Move width() into codepoint_properties.char_width.

Author: christianparpart

src/unicode/codepoint_properties.h

@@ -26,13 +26,14 @@ namespace unicode
 
 struct LIBUNICODE_PACKED codepoint_properties
 {
+    uint8_t char_width = 0;
+    uint8_t flags = 0;
     Script script = Script::Unknown;
     Grapheme_Cluster_Break grapheme_cluster_break = Grapheme_Cluster_Break::Other;
     East_Asian_Width east_asian_width = East_Asian_Width::Narrow;
     General_Category general_category = General_Category::Unassigned;
     EmojiSegmentationCategory emoji_segmentation_category = EmojiSegmentationCategory::Invalid;
-    uint8_t flags = 0;
-    uint16_t pad = 0;
+    uint8_t pad = 0;
 
     static uint8_t constexpr FlagEmoji = 0x01;                // NOLINT(readability-identifier-naming)
     static uint8_t constexpr FlagEmojiPresentation = 0x02;    // NOLINT(readability-identifier-naming)

src/unicode/codepoint_properties_loader.cpp

@@ -360,6 +360,46 @@ namespace
         }
     };
 
+    inline uint8_t toCharWidth(codepoint_properties const& properties) noexcept
+    {
+        switch (properties.general_category)
+        {
+            case General_Category::Control: // XXX really?
+            case General_Category::Enclosing_Mark:
+            case General_Category::Format:
+            case General_Category::Line_Separator:
+            // case General_Category::Modifier_Symbol:
+            case General_Category::Nonspacing_Mark:
+            case General_Category::Paragraph_Separator:
+            case General_Category::Spacing_Mark:
+            case General_Category::Surrogate: return 0;
+            default: break;
+        }
+
+        if (properties.emoji_presentation())
+            // UAX #11 §5 Recommendations:
+            //     [UTS51] emoji presentation sequences behave as though they were East Asian Wide,
+            //     regardless of their assigned East_Asian_Width property value.
+            return 2;
+
+        switch (properties.east_asian_width)
+        {
+            case East_Asian_Width::Narrow:
+            case East_Asian_Width::Ambiguous:
+            case East_Asian_Width::Halfwidth:
+            case East_Asian_Width::Neutral:
+                //.
+                return 1;
+            case East_Asian_Width::Wide:
+            case East_Asian_Width::Fullwidth:
+                //.
+                return 2;
+        }
+
+        // Should never be reached.
+        return 1;
+    }
+
     inline EmojiSegmentationCategory toEmojiSegmentationCategory(char32_t codepoint,
                                                                  codepoint_properties const& props) noexcept
     {
@@ -397,6 +437,7 @@ namespace
 
         return EmojiSegmentationCategory::Invalid;
     }
+
     class codepoint_properties_loader
     {
       public:
@@ -493,7 +534,6 @@ namespace
                 properties(codepoint).flags |= i->second;
         });
 
-
         process_properties("extracted/DerivedGeneralCategory.txt",
                            [&](char32_t codepoint, string_view value) {
                                (void) codepoint;
@@ -552,6 +592,14 @@ namespace
                     toEmojiSegmentationCategory(codepoint, properties(codepoint));
         }
         // }}}
+
+        // {{{ assign char_width
+        {
+            auto const _ = scoped_timer { _log, "Assigning char_width" };
+            for (char32_t codepoint = 0; codepoint < 0x110'000; ++codepoint)
+                properties(codepoint).char_width = toCharWidth(properties(codepoint));
+        }
+        // }}}
     }
 
     codepoint_properties_table codepoint_properties_loader::load_from_directory(

src/unicode/tablegen.cpp

@@ -81,12 +81,13 @@ void generate_cxx_properties_table(std::ostream& header,
         // clang-format off
         auto const& properties = propertiesTable[i];
         implementation << "    {"
+                       << static_cast<unsigned>(properties.char_width) << ", "
+                       << (!properties.flags ? "0"s : fmt::format("0b{:08b}", properties.flags)) << ", "
                        << fmt::format("Script::{}, ", properties.script)
                        << fmt::format("Grapheme_Cluster_Break::{}, ", properties.grapheme_cluster_break)
                        << fmt::format("East_Asian_Width::{}, ", properties.east_asian_width)
                        << fmt::format("General_Category::{}, ", properties.general_category)
-                       << fmt::format("EmojiSegmentationCategory::{}, ", properties.emoji_segmentation_category)
-                       << (!properties.flags ? "0"s : fmt::format("0b{:08b}", properties.flags))
+                       << fmt::format("EmojiSegmentationCategory::{} ", properties.emoji_segmentation_category)
                        << "},\n";
         // clang-format on
     }

src/unicode/width.cpp

@@ -18,46 +18,9 @@
 namespace unicode
 {
 
-int width(char32_t codepoint)
+int width(char32_t codepoint) noexcept
 {
-    // Small optimization to speadup US-ASCII width calculation.
-    if (0x20 <= codepoint && codepoint <= 0xA0)
-        return 1;
-
-    // TODO: make this at most one lookup
-    auto const& properties = codepoint_properties::get(codepoint);
-
-    switch (properties.general_category)
-    {
-        case General_Category::Control: // XXX really?
-        case General_Category::Enclosing_Mark:
-        case General_Category::Format:
-        case General_Category::Line_Separator:
-        // case General_Category::Modifier_Symbol:
-        case General_Category::Nonspacing_Mark:
-        case General_Category::Paragraph_Separator:
-        case General_Category::Spacing_Mark:
-        case General_Category::Surrogate: return 0;
-        default: break;
-    }
-
-    if (properties.emoji_presentation())
-        // UAX #11 §5 Recommendations:
-        //     [UTS51] emoji presentation sequences behave as though they were East Asian Wide,
-        //     regardless of their assigned East_Asian_Width property value.
-        return 2;
-
-    switch (properties.east_asian_width)
-    {
-        case East_Asian_Width::Narrow:
-        case East_Asian_Width::Ambiguous:
-        case East_Asian_Width::Halfwidth:
-        case East_Asian_Width::Neutral: return 1;
-        case East_Asian_Width::Wide:
-        case East_Asian_Width::Fullwidth: return 2;
-    }
-
-    return 1;
+    return codepoint_properties::get(codepoint).char_width;
 }
 
 } // namespace unicode

src/unicode/width.h

@@ -17,6 +17,6 @@ namespace unicode
 {
 
 /// Returns the number of text columns the given codepoint would need to be displayed.
-int width(char32_t codepoint);
+int width(char32_t codepoint) noexcept;
 
 } // namespace unicode