update v_arry for more vector-like syntax

library/search_generate.cc

@@ -308,7 +308,7 @@ class Generator : public SearchTask<input, output>
       }
       ex("w=" + tmp);
 
-      ref.erase();
+      ref.clear();
 
       /*
       vector<char>& best = ied.next();
@@ -495,4 +495,3 @@ int main(int argc, char *argv[])
   //predict();
   //run_easy();
 }
-

vowpalwabbit/autolink.cc

@@ -30,7 +30,7 @@ void predict_or_learn(autolink& b, LEARNER::single_learner& base, example& ec)
     base.predict(ec);
 
   ec.total_sum_feat_sq -= fs.sum_feat_sq;
-  fs.erase();
+  fs.clear();
   ec.indices.pop();
 }
 

vowpalwabbit/bfgs.cc

@@ -668,7 +668,7 @@ int process_pass(vw& all, bfgs& b)
       b.net_time = (int) (1000.0 * (b.t_end_global.time - b.t_start_global.time) + (b.t_end_global.millitm - b.t_start_global.millitm));
       if (!all.quiet)
         fprintf(stderr, "%-10s\t%-10.5f\t%-.5f\n", "", d_mag, b.step_size);
-      b.predictions.erase();
+      b.predictions.clear();
       update_weight(all, b.step_size);
     }
   }
@@ -727,7 +727,7 @@ int process_pass(vw& all, bfgs& b)
           fprintf(stderr, "%-10s\t%-10s\t(revise x %.1f)\t%-.5f\n",
                   "","",ratio,
                   new_step);
-        b.predictions.erase();
+        b.predictions.clear();
         update_weight(all, (float)(-b.step_size+new_step));
         b.step_size = (float)new_step;
         zero_derivative(all);
@@ -775,7 +775,7 @@ int process_pass(vw& all, bfgs& b)
           b.net_time = (int) (1000.0 * (b.t_end_global.time - b.t_start_global.time) + (b.t_end_global.millitm - b.t_start_global.millitm));
           if (!all.quiet)
             fprintf(stderr, "%-10s\t%-10.5f\t%-.5f\n", "", d_mag, b.step_size);
-          b.predictions.erase();
+          b.predictions.clear();
           update_weight(all, b.step_size);
         }
       }
@@ -811,7 +811,7 @@ int process_pass(vw& all, bfgs& b)
 
       float d_mag = direction_magnitude(all);
 
-      b.predictions.erase();
+      b.predictions.clear();
       update_weight(all, b.step_size);
       ftime(&b.t_end_global);
       b.net_time = (int) (1000.0 * (b.t_end_global.time - b.t_start_global.time) + (b.t_end_global.millitm - b.t_start_global.millitm));

vowpalwabbit/cache.cc

@@ -51,7 +51,7 @@ size_t read_cached_tag(io_buf& cache, example* ae)
   if (buf_read(cache, c, tag_size) < tag_size)
     return 0;
 
-  ae->tag.erase();
+  ae->tag.clear();
   push_many(ae->tag, c, tag_size);
   return tag_size+sizeof(tag_size);
 }

vowpalwabbit/cb.cc

@@ -28,7 +28,7 @@ bool is_test_label(CB::label& ld)
 char* bufread_label(CB::label* ld, char* c, io_buf& cache)
 {
   size_t num = *(size_t *)c;
-  ld->costs.erase();
+  ld->costs.clear();
   c += sizeof(size_t);
   size_t total = sizeof(cb_class)*num;
   if (buf_read(cache, c, total) < total)
@@ -49,7 +49,7 @@ char* bufread_label(CB::label* ld, char* c, io_buf& cache)
 size_t read_cached_label(shared_data*, void* v, io_buf& cache)
 {
   CB::label* ld = (CB::label*) v;
-  ld->costs.erase();
+  ld->costs.clear();
   char *c;
   size_t total = sizeof(size_t);
   if (buf_read(cache, c, total) < total)
@@ -87,7 +87,7 @@ void cache_label(void* v, io_buf& cache)
 void default_label(void* v)
 {
   CB::label* ld = (CB::label*) v;
-  ld->costs.erase();
+  ld->costs.clear();
 }
 
 void delete_label(void* v)

vowpalwabbit/cb_explore.cc

@@ -51,7 +51,7 @@ void predict_or_learn_first(cb_explore& data, single_learner& base, example& ec)
   else
     base.predict(ec);
 
-  probs.erase();
+  probs.clear();
   if(data.tau > 0)
   {
     float prob = 1.f/(float)data.cbcs.num_actions;
@@ -77,7 +77,7 @@ void predict_or_learn_greedy(cb_explore& data, single_learner& base, example& ec
   // TODO: pointers are copied here. What happens if base.learn/base.predict re-allocs?
   // ec.pred.a_s = probs; will restore the than free'd memory
   action_scores probs = ec.pred.a_s;
-  probs.erase();
+  probs.clear();
 
   if (is_learn)
     base.learn(ec);
@@ -98,7 +98,7 @@ void predict_or_learn_bag(cb_explore& data, single_learner& base, example& ec)
 {
   //Randomize over predictions from a base set of predictors
   action_scores probs = ec.pred.a_s;
-  probs.erase();
+  probs.clear();
 
   for(uint32_t i = 0; i < data.cbcs.num_actions; i++)
     probs.push_back({i,0.});
@@ -123,7 +123,7 @@ void predict_or_learn_bag(cb_explore& data, single_learner& base, example& ec)
 void get_cover_probabilities(cb_explore& data, single_learner& base, example& ec, v_array<action_score>& probs)
 {
   float additive_probability = 1.f / (float)data.cover_size;
-  data.preds.erase();
+  data.preds.clear();
 
   for(uint32_t i = 0; i < data.cbcs.num_actions; i++)
     probs.push_back({i,0.});
@@ -157,8 +157,8 @@ void predict_or_learn_cover(cb_explore& data, single_learner& base, example& ec)
   uint32_t num_actions = data.cbcs.num_actions;
 
   action_scores probs = ec.pred.a_s;
-  probs.erase();
-  data.cs_label.costs.erase();
+  probs.clear();
+  data.cs_label.costs.clear();
 
   for (uint32_t j = 0; j < num_actions; j++)
     data.cs_label.costs.push_back({FLT_MAX,j+1,0.,0.});
@@ -185,7 +185,7 @@ void predict_or_learn_cover(cb_explore& data, single_learner& base, example& ec)
     //Now update oracles
 
     //1. Compute loss vector
-    data.cs_label.costs.erase();
+    data.cs_label.costs.clear();
     float norm = min_prob * num_actions;
     ec.l.cb = data.cb_label;
     data.cbcs.known_cost = get_observed_cost(data.cb_label);

vowpalwabbit/cb_explore_adf.cc

@@ -153,12 +153,12 @@ void predict_or_learn_bag(cb_explore_adf& data, multi_learner& base, multi_ex& e
     num_actions--;
   if (num_actions == 0)
   {
-    preds.erase();
+    preds.clear();
     return;
   }
 
   data.action_probs.resize(num_actions);
-  data.action_probs.erase();
+  data.action_probs.clear();
   for (uint32_t i = 0; i < num_actions; i++)
     data.action_probs.push_back({ i,0. });
   vector<uint32_t>& top_actions = *data.top_actions;
@@ -218,7 +218,7 @@ void predict_or_learn_cover(cb_explore_adf& data, multi_learner& base, multi_ex&
   float additive_probability = 1.f / (float)data.cover_size;
   float min_prob = min(1.f / num_actions, 1.f / (float)sqrt(data.counter * num_actions));
   v_array<action_score>& probs = data.action_probs;
-  probs.erase();
+  probs.clear();
   for(uint32_t i = 0; i < num_actions; i++)
     probs.push_back({i,0.});
 
@@ -232,7 +232,7 @@ void predict_or_learn_cover(cb_explore_adf& data, multi_learner& base, multi_ex&
     //Create costs of each action based on online cover
     if (is_learn)
     {
-      data.cs_labels_2.costs.erase();
+      data.cs_labels_2.costs.clear();
       if (shared > 0)
         data.cs_labels_2.costs.push_back(data.cs_labels.costs[0]);
       for (uint32_t j = 0; j < num_actions; j++)

vowpalwabbit/cbify.cc

@@ -101,7 +101,7 @@ void predict_or_learn(cbify& data, single_learner& base, example& ec)
 {
   //Store the multiclass input label
   MULTICLASS::label_t ld = ec.l.multi;
-  data.cb_label.costs.erase();
+  data.cb_label.costs.clear();
   ec.l.cb = data.cb_label;
   ec.pred.a_s = data.a_s;
 
@@ -125,7 +125,7 @@ void predict_or_learn(cbify& data, single_learner& base, example& ec)
   data.cb_label.costs.push_back(cl);
   ec.l.cb = data.cb_label;
   base.learn(ec);
-  data.a_s.erase();
+  data.a_s.clear();
   data.a_s = ec.pred.a_s;
   ec.l.multi = ld;
   ec.pred.multiclass = chosen_action + 1;

vowpalwabbit/cost_sensitive.cc

@@ -42,7 +42,7 @@ bool is_test_label(label& ld)
 char* bufread_label(label* ld, char* c, io_buf& cache)
 {
   size_t num = *(size_t *)c;
-  ld->costs.erase();
+  ld->costs.clear();
   c += sizeof(size_t);
   size_t total = sizeof(wclass)*num;
   if (buf_read(cache, c, (int)total) < total)
@@ -63,7 +63,7 @@ char* bufread_label(label* ld, char* c, io_buf& cache)
 size_t read_cached_label(shared_data*, void* v, io_buf& cache)
 {
   label* ld = (label*) v;
-  ld->costs.erase();
+  ld->costs.clear();
   char *c;
   size_t total = sizeof(size_t);
   if (buf_read(cache, c, (int)total) < total)
@@ -101,7 +101,7 @@ void cache_label(void* v, io_buf& cache)
 void default_label(void* v)
 {
   label* ld = (label*) v;
-  ld->costs.erase();
+  ld->costs.clear();
 }
 
 void delete_label(void* v)
@@ -131,7 +131,7 @@ bool substring_eq(substring ss, const char* str)
 void parse_label(parser* p, shared_data*sd, void* v, v_array<substring>& words)
 {
   label* ld = (label*)v;
-  ld->costs.erase();
+  ld->costs.clear();
 
   // handle shared and label first
   if (words.size() == 1)

vowpalwabbit/csoaa.cc

@@ -237,7 +237,7 @@ void unsubtract_example(example *ec)
   features& fs = ec->feature_space[wap_ldf_namespace];
   ec->num_features -= fs.size();
   ec->total_sum_feat_sq -= fs.sum_feat_sq;
-  fs.erase();
+  fs.clear();
   ec->indices.decr();
 }
 
@@ -445,8 +445,8 @@ void do_actual_learning(ldf& data, single_learner& base, multi_ex& ec_seq_all)
   uint32_t predicted_K = start_K;
   if(data.rank)
   {
-    data.a_s.erase();
-    data.stored_preds.erase();
+    data.a_s.clear();
+    data.stored_preds.clear();
     if (start_K > 0)
       data.stored_preds.push_back(ec_seq[0]->pred.a_s);
     for (uint32_t k=start_K; k<K; k++)
@@ -486,7 +486,7 @@ void do_actual_learning(ldf& data, single_learner& base, multi_ex& ec_seq_all)
 
   if(data.rank)
   {
-    data.stored_preds[0].erase();
+    data.stored_preds[0].clear();
     if (start_K > 0)
     {
       ec_seq[0]->pred.a_s = data.stored_preds[0];

vowpalwabbit/ect.cc

@@ -236,7 +236,7 @@ void ect_train(ect& e, single_learner& base, example& ec)
 
   simple_temp.initial = 0.;
 
-  e.tournaments_won.erase();
+  e.tournaments_won.clear();
 
   uint32_t id = e.directions[mc.label - 1].winner;
   bool left = e.directions[id].left == mc.label - 1;

vowpalwabbit/ezexample.h

@@ -36,10 +36,10 @@ class ezexample
   example* get_new_example()
   { example* new_ec = VW::new_unused_example(*vw_par_ref);
     vw_par_ref->p->lp.default_label(&new_ec->l);
-    new_ec->tag.erase();
-    new_ec->indices.erase();
+    new_ec->tag.clear();
+    new_ec->indices.clear();
     for (size_t i=0; i<256; i++)
-      new_ec->feature_space[i].erase();
+      new_ec->feature_space[i].clear();
 
     new_ec->ft_offset = 0;
     new_ec->num_features = 0;
@@ -117,7 +117,7 @@ class ezexample
     for (auto ecc : example_copies)
       if (ecc->in_use && VW::is_ring_example(*vw_par_ref, ec))
         VW::finish_example(*vw_par_ref, *ecc);
-    example_copies.erase();
+    example_copies.clear();
     free(example_copies.begin());
   }
 
@@ -132,7 +132,7 @@ class ezexample
   void addns(char c)
   { if (ensure_ns_exists(c)) return;
 
-    ec->feature_space[(int)c].erase();
+    ec->feature_space[(int)c].clear();
     past_seeds.push_back(current_seed);
     current_ns = c;
     str[0] = c;
@@ -147,7 +147,7 @@ class ezexample
     else
     { if (ns_exists[(int)current_ns])
       { ec->total_sum_feat_sq -= ec->feature_space[(int)current_ns].sum_feat_sq;
-        ec->feature_space[(int)current_ns].erase();
+        ec->feature_space[(int)current_ns].clear();
         ec->num_features -= ec->feature_space[(int)current_ns].size();
 
         ns_exists[(int)current_ns] = false;
@@ -269,7 +269,7 @@ class ezexample
       for (auto ecc : example_copies)
         if (ecc->in_use)
           VW::finish_example(*vw_par_ref, *ecc);
-      example_copies.erase();
+      example_copies.clear();
     }
   }
 

vowpalwabbit/feature_group.h

@@ -251,7 +251,7 @@ struct features
   }
 
   // if one wants to add proper destructor for features, make sure to update ezexample_predict::~ezexample_predict();
-  // ~features() { ... } 
+  // ~features() { ... }
 
   inline size_t size() const { return values.size(); }
 
@@ -271,11 +271,11 @@ struct features
 
   iterator end() { return iterator(values.end(), indicies.end()); }
 
-  void erase()
+  void clear()
   { sum_feat_sq = 0.f;
-    values.erase();
-    indicies.erase();
-    space_names.erase();
+    values.clear();
+    indicies.clear();
+    space_names.clear();
   }
 
   void truncate_to(const features_value_iterator& pos)
@@ -351,4 +351,3 @@ struct features
     sum_feat_sq = src.sum_feat_sq;
   }
 };
-

vowpalwabbit/gd_mf.cc

@@ -104,7 +104,7 @@ template<class T> float mf_predict(gdmf& d, example& ec, T& weights)
   }
 
   // clear stored predictions
-  d.scalars.erase();
+  d.scalars.clear();
 
   float linear_prediction = 0.;
   // linear terms