Cleaning up url_index_private_data and in_memory_url_index_types.

components/omnibox/browser/url_index_private_data.h

 // 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.
 
 #ifndef COMPONENTS_OMNIBOX_BROWSER_URL_INDEX_PRIVATE_DATA_H_
 #define COMPONENTS_OMNIBOX_BROWSER_URL_INDEX_PRIVATE_DATA_H_
 
 #include <stddef.h>
 
 #include <set>
+#include <stack>
 #include <string>
 
 #include "base/files/file_path.h"
 #include "base/gtest_prod_util.h"
 #include "base/memory/ref_counted.h"
 #include "components/history/core/browser/history_service.h"
 #include "components/omnibox/browser/in_memory_url_index_cache.pb.h"
 #include "components/omnibox/browser/in_memory_url_index_types.h"
 #include "components/omnibox/browser/scored_history_match.h"
 
@@ skipping 176 matching lines @@
     ~HistoryItemFactorGreater();
 
     bool operator()(const HistoryID h1, const HistoryID h2);
 
    private:
     const HistoryInfoMap& history_info_map_;
   };
 
   // URL History indexing support functions.
 
-  // Composes a set of history item IDs by intersecting the set for each word
+  // Composes a vector of history item IDs by intersecting the set for each word
   // in |unsorted_words|.
-  HistoryIDSet HistoryIDSetFromWords(const String16Vector& unsorted_words);
+  HistoryIDVector HistoryIDsFromWords(const String16Vector& unsorted_words);
+
+  // Trims the candidate pool in advance of doing proper substring searching, to
+  // cap the cost of such searching. Discards the least-relevant items (based on
+  // visit stats), which are least likely to score highly in the end.  To
+  // minimize the risk of discarding a valuable URL, the candidate pool is still
+  // left two orders of magnitude larger than the final number of results
+  // returned from the HQP.
+  void TrimHistoryIdsPool(HistoryIDVector* history_ids) const;
 
   // Helper function to HistoryIDSetFromWords which composes a set of history
   // ids for the given term given in |term|.
   HistoryIDSet HistoryIDsForTerm(const base::string16& term);
 
   // Given a set of Char16s, finds words containing those characters.
   WordIDSet WordIDSetForTermChars(const Char16Set& term_chars);
 
   // Helper function for HistoryItemsForTerms().  Fills in |scored_items| from
-  // the matches listed in |history_id_set|.
-  void HistoryIdSetToScoredMatches(
-      HistoryIDSet history_id_set,
-      const base::string16& lower_raw_string,
-      const TemplateURLService* template_url_service,
-      bookmarks::BookmarkModel* bookmark_model,
-      ScoredHistoryMatches* scored_items) const;
+  // the matches listed in |history_ids|.
+  void HistoryIdsToScoredMatches(HistoryIDVector history_ids,
+                                 const base::string16& lower_raw_string,
+                                 const TemplateURLService* template_url_service,
+                                 bookmarks::BookmarkModel* bookmark_model,
+                                 ScoredHistoryMatches* scored_items) const;
 
   // Fills in |terms_to_word_starts_offsets| according to where the word starts
   // in each term.  For example, in the term "-foo" the word starts at offset 1.
   static void CalculateWordStartsOffsets(
       const String16Vector& terms,
       WordStarts* terms_to_word_starts_offsets);
 
   // Indexes one URL history item as described by |row|. Returns true if the
   // row was actually indexed. |scheme_whitelist| is used to filter
   // non-qualifying schemes.  If |history_db| is not NULL then this function
@@ skipping 10 matching lines @@
 
   // Parses and indexes the words in the URL and page title of |row| and
   // calculate the word starts in each, saving the starts in |word_starts|.
   void AddRowWordsToIndex(const history::URLRow& row,
                           RowWordStarts* word_starts);
 
   // Given a single word in |uni_word|, adds a reference for the containing
   // history item identified by |history_id| to the index.
   void AddWordToIndex(const base::string16& uni_word, HistoryID history_id);
 
-  // Creates a new entry in the word/history map for |word_id| and add
-  // |history_id| as the initial element of the word's set.
-  void AddWordHistory(const base::string16& uni_word, HistoryID history_id);
-
-  // Updates an existing entry in the word/history index by adding the
-  // |history_id| to set for |word_id| in the word_id_history_map_.
-  void UpdateWordHistory(WordID word_id, HistoryID history_id);
-
-  // Adds |word_id| to |history_id|'s entry in the history/word map,
-  // creating a new entry if one does not already exist.
-  void AddToHistoryIDWordMap(HistoryID history_id, WordID word_id);
+  // Adds a new entry to word_list. Uses previously freed positions if

[Peter Kasting, 2017/02/25 04:54:42]

Nit: word_list -> |word_list_|

[dyaroshev, 2017/02/26 01:12:17]

Done.

+  // available.
+  WordID AddNewWordToWordList(const base::string16& term);
 
   // Removes |row| and all associated words and characters from the index.
   void RemoveRowFromIndex(const history::URLRow& row);
 
   // Removes all words and characters associated with |row| from the index.
   void RemoveRowWordsFromIndex(const history::URLRow& row);
 
   // Clears |used_| for each item in the search term cache.
   void ResetSearchTermCache();
 
@@ skipping 60 matching lines @@
   // ID of the word in the word_map_. It reduces the memory overhead by
   // replacing a potentially long and repeated string with a simple index.
   String16Vector word_list_;
 
   // A list of available words slots in |word_list_|. An available word slot
   // is the index of a unused word in word_list_ vector, also referred to as
   // a WordID. As URL visits are added or modified new words may be added to
   // the index, in which case any available words are used, if any, and then
   // words are added to the end of the word_list_. When URL visits are
   // modified or deleted old words may be removed from the index, in which
   // case the slots for those words are added to available_words_ for resuse

[Alexander Yashkin, 2017/02/21 07:07:15]

reuse

[dyaroshev, 2017/02/26 01:12:17]

Done.

   // by future URL updates.
-  WordIDSet available_words_;
+  std::stack<WordID> available_words_;
 
   // A one-to-one mapping from the a word string to its slot number (i.e.
   // WordID) in the |word_list_|.
   WordMap word_map_;
 
   // A one-to-many mapping from a single character to all WordIDs of words
   // containing that character.
   CharWordIDMap char_word_map_;
 
   // A one-to-many mapping from a WordID to all HistoryIDs (the row_id as
@@ skipping 21 matching lines @@
   int saved_cache_version_;
 
   // Used for unit testing only. Records the number of candidate history items
   // at three stages in the index searching process.
   size_t pre_filter_item_count_;    // After word index is queried.
   size_t post_filter_item_count_;   // After trimming large result set.
   size_t post_scoring_item_count_;  // After performing final filter/scoring.
 };
 
 #endif  // COMPONENTS_OMNIBOX_BROWSER_URL_INDEX_PRIVATE_DATA_H_