Drop C++11 specification in Makevars (#42)

* Drop C++11 specification in Makevars * Drop C++11 specification in DESCRIPTION * NEWS * std::random_shuffle > fisher-yates
bnosac · Feb 22, 2024 · ff56ab0 · ff56ab0
1 parent 747fec4
commit ff56ab0
Show file tree

Hide file tree

Showing 4 changed files with 13 additions and 3 deletions.
diff --git a/DESCRIPTION b/DESCRIPTION
@@ -23,4 +23,3 @@ Imports: Rcpp (>= 0.11.5), utils, graphics, stats
 Suggests: udpipe, data.table
 LinkingTo: Rcpp, BH
 RoxygenNote: 7.1.2
-SystemRequirements: C++11
diff --git a/NEWS.md b/NEWS.md
@@ -2,6 +2,8 @@
 
 - Docs of starspace and starspace_dictionary: fix use of {} in itemize items
 - Remove compliance.cpp and compliance.h and move the abort/exit statements directly in the cpp files
+- Drop C++11 specification in Makevars 
+  - hereby replacing use of std::random_shuffle with the Fisher-Yates Shuffle Algorithm as indicated at https://gallery.rcpp.org/articles/stl-random-shuffle/ - as data is reshuffled, model training will give different results compared to previous versions due to different randomisations of the training data
 
 ## CHANGES IN ruimtehol VERSION 0.3.1
 

diff --git a/src/Makevars b/src/Makevars
@@ -1,4 +1,3 @@
-CXX_STD = CXX11
 PKG_LIBS = -pthread
 PKG_CPPFLAGS = -pthread -DSTRICT_R_HEADERS -DBOOST_NO_AUTO_PTR -I./Starspace/src
 

diff --git a/src/Starspace/src/model.cpp b/src/Starspace/src/model.cpp
@@ -228,7 +228,17 @@ Real EmbedModel::train(shared_ptr<InternDataHandler> data,
     int i = 0;
     for (auto& idx: indices) idx = i++;
   }
-  std::random_shuffle(indices.begin(), indices.end(), randWrapper);
+  // std::random_shuffle(indices.begin(), indices.end(), randWrapper);
+  // Rcpp::IntegerVector indices_r = Rcpp::sample(numSamples, numSamples, false, R_NilValue, true) - 1;
+  // for (auto ii = 0; ii < numSamples; ii++) {
+  //   indices[ii] = indices_r[ii];
+  // }
+  // Fisher-Yates Shuffle Algorithm - https://gallery.rcpp.org/articles/stl-random-shuffle/
+  int j;
+  for (int i = 0; i < numSamples - 1; i++) {
+    j = i + randWrapper(numSamples - i);
+    std::swap(indices[i], indices[j]);
+  }
 
   // Compute word negatives
   if (args_->trainMode == 5 || args_->trainWord) {