/* * * Copyright (c) 2002 * John Maddock * * Use, modification and distribution are subject to the * Boost Software License, Version 1.0. (See accompanying file * LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt) * */ /* * LOCATION: see http://www.boost.org for most recent version. * FILE perl_matcher_common.cpp * VERSION see * DESCRIPTION: Definitions of perl_matcher member functions that are * common to both the recursive and non-recursive versions. */ #ifndef BOOST_REGEX_V4_PERL_MATCHER_COMMON_HPP #define BOOST_REGEX_V4_PERL_MATCHER_COMMON_HPP #ifdef BOOST_MSVC #pragma warning(push) #pragma warning(disable: 4103) #endif #ifdef BOOST_HAS_ABI_HEADERS # include BOOST_ABI_PREFIX #endif #ifdef BOOST_MSVC #pragma warning(pop) #endif #ifdef __BORLANDC__ # pragma option push -w-8008 -w-8066 #endif #ifdef BOOST_MSVC # pragma warning(push) # pragma warning(disable: 4800) #endif namespace boost{ namespace re_detail{ template void perl_matcher::construct_init(const basic_regex& e, match_flag_type f) { typedef typename regex_iterator_traits::iterator_category category; typedef typename basic_regex::flag_type expression_flag_type; if(e.empty()) { // precondition failure: e is not a valid regex. std::invalid_argument ex("Invalid regular expression object"); boost::throw_exception(ex); } pstate = 0; m_match_flags = f; estimate_max_state_count(static_cast(0)); expression_flag_type re_f = re.flags(); icase = re_f & regex_constants::icase; if(!(m_match_flags & (match_perl|match_posix))) { if((re_f & (regbase::main_option_type|regbase::no_perl_ex)) == 0) m_match_flags |= match_perl; else if((re_f & (regbase::main_option_type|regbase::emacs_ex)) == (regbase::basic_syntax_group|regbase::emacs_ex)) m_match_flags |= match_perl; else m_match_flags |= match_posix; } if(m_match_flags & match_posix) { m_temp_match.reset(new match_results()); m_presult = m_temp_match.get(); } else m_presult = &m_result; #ifdef BOOST_REGEX_NON_RECURSIVE m_stack_base = 0; m_backup_state = 0; #endif // find the value to use for matching word boundaries: m_word_mask = re.get_data().m_word_mask; // find bitmask to use for matching '.': match_any_mask = static_cast((f & match_not_dot_newline) ? re_detail::test_not_newline : re_detail::test_newline); } template void perl_matcher::estimate_max_state_count(std::random_access_iterator_tag*) { // // How many states should we allow our machine to visit before giving up? // This is a heuristic: it takes the greater of O(N^2) and O(NS^2) // where N is the length of the string, and S is the number of states // in the machine. It's tempting to up this to O(N^2S) or even O(N^2S^2) // but these take unreasonably amounts of time to bale out in pathological // cases. // // Calculate NS^2 first: // static const boost::uintmax_t k = 100000; boost::uintmax_t dist = boost::re_detail::distance(base, last); if(dist == 0) dist = 1; boost::uintmax_t states = re.size(); if(states == 0) states = 1; states *= states; if((std::numeric_limits::max)() / dist < states) { max_state_count = (std::numeric_limits::max)() - 2; return; } states *= dist; if((std::numeric_limits::max)() - k < states) { max_state_count = (std::numeric_limits::max)() - 2; return; } states += k; max_state_count = states; // // Now calculate N^2: // states = dist; if((std::numeric_limits::max)() / dist < states) { max_state_count = (std::numeric_limits::max)() - 2; return; } states *= dist; if((std::numeric_limits::max)() - k < states) { max_state_count = (std::numeric_limits::max)() - 2; return; } states += k; // // N^2 can be a very large number indeed, to prevent things getting out // of control, cap the max states: // if(states > BOOST_REGEX_MAX_STATE_COUNT) states = BOOST_REGEX_MAX_STATE_COUNT; // // If (the possibly capped) N^2 is larger than our first estimate, // use this instead: // if(states > max_state_count) max_state_count = states; } template inline void perl_matcher::estimate_max_state_count(void*) { // we don't know how long the sequence is: max_state_count = BOOST_REGEX_MAX_STATE_COUNT; } #ifdef BOOST_REGEX_HAS_MS_STACK_GUARD template inline bool perl_matcher::protected_call( protected_proc_type proc) { ::boost::re_detail::concrete_protected_call > obj(this, proc); return obj.execute(); } #endif template inline bool perl_matcher::match() { #ifdef BOOST_REGEX_HAS_MS_STACK_GUARD return protected_call(&perl_matcher::match_imp); #else return match_imp(); #endif } template bool perl_matcher::match_imp() { // initialise our stack if we are non-recursive: #ifdef BOOST_REGEX_NON_RECURSIVE save_state_init init(&m_stack_base, &m_backup_state); used_block_count = BOOST_REGEX_MAX_BLOCKS; #if !defined(BOOST_NO_EXCEPTIONS) try{ #endif #endif // reset our state machine: position = base; search_base = base; state_count = 0; m_match_flags |= regex_constants::match_all; m_presult->set_size((m_match_flags & match_nosubs) ? 1 : re.mark_count(), search_base, last); m_presult->set_base(base); if(m_match_flags & match_posix) m_result = *m_presult; verify_options(re.flags(), m_match_flags); if(0 == match_prefix()) return false; return m_result[0].second == last; #if defined(BOOST_REGEX_NON_RECURSIVE) && !defined(BOOST_NO_EXCEPTIONS) } catch(...) { // unwind all pushed states, apart from anything else this // ensures that all the states are correctly destructed // not just the memory freed. while(unwind(true)){} throw; } #endif } template inline bool perl_matcher::find() { #ifdef BOOST_REGEX_HAS_MS_STACK_GUARD return protected_call(&perl_matcher::find_imp); #else return find_imp(); #endif } template bool perl_matcher::find_imp() { static matcher_proc_type const s_find_vtable[7] = { &perl_matcher::find_restart_any, &perl_matcher::find_restart_word, &perl_matcher::find_restart_line, &perl_matcher::find_restart_buf, &perl_matcher::match_prefix, &perl_matcher::find_restart_lit, &perl_matcher::find_restart_lit, }; // initialise our stack if we are non-recursive: #ifdef BOOST_REGEX_NON_RECURSIVE save_state_init init(&m_stack_base, &m_backup_state); used_block_count = BOOST_REGEX_MAX_BLOCKS; #if !defined(BOOST_NO_EXCEPTIONS) try{ #endif #endif state_count = 0; if((m_match_flags & regex_constants::match_init) == 0) { // reset our state machine: search_base = position = base; pstate = re.get_first_state(); m_presult->set_size((m_match_flags & match_nosubs) ? 1 : re.mark_count(), base, last); m_presult->set_base(base); m_match_flags |= regex_constants::match_init; } else { // start again: search_base = position = m_result[0].second; // If last match was null and match_not_null was not set then increment // our start position, otherwise we go into an infinite loop: if(((m_match_flags & match_not_null) == 0) && (m_result.length() == 0)) { if(position == last) return false; else ++position; } // reset $` start: m_presult->set_size((m_match_flags & match_nosubs) ? 1 : re.mark_count(), search_base, last); //if((base != search_base) && (base == backstop)) // m_match_flags |= match_prev_avail; } if(m_match_flags & match_posix) { m_result.set_size(re.mark_count(), base, last); m_result.set_base(base); } verify_options(re.flags(), m_match_flags); // find out what kind of expression we have: unsigned type = (m_match_flags & match_continuous) ? static_cast(regbase::restart_continue) : static_cast(re.get_restart_type()); // call the appropriate search routine: matcher_proc_type proc = s_find_vtable[type]; return (this->*proc)(); #if defined(BOOST_REGEX_NON_RECURSIVE) && !defined(BOOST_NO_EXCEPTIONS) } catch(...) { // unwind all pushed states, apart from anything else this // ensures that all the states are correctly destructed // not just the memory freed. while(unwind(true)){} throw; } #endif } template bool perl_matcher::match_prefix() { m_has_partial_match = false; m_has_found_match = false; pstate = re.get_first_state(); m_presult->set_first(position); restart = position; match_all_states(); if(!m_has_found_match && m_has_partial_match && (m_match_flags & match_partial)) { m_has_found_match = true; m_presult->set_second(last, 0, false); position = last; } #ifdef BOOST_REGEX_MATCH_EXTRA if(m_has_found_match && (match_extra & m_match_flags)) { // // we have a match, reverse the capture information: // for(unsigned i = 0; i < m_presult->size(); ++i) { typename sub_match::capture_sequence_type & seq = ((*m_presult)[i]).get_captures(); std::reverse(seq.begin(), seq.end()); } } #endif if(!m_has_found_match) position = restart; // reset search postion return m_has_found_match; } template bool perl_matcher::match_endmark() { int index = static_cast(pstate)->index; if(index > 0) { if((m_match_flags & match_nosubs) == 0) m_presult->set_second(position, index); } else if((index < 0) && (index != -4)) { // matched forward lookahead: pstate = 0; return true; } pstate = pstate->next.p; return true; } template bool perl_matcher::match_literal() { unsigned int len = static_cast(pstate)->length; const char_type* what = reinterpret_cast(static_cast(pstate) + 1); // // compare string with what we stored in // our records: for(unsigned int i = 0; i < len; ++i, ++position) { if((position == last) || (traits_inst.translate(*position, icase) != what[i])) return false; } pstate = pstate->next.p; return true; } template bool perl_matcher::match_start_line() { if(position == backstop) { if((m_match_flags & match_prev_avail) == 0) { if((m_match_flags & match_not_bol) == 0) { pstate = pstate->next.p; return true; } return false; } } else if(m_match_flags & match_single_line) return false; // check the previous value character: BidiIterator t(position); --t; if(position != last) { if(is_separator(*t) && !((*t == static_cast('\r')) && (*position == static_cast('\n'))) ) { pstate = pstate->next.p; return true; } } else if(is_separator(*t)) { pstate = pstate->next.p; return true; } return false; } template bool perl_matcher::match_end_line() { if(position != last) { if(m_match_flags & match_single_line) return false; // we're not yet at the end so *first is always valid: if(is_separator(*position)) { if((position != backstop) || (m_match_flags & match_prev_avail)) { // check that we're not in the middle of \r\n sequence BidiIterator t(position); --t; if((*t == static_cast('\r')) && (*position == static_cast('\n'))) { return false; } } pstate = pstate->next.p; return true; } } else if((m_match_flags & match_not_eol) == 0) { pstate = pstate->next.p; return true; } return false; } template bool perl_matcher::match_wild() { if(position == last) return false; if(is_separator(*position) && ((match_any_mask & static_cast(pstate)->mask) == 0)) return false; if((*position == char_type(0)) && (m_match_flags & match_not_dot_null)) return false; pstate = pstate->next.p; ++position; return true; } template bool perl_matcher::match_match() { if((m_match_flags & match_not_null) && (position == (*m_presult)[0].first)) return false; if((m_match_flags & match_all) && (position != last)) return false; if((m_match_flags & regex_constants::match_not_initial_null) && (position == search_base)) return false; m_presult->set_second(position); pstate = 0; m_has_found_match = true; if((m_match_flags & match_posix) == match_posix) { m_result.maybe_assign(*m_presult); if((m_match_flags & match_any) == 0) return false; } #ifdef BOOST_REGEX_MATCH_EXTRA if(match_extra & m_match_flags) { for(unsigned i = 0; i < m_presult->size(); ++i) if((*m_presult)[i].matched) ((*m_presult)[i]).get_captures().push_back((*m_presult)[i]); } #endif return true; } template bool perl_matcher::match_word_boundary() { bool b; // indcates whether next character is a word character if(position != last) { // prev and this character must be opposites: #if defined(BOOST_REGEX_USE_C_LOCALE) && defined(__GNUC__) && (__GNUC__ == 2) && (__GNUC_MINOR__ < 95) b = traits::isctype(*position, m_word_mask); #else b = traits_inst.isctype(*position, m_word_mask); #endif } else { b = (m_match_flags & match_not_eow) ? true : false; } if((position == backstop) && ((m_match_flags & match_prev_avail) == 0)) { if(m_match_flags & match_not_bow) b ^= true; else b ^= false; } else { --position; b ^= traits_inst.isctype(*position, m_word_mask); ++position; } if(b) { pstate = pstate->next.p; return true; } return false; // no match if we get to here... } template bool perl_matcher::match_within_word() { if(position == last) return false; // both prev and this character must be m_word_mask: bool prev = traits_inst.isctype(*position, m_word_mask); { bool b; if((position == backstop) && ((m_match_flags & match_prev_avail) == 0)) return false; else { --position; b = traits_inst.isctype(*position, m_word_mask); ++position; } if(b == prev) { pstate = pstate->next.p; return true; } } return false; } template bool perl_matcher::match_word_start() { if(position == last) return false; // can't be starting a word if we're already at the end of input if(!traits_inst.isctype(*position, m_word_mask)) return false; // next character isn't a word character if((position == backstop) && ((m_match_flags & match_prev_avail) == 0)) { if(m_match_flags & match_not_bow) return false; // no previous input } else { // otherwise inside buffer: BidiIterator t(position); --t; if(traits_inst.isctype(*t, m_word_mask)) return false; // previous character not non-word } // OK we have a match: pstate = pstate->next.p; return true; } template bool perl_matcher::match_word_end() { if((position == backstop) && ((m_match_flags & match_prev_avail) == 0)) return false; // start of buffer can't be end of word BidiIterator t(position); --t; if(traits_inst.isctype(*t, m_word_mask) == false) return false; // previous character wasn't a word character if(position == last) { if(m_match_flags & match_not_eow) return false; // end of buffer but not end of word } else { // otherwise inside buffer: if(traits_inst.isctype(*position, m_word_mask)) return false; // next character is a word character } pstate = pstate->next.p; return true; // if we fall through to here then we've succeeded } template bool perl_matcher::match_buffer_start() { if((position != backstop) || (m_match_flags & match_not_bob)) return false; // OK match: pstate = pstate->next.p; return true; } template bool perl_matcher::match_buffer_end() { if((position != last) || (m_match_flags & match_not_eob)) return false; // OK match: pstate = pstate->next.p; return true; } template bool perl_matcher::match_backref() { // // Compare with what we previously matched. // Note that this succeeds if the backref did not partisipate // in the match, this is in line with ECMAScript, but not Perl // or PCRE. // BidiIterator i = (*m_presult)[static_cast(pstate)->index].first; BidiIterator j = (*m_presult)[static_cast(pstate)->index].second; while(i != j) { if((position == last) || (traits_inst.translate(*position, icase) != traits_inst.translate(*i, icase))) return false; ++i; ++position; } pstate = pstate->next.p; return true; } template bool perl_matcher::match_long_set() { typedef typename traits::char_class_type char_class_type; // let the traits class do the work: if(position == last) return false; BidiIterator t = re_is_set_member(position, last, static_cast*>(pstate), re.get_data(), icase); if(t != position) { pstate = pstate->next.p; position = t; return true; } return false; } template bool perl_matcher::match_set() { if(position == last) return false; if(static_cast(pstate)->_map[static_cast(traits_inst.translate(*position, icase))]) { pstate = pstate->next.p; ++position; return true; } return false; } template bool perl_matcher::match_jump() { pstate = static_cast(pstate)->alt.p; return true; } template bool perl_matcher::match_combining() { if(position == last) return false; if(is_combining(traits_inst.translate(*position, icase))) return false; ++position; while((position != last) && is_combining(traits_inst.translate(*position, icase))) ++position; pstate = pstate->next.p; return true; } template bool perl_matcher::match_soft_buffer_end() { if(m_match_flags & match_not_eob) return false; BidiIterator p(position); while((p != last) && is_separator(traits_inst.translate(*p, icase)))++p; if(p != last) return false; pstate = pstate->next.p; return true; } template bool perl_matcher::match_restart_continue() { if(position == search_base) { pstate = pstate->next.p; return true; } return false; } template bool perl_matcher::match_backstep() { #ifdef BOOST_MSVC #pragma warning(push) #pragma warning(disable:4127) #endif if( ::boost::is_random_access_iterator::value) { std::ptrdiff_t maxlen = ::boost::re_detail::distance(backstop, position); if(maxlen < static_cast(pstate)->index) return false; std::advance(position, -static_cast(pstate)->index); } else { int c = static_cast(pstate)->index; while(c--) { if(position == backstop) return false; --position; } } pstate = pstate->next.p; return true; #ifdef BOOST_MSVC #pragma warning(pop) #endif } template inline bool perl_matcher::match_assert_backref() { // return true if marked sub-expression N has been matched: bool result = (*m_presult)[static_cast(pstate)->index].matched; pstate = pstate->next.p; return result; } template bool perl_matcher::match_toggle_case() { // change our case sensitivity: this->icase = static_cast(pstate)->icase; pstate = pstate->next.p; return true; } template bool perl_matcher::find_restart_any() { #ifdef BOOST_MSVC #pragma warning(push) #pragma warning(disable:4127) #endif const unsigned char* _map = re.get_map(); while(true) { // skip everything we can't match: while((position != last) && !can_start(*position, _map, (unsigned char)mask_any) ) ++position; if(position == last) { // run out of characters, try a null match if possible: if(re.can_be_null()) return match_prefix(); break; } // now try and obtain a match: if(match_prefix()) return true; if(position == last) return false; ++position; } return false; #ifdef BOOST_MSVC #pragma warning(pop) #endif } template bool perl_matcher::find_restart_word() { #ifdef BOOST_MSVC #pragma warning(push) #pragma warning(disable:4127) #endif // do search optimised for word starts: const unsigned char* _map = re.get_map(); if((m_match_flags & match_prev_avail) || (position != base)) --position; else if(match_prefix()) return true; do { while((position != last) && traits_inst.isctype(*position, m_word_mask)) ++position; while((position != last) && !traits_inst.isctype(*position, m_word_mask)) ++position; if(position == last) break; if(can_start(*position, _map, (unsigned char)mask_any) ) { if(match_prefix()) return true; } if(position == last) break; } while(true); return false; #ifdef BOOST_MSVC #pragma warning(pop) #endif } template bool perl_matcher::find_restart_line() { // do search optimised for line starts: const unsigned char* _map = re.get_map(); if(match_prefix()) return true; while(position != last) { while((position != last) && !is_separator(*position)) ++position; if(position == last) return false; ++position; if(position == last) { if(re.can_be_null() && match_prefix()) return true; return false; } if( can_start(*position, _map, (unsigned char)mask_any) ) { if(match_prefix()) return true; } if(position == last) return false; //++position; } return false; } template bool perl_matcher::find_restart_buf() { if((position == base) && ((m_match_flags & match_not_bob) == 0)) return match_prefix(); return false; } template bool perl_matcher::find_restart_lit() { #if 0 if(position == last) return false; // can't possibly match if we're at the end already unsigned type = (m_match_flags & match_continuous) ? static_cast(regbase::restart_continue) : static_cast(re.get_restart_type()); const kmp_info* info = access::get_kmp(re); int len = info->len; const char_type* x = info->pstr; int j = 0; while (position != last) { while((j > -1) && (x[j] != traits_inst.translate(*position, icase))) j = info->kmp_next[j]; ++position; ++j; if(j >= len) { if(type == regbase::restart_fixed_lit) { std::advance(position, -j); restart = position; std::advance(restart, len); m_result.set_first(position); m_result.set_second(restart); position = restart; return true; } else { restart = position; std::advance(position, -j); if(match_prefix()) return true; else { for(int k = 0; (restart != position) && (k < j); ++k, --restart) {} // dwa 10/20/2000 - warning suppression for MWCW if(restart != last) ++restart; position = restart; j = 0; //we could do better than this... } } } } if((m_match_flags & match_partial) && (position == last) && j) { // we need to check for a partial match: restart = position; std::advance(position, -j); return match_prefix(); } #endif return false; } } // namespace re_detail } // namespace boost #ifdef BOOST_MSVC # pragma warning(pop) #endif #ifdef __BORLANDC__ # pragma option pop #endif #ifdef BOOST_MSVC #pragma warning(push) #pragma warning(disable: 4103) #endif #ifdef BOOST_HAS_ABI_HEADERS # include BOOST_ABI_SUFFIX #endif #ifdef BOOST_MSVC #pragma warning(pop) #endif #endif