ext/main.cpp in feedbackmine-libsvm-ruby-0.1.0 vs ext/main.cpp in feedbackmine-libsvm-ruby-0.1.1

@@ -49,19 +49,38 @@
   struct svm_node *n;
   int i;
   int len = RARRAY(xs)->len;
   n = (struct svm_node *) calloc(sizeof(struct svm_node), len+1);
   for (i = 0; i < len; ++i) {
+    n[i].value = NUM2DBL(rb_ary_entry(xs, i));
+    n[i].index = i;
+  }
+  n[i].value = 0;
+  n[i].index = -1;
+
+  return n;
+}
+
+/*value is harcoded as 1, array is sorted features*/
+struct svm_node *rubyArrayToNodelist2(VALUE xs)
+{
+  //struct obstack xso;
+  struct svm_node *n;
+  int i;
+  int len = RARRAY(xs)->len;
+  n = (struct svm_node *) calloc(sizeof(struct svm_node), len+1);
+  for (i = 0; i < len; ++i) {
     n[i].value = 1;
     n[i].index = NUM2DBL(rb_ary_entry(xs, i));
   }
   n[i].value = 0;
   n[i].index = -1;
 
   return n;
 }
 
+
 /*
  * Serializes an SVMParameter object
  */
 VALUE svmpa_svm_dump(VALUE self, VALUE limit)
 {
@@ -537,10 +556,24 @@
   result = svm_predict(rp->m, x);
   free(x);
   return rb_float_new(result);
 }
 
+/*
+ * Predicts a value (regression or classification) based on an input vector
+ */
+static VALUE svmpm_predict2(VALUE self, VALUE xs)
+{
+  double result;
+  struct RSVM_Model *rp;
+  Data_Get_Struct(self, struct RSVM_Model, rp);
+  struct svm_node *x = rubyArrayToNodelist2(xs);
+  result = svm_predict(rp->m, x);
+  free(x);
+  return rb_float_new(result);
+}
+
 static VALUE svmpm_save(VALUE self, VALUE filename)
 {
   int result;
   struct RSVM_Model *rp;
   Data_Get_Struct(self, struct RSVM_Model, rp);
@@ -617,10 +650,23 @@
   addExample(rp, yd, fini);
   if (rp->k == 0) rp->k = RARRAY(xs)->len;
   return Qnil;
 }
 
+static VALUE svmpr_addex2(VALUE self, VALUE y, VALUE xs)
+{
+  struct RSVM_Problem *rp;
+  struct svm_node *fini;
+  double yd;
+  Data_Get_Struct(self, struct RSVM_Problem, rp);
+  yd = NUM2DBL(y);
+  fini = rubyArrayToNodelist2(xs);
+  addExample(rp, yd, fini);
+  if (rp->k == 0) rp->k = RARRAY(xs)->len;
+  return Qnil;
+}
+
 /* To be removed in next version */
 struct svm_model
 {
 	svm_parameter param;	// parameter
 	int nr_class;		// number of classes, = 2 in regression/one class svm
@@ -661,10 +707,11 @@
 
   rb_define_singleton_method(cSVMProblem, "new", (VALUE (*) (...))svmpr_new, 0);
   rb_define_method(cSVMProblem, "initialize", (VALUE (*) (...))svmpr_init, 0);
   rb_define_method(cSVMProblem, "size", (VALUE (*) (...))svmpr_size, 0);
   rb_define_method(cSVMProblem, "addExample", (VALUE (*) (...))svmpr_addex, 2);
+  rb_define_method(cSVMProblem, "addExample2", (VALUE (*) (...))svmpr_addex2, 2);
 
   rb_define_const(mSVM, "C_SVC", INT2FIX(C_SVC));
   rb_define_const(mSVM, "NU_SVC", INT2FIX(NU_SVC));
   rb_define_const(mSVM, "ONE_CLASS", INT2FIX(ONE_CLASS));
   rb_define_const(mSVM, "EPSILON_SVR", INT2FIX(EPSILON_SVR));
@@ -715,9 +762,10 @@
   rb_define_singleton_method(cSVMModel, "_alloc", (VALUE (*) (...))svmpa_allocate, 1);
   */
 
   rb_define_singleton_method(cSVMModel, "new", (VALUE (*) (...))svmpm_new, 2);
   rb_define_method(cSVMModel, "predict", (VALUE (*) (...))svmpm_predict, 1);
+  rb_define_method(cSVMModel, "predict2", (VALUE (*) (...))svmpm_predict2, 1);
   rb_define_method(cSVMModel, "predict_probability", (VALUE (*) (...))svmpm_predict_probability, 1);
   rb_define_method(cSVMModel, "predict_values", (VALUE (*) (...))svmpm_predict_values, 1);
   rb_define_method(cSVMModel, "svcount", (VALUE (*) (...))svmpm_svcount, 0);
   rb_define_method(cSVMModel, "save", (VALUE (*) (...))svmpm_save, 1);
   rb_define_singleton_method(cSVMModel, "load", (VALUE (*) (...))svmpm_load, 1);