Meta+LibUnicode: Avoid relocations for static unicode data

Previously the s_decomposition_mappings variable would refer to other
data in s_decomposition_mappings_data. This would cause thousands of
avoidable relocations at load time.

This saves about 128kB RAM for each process which uses LibUnicode.

Meta/Lagom/Tools/CodeGenerators/LibUnicode/GenerateUnicodeData.cpp

@@ -768,6 +768,13 @@ struct SpecialCasing {
  Condition condition { Condition::None };
 };
 
+struct CodePointDecompositionRaw {
+ u32 code_point { 0 };
+ CompatibilityFormattingTag tag { CompatibilityFormattingTag::Canonical };
+ size_t decomposition_index { 0 };
+ size_t decomposition_count { 0 };
+};
+
 struct CodePointDecomposition {
  u32 code_point { 0 };
  CompatibilityFormattingTag tag { CompatibilityFormattingTag::Canonical };
@@ -947,7 +954,7 @@ static constexpr Array<@mapping_type@, @size@> s_@name@_mappings { {
  generator.set("tag", mapping->tag);
  generator.set("start", String::number(mapping->decomposition_index));
  generator.set("size", String::number(mapping->decomposition_size));
- generator.append(", CompatibilityFormattingTag::@tag@, Span<u32 const> { s_decomposition_mappings_data.data() + @start@, @size@ } },");
+ generator.append(", CompatibilityFormattingTag::@tag@, @start@, @size@ },");
  } else {
  append_list_and_size(data.special_casing_indices, "&s_special_casing[{}]"sv);
  generator.append(" },");
@@ -974,7 +981,7 @@ static constexpr Array<@mapping_type@, @size@> s_@name@_mappings { {
  append_code_point_mappings("special_case"sv, "SpecialCaseMapping"sv, unicode_data.code_points_with_special_casing, [](auto const& data) { return data.special_casing_indices; });
  append_code_point_mappings("abbreviation"sv, "CodePointAbbreviation"sv, unicode_data.code_point_abbreviations.size(), [](auto const& data) { return data.abbreviation; });
 
- append_code_point_mappings("decomposition"sv, "CodePointDecomposition"sv, unicode_data.code_points_with_decomposition_mapping,
+ append_code_point_mappings("decomposition"sv, "CodePointDecompositionRaw"sv, unicode_data.code_points_with_decomposition_mapping,
  [](auto const& data) {
  return data.decomposition_mapping;
  });
@@ -1153,17 +1160,20 @@ Optional<StringView> code_point_abbreviation(u32 code_point)
  return decode_string(mapping->abbreviation);
 }
 
-Optional<CodePointDecomposition const&> code_point_decomposition(u32 code_point)
+Optional<CodePointDecomposition const> code_point_decomposition(u32 code_point)
 {
- auto const* mapping = binary_search(s_decomposition_mappings, code_point, nullptr, CodePointComparator<CodePointDecomposition> {});
+ auto const* mapping = binary_search(s_decomposition_mappings, code_point, nullptr, CodePointComparator<CodePointDecompositionRaw> {});
  if (mapping == nullptr)
  return {};
- return *mapping;
+ return CodePointDecomposition { mapping->code_point, mapping->tag, Span<u32 const> { s_decomposition_mappings_data.data() + mapping->decomposition_index, mapping->decomposition_count } };
 }
 
-Span<CodePointDecomposition const> code_point_decompositions()
+Optional<CodePointDecomposition const> code_point_decomposition_by_index(size_t index)
 {
- return s_decomposition_mappings;
+ if (index >= s_decomposition_mappings.size())
+ return {};
+ auto const& mapping = s_decomposition_mappings[index];
+ return CodePointDecomposition { mapping.code_point, mapping.tag, Span<u32 const> { s_decomposition_mappings_data.data() + mapping.decomposition_index, mapping.decomposition_count } };
 }
 )~~~");
 

Userland/Libraries/LibUnicode/Normalize.cpp

@@ -14,8 +14,8 @@
 
 namespace Unicode {
 
-Optional<CodePointDecomposition const&> __attribute__((weak)) code_point_decomposition(u32) { return {}; }
-Span<CodePointDecomposition const> __attribute__((weak)) code_point_decompositions() { return {}; }
+Optional<CodePointDecomposition const> __attribute__((weak)) code_point_decomposition(u32) { return {}; }
+Optional<CodePointDecomposition const> __attribute__((weak)) code_point_decomposition_by_index(size_t) { return {}; }
 
 NormalizationForm normalization_form_from_string(StringView form)
 {
@@ -122,7 +122,11 @@ static u32 combine_code_points(u32 a, u32 b)
 {
  Array<u32, 2> const points { a, b };
  // FIXME: Do something better than linear search to find reverse mappings.
- for (auto const& mapping : Unicode::code_point_decompositions()) {
+ for (size_t index = 0;; ++index) {
+ auto mapping_maybe = Unicode::code_point_decomposition_by_index(index);
+ if (!mapping_maybe.has_value())
+ break;
+ auto& mapping = mapping_maybe.value();
  if (mapping.tag == CompatibilityFormattingTag::Canonical && mapping.decomposition == points) {
  if (code_point_has_property(mapping.code_point, Property::Full_Composition_Exclusion))
  continue;

Userland/Libraries/LibUnicode/Normalize.h

@@ -15,8 +15,8 @@
 
 namespace Unicode {
 
-Optional<CodePointDecomposition const&> code_point_decomposition(u32 code_point);
-Span<CodePointDecomposition const> code_point_decompositions();
+Optional<CodePointDecomposition const> code_point_decomposition(u32 code_point);
+Optional<CodePointDecomposition const> code_point_decomposition_by_index(size_t index);
 
 enum class NormalizationForm {
  NFD,