LibRegex: Merge inverse string table mappings separately

Author: https://github.com/alimpfard
Commit: 69050da929
Pull-request: https://github.com/LadybirdBrowser/ladybird/pull/4246

Libraries/LibRegex/RegexByteCode.h

@@ -261,6 +261,7 @@ public:
     }
 
     FlyString get_string(size_t index) const { return m_string_table.get(index); }
+    auto const& string_table() const { return m_string_table; }
 
     void last_chunk() const = delete;
     void first_chunk() const = delete;
@@ -279,6 +280,8 @@ public:
                 }
                 m_string_table.m_table.set(entry.key, entry.value);
             }
+            for (auto const& entry : other.m_string_table.m_inverse_table)
+                m_string_table.m_inverse_table.set(entry.key, entry.value);
         }
     }
 

Libraries/LibRegex/RegexDebug.h

@@ -51,6 +51,13 @@ public:
             state.instruction_position += opcode.size();
         }
 
+        out(m_file, "String Table:\n");
+        for (auto const& entry : bytecode.string_table().m_table)
+            outln(m_file, "+ {} -> {:x}\n", entry.key, entry.value);
+        out(m_file, "Reverse String Table:\n");
+        for (auto const& entry : bytecode.string_table().m_inverse_table)
+            outln(m_file, "+ {:x} -> {}\n", entry.key, entry.value);
+
         fflush(m_file);
     }
 

Libraries/LibRegex/RegexOptimizer.cpp

@@ -999,11 +999,10 @@ void Optimizer::append_alternation(ByteCode& target, Span<ByteCode> alternatives
     if (alternatives.size() == 0)
         return;
 
-    target.merge_string_tables_from(alternatives);
-
     if (alternatives.size() == 1)
         return target.extend(move(alternatives[0]));
 
+    target.merge_string_tables_from(alternatives);
     if (all_of(alternatives, [](auto& x) { return x.is_empty(); }))
         return;
 

Tests/LibRegex/Regex.cpp

@@ -743,6 +743,8 @@ TEST_CASE(ECMA262_match)
         { "(?=)(?=\\d)"sv, "smart"sv, false },
         // Backrefs are cleared after lookaheads, the indices should be checked before lookup.
         { "(?!(b))\\1"sv, "a"sv, false },
+        // String table merge bug: inverse map should be merged regardless of available direct mappings.
+        { "((?<x>a)|(?<x>b))"sv, "aa"sv, false },
     };
 
     for (auto& test : tests) {