[LayoutNG] Don't remove segment break between Hangul characters

Spec (https://drafts.csswg.org/css-text-3/#line-break-transform)
says that, during whitespace collapsing, segment break between
Hangul characters should be converted into a space instead of
removed. This patch implements that.

Bug: 636993
Cq-Include-Trybots: luci.chromium.try:linux_layout_tests_layout_ng;luci.chromium.try:linux_layout_tests_slimming_paint_v2;master.tryserver.blink:linux_trusty_blink_rel
Change-Id: Ie61a223ea400a20166ea3ecd779904d4ae57d67a
Reviewed-on: https://chromium-review.googlesource.com/1112756Reviewed-by: Koji Ishii <kojii@chromium.org>
Reviewed-by: Emil A Eklund <eae@chromium.org>
Commit-Queue: Xiaocheng Hu <xiaochengh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#570130}

third_party/WebKit/LayoutTests/FlagExpectations/enable-blink-features=LayoutNG

@@ -146,9 +146,6 @@ crbug.com/591099 external/wpt/css/css-text/line-breaking/line-breaking-011.html
 crbug.com/591099 external/wpt/css/css-text/line-breaking/line-breaking-ic-002.html [ Pass ]
 crbug.com/591099 external/wpt/css/css-text/line-breaking/line-breaking-ic-003.html [ Pass ]
 crbug.com/591099 external/wpt/css/css-text/white-space/pre-wrap-002.html [ Pass ]
-crbug.com/591099 external/wpt/css/css-text/white-space/seg-break-transformation-010.html [ Failure ]
-crbug.com/591099 external/wpt/css/css-text/white-space/seg-break-transformation-011.html [ Failure ]
-crbug.com/591099 external/wpt/css/css-text/white-space/seg-break-transformation-012.html [ Failure ]
 crbug.com/591099 external/wpt/css/css-text/word-break/word-break-break-all-004.html [ Pass ]
 crbug.com/591099 external/wpt/css/css-transforms/transform-transformed-tr-percent-height-child.html [ Failure ]
 crbug.com/591099 external/wpt/css/css-transforms/transform3d-perspective-008.html [ Pass ]

third_party/blink/renderer/core/layout/ng/inline/ng_inline_items_builder.cc

@@ -95,19 +95,21 @@ bool ShouldRemoveNewlineSlow(const StringBuilder& before,
   if (before.Is8Bit() || after.Is8Bit())
     return false;
 
-  // Remove if East Asian Widths of both before/after the newline are Wide.
+  // Remove if East Asian Widths of both before/after the newline are Wide, and
+  // neither side is Hangul.
+  // TODO(layout-dev): Don't remove if any side is Emoji.
   if (U16_IS_TRAIL(last) && space_index >= 2) {
     UChar last_last = before[space_index - 2];
     if (U16_IS_LEAD(last_last))
       last = U16_GET_SUPPLEMENTARY(last_last, last);
   }
-  if (IsEastAsianWidthWide(last, before_style)) {
+  if (!Character::IsHangul(last) && IsEastAsianWidthWide(last, before_style)) {
     if (U16_IS_LEAD(next) && after.length() > 1) {
       UChar next_next = after[1];
       if (U16_IS_TRAIL(next_next))
         next = U16_GET_SUPPLEMENTARY(next, next_next);
     }
-    if (IsEastAsianWidthWide(next, after_style))
+    if (!Character::IsHangul(next) && IsEastAsianWidthWide(next, after_style))
       return true;
   }
 

third_party/blink/renderer/platform/text/character.cc

@@ -122,6 +122,10 @@ bool Character::IsBidiControl(UChar32 character) {
   RETURN_HAS_PROPERTY(character, kIsBidiControl);
 }
 
+bool Character::IsHangulSlow(UChar32 character) {
+  RETURN_HAS_PROPERTY(character, kIsHangul);
+}
+
 unsigned Character::ExpansionOpportunityCount(const LChar* characters,
                                               size_t length,
                                               TextDirection direction,

third_party/blink/renderer/platform/text/character.h

@@ -72,6 +72,11 @@ class PLATFORM_EXPORT Character {
            !(U_GET_GC_MASK(c) & (U_GC_M_MASK | U_GC_LM_MASK | U_GC_SK_MASK));
   }
 
+  static bool IsHangul(UChar32 c) {
+    // Below U+1100 is likely a common case.
+    return c < 0x1100 ? false : IsHangulSlow(c);
+  }
+
   static unsigned ExpansionOpportunityCount(const LChar*,
                                             size_t length,
                                             TextDirection,
@@ -182,6 +187,7 @@ class PLATFORM_EXPORT Character {
 
  private:
   static bool IsCJKIdeographOrSymbolSlow(UChar32);
+  static bool IsHangulSlow(UChar32);
 };
 
 }  // namespace blink

third_party/blink/renderer/platform/text/character_property.h

@@ -16,6 +16,7 @@ enum class CharacterProperty : CharacterPropertyType {
   kIsUprightInMixedVertical = 0x0002,
   kIsPotentialCustomElementNameChar = 0x0004,
   kIsBidiControl = 0x0008,
+  kIsHangul = 0x0010
 };
 
 inline CharacterProperty operator|(CharacterProperty a, CharacterProperty b) {

third_party/blink/renderer/platform/text/character_property_data_generator.cc

@@ -82,6 +82,7 @@ static void Generate(FILE* fp) {
   SET(kIsUprightInMixedVertical);
   SET(kIsPotentialCustomElementNameChar);
   SET(kIsBidiControl);
+  SET(kIsHangul);
 
   // Create a trie from the value array.
   UErrorCode error = U_ZERO_ERROR;

third_party/blink/renderer/platform/text/character_property_data_generator.h

@@ -230,6 +230,25 @@ static const UChar32 kIsBidiControlRanges[] = {
     0x202A, 0x202E, 0x2066, 0x2069,
 };
 
+// https://unicode.org/Public/UNIDATA/Blocks.txt
+static const UChar32 kIsHangulRanges[] = {
+    // Hangul Jamo
+    0x1100, 0x11FF,
+    // Hangul Compatibility Jamo
+    0x3130, 0x318F,
+    // Hangul Jamo Extended-A
+    0xA960, 0xA97F,
+    // Hangul Syllables
+    0xAC00, 0xD7AF,
+    // Hangul Jamo Extended-B
+    0xD7B0, 0xD7FF,
+    // Halfwidth Hangul Jamo
+    // https://www.unicode.org/charts/nameslist/c_FF00.html
+    0xFFA0, 0xFFDC,
+};
+
+static const UChar32 kIsHangulArray[] = {};
+
 }  // namespace blink
 
 #endif