Omnibox: Create Field Trial for HQP to Ignore Mid-Word Matches

chrome/browser/history/scored_history_match.h

Index: chrome/browser/history/scored_history_match.h
diff --git a/chrome/browser/history/scored_history_match.h b/chrome/browser/history/scored_history_match.h
index c10f4f3b70d881f3c5dc81cd1d5b95c685ca436b..fc2a5e8cdfb2edf4af61ad6f6cc8614cb6d4265c 100644
--- a/chrome/browser/history/scored_history_match.h
+++ b/chrome/browser/history/scored_history_match.h
@@ -38,12 +38,26 @@ struct ScoredHistoryMatch : public history::HistoryMatch {
                      BookmarkService* bookmark_service);
   ~ScoredHistoryMatch();
 
-  // Calculates a component score based on position, ordering and total
-  // substring match size using metrics recorded in |matches|. |max_length|
-  // is the length of the string against which the terms are being searched.
-  static int ScoreComponentForMatches(const TermMatches& matches,
+  // Calculates a component score based on position, ordering, word
+  // boundaries, and total substring match size using metrics recorded
+  // in |matches| and |word_starts|. |max_length| is the length of
+  // the string against which the terms are being searched.
+  // |provided_matches| should already be sorted and de-dupped, and

[Peter Kasting, 2012/12/01 01:21:44]

Nit: dupped -> duped (multiple places)

[Mark P, 2012/12/01 01:55:32]

Fixed three locations.

+  // |word_starts| must be sorted.
+  static int ScoreComponentForMatches(const TermMatches& provided_matches,

[Peter Kasting, 2012/12/01 01:21:44]

Nit: |provided_matches| here versus |matches| in the next function should be
named identically

[Mark P, 2012/12/01 01:55:32]

Done.

Also rename variable in the implementation.

+                                      const WordStarts& word_starts,
                                       size_t max_length);
 
+  // Given a set of term matches |matches| and word boundaries |word_starts|,
+  // fills in |matches_at_word_boundaries| with only the matches in |matches|
+  // that are at word boundaries.
+  // |matches| should already be sorted and de-dupped, and ||word_starts|

[Peter Kasting, 2012/12/01 01:21:44]

Nit: || -> |

[Mark P, 2012/12/01 01:55:32]

Done.

+  // must be sorted.
+  static void MakeTermMatchesOnlyAtWordBoundaries(
+      const TermMatches& matches,
+      const WordStarts& word_starts,
+      TermMatches* matches_at_word_boundaries);
+
   // Converts a raw value for some particular scoring factor into a score
   // component for that factor.  The conversion function is piecewise linear,
   // with input values provided in |value_ranks| and resulting output scores
@@ -74,7 +88,9 @@ struct ScoredHistoryMatch : public history::HistoryMatch {
   // boundaries).  |url_matches| and |title_matches| provide details
   // about where the matches in the URL and title are and what terms
   // (identified by a term number < |num_terms|) match where.
-  // |word_starts| explains where word boundaries are.
+  // |word_starts| explains where word boundaries are.  Its parts (title
+  // and url) must be sorted.  Also, |url_matches| and
+  // |titles_matches| should already be sorted and de-dupped.
   static float GetTopicalityScore(const int num_terms,
                                   const string16& url,
                                   const TermMatches& url_matches,
@@ -102,6 +118,9 @@ struct ScoredHistoryMatch : public history::HistoryMatch {
   // field trial state.
   static void InitializeNewScoringField();
 
+  // Sets only_count_matches_at_word_boundaries based on the field trial state.
+  static void InitializeOnlyCountMatchesAtWordBoundariesField();
+
   // Sets also_do_hup_like_scoring based on the field trial state.
   static void InitializeAlsoDoHUPLikeScoringField();
 
@@ -134,7 +153,7 @@ struct ScoredHistoryMatch : public history::HistoryMatch {
   static const int kMaxRawTermScore = 30;
   static float* raw_term_score_to_topicality_score;
 
-  // Allows us to determing setting for use_new_scoring_ only once.
+  // Used so we initialize static variables only once (on first use).
   static bool initialized_;
 
   // Whether to use new-scoring or old-scoring.  Set in the
@@ -145,6 +164,9 @@ struct ScoredHistoryMatch : public history::HistoryMatch {
   // class as well (see boolean below).
   static bool use_new_scoring;
 
+  // If true, we ignore all matches that are in the middle of a word.
+  static bool only_count_matches_at_word_boundaries;
+
   // If true, assign raw scores to be max(whatever it normally would be,
   // a score that's similar to the score HistoryURL provider would assign).
   // This variable is set in the constructor by examining the field trial