LibRegex: Support UTF-16 RegexStringView and improve Unicode matching

When the Unicode option is not set, regular expressions should match
based on code units; when it is set, they should match based on code
points. To do so, the regex parser must combine surrogate pairs when
the Unicode option is set. Further, RegexStringView needs to know if
the flag is set in order to return code point vs. code unit based
string lengths and substrings.

Tests/LibRegex/Regex.cpp

@@ -506,10 +506,14 @@ TEST_CASE(ECMA262_parse)
         { ",(?", regex::Error::InvalidCaptureGroup }, // #4583
         { "{1}", regex::Error::InvalidPattern },
         { "{1,2}", regex::Error::InvalidPattern },
+        { "\\uxxxx", regex::Error::NoError },
+        { "\\uxxxx", regex::Error::InvalidPattern, ECMAScriptFlags::Unicode },
+        { "\\ud83d", regex::Error::NoError, ECMAScriptFlags::Unicode },
+        { "\\ud83d\\uxxxx", regex::Error::InvalidPattern, ECMAScriptFlags::Unicode },
     };
 
     for (auto& test : tests) {
-        Regex<ECMA262> re(test.pattern);
+        Regex<ECMA262> re(test.pattern, test.flags);
         EXPECT_EQ(re.parser_result.error, test.expected_error);
         if constexpr (REGEX_DEBUG) {
             dbgln("\n");
@@ -586,6 +590,45 @@ TEST_CASE(ECMA262_match)
     }
 }
 
+TEST_CASE(ECMA262_unicode_match)
+{
+    struct _test {
+        char const* pattern;
+        char const* subject;
+        bool matches { true };
+        ECMAScriptFlags options {};
+    };
+    _test tests[] {
+        { "\\ud83d", "😀", true },
+        { "\\ud83d", "😀", false, ECMAScriptFlags::Unicode },
+        { "\\ude00", "😀", true },
+        { "\\ude00", "😀", false, ECMAScriptFlags::Unicode },
+        { "\\ud83d\\ude00", "😀", true },
+        { "\\ud83d\\ude00", "😀", true, ECMAScriptFlags::Unicode },
+        { "\\ud83d\\ud83d", "\xed\xa0\xbd\xed\xa0\xbd", true },
+        { "\\ud83d\\ud83d", "\xed\xa0\xbd\xed\xa0\xbd", true, ECMAScriptFlags::Unicode },
+    };
+
+    for (auto& test : tests) {
+        Regex<ECMA262> re(test.pattern, (ECMAScriptFlags)regex::AllFlags::Global | test.options);
+
+        auto subject = AK::utf8_to_utf16(test.subject);
+        Utf16View view { subject };
+
+        if constexpr (REGEX_DEBUG) {
+            dbgln("\n");
+            RegexDebug regex_dbg(stderr);
+            regex_dbg.print_raw_bytecode(re);
+            regex_dbg.print_header();
+            regex_dbg.print_bytecode(re);
+            dbgln("\n");
+        }
+
+        EXPECT_EQ(re.parser_result.error, Error::NoError);
+        EXPECT_EQ(re.match(view).success, test.matches);
+    }
+}
+
 TEST_CASE(replace)
 {
     struct _test {

Userland/Libraries/LibRegex/RegexByteCode.cpp

@@ -465,12 +465,13 @@ ALWAYS_INLINE ExecutionResult OpCode_Compare::execute(MatchInput const& input, M
                 return ExecutionResult::Failed_ExecuteLowPrioForks;
 
             Optional<String> str;
+            Vector<u16> utf16;
             Vector<u32> data;
             data.ensure_capacity(length);
             for (size_t i = offset; i < offset + length; ++i)
                 data.unchecked_append(m_bytecode->at(i));
 
-            auto view = input.view.construct_as_same(data, str);
+            auto view = input.view.construct_as_same(data, str, utf16);
             offset += length;
             if (!compare_string(input, state, view, had_zero_length_match))
                 return ExecutionResult::Failed_ExecuteLowPrioForks;
@@ -553,7 +554,8 @@ ALWAYS_INLINE void OpCode_Compare::compare_char(MatchInput const& input, MatchSt
 
     auto input_view = input.view.substring_view(state.string_position, 1);
     Optional<String> str;
-    auto compare_view = input_view.construct_as_same({ &ch1, 1 }, str);
+    Vector<u16> utf16;
+    auto compare_view = input_view.construct_as_same({ &ch1, 1 }, str, utf16);
     bool equal;
     if (input.regex_options & AllFlags::Insensitive)
         equal = input_view.equals_ignoring_case(compare_view);

Userland/Libraries/LibRegex/RegexMatch.h

@@ -14,6 +14,7 @@
 #include <AK/String.h>
 #include <AK/StringBuilder.h>
 #include <AK/StringView.h>
+#include <AK/Utf16View.h>
 #include <AK/Utf32View.h>
 #include <AK/Utf8View.h>
 #include <AK/Variant.h>
@@ -43,6 +44,11 @@ public:
     {
     }
 
+    RegexStringView(Utf16View view)
+        : m_view(view)
+    {
+    }
+
     RegexStringView(Utf8View view)
         : m_view(view)
     {
@@ -58,11 +64,19 @@ public:
         return m_view.get<Utf32View>();
     }
 
+    Utf16View const& u16_view() const
+    {
+        return m_view.get<Utf16View>();
+    }
+
     Utf8View const& u8_view() const
     {
         return m_view.get<Utf8View>();
     }
 
+    bool unicode() const { return m_unicode; }
+    void set_unicode(bool unicode) { m_unicode = unicode; }
+
     bool is_empty() const
     {
         return m_view.visit([](auto& view) { return view.is_empty(); });
@@ -75,12 +89,21 @@ public:
 
     size_t length() const
     {
-        return m_view.visit([](auto& view) { return view.length(); });
+        if (unicode()) {
+            return m_view.visit(
+                [](Utf16View const& view) { return view.length_in_code_points(); },
+                [](auto const& view) { return view.length(); });
+        }
+
+        return m_view.visit(
+            [](Utf16View const& view) { return view.length_in_code_units(); },
+            [](Utf8View const& view) { return view.byte_length(); },
+            [](auto const& view) { return view.length(); });
     }
 
-    RegexStringView construct_as_same(Span<u32> data, Optional<String>& optional_string_storage) const
+    RegexStringView construct_as_same(Span<u32> data, Optional<String>& optional_string_storage, Vector<u16>& optional_utf16_storage) const
     {
-        return m_view.visit(
+        auto view = m_view.visit(
             [&]<typename T>(T const&) {
                 StringBuilder builder;
                 for (auto ch : data)
@@ -90,7 +113,14 @@ public:
             },
             [&](Utf32View) {
                 return RegexStringView { Utf32View { data.data(), data.size() } };
+            },
+            [&](Utf16View) {
+                optional_utf16_storage = AK::utf32_to_utf16(Utf32View { data.data(), data.size() });
+                return RegexStringView { Utf16View { optional_utf16_storage } };
             });
+
+        view.set_unicode(unicode());
+        return view;
     }
 
     Vector<RegexStringView> lines() const
@@ -118,6 +148,21 @@ public:
                     views.empend(view);
                 return views;
             },
+            [](Utf16View view) {
+                Vector<RegexStringView> views;
+                u16 newline = '\n';
+                while (!view.is_empty()) {
+                    auto position = AK::memmem_optional(view.data(), view.length_in_code_units() * sizeof(u16), &newline, sizeof(u16));
+                    if (!position.has_value())
+                        break;
+                    auto offset = position.value() / sizeof(u16);
+                    views.empend(view.substring_view(0, offset));
+                    view = view.substring_view(offset + 1, view.length_in_code_units() - offset - 1);
+                }
+                if (!view.is_empty())
+                    views.empend(view);
+                return views;
+            },
             [](Utf8View& view) {
                 Vector<RegexStringView> views;
                 auto it = view.begin();
@@ -147,15 +192,26 @@ public:
 
     RegexStringView substring_view(size_t offset, size_t length) const
     {
-        return m_view.visit(
-            [&](auto view) { return RegexStringView { view.substring_view(offset, length) }; },
-            [&](Utf8View const& view) { return RegexStringView { view.unicode_substring_view(offset, length) }; });
+        if (unicode()) {
+            auto view = m_view.visit(
+                [&](auto view) { return RegexStringView { view.substring_view(offset, length) }; },
+                [&](Utf16View const& view) { return RegexStringView { view.unicode_substring_view(offset, length) }; },
+                [&](Utf8View const& view) { return RegexStringView { view.unicode_substring_view(offset, length) }; });
+
+            view.set_unicode(unicode());
+            return view;
+        }
+
+        auto view = m_view.visit([&](auto view) { return RegexStringView { view.substring_view(offset, length) }; });
+        view.set_unicode(unicode());
+        return view;
     }
 
     String to_string() const
     {
         return m_view.visit(
             [](StringView view) { return view.to_string(); },
+            [](Utf16View view) { return view.to_utf8(Utf16View::AllowInvalidCodeUnits::Yes); },
             [](auto& view) {
                 StringBuilder builder;
                 for (auto it = view.begin(); it != view.end(); ++it)
@@ -173,8 +229,8 @@ public:
                     return 256u + ch;
                 return ch;
             },
-            [&](auto view) -> u32 { return view[index]; },
-            [&](Utf8View& view) -> u32 {
+            [&](Utf32View& view) -> u32 { return view[index]; },
+            [&](auto& view) -> u32 {
                 size_t i = index;
                 for (auto it = view.begin(); it != view.end(); ++it, --i) {
                     if (i == 0)
@@ -188,6 +244,7 @@ public:
     {
         return m_view.visit(
             [&](Utf32View) { return to_string() == cstring; },
+            [&](Utf16View) { return to_string() == cstring; },
             [&](Utf8View const& view) { return view.as_string() == cstring; },
             [&](StringView view) { return view == cstring; });
     }
@@ -201,6 +258,7 @@ public:
     {
         return m_view.visit(
             [&](Utf32View) { return to_string() == string; },
+            [&](Utf16View) { return to_string() == string; },
             [&](Utf8View const& view) { return view.as_string() == string; },
             [&](StringView view) { return view == string; });
     }
@@ -209,6 +267,7 @@ public:
     {
         return m_view.visit(
             [&](Utf32View) { return to_string() == string; },
+            [&](Utf16View) { return to_string() == string; },
             [&](Utf8View const& view) { return view.as_string() == string; },
             [&](StringView view) { return view == string; });
     }
@@ -224,6 +283,7 @@ public:
             [&](Utf32View view) {
                 return view.length() == other.length() && __builtin_memcmp(view.code_points(), other.code_points(), view.length() * sizeof(u32)) == 0;
             },
+            [&](Utf16View) { return to_string() == RegexStringView { other }.to_string(); },
             [&](Utf8View const& view) { return view.as_string() == RegexStringView { other }.to_string(); },
             [&](StringView view) { return view == RegexStringView { other }.to_string(); });
     }
@@ -233,12 +293,25 @@ public:
         return !(*this == other);
     }
 
+    bool operator==(Utf16View const& other) const
+    {
+        return m_view.visit(
+            [&](Utf32View) { return to_string() == RegexStringView { other }.to_string(); },
+            [&](Utf16View const& view) { return view == other; },
+            [&](Utf8View const& view) { return view.as_string() == RegexStringView { other }.to_string(); },
+            [&](StringView view) { return view == RegexStringView { other }.to_string(); });
+    }
+
+    bool operator!=(Utf16View const& other) const
+    {
+        return !(*this == other);
+    }
+
     bool operator==(Utf8View const& other) const
     {
         return m_view.visit(
-            [&](Utf32View) {
-                return to_string() == other.as_string();
-            },
+            [&](Utf32View) { return to_string() == other.as_string(); },
+            [&](Utf16View) { return to_string() == other.as_string(); },
             [&](Utf8View const& view) { return view.as_string() == other.as_string(); },
             [&](StringView view) { return other.as_string() == view; });
     }
@@ -271,6 +344,9 @@ public:
             [&](Utf32View) -> bool {
                 TODO();
             },
+            [&](Utf16View) -> bool {
+                TODO();
+            },
             [&](Utf8View const& view) { return view.as_string().starts_with(str); },
             [&](StringView view) { return view.starts_with(str); });
     }
@@ -289,6 +365,7 @@ public:
                 }
                 return true;
             },
+            [&](Utf16View) -> bool { TODO(); },
             [&](Utf8View const& view) {
                 auto it = view.begin();
                 for (auto code_point : str) {
@@ -304,7 +381,8 @@ public:
     }
 
 private:
-    Variant<StringView, Utf8View, Utf32View> m_view;
+    Variant<StringView, Utf8View, Utf16View, Utf32View> m_view;
+    bool m_unicode { false };
 };
 
 class Match final {

Userland/Libraries/LibRegex/RegexMatcher.cpp

@@ -84,6 +84,10 @@ RegexResult Matcher<Parser>::match(Vector<RegexStringView> const views, Optional
     output.operations = 0;
     size_t lines_to_skip = 0;
 
+    bool unicode = input.regex_options.has_flag_set(AllFlags::Unicode);
+    for (auto& view : views)
+        const_cast<RegexStringView&>(view).set_unicode(unicode);
+
     if (input.regex_options.has_flag_set(AllFlags::Internal_Stateful)) {
         if (views.size() > 1 && input.start_offset > views.first().length()) {
             dbgln_if(REGEX_DEBUG, "Started with start={}, goff={}, skip={}", input.start_offset, input.global_offset, lines_to_skip);

Userland/Libraries/LibRegex/RegexParser.cpp

@@ -10,6 +10,7 @@
 #include <AK/String.h>
 #include <AK/StringBuilder.h>
 #include <AK/StringUtils.h>
+#include <AK/Utf16View.h>
 
 namespace regex {
 
@@ -1440,13 +1441,31 @@ bool ECMA262Parser::parse_atom_escape(ByteCode& stack, size_t& match_length_mini
 
     if (try_skip("u")) {
         if (auto code_point = read_digits(ReadDigitsInitialZeroState::Allow, true, 4); code_point.has_value()) {
-            // FIXME: The minimum length depends on the mode - should be utf8-length in u8 mode.
+            // In Unicode mode, we need to combine surrogate pairs into a single code point. But we also need to be
+            // rather forgiving if the surrogate pairs are invalid. So if a second code unit follows this code unit,
+            // but doesn't form a valid surrogate pair, insert bytecode for both code units individually.
+            Optional<u32> low_surrogate;
+            if (unicode && Utf16View::is_high_surrogate(*code_point) && try_skip("\\u")) {
+                low_surrogate = read_digits(ReadDigitsInitialZeroState::Allow, true, 4);
+                if (!low_surrogate.has_value()) {
+                    set_error(Error::InvalidPattern);
+                    return false;
+                }
+
+                if (Utf16View::is_low_surrogate(*low_surrogate)) {
+                    *code_point = Utf16View::decode_surrogate_pair(*code_point, *low_surrogate);
+                    low_surrogate.clear();
+                }
+            }
+
             match_length_minimum += 1;
-            StringBuilder builder;
-            builder.append_code_point(code_point.value());
-            // FIXME: This isn't actually correct for ECMAScript.
-            auto u8_encoded = builder.string_view();
-            stack.insert_bytecode_compare_string(u8_encoded);
+            stack.insert_bytecode_compare_values({ { CharacterCompareType::Char, (ByteCodeValueType)code_point.value() } });
+
+            if (low_surrogate.has_value()) {
+                match_length_minimum += 1;
+                stack.insert_bytecode_compare_values({ { CharacterCompareType::Char, (ByteCodeValueType)low_surrogate.value() } });
+            }
+
             return true;
         } else if (!unicode) {
             // '\u' is allowed in non-unicode mode, just matches 'u'.