components/bookmarks/browser/bookmark_index.cc

   1 // Copyright 2014 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4
   5 #include "components/bookmarks/browser/bookmark_index.h"
   6
   7 #include <algorithm>
   8 #include <functional>
   9 #include <iterator>
  10 #include <list>
  11
  12 #include "base/i18n/case_conversion.h"
  13 #include "base/logging.h"
  14 #include "base/stl_util.h"
  15 #include "base/strings/utf_offset_string_conversions.h"
  16 #include "components/bookmarks/browser/bookmark_client.h"
  17 #include "components/bookmarks/browser/bookmark_match.h"
  18 #include "components/bookmarks/browser/bookmark_node.h"
  19 #include "components/bookmarks/browser/bookmark_utils.h"
  20 #include "components/query_parser/snippet.h"
  21 #include "third_party/icu/source/common/unicode/normalizer2.h"
  22
  23 namespace bookmarks {
  24
  25 typedef BookmarkClient::NodeTypedCountPair NodeTypedCountPair;
  26 typedef BookmarkClient::NodeTypedCountPairs NodeTypedCountPairs;
  27
  28 namespace {
  29
  30 // Returns a normalized version of the UTF16 string |text|.  If it fails to
  31 // normalize the string, returns |text| itself as a best-effort.
  32 base::string16 Normalize(const base::string16& text) {
  33   UErrorCode status = U_ZERO_ERROR;
  34   const icu::Normalizer2* normalizer2 =
  35       icu::Normalizer2::getInstance(NULL, "nfkc", UNORM2_COMPOSE, status);
  36   icu::UnicodeString unicode_text(
  37       text.data(), static_cast<int32_t>(text.length()));
  38   icu::UnicodeString unicode_normalized_text;
  39   normalizer2->normalize(unicode_text, unicode_normalized_text, status);
  40   if (U_FAILURE(status))
  41     return text;
  42   return base::string16(unicode_normalized_text.getBuffer(),
  43                         unicode_normalized_text.length());
  44 }
  45
  46 // Sort functor for NodeTypedCountPairs. We sort in decreasing order of typed
  47 // count so that the best matches will always be added to the results.
  48 struct NodeTypedCountPairSortFunctor
  49     : std::binary_function<NodeTypedCountPair, NodeTypedCountPair, bool> {
  50   bool operator()(const NodeTypedCountPair& a,
  51                   const NodeTypedCountPair& b) const {
  52     return a.second > b.second;
  53   }
  54 };
  55
  56 // Extract the const Node* stored in a BookmarkClient::NodeTypedCountPair.
  57 struct NodeTypedCountPairExtractNodeFunctor
  58     : std::unary_function<NodeTypedCountPair, const BookmarkNode*> {
  59   const BookmarkNode* operator()(const NodeTypedCountPair& pair) const {
  60     return pair.first;
  61   }
  62 };
  63
  64 }  // namespace
  65
  66 BookmarkIndex::BookmarkIndex(BookmarkClient* client,
  67                              const std::string& languages)
  68     : client_(client),
  69       languages_(languages) {
  70   DCHECK(client_);
  71 }
  72
  73 BookmarkIndex::~BookmarkIndex() {
  74 }
  75
  76 void BookmarkIndex::Add(const BookmarkNode* node) {
  77   if (!node->is_url())
  78     return;
  79   std::vector<base::string16> terms =
  80       ExtractQueryWords(Normalize(node->GetTitle()));
  81   for (size_t i = 0; i < terms.size(); ++i)
  82     RegisterNode(terms[i], node);
  83   terms =
  84       ExtractQueryWords(CleanUpUrlForMatching(node->url(), languages_, NULL));
  85   for (size_t i = 0; i < terms.size(); ++i)
  86     RegisterNode(terms[i], node);
  87 }
  88
  89 void BookmarkIndex::Remove(const BookmarkNode* node) {
  90   if (!node->is_url())
  91     return;
  92
  93   std::vector<base::string16> terms =
  94       ExtractQueryWords(Normalize(node->GetTitle()));
  95   for (size_t i = 0; i < terms.size(); ++i)
  96     UnregisterNode(terms[i], node);
  97   terms =
  98       ExtractQueryWords(CleanUpUrlForMatching(node->url(), languages_, NULL));
  99   for (size_t i = 0; i < terms.size(); ++i)
 100     UnregisterNode(terms[i], node);
 101 }
 102
 103 void BookmarkIndex::GetBookmarksMatching(
 104     const base::string16& input_query,
 105     size_t max_count,
 106     query_parser::MatchingAlgorithm matching_algorithm,
 107     std::vector<BookmarkMatch>* results) {
 108   const base::string16 query = Normalize(input_query);
 109   std::vector<base::string16> terms = ExtractQueryWords(query);
 110   if (terms.empty())
 111     return;
 112
 113   NodeSet matches;
 114   for (size_t i = 0; i < terms.size(); ++i) {
 115     if (!GetBookmarksMatchingTerm(
 116             terms[i], i == 0, matching_algorithm, &matches)) {
 117       return;
 118     }
 119   }
 120
 121   Nodes sorted_nodes;
 122   SortMatches(matches, &sorted_nodes);
 123
 124   // We use a QueryParser to fill in match positions for us. It's not the most
 125   // efficient way to go about this, but by the time we get here we know what
 126   // matches and so this shouldn't be performance critical.
 127   query_parser::QueryParser parser;
 128   ScopedVector<query_parser::QueryNode> query_nodes;
 129   parser.ParseQueryNodes(query, matching_algorithm, &query_nodes.get());
 130
 131   // The highest typed counts should be at the beginning of the results vector
 132   // so that the best matches will always be included in the results. The loop
 133   // that calculates result relevance in HistoryContentsProvider::ConvertResults
 134   // will run backwards to assure higher relevance will be attributed to the
 135   // best matches.
 136   for (Nodes::const_iterator i = sorted_nodes.begin();
 137        i != sorted_nodes.end() && results->size() < max_count;
 138        ++i)
 139     AddMatchToResults(*i, &parser, query_nodes.get(), results);
 140 }
 141
 142 void BookmarkIndex::SortMatches(const NodeSet& matches,
 143                                 Nodes* sorted_nodes) const {
 144   sorted_nodes->reserve(matches.size());
 145   if (client_->SupportsTypedCountForNodes()) {
 146     NodeTypedCountPairs node_typed_counts;
 147     client_->GetTypedCountForNodes(matches, &node_typed_counts);
 148     std::sort(node_typed_counts.begin(),
 149               node_typed_counts.end(),
 150               NodeTypedCountPairSortFunctor());
 151     std::transform(node_typed_counts.begin(),
 152                    node_typed_counts.end(),
 153                    std::back_inserter(*sorted_nodes),
 154                    NodeTypedCountPairExtractNodeFunctor());
 155   } else {
 156     sorted_nodes->insert(sorted_nodes->end(), matches.begin(), matches.end());
 157   }
 158 }
 159
 160 void BookmarkIndex::AddMatchToResults(
 161     const BookmarkNode* node,
 162     query_parser::QueryParser* parser,
 163     const query_parser::QueryNodeStarVector& query_nodes,
 164     std::vector<BookmarkMatch>* results) {
 165   // Check that the result matches the query.  The previous search
 166   // was a simple per-word search, while the more complex matching
 167   // of QueryParser may filter it out.  For example, the query
 168   // ["thi"] will match the bookmark titled [Thinking], but since
 169   // ["thi"] is quoted we don't want to do a prefix match.
 170   query_parser::QueryWordVector title_words, url_words;
 171   const base::string16 lower_title =
 172       base::i18n::ToLower(Normalize(node->GetTitle()));
 173   parser->ExtractQueryWords(lower_title, &title_words);
 174   base::OffsetAdjuster::Adjustments adjustments;
 175   parser->ExtractQueryWords(
 176       CleanUpUrlForMatching(node->url(), languages_, &adjustments),
 177       &url_words);
 178   query_parser::Snippet::MatchPositions title_matches, url_matches;
 179   for (size_t i = 0; i < query_nodes.size(); ++i) {
 180     const bool has_title_matches =
 181         query_nodes[i]->HasMatchIn(title_words, &title_matches);
 182     const bool has_url_matches =
 183         query_nodes[i]->HasMatchIn(url_words, &url_matches);
 184     if (!has_title_matches && !has_url_matches)
 185       return;
 186     query_parser::QueryParser::SortAndCoalesceMatchPositions(&title_matches);
 187     query_parser::QueryParser::SortAndCoalesceMatchPositions(&url_matches);
 188   }
 189   BookmarkMatch match;
 190   if (lower_title.length() == node->GetTitle().length()) {
 191     // Only use title matches if the lowercase string is the same length
 192     // as the original string, otherwise the matches are meaningless.
 193     // TODO(mpearson): revise match positions appropriately.
 194     match.title_match_positions.swap(title_matches);
 195   }
 196   // Now that we're done processing this entry, correct the offsets of the
 197   // matches in |url_matches| so they point to offsets in the original URL
 198   // spec, not the cleaned-up URL string that we used for matching.
 199   std::vector<size_t> offsets =
 200       BookmarkMatch::OffsetsFromMatchPositions(url_matches);
 201   base::OffsetAdjuster::UnadjustOffsets(adjustments, &offsets);
 202   url_matches =
 203       BookmarkMatch::ReplaceOffsetsInMatchPositions(url_matches, offsets);
 204   match.url_match_positions.swap(url_matches);
 205   match.node = node;
 206   results->push_back(match);
 207 }
 208
 209 bool BookmarkIndex::GetBookmarksMatchingTerm(
 210     const base::string16& term,
 211     bool first_term,
 212     query_parser::MatchingAlgorithm matching_algorithm,
 213     NodeSet* matches) {
 214   Index::const_iterator i = index_.lower_bound(term);
 215   if (i == index_.end())
 216     return false;
 217
 218   if (!query_parser::QueryParser::IsWordLongEnoughForPrefixSearch(
 219       term, matching_algorithm)) {
 220     // Term is too short for prefix match, compare using exact match.
 221     if (i->first != term)
 222       return false;  // No bookmarks with this term.
 223
 224     if (first_term) {
 225       (*matches) = i->second;
 226       return true;
 227     }
 228     *matches = base::STLSetIntersection<NodeSet>(i->second, *matches);
 229   } else {
 230     // Loop through index adding all entries that start with term to
 231     // |prefix_matches|.
 232     NodeSet tmp_prefix_matches;
 233     // If this is the first term, then store the result directly in |matches|
 234     // to avoid calling stl intersection (which requires a copy).
 235     NodeSet* prefix_matches = first_term ? matches : &tmp_prefix_matches;
 236     while (i != index_.end() &&
 237            i->first.size() >= term.size() &&
 238            term.compare(0, term.size(), i->first, 0, term.size()) == 0) {
 239 #if !defined(OS_ANDROID)
 240       prefix_matches->insert(i->second.begin(), i->second.end());
 241 #else
 242       // Work around a bug in the implementation of std::set::insert in the STL
 243       // used on android (http://crbug.com/367050).
 244       for (NodeSet::const_iterator n = i->second.begin(); n != i->second.end();
 245            ++n)
 246         prefix_matches->insert(prefix_matches->end(), *n);
 247 #endif
 248       ++i;
 249     }
 250     if (!first_term)
 251       *matches = base::STLSetIntersection<NodeSet>(*prefix_matches, *matches);
 252   }
 253   return !matches->empty();
 254 }
 255
 256 std::vector<base::string16> BookmarkIndex::ExtractQueryWords(
 257     const base::string16& query) {
 258   std::vector<base::string16> terms;
 259   if (query.empty())
 260     return std::vector<base::string16>();
 261   query_parser::QueryParser parser;
 262   parser.ParseQueryWords(base::i18n::ToLower(query),
 263                          query_parser::MatchingAlgorithm::DEFAULT,
 264                          &terms);
 265   return terms;
 266 }
 267
 268 void BookmarkIndex::RegisterNode(const base::string16& term,
 269                                  const BookmarkNode* node) {
 270   index_[term].insert(node);
 271 }
 272
 273 void BookmarkIndex::UnregisterNode(const base::string16& term,
 274                                    const BookmarkNode* node) {
 275   Index::iterator i = index_.find(term);
 276   if (i == index_.end()) {
 277     // We can get here if the node has the same term more than once. For
 278     // example, a bookmark with the title 'foo foo' would end up here.
 279     return;
 280   }
 281   i->second.erase(node);
 282   if (i->second.empty())
 283     index_.erase(i);
 284 }
 285
 286 }  // namespace bookmarks