Cleanup regex lookarounds implementation and reject incompatible regex

Fixes #2487

src/regex_impl.cc

@@ -8,6 +8,7 @@
 #include "utf8_iterator.hh"
 #include "string_utils.hh"
 #include "vector.hh"
+#include "utils.hh"
 
 #include <cstring>
 
@@ -602,12 +603,17 @@ private:
 
     void validate_lookaround(NodeIndex index)
     {
+        using Lookaround = CompiledRegex::Lookaround;
         ForEachChild<>::apply(m_parsed_regex, index, [this](NodeIndex child_index) {
             auto& child = get_node(child_index);
             if (child.op != ParsedRegex::Literal and child.op != ParsedRegex::Class and
                 child.op != ParsedRegex::CharacterType and child.op != ParsedRegex::AnyChar and
                 child.op != ParsedRegex::AnyCharExceptNewLine)
                 parse_error("Lookaround can only contain literals, any chars or character classes");
+            if (child.op == ParsedRegex::Literal and
+                to_underlying(Lookaround::OpBegin) <= child.value and
+                child.value < to_underlying(Lookaround::OpEnd))
+                parse_error("Lookaround does not support literals codepoint between 0xF0000 and 0xFFFFD");
             if (child.quantifier.type != ParsedRegex::Quantifier::One)
                 parse_error("Quantifiers cannot be used in lookarounds");
             return true;
@@ -877,20 +883,22 @@ private:
     template<MatchDirection direction>
     uint32_t push_lookaround(ParsedRegex::NodeIndex index, bool ignore_case)
     {
+        using Lookaround = CompiledRegex::Lookaround;
+
         const uint32_t res = m_program.lookarounds.size();
         auto write_matcher = [this, ignore_case](ParsedRegex::NodeIndex  child) {
                 auto& character = get_node(child);
                 if (character.op == ParsedRegex::Literal)
-                    m_program.lookarounds.push_back(ignore_case ? to_lower(character.value)
-                                                                : character.value);
+                    m_program.lookarounds.push_back(
+                        static_cast<Lookaround>(ignore_case ? to_lower(character.value) : character.value));
                 else if (character.op == ParsedRegex::AnyChar)
-                    m_program.lookarounds.push_back(0xF000);
+                    m_program.lookarounds.push_back(Lookaround::AnyChar);
                 else if (character.op == ParsedRegex::AnyCharExceptNewLine)
-                    m_program.lookarounds.push_back(0xF001);
+                    m_program.lookarounds.push_back(Lookaround::AnyCharExceptNewLine);
                 else if (character.op == ParsedRegex::Class)
-                    m_program.lookarounds.push_back(0xF0001 + character.value);
+                    m_program.lookarounds.push_back(static_cast<Lookaround>(to_underlying(Lookaround::CharacterClass) + character.value));
                 else if (character.op == ParsedRegex::CharacterType)
-                    m_program.lookarounds.push_back(0xF8000 | character.value);
+                    m_program.lookarounds.push_back(static_cast<Lookaround>(to_underlying(Lookaround::CharacterType) | character.value));
                 else
                     kak_assert(false);
                 return true;
@@ -898,7 +906,7 @@ private:
 
         ForEachChild<direction>::apply(m_parsed_regex, index, write_matcher);
 
-        m_program.lookarounds.push_back((Codepoint)-1);
+        m_program.lookarounds.push_back(Lookaround::EndOfLookaround);
         return res;
     }
 
@@ -1121,8 +1129,9 @@ String dump_regex(const CompiledRegex& program)
                     name = "negative look behind (ignore case)";
 
                 String str;
-                for (auto it = program.lookarounds.begin() + inst.param; *it != -1; ++it)
-                    utf8::dump(std::back_inserter(str), *it);
+                for (auto it = program.lookarounds.begin() + inst.param;
+                     *it != CompiledRegex::Lookaround::EndOfLookaround; ++it)
+                    utf8::dump(std::back_inserter(str), to_underlying(*it));
                 res += format("{} ({})\n", name, str);
                 break;
             }

src/regex_impl.hh

@@ -8,6 +8,7 @@
 #include "utf8.hh"
 #include "utf8_iterator.hh"
 #include "vector.hh"
+#include "utils.hh"
 
 namespace Kakoune
 {
@@ -82,6 +83,17 @@ struct CompiledRegex : RefCountable, UseMemoryDomain<MemoryDomain::Regex>
         NegativeLookBehind_IgnoreCase,
     };
 
+    enum class Lookaround : Codepoint
+    {
+        OpBegin              = 0xF0000,
+        AnyChar              = 0xF0000,
+        AnyCharExceptNewLine = 0xF0001,
+        CharacterClass       = 0xF0002,
+        CharacterType        = 0xF8000,
+        OpEnd                = 0xFFFFF,
+        EndOfLookaround      = static_cast<Codepoint>(-1)
+    };
+
     struct Instruction
     {
         Op op;
@@ -98,7 +110,7 @@ struct CompiledRegex : RefCountable, UseMemoryDomain<MemoryDomain::Regex>
 
     Vector<Instruction, MemoryDomain::Regex> instructions;
     Vector<CharacterClass, MemoryDomain::Regex> character_classes;
-    Vector<Codepoint, MemoryDomain::Regex> lookarounds;
+    Vector<Lookaround, MemoryDomain::Regex> lookarounds;
     uint32_t first_backward_inst; // -1 if no backward support, 0 if only backward, >0 if both forward and backward
     uint32_t save_count;
 
@@ -522,8 +534,10 @@ private:
     template<MatchDirection look_direction, bool ignore_case>
     bool lookaround(uint32_t index, EffectiveIt pos, const ExecConfig& config) const
     {
+        using Lookaround = CompiledRegex::Lookaround;
+
         const auto end = (look_direction == MatchDirection::Forward ? config.subject_end : config.subject_begin);
-        for (auto it = m_program.lookarounds.begin() + index; *it != -1; ++it)
+        for (auto it = m_program.lookarounds.begin() + index; *it != Lookaround::EndOfLookaround; ++it)
         {
             if (pos == end)
                 return false;
@@ -531,25 +545,27 @@ private:
             if (ignore_case)
                 cp = to_lower(cp);
 
-            const Codepoint ref = *it;
-            if (ref == 0xF000)
+            const Lookaround op = *it;
+            if (op == Lookaround::AnyChar)
             {} // any character matches
-            else if (ref == 0xF001)
+            else if (op == Lookaround::AnyCharExceptNewLine)
             {
                 if (cp == '\n')
                     return false;
             }
-            else if (ref > 0xF0000 and ref < 0xF8000)
+            else if (op >= Lookaround::CharacterClass and op < Lookaround::CharacterType)
             {
-                if (not is_character_class(m_program.character_classes[ref - 0xF0001], cp))
+                auto index = to_underlying(op) - to_underlying(Lookaround::CharacterClass);
+                if (not is_character_class(m_program.character_classes[index], cp))
                     return false;
             }
-            else if (ref >= 0xF8000 and ref <= 0xFFFFD)
+            else if (op >= Lookaround::CharacterType and op < Lookaround::OpEnd)
             {
-                if (not is_ctype((CharacterType)(ref & 0xFF), cp))
+                auto ctype = static_cast<CharacterType>(to_underlying(op) & 0xFF);
+                if (not is_ctype(ctype, cp))
                     return false;
             }
-            else if (ref != cp)
+            else if (static_cast<Codepoint>(op) != cp)
                 return false;
 
             (look_direction == MatchDirection::Forward) ? ++pos : --pos;

src/utils.hh

@@ -148,6 +148,12 @@ bool skip_while_reverse(Iterator& it, const BeginIterator& begin, T condition)
     return condition(*it);
 }
 
+template<typename E>
+auto to_underlying(E value)
+{
+    return static_cast<std::underlying_type_t<E>>(value);
+}
+
 }
 
 #endif // utils_hh_INCLUDED