[core] Line-break ideographic text by character

Allow a line break to be inserted after any supported Chinese, Japanese, or Yi character in a point-placed label. Balance the lines unless non-ideographic text such as Latin letters are present.

Fixes #1223.

Author: 1ec5

cmake/core-files.cmake

@@ -451,6 +451,8 @@ set(MBGL_CORE_FILES
     src/mbgl/util/http_header.hpp
     src/mbgl/util/http_timeout.cpp
     src/mbgl/util/http_timeout.hpp
+    src/mbgl/util/i18n.cpp
+    src/mbgl/util/i18n.hpp
     src/mbgl/util/interpolate.hpp
     src/mbgl/util/intersection_tests.cpp
     src/mbgl/util/intersection_tests.hpp

package.json

@@ -23,7 +23,7 @@
     "lodash": "^4.16.4",
     "mapbox-gl-shaders": "mapbox/mapbox-gl-shaders#98a56d538b11fb331aa67a6d632d6ecd6821b007",
     "mapbox-gl-style-spec": "mapbox/mapbox-gl-style-spec#7f62a4fc9f21e619824d68abbc4b03cbc1685572",
-    "mapbox-gl-test-suite": "mapbox/mapbox-gl-test-suite#844ecd23c3314eec509dbcdaf0f956d5504e6fef",
+    "mapbox-gl-test-suite": "mapbox/mapbox-gl-test-suite#4a508b6da9bad0fe267b0ccd4440fa061da3b4f8",
     "mkdirp": "^0.5.1",
     "node-cmake": "^1.2.1",
     "request": "^2.72.0",

platform/ios/CHANGELOG.md

@@ -19,6 +19,7 @@ Mapbox welcomes participation and contributions from everyone. Please read [CONT
 * Added [quadkey](https://msdn.microsoft.com/en-us/library/bb259689.aspx) support and limited WMS support in raster tile URL templates. ([#5628](https://github.com/mapbox/mapbox-gl-native/pull/5628))
 * TileJSON manifests can now specify `"scheme": "tms"` to indicate the use of [TMS](https://en.wikipedia.org/wiki/Tile_Map_Service) coordinates. ([#2270](https://github.com/mapbox/mapbox-gl-native/pull/2270))
 * Fixed rendering artifacts and missing glyphs that occurred after viewing a large number of CJK characters on the map. ([#5908](https://github.com/mapbox/mapbox-gl-native/pull/5908))
+* Improved the line wrapping behavior of point-placed labels written in Chinese, Japanese, and Yi. ([#6828](https://github.com/mapbox/mapbox-gl-native/pull/6828))
 * `-[MGLMapView resetPosition]` now resets to the current style’s default center coordinates, zoom level, direction, and pitch, if specified. ([#6127](https://github.com/mapbox/mapbox-gl-native/pull/6127))
 * The `text-pitch-alignment` property is now supported in stylesheets for improved street label legibility on a tilted map. ([#5288](https://github.com/mapbox/mapbox-gl-native/pull/5288))
 * The `icon-text-fit` and `icon-text-fit-padding` properties are now supported in stylesheets, allowing the background of a shield to automatically resize to fit the shield’s text. ([#5334](https://github.com/mapbox/mapbox-gl-native/pull/5334))

platform/macos/CHANGELOG.md

@@ -15,6 +15,7 @@
 * GeoJSON sources specified by the stylesheet at design time now support `cluster`, `clusterMaxZoom`, and `clusterRadius` attributes for clustering point features on the base map. ([#5724](https://github.com/mapbox/mapbox-gl-native/pull/5724))
 * TileJSON manifests can now specify `"scheme": "tms"` to indicate the use of [TMS](https://en.wikipedia.org/wiki/Tile_Map_Service) coordinates. ([#2270](https://github.com/mapbox/mapbox-gl-native/pull/2270))
 * Fixed rendering artifacts and missing glyphs that occurred after viewing a large number of CJK characters on the map. ([#5908](https://github.com/mapbox/mapbox-gl-native/pull/5908))
+* Improved the line wrapping behavior of point-placed labels written in Chinese, Japanese, and Yi. ([#6828](https://github.com/mapbox/mapbox-gl-native/pull/6828))
 * Fixed an issue where the style zoom levels were not respected when deciding when to render a layer. ([#5811](https://github.com/mapbox/mapbox-gl-native/issues/5811))
 * If MGLMapView is unable to obtain or parse a style, it now calls its delegate’s `-mapViewDidFailLoadingMap:withError:` method. ([#6145](https://github.com/mapbox/mapbox-gl-native/pull/6145))
 * Added the `-[MGLMapViewDelegate mapView:didFinishLoadingStyle:]` delegate method, which offers the earliest opportunity to modify the layout or appearance of the current style before the map view is displayed to the user. ([#6636](https://github.com/mapbox/mapbox-gl-native/pull/6636))

src/mbgl/text/glyph_set.cpp

@@ -1,6 +1,7 @@
 #include <mbgl/text/glyph_set.hpp>
 #include <mbgl/platform/log.hpp>
 #include <mbgl/math/minmax.hpp>
+#include <mbgl/util/i18n.hpp>
 
 #include <cassert>
 
@@ -54,7 +55,8 @@ const Shaping GlyphSet::getShaping(const std::u32string &string, const float max
     if (shaping.positionedGlyphs.empty())
         return shaping;
 
-    lineWrap(shaping, lineHeight, maxWidth, horizontalAlign, verticalAlign, justify, translate);
+    lineWrap(shaping, lineHeight, maxWidth, horizontalAlign, verticalAlign, justify, translate,
+             util::i18n::allowsIdeographicBreaking(string));
 
     return shaping;
 }
@@ -85,9 +87,10 @@ void justifyLine(std::vector<PositionedGlyph> &positionedGlyphs, const std::map<
     }
 }
 
-void GlyphSet::lineWrap(Shaping &shaping, const float lineHeight, const float maxWidth,
-                         const float horizontalAlign, const float verticalAlign,
-                         const float justify, const Point<float> &translate) const {
+void GlyphSet::lineWrap(Shaping &shaping, const float lineHeight, float maxWidth,
+                        const float horizontalAlign, const float verticalAlign,
+                        const float justify, const Point<float> &translate,
+                        bool useBalancedIdeographicBreaking) const {
     uint32_t lastSafeBreak = 0;
 
     uint32_t lengthBeforeCurrentLine = 0;
@@ -99,6 +102,12 @@ void GlyphSet::lineWrap(Shaping &shaping, const float lineHeight, const float ma
     std::vector<PositionedGlyph> &positionedGlyphs = shaping.positionedGlyphs;
 
     if (maxWidth) {
+        if (useBalancedIdeographicBreaking) {
+            auto lastPositionedGlyph = positionedGlyphs[positionedGlyphs.size() - 1];
+            uint32_t estimatedLineCount = std::fmax(1, std::ceil(lastPositionedGlyph.x / maxWidth));
+            maxWidth = lastPositionedGlyph.x / estimatedLineCount;
+        }
+
         for (uint32_t i = 0; i < positionedGlyphs.size(); i++) {
             PositionedGlyph &shape = positionedGlyphs[i];
 
@@ -133,8 +142,9 @@ void GlyphSet::lineWrap(Shaping &shaping, const float lineHeight, const float ma
                 line++;
             }
 
-            // Spaces, plus word-breaking punctuation that often appears without surrounding spaces.
-            if (shape.glyph == 0x20 /* space */
+            // Ideographic characters, spaces, and word-breaking punctuation that often appear without surrounding spaces.
+            if (useBalancedIdeographicBreaking
+                || shape.glyph == 0x20 /* space */
                 || shape.glyph == 0x26 /* ampersand */
                 || shape.glyph == 0x2b /* plus sign */
                 || shape.glyph == 0x2d /* hyphen-minus */
@@ -143,7 +153,8 @@ void GlyphSet::lineWrap(Shaping &shaping, const float lineHeight, const float ma
                 || shape.glyph == 0xb7 /* middle dot */
                 || shape.glyph == 0x200b /* zero-width space */
                 || shape.glyph == 0x2010 /* hyphen */
-                || shape.glyph == 0x2013 /* en dash */) {
+                || shape.glyph == 0x2013 /* en dash */
+                || util::i18n::allowsIdeographicBreaking(shape.glyph)) {
                 lastSafeBreak = i;
             }
         }

src/mbgl/text/glyph_set.hpp

@@ -13,7 +13,8 @@ class GlyphSet {
                              float horizontalAlign, float verticalAlign, float justify,
                              float spacing, const Point<float> &translate) const;
     void lineWrap(Shaping &shaping, float lineHeight, float maxWidth, float horizontalAlign,
-                  float verticalAlign, float justify, const Point<float> &translate) const;
+                  float verticalAlign, float justify, const Point<float> &translate,
+                  bool useBalancedIdeographicBreaking) const;
 
 private:
     std::map<uint32_t, SDFGlyph> sdfs;

src/mbgl/util/i18n.cpp

@@ -0,0 +1,62 @@
+#include "i18n.hpp"
+
+namespace mbgl {
+namespace util {
+namespace i18n {
+
+bool allowsIdeographicBreaking(const std::u32string& string) {
+    for (uint32_t chr : string) {
+        if (!allowsIdeographicBreaking(chr)) {
+            return false;
+        }
+    }
+    return true;
+}
+
+bool allowsIdeographicBreaking(uint32_t chr) {
+    // Return early for characters outside all ideographic ranges.
+    if (chr < 0x2E80) return false;
+
+    // CJK Radicals Supplement, Kangxi Radicals, Ideographic Description Characters, CJK Symbols and Punctuation: “⺀” to “〿”
+    if (chr >= 0x2E80 && chr <= 0x303F) return true;
+
+    // Hiragana: before “ぁ” to “ゟ”
+    if (chr >= 0x3040 && chr <= 0x309F) return true;
+
+    // Katakana: “゠” to “ヿ”
+    if (chr >= 0x30A0 && chr <= 0x30FF) return true;
+
+    // CJK Strokes: “㇀” to past “㇣”
+    if (chr >= 0x31C0 && chr <= 0x31EF) return true;
+
+    // Katakana Phonetic Extensions: “ㇰ” to “ㇿ”
+    if (chr >= 0x31F0 && chr <= 0x31FF) return true;
+
+    // Enclosed CJK Letters and Months, CJK Compatibility: “㈀” to “㏿”
+    if (chr >= 0x3200 && chr <= 0x33FF) return true;
+
+    // CJK Unified Ideographs Extension A: “㐀” to past “䶵”
+    if (chr >= 0x3400 && chr <= 0x4DBF) return true;
+
+    // CJK Unified Ideographs: “一” to past “鿕”
+    if (chr >= 0x4E00 && chr <= 0x9FFF) return true;
+
+    // Yi Syllables, Yi Radicals: “ꀀ” to past “꓆”
+    if (chr >= 0xA000 && chr <= 0xA4CF) return true;
+
+    // CJK Compatibility Forms: “︰” to “﹏”
+    if (chr >= 0xFE30 && chr <= 0xFE4F) return true;
+
+    // CJK Compatibility Ideographs: “豈” to past “龎”
+    if (chr >= 0xF900 && chr <= 0xFAFF) return true;
+
+    // Halfwidth and Fullwidth Forms: before “！” to past “￮”
+    if (chr >= 0xFF00 && chr <= 0xFFEF) return true;
+
+    return false;
+}
+
+} // namespace i18n
+} // namespace util
+} // namespace mbgl
+

src/mbgl/util/i18n.hpp

@@ -0,0 +1,18 @@
+#pragma once
+
+#include <string>
+
+namespace mbgl {
+namespace util {
+namespace i18n {
+
+/// Returns whether a line break can be inserted after any character in the given string.
+/// If false, line breaking should occur on word boundaries instead.
+bool allowsIdeographicBreaking(const std::u32string& string);
+
+/// Returns whether a line break can be inserted after the character indicated by the given Unicode codepoint.
+bool allowsIdeographicBreaking(uint32_t chr);
+
+} // namespace i18n
+} // namespace util
+} // namespace mbgl