added RunningMean Class/Module

allows for fast and easy RunningMean Calculation in C++

Author: DavZim

DESCRIPTION

@@ -19,8 +19,11 @@ Imports:
     glue,
     shiny,
     plotly
+Depends:
+    Rcpp (>= 0.12.15)
 Remotes: ropenscilabs/webrockets
 LinkingTo: Rcpp
+RcppModules: RunningMeanModule
 VignetteBuilder: knitr
 Encoding: UTF-8
 RoxygenNote: 6.1.1

R/zzz.R

@@ -1,3 +1,6 @@
 .onUnload <- function(libpath) {
   library.dynam.unload("arduinor", libpath)
-}
+}
+
+Rcpp::loadModule("RunningMeanModule", TRUE)
+

src/RcppExports.cpp

@@ -65,12 +65,15 @@ BEGIN_RCPP
 END_RCPP
 }
 
+RcppExport SEXP _rcpp_module_boot_RunningMeanModule();
+
 static const R_CallMethodDef CallEntries[] = {
     {"_arduinor_ar_init", (DL_FUNC) &_arduinor_ar_init, 2},
     {"_arduinor_ar_is_open", (DL_FUNC) &_arduinor_ar_is_open, 1},
     {"_arduinor_ar_close", (DL_FUNC) &_arduinor_ar_close, 1},
     {"_arduinor_ar_read", (DL_FUNC) &_arduinor_ar_read, 4},
     {"_arduinor_ar_flush", (DL_FUNC) &_arduinor_ar_flush, 1},
+    {"_rcpp_module_boot_RunningMeanModule", (DL_FUNC) &_rcpp_module_boot_RunningMeanModule, 0},
     {NULL, NULL, 0}
 };
 

src/RunningMean.cpp

@@ -0,0 +1,50 @@
+#include "RunningMean.h"
+
+
+RunningMean::RunningMean() : n(0) {
+  stop("n (size of window) must be supplied");
+}
+
+// Initiates a new Running Mean class of size n (constant)
+RunningMean::RunningMean(int n) : n(n) {
+  if (n <= 0) stop("n (size of window) must be supplied");
+  deq.resize(n);
+  n_values = 0;
+  sum = 0;
+}
+
+// inserts a new value into the internal deque. Deletes the oldest value
+void RunningMean::insert(double v) {
+  if (n == 0) stop("n (size of window) must be larger than zero");
+  
+  // TODO: What happens if we insert a missing value? 
+  if (!R_IsNA(v) && !R_IsNaN(v)) {
+    n_values++;
+    sum += v - deq.back();
+    deq.push_front(v);
+    deq.pop_back();
+  }
+}
+
+// Returns the running mean of the values in the deque
+// if the size is 0, return NA, if the number of already inserted values (n_values)
+// is smaller than the size of the deque, use n_values. This allows the running mean 
+// to return values even if the number of values is smaller than window lengths
+// could also be set to NA if needed...
+double RunningMean::get_mean() {
+  if (n == 0) return NA_REAL;
+  if (n_values < n) return(sum / n_values);
+  return sum / n;
+}
+
+// prints the values of the deque
+void RunningMean::show() {
+  Rcout << "RunningMean object <" << static_cast<const void*>(this) << 
+    "> window size " << n << ", number of inserts " << n_values << "\n" << 
+      deq[0];
+  for (int i = 1; i < n; ++i) {
+    Rcout << " " << deq[i];
+  }
+  Rcout << "\n";
+}
+

src/RunningMean.h

@@ -0,0 +1,55 @@
+#include <Rcpp.h>
+using namespace Rcpp;
+
+/*
+ * Example R Code
+ * 
+ * # Initiate the Class with a window-length of size 3
+ * obj <- arduinor:::RunningMean$new(3)
+ * 
+ * # insert a new value
+ * obj$insert(1.03)
+ * obj$insert(1.05)
+ * obj$insert(1.04)
+ * 
+ * # inspect the object 
+ * obj
+ * 
+ * # get the running mean
+ * obj$get_mean()
+ * 
+ * # add another value, effectively removing the first insert (1.03)
+ * obj$insert(1.06)
+ * obj
+ * obj$get_mean()
+ * 
+ */
+
+// defines the RunningMean class
+class RunningMean{
+public:
+  RunningMean();
+  RunningMean(int n);
+  
+  void insert(double val);
+  double get_mean();
+  void show();
+  
+private:
+  std::deque<double> deq;
+  const int n;
+  int n_values;
+  double sum;
+};
+
+// Exposes the class
+RCPP_MODULE(RunningMeanModule) {
+  class_<RunningMean>("RunningMean")
+    .default_constructor("Default constructor")
+    .constructor<int>("Constructor with an argument") 
+    .method("insert", &RunningMean::insert)
+    .method("get_mean", &RunningMean::get_mean)
+    .method("show", &RunningMean::show)
+    ;
+}
+