#ifndef regex_impl_hh_INCLUDED #define regex_impl_hh_INCLUDED #include "unicode.hh" #include "utf8.hh" #include "utf8_iterator.hh" #include "vector.hh" #include "flags.hh" #include "ref_ptr.hh" #include namespace Kakoune { struct CompiledRegex : RefCountable { enum Op : char { Match, Literal, LiteralIgnoreCase, AnyChar, Matcher, Jump, Split_PrioritizeParent, Split_PrioritizeChild, Save, LineStart, LineEnd, WordBoundary, NotWordBoundary, SubjectBegin, SubjectEnd, LookAhead, NegativeLookAhead, LookBehind, NegativeLookBehind, }; using Offset = unsigned; explicit operator bool() const { return not bytecode.empty(); } Vector bytecode; Vector> matchers; size_t save_count; struct StartChars { bool map[256]; }; std::unique_ptr start_chars; }; CompiledRegex compile_regex(StringView re); enum class RegexExecFlags { None = 0, Search = 1 << 0, NotBeginOfLine = 1 << 1, NotEndOfLine = 1 << 2, NotBeginOfWord = 1 << 3, NotEndOfWord = 1 << 4, NotBeginOfSubject = 1 << 5, NotInitialNull = 1 << 6, AnyMatch = 1 << 7, NoSaves = 1 << 8, }; constexpr bool with_bit_ops(Meta::Type) { return true; } template class ThreadedRegexVM { public: ThreadedRegexVM(const CompiledRegex& program) : m_program{program} { kak_assert(m_program); } ThreadedRegexVM(const ThreadedRegexVM&) = delete; ThreadedRegexVM& operator=(const ThreadedRegexVM&) = delete; ~ThreadedRegexVM() { for (auto* saves : m_saves) { for (size_t i = m_program.save_count-1; i > 0; --i) saves->pos[i].~Iterator(); saves->~Saves(); } } bool exec(Iterator begin, Iterator end, RegexExecFlags flags) { m_begin = begin; m_end = end; m_flags = flags; if (flags & RegexExecFlags::NotInitialNull and m_begin == m_end) return false; Vector current_threads, next_threads; const bool no_saves = (m_flags & RegexExecFlags::NoSaves); Utf8It start{m_begin, m_begin, m_end}; const bool* start_chars = m_program.start_chars ? m_program.start_chars->map : nullptr; if (flags & RegexExecFlags::Search) to_next_start(start, end, start_chars); if (exec_from(start, no_saves ? nullptr : new_saves(nullptr), current_threads, next_threads)) return true; if (not (flags & RegexExecFlags::Search)) return false; do { to_next_start(++start, end, start_chars); if (exec_from(start, no_saves ? nullptr : new_saves(nullptr), current_threads, next_threads)) return true; } while (start != end); return false; } ArrayView captures() const { if (m_captures) return { m_captures->pos, m_program.save_count }; return {}; } private: struct Saves { int refcount; Iterator pos[1]; }; template Saves* new_saves(Iterator* pos) { kak_assert(not copy or pos != nullptr); const auto count = m_program.save_count; if (not m_free_saves.empty()) { Saves* res = m_free_saves.back(); m_free_saves.pop_back(); res->refcount = 1; if (copy) std::copy(pos, pos + count, res->pos); else std::fill(res->pos, res->pos + count, Iterator{}); return res; } void* ptr = ::operator new (sizeof(Saves) + (count-1) * sizeof(Iterator)); Saves* saves = new (ptr) Saves{1, copy ? pos[0] : Iterator{}}; for (size_t i = 1; i < count; ++i) new (&saves->pos[i]) Iterator{copy ? pos[i] : Iterator{}}; m_saves.push_back(saves); return saves; } void release_saves(Saves* saves) { if (saves and --saves->refcount == 0) m_free_saves.push_back(saves); }; struct Thread { const char* inst; Saves* saves; }; using Utf8It = utf8::iterator; enum class StepResult { Consumed, Matched, Failed }; StepResult step(const Utf8It& pos, Thread& thread, Vector& threads) { const auto prog_start = m_program.bytecode.data(); const auto prog_end = prog_start + m_program.bytecode.size(); while (true) { const Codepoint cp = pos == m_end ? 0 : *pos; const CompiledRegex::Op op = (CompiledRegex::Op)*thread.inst++; switch (op) { case CompiledRegex::Literal: if (utf8::read_codepoint(thread.inst, prog_end) == cp) return StepResult::Consumed; return StepResult::Failed; case CompiledRegex::LiteralIgnoreCase: if (utf8::read_codepoint(thread.inst, prog_end) == to_lower(cp)) return StepResult::Consumed; return StepResult::Failed; case CompiledRegex::AnyChar: return StepResult::Consumed; case CompiledRegex::Jump: thread.inst = prog_start + get_offset(thread.inst); break; case CompiledRegex::Split_PrioritizeParent: { auto parent = thread.inst + sizeof(CompiledRegex::Offset); auto child = prog_start + get_offset(thread.inst); thread.inst = parent; if (thread.saves) ++thread.saves->refcount; threads.push_back({child, thread.saves}); break; } case CompiledRegex::Split_PrioritizeChild: { auto parent = thread.inst + sizeof(CompiledRegex::Offset); auto child = prog_start + get_offset(thread.inst); thread.inst = child; if (thread.saves) ++thread.saves->refcount; threads.push_back({parent, thread.saves}); break; } case CompiledRegex::Save: { const size_t index = *thread.inst++; if (thread.saves == nullptr) break; if (thread.saves->refcount > 1) { --thread.saves->refcount; thread.saves = new_saves(thread.saves->pos); } thread.saves->pos[index] = pos.base(); break; } case CompiledRegex::Matcher: { const int matcher_id = *thread.inst++; return m_program.matchers[matcher_id](cp) ? StepResult::Consumed : StepResult::Failed; } case CompiledRegex::LineStart: if (not is_line_start(pos)) return StepResult::Failed; break; case CompiledRegex::LineEnd: if (not is_line_end(pos)) return StepResult::Failed; break; case CompiledRegex::WordBoundary: if (not is_word_boundary(pos)) return StepResult::Failed; break; case CompiledRegex::NotWordBoundary: if (is_word_boundary(pos)) return StepResult::Failed; break; case CompiledRegex::SubjectBegin: if (pos != m_begin or (m_flags & RegexExecFlags::NotBeginOfSubject)) return StepResult::Failed; break; case CompiledRegex::SubjectEnd: if (pos != m_end) return StepResult::Failed; break; case CompiledRegex::LookAhead: case CompiledRegex::NegativeLookAhead: { int count = *thread.inst++; for (auto it = pos; count and it != m_end; ++it, --count) if (*it != utf8::read(thread.inst)) break; if ((op == CompiledRegex::LookAhead and count != 0) or (op == CompiledRegex::NegativeLookAhead and count == 0)) return StepResult::Failed; thread.inst = utf8::advance(thread.inst, prog_end, CharCount{count - 1}); break; } case CompiledRegex::LookBehind: case CompiledRegex::NegativeLookBehind: { int count = *thread.inst++; for (auto it = pos-1; count and it >= m_begin; --it, --count) if (*it != utf8::read(thread.inst)) break; if ((op == CompiledRegex::LookBehind and count != 0) or (op == CompiledRegex::NegativeLookBehind and count == 0)) return StepResult::Failed; thread.inst = utf8::advance(thread.inst, prog_end, CharCount{count - 1}); break; } case CompiledRegex::Match: return StepResult::Matched; } } return StepResult::Failed; } bool exec_from(const Utf8It& start, Saves* initial_saves, Vector& current_threads, Vector& next_threads) { current_threads.push_back({m_program.bytecode.data(), initial_saves}); next_threads.clear(); bool found_match = false; for (Utf8It pos = start; pos != m_end; ++pos) { while (not current_threads.empty()) { auto thread = current_threads.back(); current_threads.pop_back(); switch (step(pos, thread, current_threads)) { case StepResult::Matched: if (not (m_flags & RegexExecFlags::Search) or // We are not at end, this is not a full match (m_flags & RegexExecFlags::NotInitialNull and pos == m_begin)) { release_saves(thread.saves); continue; } release_saves(m_captures); m_captures = thread.saves; if (m_flags & RegexExecFlags::AnyMatch) return true; found_match = true; current_threads.clear(); // remove this and lower priority threads break; case StepResult::Failed: release_saves(thread.saves); break; case StepResult::Consumed: if (contains_that(next_threads, [&](auto& t) { return t.inst == thread.inst; })) release_saves(thread.saves); else next_threads.push_back(thread); break; } } if (next_threads.empty()) return found_match; std::swap(current_threads, next_threads); std::reverse(current_threads.begin(), current_threads.end()); } if (found_match) return true; // Step remaining threads to see if they match without consuming anything else const Utf8It end{m_end, m_begin, m_end}; while (not current_threads.empty()) { auto thread = current_threads.back(); current_threads.pop_back(); if (step(end, thread, current_threads) == StepResult::Matched) { release_saves(m_captures); m_captures = thread.saves; return true; } } return false; } void to_next_start(Utf8It& start, const Iterator& end, const bool* start_chars) { if (not start_chars) return; while (start != end and *start >= 0 and *start < 256 and not start_chars[*start]) ++start; } static CompiledRegex::Offset get_offset(const char* ptr) { CompiledRegex::Offset res; memcpy(&res, ptr, sizeof(CompiledRegex::Offset)); return res; } bool is_line_start(const Utf8It& pos) const { return (pos == m_begin and not (m_flags & RegexExecFlags::NotBeginOfLine)) or *(pos-1) == '\n'; } bool is_line_end(const Utf8It& pos) const { return (pos == m_end and not (m_flags & RegexExecFlags::NotEndOfLine)) or *pos == '\n'; } bool is_word_boundary(const Utf8It& pos) const { return (pos == m_begin and not (m_flags & RegexExecFlags::NotBeginOfWord)) or (pos == m_end and not (m_flags & RegexExecFlags::NotEndOfWord)) or is_word(*(pos-1)) != is_word(*pos); } const CompiledRegex& m_program; Iterator m_begin; Iterator m_end; RegexExecFlags m_flags; Vector m_saves; Vector m_free_saves; Saves* m_captures = nullptr; }; template bool regex_match(It begin, It end, const CompiledRegex& re, RegexExecFlags flags = RegexExecFlags::None) { ThreadedRegexVM vm{re}; return vm.exec(begin, end, (RegexExecFlags)(flags & ~(RegexExecFlags::Search)) | RegexExecFlags::AnyMatch | RegexExecFlags::NoSaves); } template bool regex_match(It begin, It end, Vector& captures, const CompiledRegex& re, RegexExecFlags flags = RegexExecFlags::None) { ThreadedRegexVM vm{re}; if (vm.exec(begin, end, flags & ~(RegexExecFlags::Search))) { std::copy(vm.captures().begin(), vm.captures().end(), std::back_inserter(captures)); return true; } return false; } template bool regex_search(It begin, It end, const CompiledRegex& re, RegexExecFlags flags = RegexExecFlags::None) { ThreadedRegexVM vm{re}; return vm.exec(begin, end, flags | RegexExecFlags::Search | RegexExecFlags::AnyMatch | RegexExecFlags::NoSaves); } template bool regex_search(It begin, It end, Vector& captures, const CompiledRegex& re, RegexExecFlags flags = RegexExecFlags::None) { ThreadedRegexVM vm{re}; if (vm.exec(begin, end, flags | RegexExecFlags::Search)) { std::copy(vm.captures().begin(), vm.captures().end(), std::back_inserter(captures)); return true; } return false; } } #endif // regex_impl_hh_INCLUDED