (argc) {
+  case 3:
+  case 4:
+    Data_Get_Struct(argv[0], gsl_vector, data);
+    Data_Get_Struct(argv[1], gsl_vector_long, factor);
+    if (argc == 3) {
+      I = data->size;
+      J = NUM2INT(argv[2]);
+    } else {
+      I = NUM2INT(argv[2]);
+      J = NUM2INT(argv[3]);
+    }
+    table = (struct tamu_anova_table *) malloc(sizeof(struct tamu_anova_table));
+    *table = tamu_anova(data->data, factor->data, I, J);
+    break;
+  default:
+    rb_raise(rb_eArgError, "Wrong number of arguments (%d for 3 or 4)", argc);
+    break;
+  }
+  return Data_Wrap_Struct(klass, 0, free, table);
+}
+
+VALUE rb_tamu_anova_printtable(VALUE *vTable)
+{
+  struct tamu_anova_table *table;
+  Data_Get_Struct(vTable, struct tamu_anova_table, table);
+  tamu_anova_printtable(*table);
+  return Qtrue;
+}
+
+#endif
+
+void Init_tamu_anova(VALUE module)
+{
+#ifdef HAVE_TAMU_ANOVA_TAMU_ANOVA_H
+  VALUE mTAMU_ANOVA;
+  VALUE cTable;
+
+  mTAMU_ANOVA = rb_define_module_under(module, "TAMU_ANOVA");
+  cTable = rb_define_class_under(mTAMU_ANOVA, "Table", cGSL_Object);
+
+  rb_define_singleton_method(cTable, "alloc", rb_tamu_anova_alloc, -1);
+  rb_define_singleton_method(cTable, "oneway", rb_tamu_anova_alloc, -1);
+
+  rb_define_method(cTable, "print", rb_tamu_anova_printtable, 0);
+#endif
+}

data/ext/tensor.c

@@ -0,0 +1,38 @@
+/*
+  tensor.c
+  Ruby/GSL: Ruby extension library for GSL (GNU Scientific Library)
+    (C) Copyright 2004 by Yoshiki Tsunesada
+
+  Ruby/GSL is free software: you can redistribute it and/or modify it
+  under the terms of the GNU General Public License.
+  This library is distributed in the hope that it will be useful, but
+  WITHOUT ANY WARRANTY.
+*/
+
+/*
+  The gsl_tensor package is developed by J.Burguet, and
+  distributed separately as an add-on package.
+ */
+
+#ifdef HAVE_TENSOR_TENSOR_H
+
+#include "rb_gsl_config.h"
+#include "rb_gsl_tensor.h"
+
+#ifdef HAVE_NARRAY_H
+#include "rb_gsl_with_narray.h"
+#endif
+
+#define BASE_DOUBLE
+#include "templates_on.h"
+#include "tensor_source.c"
+#include "templates_off.h"
+#undef  BASE_DOUBLE
+
+#define BASE_INT
+#include "templates_on.h"
+#include "tensor_source.c"
+#include "templates_off.h"
+#undef  BASE_INT
+
+#endif

data/ext/tensor_source.c

@@ -0,0 +1,1121 @@
+/*
+  tensor_source.c
+  Ruby/GSL: Ruby extension library for GSL (GNU Scientific Library)
+    (C) Copyright 2004 by Yoshiki Tsunesada
+
+  Ruby/GSL is free software: you can redistribute it and/or modify it
+  under the terms of the GNU General Public License.
+  This library is distributed in the hope that it will be useful, but
+  WITHOUT ANY WARRANTY.
+*/
+
+/*
+  The tensor package is developed by Jordi Burguet-Caltell, and
+  distributed separately as an add-on package.
+  http://savannah.nongnu.org/projects/tensor/
+ */
+
+#ifdef HAVE_TENSOR_TENSOR_H
+
+#include "rb_gsl_config.h"
+#include "rb_gsl_tensor.h"
+#include "rb_gsl_common.h"
+
+#ifdef BASE_DOUBLE
+VALUE cgsl_tensor, cgsl_tensor_int;
+VALUE cgsl_tensor_view, cgsl_tensor_int_view;
+#define NUMCONV(x) NUM2DBL(x)
+#define C_TO_VALUE(x) rb_float_new(x)
+#define CHECK_TEN(x) CHECK_TENSOR(x)
+#define TEN_P(x) TENSOR_P(x)
+#define VEC_P(x) VECTOR_P(x)
+#define MAT_P(x) MATRIX_P(x)
+#else defined(BASE_INT)
+#define NUMCONV(x) FIX2INT(x)
+#define C_TO_VALUE(x) INT2FIX(x)
+#define CHECK_TEN(x) CHECK_TENSOR_INT(x)
+#define TEN_P(x) TENSOR_INT_P(x)
+#define VEC_P(x) VECTOR_INT_P(x)
+#define MAT_P(x) MATRIX_INT_P(x)
+#endif
+
+GSL_TYPE(rbgsl_tensor)* FUNCTION(rbgsl_tensor,alloc)(const unsigned int rank, 
+						     const size_t dimension)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  t = ALLOC(GSL_TYPE(rbgsl_tensor));
+  t->tensor = FUNCTION(tensor,alloc)(rank, dimension);
+  if (rank == 0) 
+    t->indices = gsl_permutation_alloc(1);
+  else
+    t->indices = gsl_permutation_alloc(rank);
+  return t;
+}
+
+GSL_TYPE(rbgsl_tensor)* FUNCTION(rbgsl_tensor,calloc)(const unsigned int rank, 
+						      const size_t dimension)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  t = ALLOC(GSL_TYPE(rbgsl_tensor));
+  t->tensor = FUNCTION(tensor,calloc)(rank, dimension);
+  if (rank == 0) 
+    t->indices = gsl_permutation_alloc(1);
+  else
+    t->indices = gsl_permutation_alloc(rank);
+  return t;
+}
+
+GSL_TYPE(rbgsl_tensor)* FUNCTION(rbgsl_tensor,copy)(const GSL_TYPE(rbgsl_tensor) *t)
+{
+  GSL_TYPE(rbgsl_tensor) *tnew;
+  tnew = ALLOC(GSL_TYPE(rbgsl_tensor));
+  if (t->tensor->rank == 0) 
+    tnew->indices = gsl_permutation_alloc(1);
+  else
+    tnew->indices = gsl_permutation_alloc(t->tensor->rank);
+  tnew->tensor = FUNCTION(tensor,copy)(t->tensor);
+  return tnew;
+}
+
+void FUNCTION(rbgsl_tensor,free)(GSL_TYPE(rbgsl_tensor) *t)
+{
+  gsl_permutation_free(t->indices);
+  FUNCTION(tensor,free)(t->tensor);
+  free((GSL_TYPE(rbgsl_tensor) *) t);
+}
+
+void FUNCTION(rbgsl_tensor,free2)(GSL_TYPE(rbgsl_tensor) *t)
+{
+  gsl_permutation_free(t->indices);
+  free((GSL_TYPE(tensor)*) t->tensor);
+  free((GSL_TYPE(rbgsl_tensor) *) t);
+}
+
+/* singleton methods */
+static VALUE FUNCTION(rb_tensor,new)(int argc, VALUE *argv, VALUE klass)
+{
+  unsigned int rank;
+  size_t dim;
+  GSL_TYPE(rbgsl_tensor) *t = NULL;
+  switch (argc) {
+  case 2:
+    rank = FIX2UINT(argv[0]);
+    dim = (size_t) FIX2UINT(argv[1]);
+    t = FUNCTION(rbgsl_tensor,alloc)(rank, dim);
+    break;
+  default:
+    rb_raise(rb_eArgError, "wrong number of arguments (%d for 2, rank and dimension)",
+	     argc);
+    break;
+  }
+  return Data_Wrap_Struct(GSL_TYPE(cgsl_tensor), 0, FUNCTION(rbgsl_tensor,free), t);
+}
+
+static VALUE FUNCTION(rb_tensor,calloc)(VALUE klass, VALUE r, VALUE s)
+{
+  unsigned int rank;
+  size_t dim;
+  GSL_TYPE(rbgsl_tensor) *t = NULL;
+  rank = FIX2UINT(r);
+  dim = (size_t) FIX2UINT(s);
+  t = FUNCTION(rbgsl_tensor,calloc)(rank, dim);
+  return Data_Wrap_Struct(GSL_TYPE(cgsl_tensor), 0, FUNCTION(rbgsl_tensor,free), t);
+}
+
+static VALUE FUNCTION(rb_tensor,copy_singleton)(VALUE klass, VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t = NULL, *tnew = NULL;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  tnew = FUNCTION(rbgsl_tensor,copy(t));
+  return Data_Wrap_Struct(GSL_TYPE(cgsl_tensor), 0, FUNCTION(rbgsl_tensor,free), tnew);
+}
+
+static VALUE FUNCTION(rb_tensor,memcpy_singleton)(VALUE klass, VALUE a, VALUE b)
+{
+  GSL_TYPE(rbgsl_tensor) *dst, *src;
+  CHECK_TEN(b);
+  Data_Get_Struct(a, GSL_TYPE(rbgsl_tensor), dst);
+  Data_Get_Struct(b, GSL_TYPE(rbgsl_tensor), src);
+  return INT2FIX(FUNCTION(tensor,memcpy)(dst->tensor, src->tensor));
+}
+
+static VALUE FUNCTION(rb_tensor,swap_singleton)(VALUE klass, VALUE a, VALUE b)
+{
+  GSL_TYPE(rbgsl_tensor) *t1, *t2;
+  CHECK_TEN(b);
+  Data_Get_Struct(a, GSL_TYPE(rbgsl_tensor), t1);
+  Data_Get_Struct(b, GSL_TYPE(rbgsl_tensor), t2);
+  return INT2FIX(FUNCTION(tensor,swap)(t1->tensor, t2->tensor));
+}
+
+/*****/
+static VALUE FUNCTION(rb_tensor,copy)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t = NULL, *tnew = NULL;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  tnew = FUNCTION(rbgsl_tensor,copy)(t);
+  return Data_Wrap_Struct(GSL_TYPE(cgsl_tensor), 0, FUNCTION(rbgsl_tensor,free), tnew);
+}
+
+static VALUE FUNCTION(rb_tensor,set_zero)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t = NULL;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  FUNCTION(tensor,set_zero)(t->tensor);
+  return obj;
+}
+
+static VALUE FUNCTION(rb_tensor,set_all)(VALUE obj, VALUE xx)
+{
+  GSL_TYPE(rbgsl_tensor) *t = NULL;
+  BASE x;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  x = NUMCONV(xx);
+  FUNCTION(tensor,set_all)(t->tensor, x);
+  return obj;
+}
+
+static void rb_tensor_get_indices_array(tensor_indices *v, VALUE ary);
+static void rbgsl_tensor_get_indices(int argc, VALUE *argv, tensor_indices *indices, 
+				     size_t *n);
+#ifdef BASE_DOUBLE
+static void rb_tensor_get_indices_array(tensor_indices *v, VALUE ary)
+{
+  size_t i, nn;
+  nn = (size_t) GSL_MIN_INT((int) v->size, (int) RARRAY(ary)->len);
+  for (i = 0; i < nn; i++)
+    v->data[i] = FIX2UINT(rb_ary_entry(ary, i));
+}
+
+static void rbgsl_tensor_get_indices(int argc, VALUE *argv, 
+				     tensor_indices *indices, size_t *n)
+{
+  size_t i;
+  for (i = 0; i < indices->size; i++) indices->data[i] = 0;
+  switch (argc) {
+  case 1:
+    switch (TYPE(argv[0])) {
+    case T_ARRAY:
+      *n = (size_t) GSL_MIN_INT((int) indices->size, (int) RARRAY(argv[0])->len);
+      rb_tensor_get_indices_array(indices, argv[0]);
+      break;
+    case T_FIXNUM:
+      *n = 1;
+      indices->data[0] = FIX2INT(argv[0]);
+      break;
+    default:
+      rb_raise(rb_eTypeError, "wrong argument type %s (Array expected)",
+	       rb_class2name(CLASS_OF(argv[0])));
+      break;
+    }
+    break;
+  default:
+    *n = (size_t) GSL_MIN_INT(argc, (int) indices->size);
+    for (i = 0; i < *n; i++) {
+      CHECK_FIXNUM(argv[i]);
+      indices->data[i] = FIX2INT(argv[i]);
+    }
+    break;
+  }
+}
+#endif
+
+size_t FUNCTION(tensor,position)(const size_t * indices,
+				     const GSL_TYPE(tensor) * t);
+static VALUE FUNCTION(rb_tensor,position)(int argc, VALUE *argv, VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t = NULL;
+  size_t n, position;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);  
+  rbgsl_tensor_get_indices(argc, argv, t->indices, &n);  
+  position = (size_t) FUNCTION(tensor,position)(t->indices->data,t->tensor);
+  return INT2FIX(position);
+}
+static VALUE FUNCTION(rb_tensor,subtensor)(int argc, VALUE *argv, VALUE obj);
+static VALUE FUNCTION(rb_tensor,get)(int argc, VALUE *argv, VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t = NULL;
+  BASE x;
+  size_t n;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);  
+  rbgsl_tensor_get_indices(argc, argv, t->indices, &n);  
+  if (n < t->tensor->rank) {
+    return FUNCTION(rb_tensor,subtensor)(argc, argv, obj);
+  } else {
+    x = FUNCTION(tensor,get)(t->tensor, t->indices->data);
+    return C_TO_VALUE(x);
+  }
+  return Qnil;
+}
+
+static VALUE FUNCTION(rb_tensor,set)(int argc, VALUE *argv, VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t = NULL;
+  size_t n;
+  BASE x;
+  if (argc < 2) 
+    rb_raise(rb_eArgError, "wrong number of arguments (%d for >= 2)", argc);
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);  
+  rbgsl_tensor_get_indices(argc-1, argv, t->indices, &n);
+  x = NUMCONV(argv[argc-1]);
+  FUNCTION(tensor,set)(t->tensor, t->indices->data, x);
+  return obj;
+}
+
+static VALUE FUNCTION(rb_tensor,fread)(VALUE obj, VALUE io)
+{
+  GSL_TYPE(rbgsl_tensor) *t = NULL;
+  FILE *f = NULL;
+  int status, flag = 0;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  f = rb_gsl_open_readfile(io, &flag);
+  status = FUNCTION(tensor,fread)(f, t->tensor);
+  if (flag == 1) fclose(f);
+  return INT2FIX(status);
+}
+
+static VALUE FUNCTION(rb_tensor,fwrite)(VALUE obj, VALUE io)
+{
+  GSL_TYPE(rbgsl_tensor) *t = NULL;
+  FILE *f = NULL;
+  int status, flag = 0;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  f = rb_gsl_open_writefile(io, &flag);
+  status = FUNCTION(tensor,fwrite)(f, t->tensor);
+  if (flag == 1) fclose(f);
+  return INT2FIX(status);
+}
+
+static VALUE FUNCTION(rb_tensor,fprintf)(int argc, VALUE *argv, VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *h = NULL;
+  FILE *fp = NULL;
+  int status, flag = 0;
+  if (argc != 1 && argc != 2) 
+    rb_raise(rb_eArgError, "wrong number of arguments (%d for 1 or 2)", argc);
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), h);
+  fp = rb_gsl_open_writefile(argv[0], &flag);
+  switch (argc) {
+  case 2:
+    if (TYPE(argv[1]) == T_STRING)
+      status = FUNCTION(tensor,fprintf)(fp, h->tensor, STR2CSTR(argv[1]));
+    else
+      rb_raise(rb_eTypeError, "argv 2 String expected");
+    break;
+  default:
+    status = FUNCTION(tensor,fprintf)(fp, h->tensor, OUT_FORMAT);
+    break;
+  }
+  if (flag == 1) fclose(fp);
+  return INT2FIX(status);
+}
+
+static VALUE FUNCTION(rb_tensor,printf)(int argc, VALUE *argv, VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *h = NULL;
+  int status;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), h);
+  if (argc == 1) {
+    if (TYPE(argv[0]) != T_STRING) 
+      rb_raise(rb_eTypeError, "String expected");
+    else
+      status = FUNCTION(tensor,fprintf)(stdout, h->tensor, STR2CSTR(argv[0]));
+  } else {
+    status = FUNCTION(tensor,fprintf)(stdout, h->tensor, OUT_FORMAT);
+  }
+  return INT2FIX(status);
+}
+
+static VALUE FUNCTION(rb_tensor,fscanf)(VALUE obj, VALUE io)
+{
+  GSL_TYPE(rbgsl_tensor) *h = NULL;
+  FILE *fp = NULL;
+  int status, flag = 0;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), h);
+  fp = rb_gsl_open_readfile(io, &flag);
+  status = FUNCTION(tensor,fscanf)(fp, h->tensor);
+  if (flag == 1) fclose(fp);
+  return INT2FIX(status);
+}
+
+static VALUE FUNCTION(rb_tensor,swap_indices)(VALUE obj, VALUE ii, VALUE jj)
+{
+  GSL_TYPE(rbgsl_tensor) *t, *tnew;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  tnew = ALLOC(GSL_TYPE(rbgsl_tensor));
+  if (t->tensor->rank == 0) 
+    tnew->indices = gsl_permutation_alloc(1);
+  else
+    tnew->indices = gsl_permutation_alloc(t->tensor->rank);
+  tnew->tensor = FUNCTION(tensor,swap_indices)(t->tensor, FIX2INT(ii), FIX2INT(jj));
+  return Data_Wrap_Struct(GSL_TYPE(cgsl_tensor), 0, FUNCTION(rbgsl_tensor,free), tnew);
+}
+
+static VALUE FUNCTION(rb_tensor,max)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  return C_TO_VALUE(FUNCTION(tensor,max)(t->tensor));
+}
+
+static VALUE FUNCTION(rb_tensor,min)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  return C_TO_VALUE(FUNCTION(tensor,min)(t->tensor));
+}
+
+static VALUE FUNCTION(rb_tensor,minmax)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  BASE min, max;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  FUNCTION(tensor,minmax)(t->tensor, &min, &max);
+  return rb_ary_new3(2, C_TO_VALUE(min), C_TO_VALUE(max));
+}
+
+static VALUE FUNCTION(rb_tensor,max_index)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  FUNCTION(tensor,max_index)(t->tensor, t->indices->data);
+  return Data_Wrap_Struct(cgsl_index, 0, NULL, t->indices);
+}
+
+static VALUE FUNCTION(rb_tensor,min_index)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  FUNCTION(tensor,min_index)(t->tensor, t->indices->data);
+  return Data_Wrap_Struct(cgsl_index, 0, NULL, t->indices);
+}
+
+static VALUE FUNCTION(rb_tensor,minmax_index)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  gsl_permutation *min, *max;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  if (t->tensor->rank == 0) {
+    min = gsl_permutation_alloc(1);
+    max = gsl_permutation_alloc(1);
+  } else {
+    min = gsl_permutation_alloc(t->tensor->rank);
+    max = gsl_permutation_alloc(t->tensor->rank);
+  }
+  FUNCTION(tensor,minmax_index)(t->tensor, min->data, max->data);
+  return rb_ary_new3(2,
+		     Data_Wrap_Struct(cgsl_index, 0, gsl_permutation_free, min), 
+		     Data_Wrap_Struct(cgsl_index, 0, gsl_permutation_free, max));
+}
+
+static VALUE FUNCTION(rb_tensor,isnull)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  return INT2FIX(FUNCTION(tensor,isnull)(t->tensor));
+}
+
+static VALUE FUNCTION(rb_tensor,isnull2)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  int status;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  status = FUNCTION(tensor,isnull)(t->tensor);
+  if (status) return Qtrue;
+  else return Qfalse;
+}
+
+static VALUE FUNCTION(rb_tensor,oper)(VALUE obj, VALUE bb,
+					   int flag)
+{
+  GSL_TYPE(rbgsl_tensor) *a, *b, *anew;
+  BASE x;
+  int (*f)(GSL_TYPE(tensor)*, const GSL_TYPE(tensor)*);
+  int (*f2)(GSL_TYPE(tensor)*, const double);
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), a);
+  anew = FUNCTION(rbgsl_tensor,copy)(a);
+  if (TEN_P(bb)) {
+    Data_Get_Struct(bb, GSL_TYPE(rbgsl_tensor), b);
+    switch (flag) {
+    case TENSOR_ADD: f = FUNCTION(&tensor,add); break;
+    case TENSOR_SUB: f = FUNCTION(&tensor,sub); break;
+    case TENSOR_MUL_ELEMENTS: f = FUNCTION(&tensor,mul_elements); break;
+    case TENSOR_DIV_ELEMENTS: f = FUNCTION(&tensor,div_elements); break;
+    default: rb_raise(rb_eRuntimeError, "unknown operation"); break;
+    } 
+    (*f)(anew->tensor, b->tensor);
+  } else {
+    switch (flag) {
+    case TENSOR_ADD:
+    case TENSOR_ADD_CONSTANT: 
+      x = NUMCONV(bb);
+      f2 = FUNCTION(&tensor,add_constant); 
+      break;
+    case TENSOR_SUB:
+      x = -NUMCONV(bb);
+      f2 = FUNCTION(&tensor,add_constant); 
+      break;
+    case TENSOR_ADD_DIAGONAL: f2 = FUNCTION(&tensor,add_diagonal); break;
+    case TENSOR_MUL_ELEMENTS:
+    case TENSOR_SCALE:
+      x = NUMCONV(bb); 
+      f2 = FUNCTION(&tensor,scale); 
+      break;
+    case TENSOR_DIV_ELEMENTS: 
+      x = 1.0/NUMCONV(bb); 
+      f2 = FUNCTION(&tensor,scale); 
+      break;
+    default: rb_raise(rb_eRuntimeError, "unknown operation"); break;
+    }
+    (*f2)(anew->tensor, x);
+  }
+  return Data_Wrap_Struct(GSL_TYPE(cgsl_tensor), 0, FUNCTION(rbgsl_tensor,free), anew);
+}
+
+
+static VALUE FUNCTION(rb_tensor,oper_bang)(VALUE obj, VALUE bb, int flag)
+{
+  GSL_TYPE(rbgsl_tensor) *a, *b;
+  BASE x;
+  int (*f)(GSL_TYPE(tensor)*, const GSL_TYPE(tensor)*);
+  int (*f2)(GSL_TYPE(tensor)*, const double);
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), a);
+  if (TEN_P(bb)) {
+    Data_Get_Struct(bb, GSL_TYPE(rbgsl_tensor), b);
+    switch (flag) {
+    case TENSOR_ADD: f = FUNCTION(&tensor,add); break;
+    case TENSOR_SUB: f = FUNCTION(&tensor,sub); break;
+    case TENSOR_MUL_ELEMENTS: f = FUNCTION(&tensor,mul_elements); break;
+    case TENSOR_DIV_ELEMENTS: f = FUNCTION(&tensor,div_elements); break;
+    default: rb_raise(rb_eRuntimeError, "unknown operation"); break;
+    } 
+    (*f)(a->tensor, b->tensor);
+  } else {
+    switch (flag) {
+    case TENSOR_ADD:
+    case TENSOR_ADD_CONSTANT: 
+      x = NUMCONV(bb);
+      f2 = FUNCTION(&tensor,add_constant); 
+      break;
+    case TENSOR_SUB:
+      x = -NUMCONV(bb);
+      f2 = FUNCTION(&tensor,add_constant); 
+      break;
+    case TENSOR_ADD_DIAGONAL: f2 = FUNCTION(&tensor,add_diagonal); break;
+    case TENSOR_MUL_ELEMENTS:
+    case TENSOR_SCALE:
+      x = NUMCONV(bb); 
+      f2 = FUNCTION(&tensor,scale); 
+      break;
+    case TENSOR_DIV_ELEMENTS: 
+      x = 1.0/NUMCONV(bb); 
+      f2 = FUNCTION(&tensor,scale); 
+      break;
+    default: rb_raise(rb_eRuntimeError, "unknown operation"); break;
+    }
+    (*f2)(a->tensor, x);
+  }
+  return obj;
+}
+
+static VALUE FUNCTION(rb_tensor,add)(VALUE obj, VALUE bb)
+{
+  return FUNCTION(rb_tensor,oper)(obj, bb, TENSOR_ADD);
+}
+
+static VALUE FUNCTION(rb_tensor,sub)(VALUE obj, VALUE bb)
+{
+  return FUNCTION(rb_tensor,oper)(obj, bb, TENSOR_SUB);
+}
+
+static VALUE FUNCTION(rb_tensor,mul_elements)(VALUE obj, VALUE bb)
+{
+  return FUNCTION(rb_tensor,oper)(obj, bb, TENSOR_MUL_ELEMENTS);
+}
+
+static VALUE FUNCTION(rb_tensor,div_elements)(VALUE obj, VALUE bb)
+{
+  return FUNCTION(rb_tensor,oper)(obj, bb, TENSOR_DIV_ELEMENTS);
+}
+
+static VALUE FUNCTION(rb_tensor,add_constant)(VALUE obj, VALUE bb)
+{
+  return FUNCTION(rb_tensor,oper)(obj, bb, TENSOR_ADD_CONSTANT);
+}
+
+static VALUE FUNCTION(rb_tensor,add_diagonal)(VALUE obj, VALUE bb)
+{
+  return FUNCTION(rb_tensor,oper)(obj, bb, TENSOR_ADD_DIAGONAL);
+}
+
+static VALUE FUNCTION(rb_tensor,scale)(VALUE obj, VALUE bb)
+{
+  return FUNCTION(rb_tensor,oper)(obj, bb, TENSOR_SCALE);
+}
+
+/***/
+static VALUE FUNCTION(rb_tensor,add_bang)(VALUE obj, VALUE bb)
+{
+  return FUNCTION(rb_tensor,oper_bang)(obj, bb, TENSOR_ADD);
+}
+
+static VALUE FUNCTION(rb_tensor,sub_bang)(VALUE obj, VALUE bb)
+{
+  return FUNCTION(rb_tensor,oper_bang)(obj, bb, TENSOR_SUB);
+}
+
+static VALUE FUNCTION(rb_tensor,mul_elements_bang)(VALUE obj, VALUE bb)
+{
+  return FUNCTION(rb_tensor,oper_bang)(obj, bb, TENSOR_MUL_ELEMENTS);
+}
+
+static VALUE FUNCTION(rb_tensor,div_elements_bang)(VALUE obj, VALUE bb)
+{
+  return FUNCTION(rb_tensor,oper_bang)(obj, bb, TENSOR_DIV_ELEMENTS);
+}
+
+static VALUE FUNCTION(rb_tensor,add_constant_bang)(VALUE obj, VALUE bb)
+{
+  return FUNCTION(rb_tensor,oper_bang)(obj, bb, TENSOR_ADD_CONSTANT);
+}
+
+static VALUE FUNCTION(rb_tensor,add_diagonal_bang)(VALUE obj, VALUE bb)
+{
+  return FUNCTION(rb_tensor,oper_bang)(obj, bb, TENSOR_ADD_DIAGONAL);
+}
+
+static VALUE FUNCTION(rb_tensor,scale_bang)(VALUE obj, VALUE bb)
+{
+  return FUNCTION(rb_tensor,oper_bang)(obj, bb, TENSOR_SCALE);
+}
+
+/*****/
+static VALUE FUNCTION(rb_tensor,product_singleton)(VALUE obj, VALUE aa, VALUE bb)
+{
+  GSL_TYPE(rbgsl_tensor) *a, *b, *c;
+  CHECK_TEN(aa);
+  CHECK_TEN(bb);
+  Data_Get_Struct(aa, GSL_TYPE(rbgsl_tensor), a);
+  switch (TYPE(bb)) {
+  case T_FIXNUM:
+  case T_BIGNUM:
+  case T_FLOAT:
+    return FUNCTION(rb_tensor,mul_elements)(aa, bb);
+    break;
+  default:
+    Data_Get_Struct(bb, GSL_TYPE(rbgsl_tensor), b);
+    c = ALLOC(GSL_TYPE(rbgsl_tensor));
+    c->tensor = FUNCTION(tensor,product(a->tensor, b->tensor));
+    if (c->tensor->rank == 0)
+      c->indices = gsl_permutation_alloc(1);
+    else
+      c->indices = gsl_permutation_alloc(c->tensor->rank);
+    return Data_Wrap_Struct(GSL_TYPE(cgsl_tensor), 0, FUNCTION(rbgsl_tensor,free), c);
+    break;
+  }
+}
+
+static VALUE FUNCTION(rb_tensor,product)(VALUE obj, VALUE bb)
+{
+  GSL_TYPE(rbgsl_tensor) *a, *b, *c;
+  switch (TYPE(bb)) {
+  case T_FIXNUM:
+  case T_BIGNUM:
+  case T_FLOAT:
+    return FUNCTION(rb_tensor,mul_elements)(obj, bb);
+    break;
+  default:
+    CHECK_TEN(bb);
+    Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), a);
+    Data_Get_Struct(bb, GSL_TYPE(rbgsl_tensor), b);
+    c = ALLOC(GSL_TYPE(rbgsl_tensor));
+    c->tensor = FUNCTION(tensor,product(a->tensor, b->tensor));
+    if (c->tensor->rank == 0)
+      c->indices = gsl_permutation_alloc(1);
+    else
+      c->indices = gsl_permutation_alloc(c->tensor->rank);
+    return Data_Wrap_Struct(GSL_TYPE(cgsl_tensor), 0, FUNCTION(rbgsl_tensor,free), c);
+    break;
+  }
+}
+
+static VALUE FUNCTION(rb_tensor,contract)(VALUE obj, VALUE ii, VALUE jj)
+{
+  GSL_TYPE(rbgsl_tensor) *t, *tnew;
+  size_t rank;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  tnew = ALLOC(GSL_TYPE(rbgsl_tensor));
+  tnew->tensor = FUNCTION(tensor,contract)(t->tensor, FIX2INT(ii), FIX2INT(jj));
+  if (tnew->tensor->rank == 0) rank = 1;
+  else rank = tnew->tensor->rank;
+  tnew->indices = gsl_permutation_alloc(rank);
+  return Data_Wrap_Struct(GSL_TYPE(cgsl_tensor), 0, FUNCTION(rbgsl_tensor,free), tnew);
+}
+
+static VALUE FUNCTION(rb_tensor,size)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  return INT2FIX(t->tensor->size);
+}
+
+static VALUE FUNCTION(rb_tensor,rank)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  return INT2FIX(t->tensor->rank);
+}
+
+static VALUE FUNCTION(rb_tensor,dimension)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  return INT2FIX(t->tensor->dimension);
+}
+
+static VALUE FUNCTION(rb_tensor,data)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  QUALIFIED_VIEW(gsl_vector,view) *v = NULL;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  v = FUNCTION(rb_gsl_make_vector,view)(t->tensor->data, t->tensor->size, 1);
+  return Data_Wrap_Struct(QUALIFIED_VIEW(cgsl_vector,view), 0, free, v);
+}
+
+static VALUE FUNCTION(rb_tensor,2matrix)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  GSL_TYPE(gsl_matrix) *m = NULL;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  m = (GSL_TYPE(gsl_matrix)*) FUNCTION(tensor,2matrix)(t->tensor);
+  return Data_Wrap_Struct(QUALIFIED_VIEW(cgsl_matrix,view), 0, FUNCTION(gsl_matrix,free), m);
+}
+
+static VALUE FUNCTION(rb_tensor,2vector)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  GSL_TYPE(gsl_vector) *v = NULL;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  v = (GSL_TYPE(gsl_vector) *) FUNCTION(tensor,2vector)(t->tensor);
+  return Data_Wrap_Struct(QUALIFIED_VIEW(cgsl_vector,view), 0, FUNCTION(gsl_vector,free), v);
+}
+
+static VALUE FUNCTION(rb_tensor,to_v)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  GSL_TYPE(gsl_vector) *v;
+
+  v = FUNCTION(gsl_vector,alloc)(t->tensor->size);
+  memcpy(v->data, t->tensor->data, sizeof(BASE)*v->size);
+  return Data_Wrap_Struct(GSL_TYPE(cgsl_vector), 0, FUNCTION(gsl_vector,free), v);
+}
+
+/*
+  Creates a subtensor slicing the existing tensor.
+  NOTE: no new data region is malloced.
+  t: Tensor
+  rank: rank of the tensor created
+*/
+GSL_TYPE(tensor) FUNCTION(tensor,subtensor)(const GSL_TYPE(tensor) *t,
+						     const unsigned int rank,
+						     size_t *indices)
+{
+  GSL_TYPE(tensor) tnew;
+  size_t position;
+  tnew.rank = rank;
+  tnew.dimension = t->dimension;
+  tnew.size =  quick_pow(t->dimension, rank);
+  position = FUNCTION(tensor,position)(indices, t);
+  if (position >= t->size)
+    rb_raise(rb_eRangeError, "wrong indices given");
+  tnew.data = t->data + position;
+  return tnew;
+}
+
+static VALUE FUNCTION(rb_tensor,subtensor)(int argc, VALUE *argv, VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t, *tnew;
+  unsigned int rank;
+  size_t n;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  /* n: number of indices given */
+  rbgsl_tensor_get_indices(argc, argv, t->indices, &n);  
+  rank = t->tensor->rank - n; 
+  tnew = ALLOC(GSL_TYPE(rbgsl_tensor));
+  tnew->tensor = (GSL_TYPE(tensor)*) malloc(sizeof(GSL_TYPE(tensor)));
+  *(tnew->tensor) = FUNCTION(tensor,subtensor)(t->tensor, rank, t->indices->data);
+  if (rank == 0) 
+    tnew->indices = gsl_permutation_alloc(1);
+  else
+    tnew->indices = gsl_permutation_alloc(rank);
+  return Data_Wrap_Struct(QUALIFIED_VIEW(cgsl_tensor,view), 0, FUNCTION(rbgsl_tensor,free2), tnew);	       
+}
+
+#ifdef BASE_DOUBLE
+#define SHOW_ELM 6
+#define PRINTF_FORMAT "%4.3e "
+#else
+#define SHOW_ELM 15
+#define PRINTF_FORMAT "%d "
+#endif
+static VALUE FUNCTION(rb_tensor,to_s)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  QUALIFIED_VIEW(gsl_matrix,view) matrix;
+  QUALIFIED_VIEW(gsl_vector,view) vector;
+  GSL_TYPE(gsl_matrix) *m;
+  GSL_TYPE(gsl_vector) *v;
+  char buf[16];
+  size_t i, j;
+  VALUE str;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  str = rb_str_new2("[ ");
+  switch (t->tensor->rank) {
+  case 2:
+    matrix.matrix.data = t->tensor->data;
+    matrix.matrix.size1 = t->tensor->dimension;
+    matrix.matrix.size2 = t->tensor->dimension;
+    matrix.matrix.tda = t->tensor->dimension;
+    matrix.matrix.block = 0;
+    matrix.matrix.owner = 0;
+    m = &(matrix.matrix);
+    for (i = 0; i < m->size1; i++) {
+      if (i != 0) {
+	strcpy(buf, "  ");
+	rb_str_cat(str, buf, strlen(buf));
+      }
+      for (j = 0; j < m->size2; j++) {
+	sprintf(buf, PRINTF_FORMAT, FUNCTION(gsl_matrix,get)(m, i, j));
+	rb_str_cat(str, buf, strlen(buf));
+	if (j == SHOW_ELM) {
+	  strcpy(buf, "... ");
+	  rb_str_cat(str, buf, strlen(buf));
+	  break;
+	}
+      }
+      if (i == 6) {
+	strcpy(buf, "\n  ... ]");
+	rb_str_cat(str, buf, strlen(buf));
+	break;
+      }
+      if (i == m->size1 - 1) {
+	strcpy(buf, "]");
+	rb_str_cat(str, buf, strlen(buf));
+      } else {
+	strcpy(buf, "\n");
+	rb_str_cat(str, buf, strlen(buf));
+      }
+    }
+    return str;
+    break;
+  default:
+    vector.vector.data = t->tensor->data;
+    vector.vector.stride = 1;
+    vector.vector.size = t->tensor->size;
+    vector.vector.owner = 0;
+    vector.vector.block = 0;
+    v = &(vector.vector);
+    sprintf(buf,  PRINTF_FORMAT, FUNCTION(gsl_vector,get)(v, 0));
+    rb_str_cat(str, buf, strlen(buf));
+    for (i = 1; i < v->size; i++) {
+      sprintf(buf,  PRINTF_FORMAT, FUNCTION(gsl_vector,get)(v, i));
+      rb_str_cat(str, buf, strlen(buf));
+      if (i == SHOW_ELM && i != v->size-1) {
+        strcpy(buf, "... ");
+        rb_str_cat(str, buf, strlen(buf));
+        break;
+      }
+    }
+    sprintf(buf, "]");
+    rb_str_cat(str, buf, strlen(buf));
+    return str;
+    break;
+  }
+}
+#undef SHOW_ELM
+#undef PRINTF_FORMAT
+
+static VALUE FUNCTION(rb_tensor,inspect)(VALUE obj)
+{
+  VALUE str;
+  char buf[64];
+  sprintf(buf, "%s\n", rb_class2name(CLASS_OF(obj)));
+  str = rb_str_new2(buf);
+  return rb_str_concat(str, FUNCTION(rb_tensor,to_s)(obj));
+}
+
+VALUE FUNCTION(rb_tensor,equal)(int argc, VALUE *argv, VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *a, *b;
+  GSL_TYPE(gsl_matrix) *m;
+  GSL_TYPE(gsl_vector) *v;
+  VALUE other;
+  double eps = 1e-10;
+  size_t i;
+  switch (argc) {
+  case 2:
+    other = argv[0];
+    eps = NUM2DBL(argv[1]);
+    break;
+  case 1:
+    other = argv[0];
+    break;
+  default:
+    rb_raise(rb_eArgError, "wrong number of arguments (%d for 1 or 2)", argc);
+  }
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), a);
+  if (TEN_P(other)) {
+    Data_Get_Struct(other, GSL_TYPE(rbgsl_tensor), b);
+    if (a->tensor->rank != b->tensor->rank) return Qfalse;
+    if (a->tensor->dimension != b->tensor->dimension) return Qfalse;
+    if (a->tensor->size != b->tensor->size) return Qfalse;
+    for (i = 0; i < a->tensor->size; i++)
+      if (fabs(a->tensor->data[i]-b->tensor->data[i]) > eps) 
+	return Qfalse;
+    return Qtrue;
+  } else if (MAT_P(other)) {
+    if (a->tensor->rank != 2) return Qfalse;
+    Data_Get_Struct(other, GSL_TYPE(gsl_matrix), m);
+    if (a->tensor->dimension != m->size1 || a->tensor->dimension != m->size2)
+      return Qfalse;
+    for (i = 0; i < a->tensor->size; i++)
+      if (fabs(a->tensor->data[i]-m->data[i]) > eps) 
+	return Qfalse;
+    return Qtrue;
+  } else if (VEC_P(other)) {
+    Data_Get_Struct(other, GSL_TYPE(gsl_vector), v);
+    if (a->tensor->size != v->size) return Qfalse;
+    for (i = 0; i < a->tensor->size; i++)
+      if (fabs(a->tensor->data[i]-v->data[i]) > eps) 
+	return Qfalse;
+    return Qtrue;
+  } else {
+    rb_raise(rb_eTypeError, "wrong argument type %s (Tensor, Matrix or Vector expected)", 
+	     rb_class2name(CLASS_OF(other)));
+  }
+}
+
+static VALUE FUNCTION(rb_tensor,uplus)(VALUE obj)
+{
+  return obj;
+}
+
+static VALUE FUNCTION(rb_tensor,uminus)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t, *tnew;
+  size_t i;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  tnew = FUNCTION(rbgsl_tensor,copy)(t);
+  for (i = 0; i < tnew->tensor->size; i++)
+    tnew->tensor->data[i] *= -1;
+  return Data_Wrap_Struct(GSL_TYPE(cgsl_tensor), 0, FUNCTION(rbgsl_tensor,free), tnew);
+}
+
+static VALUE FUNCTION(rb_tensor,coerce)(VALUE obj, VALUE other)
+{
+  GSL_TYPE(rbgsl_tensor) *t, *tnew;
+  VALUE tt;
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  switch (TYPE(other)) {
+  case T_FLOAT:
+  case T_FIXNUM:
+  case T_BIGNUM:
+    tnew = FUNCTION(rbgsl_tensor,alloc)(t->tensor->rank, t->tensor->dimension);
+    FUNCTION(tensor,set_all)(tnew->tensor, NUMCONV(other));
+    tt = Data_Wrap_Struct(GSL_TYPE(cgsl_tensor), 0, FUNCTION(rbgsl_tensor,free), tnew);
+    return rb_ary_new3(2, tt, obj);
+    break;
+  default:
+    rb_raise(rb_eRuntimeError, "undefined operation with %s", 
+	     rb_class2name(CLASS_OF(other)));
+    break;
+  }
+}
+
+static VALUE FUNCTION(rb_tensor,info)(VALUE obj)
+{
+  GSL_TYPE(rbgsl_tensor) *t;
+  char buf[256];
+  Data_Get_Struct(obj, GSL_TYPE(rbgsl_tensor), t);
+  sprintf(buf, "Class:      %s\n", rb_class2name(CLASS_OF(obj)));
+  //  sprintf(buf, "%sSuperClass: %s\n", buf, rb_class2name(RCLASS(CLASS_OF(obj))->super));
+  sprintf(buf, "%sRank:       %d\n", buf, (int) t->tensor->rank);
+  sprintf(buf, "%sDimension:  %d\n", buf, (int) t->tensor->dimension);
+  sprintf(buf, "%sSize:       %d\n", buf, (int) t->tensor->size);
+  return rb_str_new2(buf);
+}
+
+void FUNCTION(Init_tensor,init)(VALUE module)
+{
+#ifdef BASE_DOUBLE
+  cgsl_tensor = rb_define_class_under(module, "Tensor", cGSL_Object);
+  cgsl_tensor_int = rb_define_class_under(cgsl_tensor, "Int", cGSL_Object);
+  cgsl_tensor_view = rb_define_class_under(cgsl_tensor, "View", cgsl_tensor);
+  cgsl_tensor_int_view = rb_define_class_under(cgsl_tensor_int, "View",
+					       cgsl_tensor_int);
+  /*
+  cgsl_index = rb_define_class_under(cgsl_tensor, "Index", 
+  cgsl_permutation);*/
+#endif
+
+  rb_define_singleton_method(GSL_TYPE(cgsl_tensor), "new",
+			     FUNCTION(rb_tensor,new), -1);
+  rb_define_singleton_method(GSL_TYPE(cgsl_tensor), "[]",
+			     FUNCTION(rb_tensor,new), -1);
+  rb_define_singleton_method(GSL_TYPE(cgsl_tensor), "alloc",
+			     FUNCTION(rb_tensor,new), -1);
+  rb_define_singleton_method(GSL_TYPE(cgsl_tensor), "calloc",
+			     FUNCTION(rb_tensor,calloc), 2);
+  rb_define_singleton_method(GSL_TYPE(cgsl_tensor), "copy",
+			     FUNCTION(rb_tensor,copy_singleton), 1);
+  rb_define_singleton_method(GSL_TYPE(cgsl_tensor), "memcpy",
+			     FUNCTION(rb_tensor,memcpy_singleton), 2);
+  rb_define_singleton_method(GSL_TYPE(cgsl_tensor), "swap",
+			     FUNCTION(rb_tensor,swap_singleton), 2);
+
+  /*****/
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "copy",
+			     FUNCTION(rb_tensor,copy), 0);
+  rb_define_alias(GSL_TYPE(cgsl_tensor), "clone", "copy");
+  rb_define_alias(GSL_TYPE(cgsl_tensor), "duplicate", "copy");
+  rb_define_method(GSL_TYPE(cgsl_tensor), "set_zero",
+			     FUNCTION(rb_tensor,set_zero), 0);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "set_all",
+			     FUNCTION(rb_tensor,set_all), 1);
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "position",
+		   FUNCTION(rb_tensor,position), -1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "get",
+			     FUNCTION(rb_tensor,get), -1);
+  rb_define_alias(GSL_TYPE(cgsl_tensor), "[]", "get");
+  rb_define_method(GSL_TYPE(cgsl_tensor), "set",
+			     FUNCTION(rb_tensor,set), -1);
+  rb_define_alias(GSL_TYPE(cgsl_tensor), "[]=", "set");
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "fread",
+			     FUNCTION(rb_tensor,fread), 1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "fwrite",
+			     FUNCTION(rb_tensor,fwrite), 1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "fprintf",
+			     FUNCTION(rb_tensor,fprintf), -1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "printf",
+			     FUNCTION(rb_tensor,printf), -1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "fscanf",
+			     FUNCTION(rb_tensor,fscanf), 1);
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "swap_indices",
+			     FUNCTION(rb_tensor,swap_indices), 2);
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "max",
+			     FUNCTION(rb_tensor,max), 0);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "min",
+			     FUNCTION(rb_tensor,min), 0);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "minmax",
+			     FUNCTION(rb_tensor,minmax), 0);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "max_index",
+			     FUNCTION(rb_tensor,max_index), 0);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "min_index",
+			     FUNCTION(rb_tensor,min_index), 0);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "minmax_index",
+			     FUNCTION(rb_tensor,minmax_index), 0);
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "isnull",
+			     FUNCTION(rb_tensor,isnull), 0);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "isnull?",
+			     FUNCTION(rb_tensor,isnull2), 0);
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "add",
+			     FUNCTION(rb_tensor,add), 1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "sub",
+			     FUNCTION(rb_tensor,sub), 1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "mul_elements",
+			     FUNCTION(rb_tensor,mul_elements), 1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "div_elements",
+			     FUNCTION(rb_tensor,div_elements), 1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "add_constant",
+			     FUNCTION(rb_tensor,add_constant), 1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "add_diagonal",
+			     FUNCTION(rb_tensor,add_diagonal), 1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "scale",
+			     FUNCTION(rb_tensor,scale), 1);
+  rb_define_singleton_method(GSL_TYPE(cgsl_tensor), "product",
+			     FUNCTION(rb_tensor,product_singleton), 2);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "product",
+			     FUNCTION(rb_tensor,product), 1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "contract",
+			     FUNCTION(rb_tensor,contract), 2);
+
+  rb_define_alias(GSL_TYPE(cgsl_tensor), "+", "add");
+  rb_define_alias(GSL_TYPE(cgsl_tensor), "-", "sub");
+  /*  rb_define_alias(GSL_TYPE(cgsl_tensor), "*", "mul_elements");*/
+  rb_define_alias(GSL_TYPE(cgsl_tensor), "/", "div_elements");
+  rb_define_alias(GSL_TYPE(cgsl_tensor), "*", "product");
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "add!",
+			     FUNCTION(rb_tensor,add_bang), 1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "sub!",
+			     FUNCTION(rb_tensor,sub_bang), 1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "mul_elements!",
+			     FUNCTION(rb_tensor,mul_elements_bang), 1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "div_elements!",
+			     FUNCTION(rb_tensor,div_elements_bang), 1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "add_constant!",
+			     FUNCTION(rb_tensor,add_constant_bang), 1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "add_diagonal!",
+			     FUNCTION(rb_tensor,add_diagonal_bang), 1);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "scale!",
+			     FUNCTION(rb_tensor,scale_bang), 1);
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "+@",
+			     FUNCTION(rb_tensor,uplus), 0);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "-@",
+			     FUNCTION(rb_tensor,uminus), 0);
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "size",
+			     FUNCTION(rb_tensor,size), 0);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "rank",
+			     FUNCTION(rb_tensor,rank), 0);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "dimension",
+			     FUNCTION(rb_tensor,dimension), 0);
+  rb_define_alias(GSL_TYPE(cgsl_tensor), "dim", "dimension");
+  rb_define_method(GSL_TYPE(cgsl_tensor), "data",
+			     FUNCTION(rb_tensor,data), 0);
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "to_v",
+			     FUNCTION(rb_tensor,to_v), 0);
+  rb_define_alias(GSL_TYPE(cgsl_tensor), "to_gv", "to_v");
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "to_vector",
+			     FUNCTION(rb_tensor,2vector), 0);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "to_matrix",
+			     FUNCTION(rb_tensor,2matrix), 0);
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "subtensor",
+			     FUNCTION(rb_tensor,subtensor), -1);
+  rb_define_alias(GSL_TYPE(cgsl_tensor), "view", "subtensor");
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "to_s",
+			     FUNCTION(rb_tensor,to_s), 0);
+  rb_define_method(GSL_TYPE(cgsl_tensor), "inspect",
+			     FUNCTION(rb_tensor,inspect), 0);
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "equal?",
+			     FUNCTION(rb_tensor,equal), -1);
+  rb_define_alias(GSL_TYPE(cgsl_tensor), "==", "equal?");
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "coerce",
+			     FUNCTION(rb_tensor,coerce), 1);
+
+  rb_define_method(GSL_TYPE(cgsl_tensor), "info", 
+		   FUNCTION(rb_tensor,info), 0);
+}
+
+#undef NUMCONV
+#undef C_TO_VALUE
+#undef CHECK_TEN
+#undef TEN_P
+#undef VEC_P
+#undef MAT_P
+
+#endif