This is a reapply of

http://src.chromium.org/viewvc/chrome?view=rev&revision=102006

fix know issues in RenderText

1. add tests.
2. change SelectWord() to use BreakIterator, so it works for Chinese and Complex script.
3. DELETE/ReplaceChar delete/replace a whole grapheme. ReplaceTextInternal should only replace one grapheme when there is no selection.
4. pointing to position outside of text returns HOME/END position.
5. based on Chrome Linux omnibox and gedit, given 
"abc|   def", double click should select "   " instead of "abc". Change test expectation.

BUG=90426
TEST=compile with touchui=1 test omnibox. run views_unittests.NativeTextfieldViewsTest/TextfieldViewsModelTest
Review URL: http://codereview.chromium.org/7841056

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@102160 0039d316-1c4b-4281-b951-d872f2087c98

base/i18n/break_iterator.cc

@@ -92,6 +92,27 @@ bool BreakIterator::IsWord() const {
   return (break_type_ == BREAK_WORD && status != UBRK_WORD_NONE);
 }
 
+bool BreakIterator::IsEndOfWord(size_t position) const {
+  if (break_type_ != BREAK_WORD)
+    return false;
+
+  UBreakIterator* iter = static_cast<UBreakIterator*>(iter_);
+  UBool boundary = ubrk_isBoundary(iter, static_cast<int32_t>(position));
+  int32_t status = ubrk_getRuleStatus(iter);
+  return (!!boundary && status != UBRK_WORD_NONE);
+}
+
+bool BreakIterator::IsStartOfWord(size_t position) const {
+  if (break_type_ != BREAK_WORD)
+    return false;
+
+  UBreakIterator* iter = static_cast<UBreakIterator*>(iter_);
+  UBool boundary = ubrk_isBoundary(iter, static_cast<int32_t>(position));
+  ubrk_next(iter);
+  int32_t next_status = ubrk_getRuleStatus(iter);
+  return (!!boundary && next_status != UBRK_WORD_NONE);
+}
+
 string16 BreakIterator::GetString() const {
   DCHECK(prev_ != npos && pos_ != npos);
   return string_.substr(prev_, pos_ - prev_);

base/i18n/break_iterator.h

@@ -88,6 +88,12 @@ class BASE_I18N_EXPORT BreakIterator {
   // this distinction doesn't apply and it always retuns false.
   bool IsWord() const;
 
+  // Under BREAK_WORD mode, returns true if |position| is at the end of word or
+  // at the start of word. It always retuns false under BREAK_LINE and
+  // BREAK_NEWLINE modes.
+  bool IsEndOfWord(size_t position) const;
+  bool IsStartOfWord(size_t position) const;
+
   // Returns the string between prev() and pos().
   // Advance() must have been called successfully at least once for pos() to
   // have advanced to somewhere useful.

ui/gfx/render_text.cc

@@ -247,6 +247,12 @@ bool RenderText::MoveCursorTo(const SelectionModel& selection_model) {
     sel.set_caret_pos(end.caret_pos());
     sel.set_caret_placement(end.caret_placement());
   }
+
+  if (!IsCursorablePosition(sel.selection_start()) ||
+      !IsCursorablePosition(sel.selection_end()) ||
+      !IsCursorablePosition(sel.caret_pos()))
+    return false;
+
   bool changed = !sel.Equals(selection_model_);
   SetSelectionModel(sel);
   return changed;
@@ -281,41 +287,28 @@ void RenderText::SelectAll() {
   SetSelectionModel(sel);
 }
 
-// TODO(xji): it does not work for languages do not use space as word breaker,
-// such as Chinese. Should use BreakIterator.
 void RenderText::SelectWord() {
-  size_t selection_start = GetSelectionStart();
   size_t cursor_position = GetCursorPosition();
-  // First we setup selection_start_ and selection_end_. There are so many cases
-  // because we try to emulate what select-word looks like in a gtk textfield.
-  // See associated testcase for different cases.
-  if (cursor_position > 0 && cursor_position < text().length()) {
-    if (u_isalnum(text()[cursor_position])) {
-      selection_start = cursor_position;
-      cursor_position++;
-    } else
-      selection_start = cursor_position - 1;
-  } else if (cursor_position == 0) {
-    selection_start = cursor_position;
-    if (text().length() > 0)
-      cursor_position++;
-  } else {
-    selection_start = cursor_position - 1;
-  }
 
-  // Now we move selection_start_ to beginning of selection. Selection boundary
-  // is defined as the position where we have alpha-num character on one side
-  // and non-alpha-num char on the other side.
-  for (; selection_start > 0; selection_start--) {
-    if (IsPositionAtWordSelectionBoundary(selection_start))
+  base::i18n::BreakIterator iter(text(), base::i18n::BreakIterator::BREAK_WORD);
+  bool success = iter.Init();
+  DCHECK(success);
+  if (!success)
+    return;
+
+  size_t selection_start = cursor_position;
+  for (; selection_start != 0; --selection_start) {
+    if (iter.IsStartOfWord(selection_start) ||
+        iter.IsEndOfWord(selection_start))
       break;
   }
 
-  // Now we move selection_end_ to end of selection. Selection boundary
-  // is defined as the position where we have alpha-num character on one side
-  // and non-alpha-num char on the other side.
-  for (; cursor_position < text().length(); cursor_position++) {
-    if (IsPositionAtWordSelectionBoundary(cursor_position))
+  if (selection_start == cursor_position)
+    ++cursor_position;
+
+  for (; cursor_position < text().length(); ++cursor_position) {
+    if (iter.IsEndOfWord(cursor_position) ||
+        iter.IsStartOfWord(cursor_position))
       break;
   }
 
@@ -462,6 +455,10 @@ const Rect& RenderText::GetUpdatedCursorBounds() {
   return cursor_bounds_;
 }
 
+size_t RenderText::GetIndexOfNextGrapheme(size_t position) {
+  return IndexOfAdjacentGrapheme(position, true);
+}
+
 RenderText::RenderText()
     : text_(),
       selection_model_(),
@@ -557,8 +554,7 @@ SelectionModel RenderText::RightEndSelectionModel() {
 }
 
 size_t RenderText::GetIndexOfPreviousGrapheme(size_t position) {
-  // TODO(msw): Handle complex script.
-  return std::max(static_cast<long>(position - 1), static_cast<long>(0));
+  return IndexOfAdjacentGrapheme(position, false);
 }
 
 std::vector<Rect> RenderText::GetSubstringBounds(size_t from, size_t to) {
@@ -632,13 +628,10 @@ void RenderText::MoveCursorTo(size_t position, bool select) {
   SelectionModel::CaretPlacement placement = (caret_pos == cursor) ?
       SelectionModel::LEADING : SelectionModel::TRAILING;
   size_t selection_start = select ? GetSelectionStart() : cursor;
-  SelectionModel sel(selection_start, cursor, caret_pos, placement);
-  SetSelectionModel(sel);
-}
-
-bool RenderText::IsPositionAtWordSelectionBoundary(size_t pos) {
-  return pos == 0 || (u_isalnum(text()[pos - 1]) && !u_isalnum(text()[pos])) ||
-      (!u_isalnum(text()[pos - 1]) && u_isalnum(text()[pos]));
+  if (IsCursorablePosition(cursor)) {
+    SelectionModel sel(selection_start, cursor, caret_pos, placement);
+    SetSelectionModel(sel);
+  }
 }
 
 void RenderText::UpdateCachedBoundsAndOffset() {

ui/gfx/render_text.h

@@ -193,7 +193,9 @@ class UI_EXPORT RenderText {
   // Set the selection_model_ to the value of |selection|.
   // The selection model components are modified if invalid.
   // Returns true if the cursor position or selection range changed.
-  // TODO(xji): need to check the cursor is set at grapheme boundary.
+  // If |selectin_start_| or |selection_end_| or |caret_pos_| in
+  // |selection_model| is not a cursorable position (not on grapheme boundary),
+  // it is a NO-OP and returns false.
   bool MoveCursorTo(const SelectionModel& selection_model);
 
   // Move the cursor to the position associated with the clicked point.
@@ -254,6 +256,9 @@ class UI_EXPORT RenderText {
   // Subsequent text, cursor, or bounds changes may invalidate returned values.
   const Rect& GetUpdatedCursorBounds();
 
+  // Get the logical index of the grapheme following the argument |position|.
+  virtual size_t GetIndexOfNextGrapheme(size_t position);
+
  protected:
   RenderText();
 
@@ -288,6 +293,10 @@ class UI_EXPORT RenderText {
   // TODO(msw) Re-evaluate this function's necessity and signature.
   virtual std::vector<Rect> GetSubstringBounds(size_t from, size_t to);
 
+  // Return true if cursor can appear in front of the character at |position|,
+  // which means it is a grapheme boundary or the first character in the text.
+  virtual bool IsCursorablePosition(size_t position) = 0;
+
   // Apply composition style (underline) to composition range and selection
   // style (foreground) to selection range.
   void ApplyCompositionAndSelectionStyles(StyleRanges* style_ranges) const;
@@ -305,16 +314,20 @@ class UI_EXPORT RenderText {
   FRIEND_TEST_ALL_PREFIXES(RenderTextTest, ApplyStyleRange);
   FRIEND_TEST_ALL_PREFIXES(RenderTextTest, StyleRangesAdjust);
 
+  // Return an index belonging to the |next| or previous logical grapheme.
+  // The return value is bounded by 0 and the text length, inclusive.
+  virtual size_t IndexOfAdjacentGrapheme(size_t index, bool next) = 0;
+
   // Sets the selection model, the argument is assumed to be valid.
   void SetSelectionModel(const SelectionModel& selection_model);
 
   // Set the cursor to |position|, with the caret trailing the previous
   // grapheme, or if there is no previous grapheme, leading the cursor position.
   // If |select| is false, the selection start is moved to the same position.
+  // If the |position| is not a cursorable position (not on grapheme boundary),
+  // it is a NO-OP.
   void MoveCursorTo(size_t position, bool select);
 
-  bool IsPositionAtWordSelectionBoundary(size_t pos);
-
   // Update the cached bounds and display offset to ensure that the current
   // cursor is within the visible display area.
   void UpdateCachedBoundsAndOffset();

ui/gfx/render_text_linux.cc

@@ -123,13 +123,19 @@ void RenderTextLinux::Draw(Canvas* canvas) {
 }
 
 SelectionModel RenderTextLinux::FindCursorPosition(const Point& point) {
-  // TODO(xji): when points outside of text, return HOME/END position.
   PangoLayout* layout = EnsureLayout();
 
   if (text().empty())
     return SelectionModel(0, 0, SelectionModel::LEADING);
 
   Point p(ToTextPoint(point));
+
+  // When the point is outside of text, return HOME/END position.
+  if (p.x() < 0)
+    return LeftEndSelectionModel();
+  else if (p.x() > GetStringWidth())
+    return RightEndSelectionModel();
+
   int caret_pos, trailing;
   pango_layout_xy_to_index(layout, p.x() * PANGO_SCALE, p.y() * PANGO_SCALE,
                            &caret_pos, &trailing);
@@ -212,7 +218,7 @@ SelectionModel RenderTextLinux::LeftEndSelectionModel() {
         return SelectionModel(text().length(), caret, SelectionModel::LEADING);
       } else {  // RTL.
         size_t caret = Utf16IndexOfAdjacentGrapheme(item->offset + item->length,
-                                                    PREVIOUS);
+                                                    false);
         return SelectionModel(text().length(), caret, SelectionModel::TRAILING);
       }
     }
@@ -227,7 +233,7 @@ SelectionModel RenderTextLinux::RightEndSelectionModel() {
       PangoItem* item = last_visual_run->item;
       if (item->analysis.level % 2 == 0) {  // LTR.
         size_t caret = Utf16IndexOfAdjacentGrapheme(item->offset + item->length,
-                                                    PREVIOUS);
+                                                    false);
         return SelectionModel(text().length(), caret, SelectionModel::TRAILING);
       } else {  // RTL.
         size_t caret = Utf8IndexToUtf16Index(item->offset);
@@ -238,16 +244,18 @@ SelectionModel RenderTextLinux::RightEndSelectionModel() {
   return SelectionModel(0, 0, SelectionModel::LEADING);
 }
 
-size_t RenderTextLinux::GetIndexOfPreviousGrapheme(size_t position) {
+bool RenderTextLinux::IsCursorablePosition(size_t position) {
+  if (position == 0 && text().empty())
+    return true;
+
   EnsureLayout();
-  size_t index = Utf16IndexToUtf8Index(position);
-  return Utf16IndexOfAdjacentGrapheme(index, PREVIOUS);
+  return (position < static_cast<size_t>(num_log_attrs_) &&
+          log_attrs_[position].is_cursor_position);
 }
 
-size_t RenderTextLinux::GetIndexOfNextGrapheme(size_t position) {
+size_t RenderTextLinux::IndexOfAdjacentGrapheme(size_t index, bool next) {
   EnsureLayout();
-  size_t index = Utf16IndexToUtf8Index(position);
-  return Utf16IndexOfAdjacentGrapheme(index, NEXT);
+  return Utf16IndexOfAdjacentGrapheme(Utf16IndexToUtf8Index(index), next);
 }
 
 GSList* RenderTextLinux::GetRunContainingPosition(size_t position) const {
@@ -266,12 +274,12 @@ GSList* RenderTextLinux::GetRunContainingPosition(size_t position) const {
 
 size_t RenderTextLinux::Utf8IndexOfAdjacentGrapheme(
     size_t utf8_index_of_current_grapheme,
-    RelativeLogicalPosition pos) const {
+    bool next) const {
   const char* ch = layout_text_ + utf8_index_of_current_grapheme;
   int char_offset = static_cast<int>(g_utf8_pointer_to_offset(layout_text_,
                                                               ch));
   int start_char_offset = char_offset;
-  if (pos == PREVIOUS) {
+  if (!next) {
     if (char_offset > 0) {
       do {
         --char_offset;
@@ -292,23 +300,23 @@ size_t RenderTextLinux::Utf8IndexOfAdjacentGrapheme(
 
 size_t RenderTextLinux::Utf16IndexOfAdjacentGrapheme(
     size_t utf8_index_of_current_grapheme,
-    RelativeLogicalPosition pos) const {
+    bool next) const {
   size_t utf8_index = Utf8IndexOfAdjacentGrapheme(
-      utf8_index_of_current_grapheme, pos);
+      utf8_index_of_current_grapheme, next);
   return Utf8IndexToUtf16Index(utf8_index);
 }
 
 SelectionModel RenderTextLinux::FirstSelectionModelInsideRun(
     const PangoItem* item) const {
   size_t caret = Utf8IndexToUtf16Index(item->offset);
-  size_t cursor = Utf16IndexOfAdjacentGrapheme(item->offset, NEXT);
+  size_t cursor = Utf16IndexOfAdjacentGrapheme(item->offset, true);
   return SelectionModel(cursor, caret, SelectionModel::TRAILING);
 }
 
 SelectionModel RenderTextLinux::LastSelectionModelInsideRun(
     const PangoItem* item) const {
   size_t caret = Utf16IndexOfAdjacentGrapheme(item->offset + item->length,
-                                              PREVIOUS);
+                                              false);
   return SelectionModel(caret, caret, SelectionModel::LEADING);
 }
 

ui/gfx/render_text_linux.h

@@ -39,28 +39,22 @@ class RenderTextLinux : public RenderText {
                                                 BreakType break_type) OVERRIDE;
   virtual SelectionModel LeftEndSelectionModel() OVERRIDE;
   virtual SelectionModel RightEndSelectionModel() OVERRIDE;
-  virtual size_t GetIndexOfPreviousGrapheme(size_t position) OVERRIDE;
+  virtual bool IsCursorablePosition(size_t position) OVERRIDE;
 
  private:
-  enum RelativeLogicalPosition {
-    PREVIOUS,
-    NEXT
-  };
-
-  // Get the logical start index of the next grapheme after |position|.
-  size_t GetIndexOfNextGrapheme(size_t position);
+  virtual size_t IndexOfAdjacentGrapheme(size_t index, bool next) OVERRIDE;
 
   // Returns the run that contains |position|. Return NULL if not found.
   GSList* GetRunContainingPosition(size_t position) const;
 
   // Given |utf8_index_of_current_grapheme|, returns the UTF8 or UTF16 index of
-  // next grapheme in the text if |pos| is NEXT, otherwise, returns the index of
-  // previous grapheme. Returns 0 if there is no previous grapheme, and returns
-  // the |text_| length if there is no next grapheme.
+  // next grapheme in the text if |next| is true, otherwise, returns the index
+  // of previous grapheme. Returns 0 if there is no previous grapheme, and
+  // returns the |text_| length if there is no next grapheme.
   size_t Utf8IndexOfAdjacentGrapheme(size_t utf8_index_of_current_grapheme,
-                                     RelativeLogicalPosition pos) const;
+                                     bool next) const;
   size_t Utf16IndexOfAdjacentGrapheme(size_t utf8_index_of_current_grapheme,
-                                      RelativeLogicalPosition pos) const;
+                                      bool next) const;
 
   // Given a |run|, returns the SelectionModel that contains the logical first
   // or last caret position inside (not at a boundary of) the run.

ui/gfx/render_text_win.cc

@@ -237,10 +237,6 @@ SelectionModel RenderTextWin::RightEndSelectionModel() {
   return SelectionModel(cursor, caret, placement);
 }
 
-size_t RenderTextWin::GetIndexOfPreviousGrapheme(size_t position) {
-  return IndexOfAdjacentGrapheme(position, false);
-}
-
 std::vector<Rect> RenderTextWin::GetSubstringBounds(size_t from, size_t to) {
   ui::Range range(from, to);
   DCHECK(ui::Range(0, text().length()).Contains(range));
@@ -295,6 +291,41 @@ std::vector<Rect> RenderTextWin::GetSubstringBounds(size_t from, size_t to) {
   return bounds;
 }
 
+bool RenderTextWin::IsCursorablePosition(size_t position) {
+  if (position == 0 || position == text().length())
+    return true;
+
+  size_t run_index = GetRunContainingPosition(position);
+  if (run_index >= runs_.size())
+    return false;
+
+  internal::TextRun* run = runs_[run_index];
+  size_t start = run->range.start();
+  if (position == start)
+    return true;
+  return run->logical_clusters[position - start] !=
+         run->logical_clusters[position - start - 1];
+}
+
+size_t RenderTextWin::IndexOfAdjacentGrapheme(size_t index, bool next) {
+  size_t run_index = GetRunContainingPosition(index);
+  internal::TextRun* run = run_index < runs_.size() ? runs_[run_index] : NULL;
+  long start = run ? run->range.start() : 0;
+  long length = run ? run->range.length() : text().length();
+  long ch = index - start;
+  WORD cluster = run ? run->logical_clusters[ch] : 0;
+
+  if (!next) {
+    do {
+      ch--;
+    } while (ch >= 0 && run && run->logical_clusters[ch] == cluster);
+  } else {
+    while (ch < length && run && run->logical_clusters[ch] == cluster)
+      ch++;
+  }
+  return std::max(static_cast<long>(std::min(ch, length) + start), 0L);
+}
+
 void RenderTextWin::ItemizeLogicalText() {
   text_is_dirty_ = false;
   STLDeleteContainerPointers(runs_.begin(), runs_.end());
@@ -461,34 +492,16 @@ size_t RenderTextWin::GetRunContainingPoint(const Point& point) const {
   return run;
 }
 
-size_t RenderTextWin::IndexOfAdjacentGrapheme(size_t index, bool next) const {
-  size_t run_index = GetRunContainingPosition(index);
-  internal::TextRun* run = run_index < runs_.size() ? runs_[run_index] : NULL;
-  long start = run ? run->range.start() : 0;
-  long length = run ? run->range.length() : text().length();
-  long ch = index - start;
-  WORD cluster = run ? run->logical_clusters[ch] : 0;
-
-  if (!next) {
-    do {
-      ch--;
-    } while (ch >= 0 && run && run->logical_clusters[ch] == cluster);
-  } else {
-    while (ch < length && run && run->logical_clusters[ch] == cluster)
-      ch++;
-  }
-  return std::max(static_cast<long>(std::min(ch, length) + start), 0L);
-}
 
 SelectionModel RenderTextWin::FirstSelectionModelInsideRun(
-    internal::TextRun* run) const {
+    internal::TextRun* run) {
   size_t caret = run->range.start();
   size_t cursor = IndexOfAdjacentGrapheme(caret, true);
   return SelectionModel(cursor, caret, SelectionModel::TRAILING);
 }
 
 SelectionModel RenderTextWin::LastSelectionModelInsideRun(
-    internal::TextRun* run) const {
+    internal::TextRun* run) {
   size_t caret = IndexOfAdjacentGrapheme(run->range.end(), false);
   return SelectionModel(caret, caret, SelectionModel::LEADING);
 }

ui/gfx/render_text_win.h

@@ -72,10 +72,12 @@ class RenderTextWin : public RenderText {
                                                 BreakType break_type) OVERRIDE;
   virtual SelectionModel LeftEndSelectionModel() OVERRIDE;
   virtual SelectionModel RightEndSelectionModel() OVERRIDE;
-  virtual size_t GetIndexOfPreviousGrapheme(size_t position) OVERRIDE;
   virtual std::vector<Rect> GetSubstringBounds(size_t from, size_t to) OVERRIDE;
+  virtual bool IsCursorablePosition(size_t position) OVERRIDE;
 
  private:
+  virtual size_t IndexOfAdjacentGrapheme(size_t index, bool next) OVERRIDE;
+
   void ItemizeLogicalText();
   void LayoutVisualText(HDC hdc);
 
@@ -84,15 +86,11 @@ class RenderTextWin : public RenderText {
   size_t GetRunContainingPosition(size_t position) const;
   size_t GetRunContainingPoint(const Point& point) const;
 
-  // Return an index belonging to the |next| or previous logical grapheme.
-  // The return value is bounded by 0 and the text length, inclusive.
-  size_t IndexOfAdjacentGrapheme(size_t index, bool next) const;
-
   // Given a |run|, returns the SelectionModel that contains the logical first
   // or last caret position inside (not at a boundary of) the run.
   // The returned value represents a cursor/caret position without a selection.
-  SelectionModel FirstSelectionModelInsideRun(internal::TextRun*) const;
-  SelectionModel LastSelectionModelInsideRun(internal::TextRun*) const;
+  SelectionModel FirstSelectionModelInsideRun(internal::TextRun*);
+  SelectionModel LastSelectionModelInsideRun(internal::TextRun*);
 
   // Get the selection model visually left/right of |selection| by one grapheme.
   // The returned value represents a cursor/caret position without a selection.

views/controls/textfield/textfield_views_model.cc

@@ -320,7 +320,10 @@ void TextfieldViewsModel::Append(const string16& text) {
   if (HasCompositionText())
     ConfirmCompositionText();
   size_t save = GetCursorPosition();
-  MoveCursorRight(gfx::LINE_BREAK, false);
+  if (render_text_->GetTextDirection() == base::i18n::LEFT_TO_RIGHT)
+    MoveCursorRight(gfx::LINE_BREAK, false);
+  else
+    MoveCursorLeft(gfx::LINE_BREAK, false);
   InsertText(text);
   render_text_->SetCursorPosition(save);
   ClearSelection();
@@ -338,7 +341,9 @@ bool TextfieldViewsModel::Delete() {
   }
   if (GetText().length() > GetCursorPosition()) {
     size_t cursor_position = GetCursorPosition();
-    ExecuteAndRecordDelete(cursor_position, cursor_position + 1, true);
+    size_t next_grapheme_index =
+        render_text_->GetIndexOfNextGrapheme(cursor_position);
+    ExecuteAndRecordDelete(cursor_position, next_grapheme_index, true);
     return true;
   }
   return false;
@@ -656,7 +661,7 @@ void TextfieldViewsModel::ReplaceTextInternal(const string16& text,
   } else if (!HasSelection()) {
     size_t cursor = GetCursorPosition();
     gfx::SelectionModel sel(render_text_->selection_model());
-    sel.set_selection_start(cursor + text.length());
+    sel.set_selection_start(render_text_->GetIndexOfNextGrapheme(cursor));
     render_text_->MoveCursorTo(sel);
   }
   // Edit history is recorded in InsertText.