3 files changed, 58 insertions, 22 deletions

diff --git a/base/i18n/break_iterator.cc b/base/i18n/break_iterator.cc
index e1b5e29..edc8950 100644
--- a/base/i18n/break_iterator.cc
+++ b/base/i18n/break_iterator.cc
@@ -1,4 +1,4 @@
-// Copyright (c) 2009 The Chromium Authors. All rights reserved.
+// Copyright (c) 2011 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.
 
@@ -33,7 +33,7 @@ bool BreakIterator::Init() {
     case BREAK_WORD:
       break_type = UBRK_WORD;
       break;
-    case BREAK_SPACE:
+    case BREAK_LINE:
     case BREAK_NEWLINE:
       break_type = UBRK_LINE;
       break;
@@ -59,7 +59,7 @@ bool BreakIterator::Advance() {
   prev_ = pos_;
   switch (break_type_) {
     case BREAK_WORD:
-    case BREAK_SPACE:
+    case BREAK_LINE:
       pos = ubrk_next(static_cast<UBreakIterator*>(iter_));
       if (pos == UBRK_DONE) {
         pos_ = npos;
diff --git a/base/i18n/break_iterator.h b/base/i18n/break_iterator.h
index 9de7ac7..f64a1e1 100644
--- a/base/i18n/break_iterator.h
+++ b/base/i18n/break_iterator.h
@@ -1,4 +1,4 @@
-// Copyright (c) 2010 The Chromium Authors. All rights reserved.
+// Copyright (c) 2011 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.
 
@@ -12,19 +12,28 @@
 // The BreakIterator class iterates through the words, word breaks, and
 // line breaks in a UTF-16 string.
 //
-// It provides several modes, BREAK_WORD, BREAK_SPACE, and BREAK_NEWLINE,
+// It provides several modes, BREAK_WORD, BREAK_LINE, and BREAK_NEWLINE,
 // which modify how characters are aggregated into the returned string.
 //
 // Under BREAK_WORD mode, once a word is encountered any non-word
 // characters are not included in the returned string (e.g. in the
 // UTF-16 equivalent of the string " foo bar! ", the word breaks are at
 // the periods in ". .foo. .bar.!. .").
+// Note that Chinese/Japanese/Thai do not use spaces between words so that
+// boundaries can fall in the middle of a continuous run of non-space /
+// non-punctuation characters.
 //
-// Under BREAK_SPACE mode, once a word is encountered, any non-word
-// characters are included in the returned string, breaking only when a
-// space-equivalent character is encountered (e.g. in the
-// UTF16-equivalent of the string " foo bar! ", the word breaks are at
-//   the periods in ". .foo .bar! .").
+// Under BREAK_LINE mode, once a line breaking opportunity is encountered,
+// any non-word  characters are included in the returned string, breaking
+// only when a space-equivalent character or a line breaking opportunity
+// is encountered (e.g. in the UTF16-equivalent of the string " foo bar! ",
+// the breaks are at the periods in ". .foo .bar! .").
+//
+// Note that lines can be broken at any character/syllable/grapheme cluster
+// boundary in Chinese/Japanese/Korean and at word boundaries in Thai
+// (Thai does not use spaces between words). Therefore, this is NOT the same
+// as breaking only at space-equivalent characters where its former
+// name (BREAK_SPACE) implied.
 //
 // Under BREAK_NEWLINE mode, all characters are included in the returned
 // string, breking only when a newline-equivalent character is encountered
@@ -48,7 +57,11 @@ class BreakIterator {
  public:
   enum BreakType {
     BREAK_WORD,
-    BREAK_SPACE,
+    BREAK_LINE,
+    // TODO(jshin): Remove this after reviewing call sites.
+    // If call sites really need break only on space-like characters
+    // implement it separately.
+    BREAK_SPACE = BREAK_LINE,
     BREAK_NEWLINE,
   };
 
@@ -75,7 +88,7 @@ class BreakIterator {
 
   // Under BREAK_WORD mode, returns true if the break we just hit is the
   // end of a word. (Otherwise, the break iterator just skipped over e.g.
-  // whitespace or punctuation.)  Under BREAK_SPACE and BREAK_NEWLINE modes,
+  // whitespace or punctuation.)  Under BREAK_LINE and BREAK_NEWLINE modes,
   // this distinction doesn't apply and it always retuns false.
   bool IsWord() const;
 
diff --git a/chrome/browser/ui/views/download_item_view.cc b/chrome/browser/ui/views/download_item_view.cc
index b644718..d119264 100644
--- a/chrome/browser/ui/views/download_item_view.cc
+++ b/chrome/browser/ui/views/download_item_view.cc
@@ -8,6 +8,7 @@
 
 #include "base/callback.h"
 #include "base/file_path.h"
+#include "base/i18n/break_iterator.h"
 #include "base/i18n/rtl.h"
 #include "base/metrics/histogram.h"
 #include "base/string_util.h"
@@ -28,6 +29,7 @@
 #include "ui/gfx/canvas_skia.h"
 #include "ui/gfx/color_utils.h"
 #include "ui/gfx/image.h"
+#include "unicode/uchar.h"
 #include "views/controls/button/native_button.h"
 #include "views/controls/menu/menu_2.h"
 #include "views/widget/root_view.h"
@@ -1047,28 +1049,49 @@ void DownloadItemView::SizeLabelToMinWidth() {
 
   gfx::Size size;
   int min_width = -1;
-  size_t sp_index = text.find(L" ");
-  while (sp_index != std::wstring::npos) {
-    text.replace(sp_index, 1, L"\n");
-    dangerous_download_label_->SetText(text);
+  string16 text16 = WideToUTF16(text);
+  // Using BREAK_WORD can work in most cases, but it can also break
+  // lines where it should not. Using BREAK_LINE is safer although
+  // slower for Chinese/Japanese. This is not perf-critical at all, though.
+  base::BreakIterator iter(&text16, base::BreakIterator::BREAK_LINE);
+  bool status = iter.Init();
+  DCHECK(status);
+
+  string16 current_text = text16;
+  string16 prev_text = text16;
+  while (iter.Advance()) {
+    size_t pos = iter.pos();
+    if (pos >= text16.length())
+      break;
+    // This can be a low surrogate codepoint, but u_isUWhiteSpace will
+    // return false and inserting a new line after a surrogate pair
+    // is perfectly ok.
+    char16 line_end_char = text16[pos - 1];
+    if (u_isUWhiteSpace(line_end_char))
+      current_text.replace(pos - 1, 1, 1, char16('\n'));
+    else
+      current_text.insert(pos, 1, char16('\n'));
+    dangerous_download_label_->SetText(UTF16ToWide(current_text));
     size = dangerous_download_label_->GetPreferredSize();
 
     if (min_width == -1)
       min_width = size.width();
 
     // If the width is growing again, it means we passed the optimal width spot.
-    if (size.width() > min_width)
+    if (size.width() > min_width) {
+      dangerous_download_label_->SetText(UTF16ToWide(prev_text));
       break;
-    else
+    } else {
       min_width = size.width();
+    }
 
     // Restore the string.
-    text.replace(sp_index, 1, L" ");
-
-    sp_index = text.find(L" ", sp_index + 1);
+    prev_text = current_text;
+    current_text = text16;
   }
 
-  // If we have a line with no space, we won't cut it.
+  // If we have a line with no line breaking opportunity (which is very
+  // unlikely), we won't cut it.
   if (min_width == -1)
     size = dangerous_download_label_->GetPreferredSize();