Improvements to tab title prefix eliding as per email discussions.

chrome/browser/ui/title_prefix_matcher.cc

Index: chrome/browser/ui/title_prefix_matcher.cc
===================================================================
--- chrome/browser/ui/title_prefix_matcher.cc	(revision 80442)
+++ chrome/browser/ui/title_prefix_matcher.cc	(working copy)
@@ -7,6 +7,7 @@
 #include "base/hash_tables.h"
 #include "base/i18n/break_iterator.h"
 #include "base/logging.h"
+#include "base/utf_string_conversions.h"
 
 namespace {
 // We use this value to identify that we have already seen the title associated
@@ -14,14 +15,23 @@
 const size_t kPreviouslySeenIndex = 0xFFFFFFFF;
 }
 
-TitlePrefixMatcher::TitleInfo::TitleInfo(const string16* title,
-                                         int caller_value)
+// static
+const int TitlePrefixMatcher::kCommonCharsToShow = 4;
+const size_t TitlePrefixMatcher::kMinElidingLength =
+    TitlePrefixMatcher::kCommonCharsToShow + 3;
+
+TitlePrefixMatcher::TitleInfo::TitleInfo(
+    const string16* title, const GURL& url, int caller_value)
     : title(title),
+      url(url),
       prefix_length(0),
       caller_value(caller_value) {
   DCHECK(title != NULL);
 }
 
+TitlePrefixMatcher::TitleInfo::~TitleInfo() {
+}
+
 // static
 void TitlePrefixMatcher::CalculatePrefixLengths(
     std::vector<TitleInfo>* title_infos) {
@@ -41,12 +51,12 @@
   for (size_t i = 0; i < title_infos->size(); ++i) {
     // We use pairs to test existence and insert in one shot.
     std::pair<base::hash_map<string16, size_t>::iterator, bool> insert_result =
-        existing_title.insert(std::make_pair(*title_infos->at(i).title, i));
+        existing_title.insert(std::make_pair(*(*title_infos)[i].title, i));
     if (!insert_result.second) {
       // insert_result.second is false when we insert a duplicate in the set.
       // insert_result.first is a map iterator and thus
       // insert_result.first->first is the string title key of the map.
-      DCHECK(*title_infos->at(i).title == insert_result.first->first);
+      DCHECK_EQ(*(*title_infos)[i].title, insert_result.first->first);
       duplicate_titles.insert(i);
       // insert_result.first->second is the value of the title index and if it's
       // not kPreviouslySeenIndex yet, we must remember it as a duplicate too.
@@ -64,7 +74,11 @@
     // Duplicate titles are not to be included in this process.
     if (duplicate_titles.find(i) != duplicate_titles.end())
       continue;
-    const string16* title = title_infos->at(i).title;
+    const TitleInfo& title_info = (*title_infos)[i];
+    const string16* title = title_info.title;
+    // We prefix the hostname at the beginning, so that we only group
+    // titles that are from the same hostname.
+    string16 hostname = ASCIIToUTF16(title_info.url.host());
     // We only create prefixes at word boundaries.
     base::BreakIterator iter(title, base::BreakIterator::BREAK_WORD);
     // We ignore this title if we can't break it into words, or if it only
@@ -76,7 +90,7 @@
     // previous word and more easily ignore the last word while iterating.
     while (iter.Advance()) {
       if (iter.IsWord())
-        prefixes[title->substr(0, iter.prev())].push_back(i);
+        prefixes[hostname + title->substr(0, iter.prev())].push_back(i);
     }
   }
 
@@ -86,10 +100,16 @@
        prefixes.begin(); iter != prefixes.end(); ++iter) {
     // iter->first is the prefix string, iter->second is a vector of indices.
     if (iter->second.size() > 1) {
-      size_t prefix_length = iter->first.size();
+      // We need to subtract the hostname size since we added it to the prefix.
+      const TitleInfo& first_title_info = (*title_infos)[iter->second[0]];
+      DCHECK_GE(iter->first.size(), first_title_info.url.host().size());
+      size_t prefix_length = iter->first.size() -
+          first_title_info.url.host().size();
       for (size_t i = 0; i < iter->second.size(); ++i){
-        if (title_infos->at(iter->second[i]).prefix_length < prefix_length)
-          title_infos->at(iter->second[i]).prefix_length = prefix_length;
+        TitleInfo& title_info = (*title_infos)[iter->second[i]];
+        DCHECK_EQ(first_title_info.url.host(), title_info.url.host());
+        if (title_info.prefix_length < prefix_length)
+          title_info.prefix_length = prefix_length;
       }
     }
   }