Generating autocomplete results with and without word breaks in the Omnibox.

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>
@@ skipping 32 matching lines @@
 // this class is for exclusive use by the InMemoryURLIndex class there should
 // be no calls from any other class.
 //
 // All public member functions are called on the main thread unless otherwise
 // annotated.
 class URLIndexPrivateData
     : public base::RefCountedThreadSafe<URLIndexPrivateData> {
  public:
   URLIndexPrivateData();
 
-  // Given a base::string16 in |term_string|, scans the history index and
-  // returns a vector with all scored, matching history items. The
-  // |term_string| is broken down into individual terms (words), each of which
-  // must occur in the candidate history item's URL or page title for the item
-  // to qualify; however, the terms do not necessarily have to be adjacent. We
-  // also allow breaking |term_string| at |cursor_position| (if
-  // set). Once we have a set of candidates, they are filtered to ensure
-  // that all |term_string| terms, as separated by whitespace and the
-  // cursor (if set), occur within the candidate's URL or page title.
-  // Scores are then calculated on no more than |kItemsToScoreLimit|
-  // candidates, as the scoring of such a large number of candidates may
-  // cause perceptible typing response delays in the omnibox. This is
-  // likely to occur for short omnibox terms such as 'h' and 'w' which
+  // Given |term_string|, scans the history index and returns a vector with

[Mark P, 2016/08/24 20:15:03]

nit: the "a" here that was before "|term_string|" reads fine.  Please put it
back.
(I think the difference is that |term_string| sounds singular whereas in the
other place you put a similar comment, the variable is |lower_words|, which
sound plural.)

[Lavar Askew, 2016/08/24 20:23:50]

Done.

+  // all scored, matching history items. The |term_string| is broken down into
+  // individual terms (words), each of which must occur in the candidate
+  // history item's URL or page title for the item to qualify; however, the
+  // terms do not necessarily have to be adjacent. We also allow breaking
+  // |term_string| at |cursor_position| (if set). Once we have a set of
+  // candidates, they are filtered to ensure that all |term_string| terms, as
+  // separated by whitespace and the cursor (if set), occur within the
+  // candidate's URL or page title.  Scores are then calculated on no more than
+  // |kItemsToScoreLimit| candidates, as the scoring of such a large number of
+  // candidates may cause perceptible typing response delays in the omnibox.
+  // This is likely to occur for short omnibox terms such as 'h' and 'w' which
   // will be found in nearly all history candidates. Results are sorted by
   // descending score. The full results set (i.e. beyond the
   // |kItemsToScoreLimit| limit) will be retained and used for subsequent calls
   // to this function. In total, |max_matches| of items will be returned in the
   // |ScoredHistoryMatches| vector.
   ScoredHistoryMatches HistoryItemsForTerms(
       base::string16 term_string,
       size_t cursor_position,
       size_t max_matches,
       bookmarks::BookmarkModel* bookmark_model,
@@ skipping 243 matching lines @@
       const in_memory_url_index::InMemoryURLIndexCacheItem& cache);
   bool RestoreHistoryInfoMap(
       const in_memory_url_index::InMemoryURLIndexCacheItem& cache);
   bool RestoreWordStartsMap(
       const in_memory_url_index::InMemoryURLIndexCacheItem& cache);
 
   // Determines if |gurl| has a whitelisted scheme and returns true if so.
   static bool URLSchemeIsWhitelisted(const GURL& gurl,
                                      const std::set<std::string>& whitelist);
 
+  // Support functions for HistoryItemsForTerms
+
+  // Returns a vector of individual words from |search_string|, in order of
+  // appearance, duplicates allowed.
+  String16Vector ExtractIndividualWordVector(
+      const base::string16& search_string);
+
+  // Given |lower_words|, scans the history index and returns the set of
+  // matching history item IDs.  Will return an empty HistoryIDSet if
+  // |lower_words| is empty.
+  HistoryIDSet GetHistoryIDSet(const String16Vector& lower_words);
+
+  // This function is responsible for splitting |search_string| into a vector
+  // of terms based on 'true' whitespace as opposed to escaped whitespace,
+  // e.g., When the user types "colspec=ID%20Mstone Release" we get two
+  //'terms': "colspec=id%20mstone" and "release".  Each URL in the user's
+  // history is scored against the vector of terms.  The matches are returned
+  // sorted based on highest score.
+  ScoredHistoryMatches GetScoredItemsForSearchString(
+      const base::string16& search_string,
+      const HistoryIDSet& history_id_set,
+      const size_t max_matches,
+      bookmarks::BookmarkModel* bookmark_model,
+      TemplateURLService* template_url_service);
+
   // Cache of search terms.
   SearchTermCacheMap search_term_cache_;
 
   // Start of data members that are cached -------------------------------------
 
   // The version of the cache file most recently used to restore this instance
   // of the private data. If the private data was rebuilt from the history
   // database this will be 0.
   int restored_cache_version_;
 
@@ skipping 48 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_