Cleaning up url_index_private_data and in_memory_url_index_types.

components/omnibox/browser/url_index_private_data.cc

 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "components/omnibox/browser/url_index_private_data.h"
 
 #include <stdint.h>
 
 #include <functional>
 #include <iterator>
@@ skipping 191 matching lines @@
     if (lower_words.empty())
       continue;
     HistoryIDSet history_id_set = HistoryIDSetFromWords(lower_words);
     pre_filter_item_count_ += history_id_set.size();
     // Trim the candidate pool if it is large. Note that we do not filter out
     // items that do not contain the search terms as proper substrings --
     // doing so is the performance-costly operation we are trying to avoid in
     // order to maintain omnibox responsiveness.
     const size_t kItemsToScoreLimit = 500;
     if (history_id_set.size() > kItemsToScoreLimit) {
-      HistoryIDVector history_ids;
-      std::copy(history_id_set.begin(), history_id_set.end(),
-                std::back_inserter(history_ids));
+      HistoryIDVector history_ids(history_id_set.begin(), history_id_set.end());
+
       // Trim down the set by sorting by typed-count, visit-count, and last
       // visit.
       HistoryItemFactorGreater item_factor_functor(history_info_map_);
-      std::partial_sort(history_ids.begin(),
-                        history_ids.begin() + kItemsToScoreLimit,
-                        history_ids.end(), item_factor_functor);
-      history_id_set.clear();
-      std::copy(history_ids.begin(), history_ids.begin() + kItemsToScoreLimit,
-                std::inserter(history_id_set, history_id_set.end()));
+      std::nth_element(history_ids.begin(),
+                       history_ids.begin() + kItemsToScoreLimit,
+                       history_ids.end(), item_factor_functor);
+      history_id_set = {history_ids.begin(),
+                        history_ids.begin() + kItemsToScoreLimit};
       post_filter_item_count_ += history_id_set.size();
     } else {
       post_filter_item_count_ += pre_filter_item_count_;
     }
     ScoredHistoryMatches temp_scored_items;
     HistoryIdSetToScoredMatches(history_id_set, lower_raw_string,
                                 template_url_service, bookmark_model,
                                 &temp_scored_items);
     scored_items.insert(scored_items.end(), temp_scored_items.begin(),
                         temp_scored_items.end());
@@ skipping 243 matching lines @@
   //    post_scoring_item_count_
 }
 
 bool URLIndexPrivateData::Empty() const {
   return history_info_map_.empty();
 }
 
 void URLIndexPrivateData::Clear() {
   last_time_rebuilt_from_history_ = base::Time();
   word_list_.clear();
-  available_words_.clear();
+  available_words_ = {};
   word_map_.clear();
   char_word_map_.clear();
   word_id_history_map_.clear();
   history_id_word_map_.clear();
   history_info_map_.clear();
   word_starts_map_.clear();
 }
 
 URLIndexPrivateData::~URLIndexPrivateData() {}
 
 HistoryIDSet URLIndexPrivateData::HistoryIDSetFromWords(
     const String16Vector& unsorted_words) {
   SCOPED_UMA_HISTOGRAM_TIMER("Omnibox.HistoryQuickHistoryIDSetFromWords");
   // Break the terms down into individual terms (words), get the candidate
   // set for each term, and intersect each to get a final candidate list.
   // Note that a single 'term' from the user's perspective might be
   // a string like "http://www.somewebsite.com" which, from our perspective,
   // is four words: 'http', 'www', 'somewebsite', and 'com'.
   HistoryIDSet history_id_set;
   String16Vector words(unsorted_words);
   // Sort the words into the longest first as such are likely to narrow down
   // the results quicker. Also, single character words are the most expensive
   // to process so save them for last.
   std::sort(words.begin(), words.end(), LengthGreater);
   for (String16Vector::iterator iter = words.begin(); iter != words.end();

[dyaroshev, 2017/02/17 20:44:38]

set_intersection.

        ++iter) {
-    base::string16 uni_word = *iter;
-    HistoryIDSet term_history_set = HistoryIDsForTerm(uni_word);
-    if (term_history_set.empty()) {
-      history_id_set.clear();
-      break;
-    }
-    if (iter == words.begin()) {
-      history_id_set.swap(term_history_set);
-    } else {
-      HistoryIDSet new_history_id_set = base::STLSetIntersection<HistoryIDSet>(
-          history_id_set, term_history_set);
-      history_id_set.swap(new_history_id_set);
-    }
+    HistoryIDSet term_history_set = HistoryIDsForTerm(*iter);
+    if (term_history_set.empty())
+      return HistoryIDSet();
+
+    history_id_set = (iter == words.begin())
+                         ? std::move(term_history_set)
+                         : base::STLSetIntersection<HistoryIDSet>(
+                               history_id_set, term_history_set);
   }
   return history_id_set;
 }
 
 HistoryIDSet URLIndexPrivateData::HistoryIDsForTerm(
     const base::string16& term) {
   if (term.empty())
     return HistoryIDSet();
 
   // TODO(mrossetti): Consider optimizing for very common terms such as
@@ skipping 48 matching lines @@
 
     // Reduce the word set with any leftover, unprocessed characters.
     if (!unique_chars.empty()) {
       WordIDSet leftover_set(WordIDSetForTermChars(unique_chars));
       // We might come up empty on the leftovers.
       if (leftover_set.empty()) {
         search_term_cache_[term] = SearchTermCacheItem();
         return HistoryIDSet();
       }
       // Or there may not have been a prefix from which to start.
-      if (prefix_chars.empty()) {
-        word_id_set.swap(leftover_set);
-      } else {
-        WordIDSet new_word_id_set = base::STLSetIntersection<WordIDSet>(
-            word_id_set, leftover_set);
-        word_id_set.swap(new_word_id_set);
-      }
+      word_id_set = prefix_chars.empty() ? std::move(leftover_set)
+                                         : base::STLSetIntersection<WordIDSet>(
+                                               word_id_set, leftover_set);
     }
 
     // We must filter the word list because the resulting word set surely
     // contains words which do not have the search term as a proper subset.
     for (WordIDSet::iterator word_set_iter = word_id_set.begin();

[dyaroshev, 2017/02/17 20:44:38]

erase_if

[Peter Kasting, 2017/02/18 00:40:40]

How come you ended up reverting the changes here?

[dyaroshev, 2017/02/18 00:57:36]

I reverted flat_set to std::set, which has const keys. remove_if is a mutating
algorithm.

[Peter Kasting, 2017/02/18 01:46:30]

Seems like the issue is that std::set is an inherently sorted container, so an
algorithm like "remove_if" doesn't really make sense on it at all.

Conceptually, that worries me about flat_set too -- if it lets you "un-sort" it
via remove_if, that seems kinda like a contract violation.

Maybe the answer is to expose erase_if in the flat_set API itself (rather than
as a helper in base/)?  I suppose the main idea of putting this in base/ itself
might need specializations for things like set<>.

          word_set_iter != word_id_set.end(); ) {
       if (word_list_[*word_set_iter].find(term) == base::string16::npos)
         word_set_iter = word_id_set.erase(word_set_iter);
       else
         ++word_set_iter;
     }
   } else {
     word_id_set = WordIDSetForTermChars(Char16SetFromString16(term));
   }
 
   // If any words resulted then we can compose a set of history IDs by unioning
   // the sets from each word.

[Peter Kasting, 2017/02/18 00:40:40]

And here?

[dyaroshev, 2017/02/18 00:57:36]

Putting in a vector and then inserting makes sense mostly for flat_set.

[Peter Kasting, 2017/02/18 01:46:30]

Right, but if it's perf-neutral on std::set and perf-positive on flat_set, it
makes some sense to do as a preparatory change, so the functional change
afterwards is as small as possible.

And since std::set does have a range-based insert(), maybe this would even be
better than perf-neutral for std::set?

   HistoryIDSet history_id_set;
   if (!word_id_set.empty()) {
-    for (WordIDSet::iterator word_id_iter = word_id_set.begin();
-         word_id_iter != word_id_set.end(); ++word_id_iter) {
-      WordID word_id = *word_id_iter;
+    for (WordID word_id : word_id_set) {
       WordIDHistoryMap::iterator word_iter = word_id_history_map_.find(word_id);
       if (word_iter != word_id_history_map_.end()) {
         HistoryIDSet& word_history_id_set(word_iter->second);
         history_id_set.insert(word_history_id_set.begin(),
                               word_history_id_set.end());
       }
     }
   }
 
   // Record a new cache entry for this word if the term is longer than
   // a single character.
   if (term_length > 1)
     search_term_cache_[term] = SearchTermCacheItem(word_id_set, history_id_set);
 
   return history_id_set;
 }
 
 WordIDSet URLIndexPrivateData::WordIDSetForTermChars(
     const Char16Set& term_chars) {
   WordIDSet word_id_set;
+
   for (Char16Set::const_iterator c_iter = term_chars.begin();

[dyaroshev, 2017/02/17 20:44:37]

set_intersection.

        c_iter != term_chars.end(); ++c_iter) {
     CharWordIDMap::iterator char_iter = char_word_map_.find(*c_iter);
     if (char_iter == char_word_map_.end()) {
       // A character was not found so there are no matching results: bail.
       word_id_set.clear();
       break;
     }
     WordIDSet& char_word_id_set(char_iter->second);
     // It is possible for there to no longer be any words associated with
     // a particular character. Give up in that case.
     if (char_word_id_set.empty()) {
       word_id_set.clear();
       break;
     }
 
     if (c_iter == term_chars.begin()) {
       // First character results becomes base set of results.
       word_id_set = char_word_id_set;
     } else {
       // Subsequent character results get intersected in.
-      WordIDSet new_word_id_set = base::STLSetIntersection<WordIDSet>(
-          word_id_set, char_word_id_set);
-      word_id_set.swap(new_word_id_set);
+      word_id_set =
+          base::STLSetIntersection<WordIDSet>(word_id_set, char_word_id_set);
     }
   }
+
   return word_id_set;
 }
 
 void URLIndexPrivateData::HistoryIdSetToScoredMatches(
     HistoryIDSet history_id_set,
     const base::string16& lower_raw_string,
     const TemplateURLService* template_url_service,
     bookmarks::BookmarkModel* bookmark_model,
     ScoredHistoryMatches* scored_items) const {
   if (history_id_set.empty())
@@ skipping 17 matching lines @@
   // are some form of whitespace), but this is such a rare edge case that it's
   // not worth the time.
   if (lower_raw_terms.empty())
     return;
 
   WordStarts lower_terms_to_word_starts_offsets;
   CalculateWordStartsOffsets(lower_raw_terms,
                              &lower_terms_to_word_starts_offsets);
 
   // Filter bad matches and other matches we don't want to display.
   for (auto it = history_id_set.begin();;) {

[Peter Kasting, 2017/02/18 00:40:40]

And here?

[dyaroshev, 2017/02/18 00:57:36]

Same story with remove_if.

     it = std::find_if(it, history_id_set.end(),
                       [this, template_url_service](const HistoryID history_id) {
                         return ShouldFilter(history_id, template_url_service);
                       });
     if (it == history_id_set.end())
       break;
     it = history_id_set.erase(it);
   }
 
   // Score the matches.
@@ skipping 105 matching lines @@
   HistoryID history_id = static_cast<HistoryID>(row.id());
   // Split URL into individual, unique words then add in the title words.
   const GURL& gurl(row.url());
   const base::string16& url =
       bookmarks::CleanUpUrlForMatching(gurl, nullptr);
   String16Set url_words = String16SetFromString16(url,
       word_starts ? &word_starts->url_word_starts_ : nullptr);
   const base::string16& title = bookmarks::CleanUpTitleForMatching(row.title());
   String16Set title_words = String16SetFromString16(title,
       word_starts ? &word_starts->title_word_starts_ : nullptr);
-  String16Set words = base::STLSetUnion<String16Set>(url_words, title_words);
-  for (String16Set::iterator word_iter = words.begin();
-       word_iter != words.end(); ++word_iter)
-    AddWordToIndex(*word_iter, history_id);
+  for (const auto& word :
+       base::STLSetUnion<String16Set>(url_words, title_words))
+    AddWordToIndex(word, history_id);
 
   search_term_cache_.clear();  // Invalidate the term cache.
 }
 
 void URLIndexPrivateData::AddWordToIndex(const base::string16& term,
                                          HistoryID history_id) {
-  WordMap::iterator word_pos = word_map_.find(term);
-  if (word_pos != word_map_.end())
-    UpdateWordHistory(word_pos->second, history_id);
-  else
-    AddWordHistory(term, history_id);
+  WordMap::iterator word_pos = word_map_.lower_bound(term);
+
+  // Adding a new word (i.e. a word that is not already in the word index).
+  if (word_pos->first != term) {
+    word_pos =
+        word_map_.emplace_hint(word_pos, term, AddNewWordToWordList(term));
+
+    // For each character in the newly added word add the word to the character
+    // index.
+    for (base::char16 uni_char : Char16SetFromString16(term))

[dyaroshev, 2017/02/17 20:44:38]

parallel_set.

+      char_word_map_[uni_char].insert(word_pos->second);
+  }
+
+  DCHECK_EQ(word_pos->first, term);
+
+  word_id_history_map_[word_pos->second].insert(history_id);

[Peter Kasting, 2017/02/18 00:40:40]

This looks functionally different than the old code, in that the old code reset
the set here to have one entry, while this just adds the new entry to the
existing set.  Does that matter?

[dyaroshev, 2017/02/18 00:57:36]

No - previous code reseted set only for new entries, for old entries it
inserted. I united both pieces of code to just insert because for a new entry
set would be empty.

+  history_id_word_map_[history_id].insert(word_pos->second);
 }
 
-void URLIndexPrivateData::AddWordHistory(const base::string16& term,
-                                         HistoryID history_id) {
+WordID URLIndexPrivateData::AddNewWordToWordList(const base::string16& term) {
   WordID word_id = word_list_.size();
   if (available_words_.empty()) {
     word_list_.push_back(term);
-  } else {
-    word_id = *(available_words_.begin());
-    word_list_[word_id] = term;
-    available_words_.erase(word_id);
+    return word_id;
   }
-  word_map_[term] = word_id;
 
-  HistoryIDSet history_id_set;
-  history_id_set.insert(history_id);
-  word_id_history_map_[word_id] = history_id_set;
-  AddToHistoryIDWordMap(history_id, word_id);
-
-  // For each character in the newly added word (i.e. a word that is not
-  // already in the word index), add the word to the character index.
-  Char16Set characters = Char16SetFromString16(term);
-  for (Char16Set::iterator uni_char_iter = characters.begin();
-       uni_char_iter != characters.end(); ++uni_char_iter) {
-    base::char16 uni_char = *uni_char_iter;
-    CharWordIDMap::iterator char_iter = char_word_map_.find(uni_char);
-    if (char_iter != char_word_map_.end()) {
-      // Update existing entry in the char/word index.
-      WordIDSet& word_id_set(char_iter->second);
-      word_id_set.insert(word_id);
-    } else {
-      // Create a new entry in the char/word index.
-      WordIDSet word_id_set;
-      word_id_set.insert(word_id);
-      char_word_map_[uni_char] = word_id_set;
-    }
-  }
-}
-
-void URLIndexPrivateData::UpdateWordHistory(WordID word_id,
-                                            HistoryID history_id) {
-  WordIDHistoryMap::iterator history_pos = word_id_history_map_.find(word_id);
-  DCHECK(history_pos != word_id_history_map_.end());
-  HistoryIDSet& history_id_set(history_pos->second);
-  history_id_set.insert(history_id);
-  AddToHistoryIDWordMap(history_id, word_id);
-}
-
-void URLIndexPrivateData::AddToHistoryIDWordMap(HistoryID history_id,
-                                                WordID word_id) {
-  HistoryIDWordMap::iterator iter = history_id_word_map_.find(history_id);
-  if (iter != history_id_word_map_.end()) {
-    WordIDSet& word_id_set(iter->second);
-    word_id_set.insert(word_id);
-  } else {
-    WordIDSet word_id_set;
-    word_id_set.insert(word_id);
-    history_id_word_map_[history_id] = word_id_set;
-  }
+  word_id = available_words_.top();
+  available_words_.pop();

[Peter Kasting, 2017/02/18 00:40:40]

Doesn't this need to actually place |term| in the |word_list_|?

[dyaroshev, 2017/02/18 00:57:36]

Yes. Tests that I ran didn't catch it. Sorry.

+  return word_id;
 }
 
 void URLIndexPrivateData::RemoveRowFromIndex(const history::URLRow& row) {
   RemoveRowWordsFromIndex(row);
   HistoryID history_id = static_cast<HistoryID>(row.id());
   history_info_map_.erase(history_id);
   word_starts_map_.erase(history_id);
 }
 
 void URLIndexPrivateData::RemoveRowWordsFromIndex(const history::URLRow& row) {
   // Remove the entries in history_id_word_map_ and word_id_history_map_ for
   // this row.
   HistoryID history_id = static_cast<HistoryID>(row.id());
   WordIDSet word_id_set = history_id_word_map_[history_id];
   history_id_word_map_.erase(history_id);
 
   // Reconcile any changes to word usage.
-  for (WordIDSet::iterator word_id_iter = word_id_set.begin();
-       word_id_iter != word_id_set.end(); ++word_id_iter) {
-    WordID word_id = *word_id_iter;
-    word_id_history_map_[word_id].erase(history_id);
-    if (!word_id_history_map_[word_id].empty())
-      continue;  // The word is still in use.
+  for (WordID word_id : word_id_set) {
+    auto word_id_history_map_iter = word_id_history_map_.find(word_id);
+    DCHECK(word_id_history_map_iter != word_id_history_map_.end());
+
+    word_id_history_map_iter->second.erase(history_id);
+    if (!word_id_history_map_iter->second.empty())
+      continue;
 
     // The word is no longer in use. Reconcile any changes to character usage.
     base::string16 word = word_list_[word_id];
-    Char16Set characters = Char16SetFromString16(word);
-    for (Char16Set::iterator uni_char_iter = characters.begin();
-         uni_char_iter != characters.end(); ++uni_char_iter) {
-      base::char16 uni_char = *uni_char_iter;
-      char_word_map_[uni_char].erase(word_id);
-      if (char_word_map_[uni_char].empty())
-        char_word_map_.erase(uni_char);  // No longer in use.
+    for (base::char16 uni_char : Char16SetFromString16(word)) {

[dyaroshev, 2017/02/17 20:44:38]

parallel_set.

+      auto char_word_map_iter = char_word_map_.find(uni_char);
+      char_word_map_iter->second.erase(word_id);
+      if (char_word_map_iter->second.empty())
+        char_word_map_.erase(char_word_map_iter);
     }
 
     // Complete the removal of references to the word.
-    word_id_history_map_.erase(word_id);
+    word_id_history_map_.erase(word_id_history_map_iter);
     word_map_.erase(word);
     word_list_[word_id] = base::string16();
-    available_words_.insert(word_id);
+    available_words_.push(word_id);
   }
 }
 
 void URLIndexPrivateData::ResetSearchTermCache() {
-  for (SearchTermCacheMap::iterator iter = search_term_cache_.begin();
-       iter != search_term_cache_.end(); ++iter)
-    iter->second.used_ = false;
+  for (auto& item : search_term_cache_)
+    item.second.used_ = false;
 }
 
 bool URLIndexPrivateData::SaveToFile(const base::FilePath& file_path) {
   base::TimeTicks beginning_time = base::TimeTicks::Now();
   InMemoryURLIndexCacheItem index_cache;
   SavePrivateData(&index_cache);
   std::string data;
   if (!index_cache.SerializeToString(&data)) {
     LOG(WARNING) << "Failed to serialize the InMemoryURLIndex cache.";
     return false;
@@ skipping 24 matching lines @@
   SaveWordIDHistoryMap(cache);
   SaveHistoryInfoMap(cache);
   SaveWordStartsMap(cache);
 }
 
 void URLIndexPrivateData::SaveWordList(InMemoryURLIndexCacheItem* cache) const {
   if (word_list_.empty())
     return;
   WordListItem* list_item = cache->mutable_word_list();
   list_item->set_word_count(word_list_.size());
-  for (String16Vector::const_iterator iter = word_list_.begin();
-       iter != word_list_.end(); ++iter)
-    list_item->add_word(base::UTF16ToUTF8(*iter));
+  for (const base::string16& word : word_list_)
+    list_item->add_word(base::UTF16ToUTF8(word));
 }
 
 void URLIndexPrivateData::SaveWordMap(InMemoryURLIndexCacheItem* cache) const {
   if (word_map_.empty())
     return;
   WordMapItem* map_item = cache->mutable_word_map();
   map_item->set_item_count(word_map_.size());
-  for (WordMap::const_iterator iter = word_map_.begin();
-       iter != word_map_.end(); ++iter) {
+  for (const auto& elem : word_map_) {
     WordMapEntry* map_entry = map_item->add_word_map_entry();
-    map_entry->set_word(base::UTF16ToUTF8(iter->first));
-    map_entry->set_word_id(iter->second);
+    map_entry->set_word(base::UTF16ToUTF8(elem.first));
+    map_entry->set_word_id(elem.second);
   }
 }
 
 void URLIndexPrivateData::SaveCharWordMap(
     InMemoryURLIndexCacheItem* cache) const {
   if (char_word_map_.empty())
     return;
   CharWordMapItem* map_item = cache->mutable_char_word_map();
   map_item->set_item_count(char_word_map_.size());
-  for (CharWordIDMap::const_iterator iter = char_word_map_.begin();
-       iter != char_word_map_.end(); ++iter) {
+  for (const auto& entry : char_word_map_) {
     CharWordMapEntry* map_entry = map_item->add_char_word_map_entry();
-    map_entry->set_char_16(iter->first);
-    const WordIDSet& word_id_set(iter->second);
+    map_entry->set_char_16(entry.first);
+    const WordIDSet& word_id_set(entry.second);
     map_entry->set_item_count(word_id_set.size());
-    for (WordIDSet::const_iterator set_iter = word_id_set.begin();
-         set_iter != word_id_set.end(); ++set_iter)
-      map_entry->add_word_id(*set_iter);
+    for (WordID word_id : word_id_set)
+      map_entry->add_word_id(word_id);
   }
 }
 
 void URLIndexPrivateData::SaveWordIDHistoryMap(
     InMemoryURLIndexCacheItem* cache) const {
   if (word_id_history_map_.empty())
     return;
   WordIDHistoryMapItem* map_item = cache->mutable_word_id_history_map();
   map_item->set_item_count(word_id_history_map_.size());
-  for (WordIDHistoryMap::const_iterator iter = word_id_history_map_.begin();
-       iter != word_id_history_map_.end(); ++iter) {
+  for (const auto& entry : word_id_history_map_) {
     WordIDHistoryMapEntry* map_entry =
         map_item->add_word_id_history_map_entry();
-    map_entry->set_word_id(iter->first);
-    const HistoryIDSet& history_id_set(iter->second);
+    map_entry->set_word_id(entry.first);
+    const HistoryIDSet& history_id_set(entry.second);
     map_entry->set_item_count(history_id_set.size());
-    for (HistoryIDSet::const_iterator set_iter = history_id_set.begin();
-         set_iter != history_id_set.end(); ++set_iter)
-      map_entry->add_history_id(*set_iter);
+    for (HistoryID history_id : history_id_set)
+      map_entry->add_history_id(history_id);
   }
 }
 
 void URLIndexPrivateData::SaveHistoryInfoMap(
     InMemoryURLIndexCacheItem* cache) const {
   if (history_info_map_.empty())
     return;
   HistoryInfoMapItem* map_item = cache->mutable_history_info_map();
   map_item->set_item_count(history_info_map_.size());
-  for (HistoryInfoMap::const_iterator iter = history_info_map_.begin();
-       iter != history_info_map_.end(); ++iter) {
+  for (const auto& entry : history_info_map_) {
     HistoryInfoMapEntry* map_entry = map_item->add_history_info_map_entry();
-    map_entry->set_history_id(iter->first);
-    const history::URLRow& url_row(iter->second.url_row);
+    map_entry->set_history_id(entry.first);
+    const history::URLRow& url_row(entry.second.url_row);
     // Note: We only save information that contributes to the index so there
     // is no need to save search_term_cache_ (not persistent).
     map_entry->set_visit_count(url_row.visit_count());
     map_entry->set_typed_count(url_row.typed_count());
     map_entry->set_last_visit(url_row.last_visit().ToInternalValue());
     map_entry->set_url(url_row.url().spec());
     map_entry->set_title(base::UTF16ToUTF8(url_row.title()));
-    const VisitInfoVector& visits(iter->second.visits);
-    for (VisitInfoVector::const_iterator visit_iter = visits.begin();
-         visit_iter != visits.end(); ++visit_iter) {
+    for (const auto& visit : entry.second.visits) {
       HistoryInfoMapEntry_VisitInfo* visit_info = map_entry->add_visits();
-      visit_info->set_visit_time(visit_iter->first.ToInternalValue());
-      visit_info->set_transition_type(visit_iter->second);
+      visit_info->set_visit_time(visit.first.ToInternalValue());
+      visit_info->set_transition_type(visit.second);
     }
   }
 }
 
 void URLIndexPrivateData::SaveWordStartsMap(
     InMemoryURLIndexCacheItem* cache) const {
   if (word_starts_map_.empty())
     return;
   // For unit testing: Enable saving of the cache as an earlier version to
   // allow testing of cache file upgrading in ReadFromFile().
   // TODO(mrossetti): Instead of intruding on production code with this kind of
   // test harness, save a copy of an older version cache with known results.
   // Implement this when switching the caching over to SQLite.
   if (saved_cache_version_ < 1)
     return;
 
   WordStartsMapItem* map_item = cache->mutable_word_starts_map();
   map_item->set_item_count(word_starts_map_.size());
-  for (WordStartsMap::const_iterator iter = word_starts_map_.begin();
-       iter != word_starts_map_.end(); ++iter) {
+  for (const auto& entry : word_starts_map_) {
     WordStartsMapEntry* map_entry = map_item->add_word_starts_map_entry();
-    map_entry->set_history_id(iter->first);
-    const RowWordStarts& word_starts(iter->second);
-    for (WordStarts::const_iterator i = word_starts.url_word_starts_.begin();
-         i != word_starts.url_word_starts_.end(); ++i)
-      map_entry->add_url_word_starts(*i);
-    for (WordStarts::const_iterator i = word_starts.title_word_starts_.begin();
-         i != word_starts.title_word_starts_.end(); ++i)
-      map_entry->add_title_word_starts(*i);
+    map_entry->set_history_id(entry.first);
+    const RowWordStarts& word_starts(entry.second);
+    for (auto url_word_start : word_starts.url_word_starts_)
+      map_entry->add_url_word_starts(url_word_start);
+    for (auto title_word_start : word_starts.title_word_starts_)
+      map_entry->add_title_word_starts(title_word_start);
   }
 }
 
 bool URLIndexPrivateData::RestorePrivateData(
     const InMemoryURLIndexCacheItem& cache) {
   last_time_rebuilt_from_history_ =
       base::Time::FromInternalValue(cache.last_rebuild_timestamp());
   const base::TimeDelta rebuilt_ago =
       base::Time::Now() - last_time_rebuilt_from_history_;
   if ((rebuilt_ago > base::TimeDelta::FromDays(7)) ||
       (rebuilt_ago < base::TimeDelta::FromDays(-1))) {
     // Cache is more than a week old or, somehow, from some time in the future.
     // It's probably a good time to rebuild the index from history to
-    // allow synced entries to now appear, expired entries to disappear, etc.
+    // allow synced entrys to now appear, expired entries to disappear, etc.

[dyaroshev, 2017/02/17 20:17:23]

Fix with next patch!

     // Allow one day in the future to make the cache not rebuild on simple
     // system clock changes such as time zone changes.
     return false;
   }
   if (cache.has_version()) {
     if (cache.version() < kCurrentCacheFileVersion) {
       // Don't try to restore an old format cache file.  (This will cause
       // the InMemoryURLIndex to schedule rebuilding the URLIndexPrivateData
       // from history.)
       return false;
     }
     restored_cache_version_ = cache.version();
   }
   return RestoreWordList(cache) && RestoreWordMap(cache) &&
       RestoreCharWordMap(cache) && RestoreWordIDHistoryMap(cache) &&
       RestoreHistoryInfoMap(cache) && RestoreWordStartsMap(cache);
 }
 
 bool URLIndexPrivateData::RestoreWordList(
     const InMemoryURLIndexCacheItem& cache) {
   if (!cache.has_word_list())
     return false;
   const WordListItem& list_item(cache.word_list());
   uint32_t expected_item_count = list_item.word_count();
   uint32_t actual_item_count = list_item.word_size();
   if (actual_item_count == 0 || actual_item_count != expected_item_count)
     return false;
   const RepeatedPtrField<std::string>& words(list_item.word());
-  for (RepeatedPtrField<std::string>::const_iterator iter = words.begin();
-       iter != words.end(); ++iter)
-    word_list_.push_back(base::UTF8ToUTF16(*iter));
+  word_list_.reserve(words.size());
+  std::transform(
+      words.begin(), words.end(), std::back_inserter(word_list_),
+      [](const std::string& word) { return base::UTF8ToUTF16(word); });
   return true;
 }
 
 bool URLIndexPrivateData::RestoreWordMap(
     const InMemoryURLIndexCacheItem& cache) {
   if (!cache.has_word_map())
     return false;
   const WordMapItem& list_item(cache.word_map());
   uint32_t expected_item_count = list_item.item_count();
   uint32_t actual_item_count = list_item.word_map_entry_size();
   if (actual_item_count == 0 || actual_item_count != expected_item_count)
     return false;
-  const RepeatedPtrField<WordMapEntry>& entries(list_item.word_map_entry());
-  for (RepeatedPtrField<WordMapEntry>::const_iterator iter = entries.begin();
-       iter != entries.end(); ++iter)
-    word_map_[base::UTF8ToUTF16(iter->word())] = iter->word_id();
+  for (const auto& entry : list_item.word_map_entry())
+    word_map_[base::UTF8ToUTF16(entry.word())] = entry.word_id();

[dyaroshev, 2017/02/17 20:44:37]

parallel_set.

+
   return true;
 }
 
 bool URLIndexPrivateData::RestoreCharWordMap(
     const InMemoryURLIndexCacheItem& cache) {
   if (!cache.has_char_word_map())
     return false;
   const CharWordMapItem& list_item(cache.char_word_map());
   uint32_t expected_item_count = list_item.item_count();
   uint32_t actual_item_count = list_item.char_word_map_entry_size();
   if (actual_item_count == 0 || actual_item_count != expected_item_count)
     return false;
-  const RepeatedPtrField<CharWordMapEntry>&
-      entries(list_item.char_word_map_entry());
-  for (RepeatedPtrField<CharWordMapEntry>::const_iterator iter =
-       entries.begin(); iter != entries.end(); ++iter) {
-    expected_item_count = iter->item_count();
-    actual_item_count = iter->word_id_size();
+
+  for (const auto& entry : list_item.char_word_map_entry()) {
+    expected_item_count = entry.item_count();
+    actual_item_count = entry.word_id_size();
     if (actual_item_count == 0 || actual_item_count != expected_item_count)
       return false;
-    base::char16 uni_char = static_cast<base::char16>(iter->char_16());
-    WordIDSet word_id_set;
-    const RepeatedField<int32_t>& word_ids(iter->word_id());
-    for (RepeatedField<int32_t>::const_iterator jiter = word_ids.begin();
-         jiter != word_ids.end(); ++jiter)
-      word_id_set.insert(*jiter);
-    char_word_map_[uni_char] = word_id_set;
+    base::char16 uni_char = static_cast<base::char16>(entry.char_16());
+    const RepeatedField<int32_t>& word_ids(entry.word_id());
+    char_word_map_[uni_char] = {word_ids.begin(), word_ids.end()};
   }
   return true;
 }
 
 bool URLIndexPrivateData::RestoreWordIDHistoryMap(
     const InMemoryURLIndexCacheItem& cache) {
   if (!cache.has_word_id_history_map())
     return false;
   const WordIDHistoryMapItem& list_item(cache.word_id_history_map());
   uint32_t expected_item_count = list_item.item_count();
   uint32_t actual_item_count = list_item.word_id_history_map_entry_size();
   if (actual_item_count == 0 || actual_item_count != expected_item_count)
     return false;
-  const RepeatedPtrField<WordIDHistoryMapEntry>&
-      entries(list_item.word_id_history_map_entry());
-  for (RepeatedPtrField<WordIDHistoryMapEntry>::const_iterator iter =
-       entries.begin(); iter != entries.end(); ++iter) {
-    expected_item_count = iter->item_count();
-    actual_item_count = iter->history_id_size();
+  for (const auto& entry : list_item.word_id_history_map_entry()) {
+    expected_item_count = entry.item_count();
+    actual_item_count = entry.history_id_size();
     if (actual_item_count == 0 || actual_item_count != expected_item_count)
       return false;
-    WordID word_id = iter->word_id();
-    HistoryIDSet history_id_set;
-    const RepeatedField<int64_t>& history_ids(iter->history_id());
-    for (RepeatedField<int64_t>::const_iterator jiter = history_ids.begin();
-         jiter != history_ids.end(); ++jiter) {
-      history_id_set.insert(*jiter);
-      AddToHistoryIDWordMap(*jiter, word_id);
-    }
-    word_id_history_map_[word_id] = history_id_set;
+    WordID word_id = entry.word_id();
+    const RepeatedField<int64_t>& history_ids(entry.history_id());
+    word_id_history_map_[word_id] = {history_ids.begin(), history_ids.end()};
+    for (HistoryID history_id : history_ids)

[dyaroshev, 2017/02/17 20:44:38]

parallel_set.

+      history_id_word_map_[history_id].insert(word_id);
   }
   return true;
 }
 
 bool URLIndexPrivateData::RestoreHistoryInfoMap(
     const InMemoryURLIndexCacheItem& cache) {
   if (!cache.has_history_info_map())
     return false;
   const HistoryInfoMapItem& list_item(cache.history_info_map());
   uint32_t expected_item_count = list_item.item_count();
   uint32_t actual_item_count = list_item.history_info_map_entry_size();
   if (actual_item_count == 0 || actual_item_count != expected_item_count)
     return false;
-  const RepeatedPtrField<HistoryInfoMapEntry>&
-      entries(list_item.history_info_map_entry());
-  for (RepeatedPtrField<HistoryInfoMapEntry>::const_iterator iter =
-       entries.begin(); iter != entries.end(); ++iter) {
-    HistoryID history_id = iter->history_id();
-    GURL url(iter->url());
-    history::URLRow url_row(url, history_id);
-    url_row.set_visit_count(iter->visit_count());
-    url_row.set_typed_count(iter->typed_count());
-    url_row.set_last_visit(base::Time::FromInternalValue(iter->last_visit()));
-    if (iter->has_title()) {
-      base::string16 title(base::UTF8ToUTF16(iter->title()));
-      url_row.set_title(title);
-    }
-    history_info_map_[history_id].url_row = url_row;
+
+  for (const auto& entry : list_item.history_info_map_entry()) {
+    HistoryID history_id = entry.history_id();
+    history::URLRow url_row(GURL(entry.url()), history_id);
+    url_row.set_visit_count(entry.visit_count());
+    url_row.set_typed_count(entry.typed_count());
+    url_row.set_last_visit(base::Time::FromInternalValue(entry.last_visit()));
+    if (entry.has_title())
+      url_row.set_title(base::UTF8ToUTF16(entry.title()));
+    history_info_map_[history_id].url_row = std::move(url_row);
 
     // Restore visits list.
     VisitInfoVector visits;
-    visits.reserve(iter->visits_size());
-    for (int i = 0; i < iter->visits_size(); ++i) {
-      visits.push_back(std::make_pair(
-          base::Time::FromInternalValue(iter->visits(i).visit_time()),
-          ui::PageTransitionFromInt(iter->visits(i).transition_type())));
+    visits.reserve(entry.visits_size());
+    for (const auto& entry_visit : entry.visits()) {
+      visits.emplace_back(
+          base::Time::FromInternalValue(entry_visit.visit_time()),
+          ui::PageTransitionFromInt(entry_visit.transition_type()));
     }
-    history_info_map_[history_id].visits = visits;
+    history_info_map_[history_id].visits = std::move(visits);
   }
   return true;
 }
 
 bool URLIndexPrivateData::RestoreWordStartsMap(
     const InMemoryURLIndexCacheItem& cache) {
   // Note that this function must be called after RestoreHistoryInfoMap() has
   // been run as the word starts may have to be recalculated from the urls and
   // page titles.
   if (cache.has_word_starts_map()) {
     const WordStartsMapItem& list_item(cache.word_starts_map());
     uint32_t expected_item_count = list_item.item_count();
     uint32_t actual_item_count = list_item.word_starts_map_entry_size();
     if (actual_item_count == 0 || actual_item_count != expected_item_count)
       return false;
-    const RepeatedPtrField<WordStartsMapEntry>&
-        entries(list_item.word_starts_map_entry());
-    for (RepeatedPtrField<WordStartsMapEntry>::const_iterator iter =
-         entries.begin(); iter != entries.end(); ++iter) {
-      HistoryID history_id = iter->history_id();
+    for (const auto& entry : list_item.word_starts_map_entry()) {
+      HistoryID history_id = entry.history_id();
       RowWordStarts word_starts;
       // Restore the URL word starts.
-      const RepeatedField<int32_t>& url_starts(iter->url_word_starts());
-      for (RepeatedField<int32_t>::const_iterator jiter = url_starts.begin();
-           jiter != url_starts.end(); ++jiter)
-        word_starts.url_word_starts_.push_back(*jiter);
+      const RepeatedField<int32_t>& url_starts(entry.url_word_starts());
+      word_starts.url_word_starts_ = {url_starts.begin(), url_starts.end()};
+
       // Restore the page title word starts.
-      const RepeatedField<int32_t>& title_starts(iter->title_word_starts());
-      for (RepeatedField<int32_t>::const_iterator jiter = title_starts.begin();
-           jiter != title_starts.end(); ++jiter)
-        word_starts.title_word_starts_.push_back(*jiter);
-      word_starts_map_[history_id] = word_starts;
+      const RepeatedField<int32_t>& title_starts(entry.title_word_starts());
+      word_starts.title_word_starts_ = {title_starts.begin(),
+                                        title_starts.end()};
+
+      word_starts_map_[history_id] = std::move(word_starts);
     }
   } else {
     // Since the cache did not contain any word starts we must rebuild then from
     // the URL and page titles.
-    for (HistoryInfoMap::const_iterator iter = history_info_map_.begin();
-         iter != history_info_map_.end(); ++iter) {
+    for (const auto& entry : history_info_map_) {

[dyaroshev, 2017/02/17 20:44:38]

parallel_set.

       RowWordStarts word_starts;
-      const history::URLRow& row(iter->second.url_row);
+      const history::URLRow& row(entry.second.url_row);
       const base::string16& url =
           bookmarks::CleanUpUrlForMatching(row.url(), nullptr);
       String16VectorFromString16(url, false, &word_starts.url_word_starts_);
       const base::string16& title =
           bookmarks::CleanUpTitleForMatching(row.title());
       String16VectorFromString16(title, false, &word_starts.title_word_starts_);
-      word_starts_map_[iter->first] = word_starts;
+      word_starts_map_[entry.first] = std::move(word_starts);
     }
   }
   return true;
 }
 
 // static
 bool URLIndexPrivateData::URLSchemeIsWhitelisted(
     const GURL& gurl,
     const std::set<std::string>& whitelist) {
   return whitelist.find(gurl.scheme()) != whitelist.end();
@@ skipping 62 matching lines @@
   // First cut: typed count, visit count, recency.
   // TODO(mrossetti): This is too simplistic. Consider an approach which ranks
   // recently visited (within the last 12/24 hours) as highly important. Get
   // input from mpearson.
   if (r1.typed_count() != r2.typed_count())
     return (r1.typed_count() > r2.typed_count());
   if (r1.visit_count() != r2.visit_count())
     return (r1.visit_count() > r2.visit_count());
   return (r1.last_visit() > r2.last_visit());
 }