@@ -65,81 +65,114 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
_S_opcode_accept,
};
- template<typename _CharT, typename _TraitsT>
- class _State
+ struct _State_base
+ {
+ _Opcode _M_opcode; // type of outgoing transition
+ _StateIdT _M_next; // outgoing transition
+ union // Since they are mutually exclusive.
{
- public:
- typedef _Matcher<_CharT> _MatcherT;
-
- _Opcode _M_opcode; // type of outgoing transition
- _StateIdT _M_next; // outgoing transition
- union // Since they are mutually exclusive.
+ size_t _M_subexpr; // for _S_opcode_subexpr_*
+ size_t _M_backref_index; // for _S_opcode_backref
+ struct
{
- size_t _M_subexpr; // for _S_opcode_subexpr_*
- size_t _M_backref_index; // for _S_opcode_backref
- struct
- {
- // for _S_opcode_alternative.
- _StateIdT _M_quant_index;
- // for _S_opcode_alternative or _S_opcode_subexpr_lookahead
- _StateIdT _M_alt;
- // for _S_opcode_word_boundary or _S_opcode_subexpr_lookahead or
- // quantifiers(ungreedy if set true)
- bool _M_neg;
- };
+ // for _S_opcode_alternative.
+ _StateIdT _M_quant_index;
+ // for _S_opcode_alternative or _S_opcode_subexpr_lookahead
+ _StateIdT _M_alt;
+ // for _S_opcode_word_boundary or _S_opcode_subexpr_lookahead or
+ // quantifiers (ungreedy if set true)
+ bool _M_neg;
};
- _MatcherT _M_matches; // for _S_opcode_match
+ };
- explicit _State(_Opcode __opcode)
- : _M_opcode(__opcode), _M_next(_S_invalid_state_id)
- { }
+ explicit _State_base(_Opcode __opcode)
+ : _M_opcode(__opcode), _M_next(_S_invalid_state_id)
+ { }
+
+ protected:
+ ~_State_base() = default;
+ public:
#ifdef _GLIBCXX_DEBUG
- std::ostream&
- _M_print(std::ostream& ostr) const;
+ std::ostream&
+ _M_print(std::ostream& ostr) const;
- // Prints graphviz dot commands for state.
- std::ostream&
- _M_dot(std::ostream& __ostr, _StateIdT __id) const;
+ // Prints graphviz dot commands for state.
+ std::ostream&
+ _M_dot(std::ostream& __ostr, _StateIdT __id) const;
#endif
- };
+ };
template<typename _CharT, typename _TraitsT>
- class _NFA
- : public std::vector<_State<_CharT, _TraitsT>>
+ struct _State : _State_base
{
- public:
- typedef _State<_CharT, _TraitsT> _StateT;
- typedef const _Matcher<_CharT>& _MatcherT;
- typedef size_t _SizeT;
- typedef regex_constants::syntax_option_type _FlagT;
-
- _NFA(_FlagT __f)
- : _M_flags(__f), _M_start_state(0), _M_subexpr_count(0),
- _M_quant_count(0), _M_has_backref(false)
- { }
+ typedef _Matcher<_CharT> _MatcherT;
- _FlagT
- _M_options() const
- { return _M_flags; }
+ _MatcherT _M_matches; // for _S_opcode_match
- _StateIdT
- _M_start() const
- { return _M_start_state; }
+ explicit _State(_Opcode __opcode) : _State_base(__opcode) { }
+ };
- const _StateSet&
- _M_final_states() const
- { return _M_accepting_states; }
+ struct _NFA_base
+ {
+ typedef size_t _SizeT;
+ typedef regex_constants::syntax_option_type _FlagT;
+
+ explicit
+ _NFA_base(_FlagT __f)
+ : _M_flags(__f), _M_start_state(0), _M_subexpr_count(0),
+ _M_quant_count(0), _M_has_backref(false)
+ { }
+
+ _NFA_base(_NFA_base&&) = default;
+
+ protected:
+ ~_NFA_base() = default;
+
+ public:
+ _FlagT
+ _M_options() const
+ { return _M_flags; }
+
+ _StateIdT
+ _M_start() const
+ { return _M_start_state; }
+
+ const _StateSet&
+ _M_final_states() const
+ { return _M_accepting_states; }
+
+ _SizeT
+ _M_sub_count() const
+ { return _M_subexpr_count; }
+
+ std::vector<size_t> _M_paren_stack;
+ _StateSet _M_accepting_states;
+ _FlagT _M_flags;
+ _StateIdT _M_start_state;
+ _SizeT _M_subexpr_count;
+ _SizeT _M_quant_count;
+ bool _M_has_backref;
+ };
+
+ template<typename _CharT, typename _TraitsT>
+ struct _NFA
+ : _NFA_base, std::vector<_State<_CharT, _TraitsT>>
+ {
+ typedef _State<_CharT, _TraitsT> _StateT;
+ typedef _Matcher<_CharT> _MatcherT;
+
+ using _NFA_base::_NFA_base;
- _SizeT
- _M_sub_count() const
- { return _M_subexpr_count; }
+ // for performance reasons _NFA objects should only be moved not copied
+ _NFA(const _NFA&) = delete;
+ _NFA(_NFA&&) = default;
_StateIdT
_M_insert_accept()
{
auto __ret = _M_insert_state(_StateT(_S_opcode_accept));
- _M_accepting_states.insert(__ret);
+ this->_M_accepting_states.insert(__ret);
return __ret;
}
@@ -149,38 +182,38 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
_StateT __tmp(_S_opcode_alternative);
// It labels every quantifier to make greedy comparison easier in BFS
// approach.
- __tmp._M_quant_index = _M_quant_count++;
+ __tmp._M_quant_index = this->_M_quant_count++;
__tmp._M_next = __next;
__tmp._M_alt = __alt;
__tmp._M_neg = __neg;
- return _M_insert_state(__tmp);
+ return _M_insert_state(std::move(__tmp));
}
_StateIdT
_M_insert_matcher(_MatcherT __m)
{
_StateT __tmp(_S_opcode_match);
- __tmp._M_matches = __m;
- return _M_insert_state(__tmp);
+ __tmp._M_matches = std::move(__m);
+ return _M_insert_state(std::move(__tmp));
}
_StateIdT
_M_insert_subexpr_begin()
{
- auto __id = _M_subexpr_count++;
- _M_paren_stack.push_back(__id);
+ auto __id = this->_M_subexpr_count++;
+ this->_M_paren_stack.push_back(__id);
_StateT __tmp(_S_opcode_subexpr_begin);
__tmp._M_subexpr = __id;
- return _M_insert_state(__tmp);
+ return _M_insert_state(std::move(__tmp));
}
_StateIdT
_M_insert_subexpr_end()
{
_StateT __tmp(_S_opcode_subexpr_end);
- __tmp._M_subexpr = _M_paren_stack.back();
- _M_paren_stack.pop_back();
- return _M_insert_state(__tmp);
+ __tmp._M_subexpr = this->_M_paren_stack.back();
+ this->_M_paren_stack.pop_back();
+ return _M_insert_state(std::move(__tmp));
}
_StateIdT
@@ -199,7 +232,7 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
{
_StateT __tmp(_S_opcode_word_boundary);
__tmp._M_neg = __neg;
- return _M_insert_state(__tmp);
+ return _M_insert_state(std::move(__tmp));
}
_StateIdT
@@ -208,7 +241,7 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
_StateT __tmp(_S_opcode_subexpr_lookahead);
__tmp._M_alt = __alt;
__tmp._M_neg = __neg;
- return _M_insert_state(__tmp);
+ return _M_insert_state(std::move(__tmp));
}
_StateIdT
@@ -218,7 +251,7 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
_StateIdT
_M_insert_state(_StateT __s)
{
- this->push_back(__s);
+ this->push_back(std::move(__s));
return this->size()-1;
}
@@ -230,14 +263,6 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
std::ostream&
_M_dot(std::ostream& __ostr) const;
#endif
-
- std::vector<size_t> _M_paren_stack;
- _StateSet _M_accepting_states;
- _FlagT _M_flags;
- _StateIdT _M_start_state;
- _SizeT _M_subexpr_count;
- _SizeT _M_quant_count;
- bool _M_has_backref;
};
/// Describes a sequence of one or more %_State, its current start
@@ -251,7 +276,7 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
public:
_StateSeq(_RegexT& __nfa, _StateIdT __s)
- : _StateSeq(__nfa, __s, __s)
+ : _M_nfa(__nfa), _M_start(__s), _M_end(__s)
{ }
_StateSeq(_RegexT& __nfa, _StateIdT __s, _StateIdT __end)
@@ -35,118 +35,116 @@ namespace __detail
_GLIBCXX_BEGIN_NAMESPACE_VERSION
#ifdef _GLIBCXX_DEBUG
- template<typename _CharT, typename _TraitsT>
- std::ostream& _State<_CharT, _TraitsT>::
- _M_print(std::ostream& ostr) const
+ std::ostream&
+ _State_base::_M_print(std::ostream& ostr) const
+ {
+ switch (_M_opcode)
{
- switch (_M_opcode)
- {
- case _S_opcode_alternative:
- ostr << "alt next=" << _M_next << " alt=" << _M_alt;
- break;
- case _S_opcode_subexpr_begin:
- ostr << "subexpr begin next=" << _M_next << " index=" << _M_subexpr;
- break;
- case _S_opcode_subexpr_end:
- ostr << "subexpr end next=" << _M_next << " index=" << _M_subexpr;
- break;
- case _S_opcode_backref:
- ostr << "backref next=" << _M_next << " index=" << _M_backref_index;
- break;
- case _S_opcode_match:
- ostr << "match next=" << _M_next;
- break;
- case _S_opcode_accept:
- ostr << "accept next=" << _M_next;
- break;
- default:
- ostr << "unknown next=" << _M_next;
- break;
- }
- return ostr;
+ case _S_opcode_alternative:
+ ostr << "alt next=" << _M_next << " alt=" << _M_alt;
+ break;
+ case _S_opcode_subexpr_begin:
+ ostr << "subexpr begin next=" << _M_next << " index=" << _M_subexpr;
+ break;
+ case _S_opcode_subexpr_end:
+ ostr << "subexpr end next=" << _M_next << " index=" << _M_subexpr;
+ break;
+ case _S_opcode_backref:
+ ostr << "backref next=" << _M_next << " index=" << _M_backref_index;
+ break;
+ case _S_opcode_match:
+ ostr << "match next=" << _M_next;
+ break;
+ case _S_opcode_accept:
+ ostr << "accept next=" << _M_next;
+ break;
+ default:
+ ostr << "unknown next=" << _M_next;
+ break;
}
+ return ostr;
+ }
// Prints graphviz dot commands for state.
- template<typename _CharT, typename _TraitsT>
- std::ostream& _State<_CharT, _TraitsT>::
- _M_dot(std::ostream& __ostr, _StateIdT __id) const
+ std::ostream&
+ _State_base::_M_dot(std::ostream& __ostr, _StateIdT __id) const
+ {
+ switch (_M_opcode)
{
- switch (_M_opcode)
- {
- case _S_opcode_alternative:
- __ostr << __id << " [label=\"" << __id << "\\nALT\"];\n"
- << __id << " -> " << _M_next
- << " [label=\"epsilon\", tailport=\"s\"];\n"
- << __id << " -> " << _M_alt
- << " [label=\"epsilon\", tailport=\"n\"];\n";
- break;
- case _S_opcode_backref:
- __ostr << __id << " [label=\"" << __id << "\\nBACKREF "
- << _M_subexpr << "\"];\n"
- << __id << " -> " << _M_next << " [label=\"<match>\"];\n";
- break;
- case _S_opcode_line_begin_assertion:
- __ostr << __id << " [label=\"" << __id << "\\nLINE_BEGIN \"];\n"
- << __id << " -> " << _M_next << " [label=\"epsilon\"];\n";
- break;
- case _S_opcode_line_end_assertion:
- __ostr << __id << " [label=\"" << __id << "\\nLINE_END \"];\n"
- << __id << " -> " << _M_next << " [label=\"epsilon\"];\n";
- break;
- case _S_opcode_word_boundary:
- __ostr << __id << " [label=\"" << __id << "\\nWORD_BOUNDRY "
- << _M_neg << "\"];\n"
- << __id << " -> " << _M_next << " [label=\"epsilon\"];\n";
- break;
- case _S_opcode_subexpr_lookahead:
- __ostr << __id << " [label=\"" << __id << "\\nLOOK_AHEAD\"];\n"
- << __id << " -> " << _M_next
- << " [label=\"epsilon\", tailport=\"s\"];\n"
- << __id << " -> " << _M_alt
- << " [label=\"<assert>\", tailport=\"n\"];\n";
- break;
- case _S_opcode_subexpr_begin:
- __ostr << __id << " [label=\"" << __id << "\\nSBEGIN "
- << _M_subexpr << "\"];\n"
- << __id << " -> " << _M_next << " [label=\"epsilon\"];\n";
- break;
- case _S_opcode_subexpr_end:
- __ostr << __id << " [label=\"" << __id << "\\nSEND "
- << _M_subexpr << "\"];\n"
- << __id << " -> " << _M_next << " [label=\"epsilon\"];\n";
- break;
- case _S_opcode_dummy:
- break;
- case _S_opcode_match:
- __ostr << __id << " [label=\"" << __id << "\\nMATCH\"];\n"
- << __id << " -> " << _M_next << " [label=\"<match>\"];\n";
- break;
- case _S_opcode_accept:
- __ostr << __id << " [label=\"" << __id << "\\nACC\"];\n" ;
- break;
- default:
- _GLIBCXX_DEBUG_ASSERT(false);
- break;
- }
- return __ostr;
+ case _S_opcode_alternative:
+ __ostr << __id << " [label=\"" << __id << "\\nALT\"];\n"
+ << __id << " -> " << _M_next
+ << " [label=\"epsilon\", tailport=\"s\"];\n"
+ << __id << " -> " << _M_alt
+ << " [label=\"epsilon\", tailport=\"n\"];\n";
+ break;
+ case _S_opcode_backref:
+ __ostr << __id << " [label=\"" << __id << "\\nBACKREF "
+ << _M_subexpr << "\"];\n"
+ << __id << " -> " << _M_next << " [label=\"<match>\"];\n";
+ break;
+ case _S_opcode_line_begin_assertion:
+ __ostr << __id << " [label=\"" << __id << "\\nLINE_BEGIN \"];\n"
+ << __id << " -> " << _M_next << " [label=\"epsilon\"];\n";
+ break;
+ case _S_opcode_line_end_assertion:
+ __ostr << __id << " [label=\"" << __id << "\\nLINE_END \"];\n"
+ << __id << " -> " << _M_next << " [label=\"epsilon\"];\n";
+ break;
+ case _S_opcode_word_boundary:
+ __ostr << __id << " [label=\"" << __id << "\\nWORD_BOUNDRY "
+ << _M_neg << "\"];\n"
+ << __id << " -> " << _M_next << " [label=\"epsilon\"];\n";
+ break;
+ case _S_opcode_subexpr_lookahead:
+ __ostr << __id << " [label=\"" << __id << "\\nLOOK_AHEAD\"];\n"
+ << __id << " -> " << _M_next
+ << " [label=\"epsilon\", tailport=\"s\"];\n"
+ << __id << " -> " << _M_alt
+ << " [label=\"<assert>\", tailport=\"n\"];\n";
+ break;
+ case _S_opcode_subexpr_begin:
+ __ostr << __id << " [label=\"" << __id << "\\nSBEGIN "
+ << _M_subexpr << "\"];\n"
+ << __id << " -> " << _M_next << " [label=\"epsilon\"];\n";
+ break;
+ case _S_opcode_subexpr_end:
+ __ostr << __id << " [label=\"" << __id << "\\nSEND "
+ << _M_subexpr << "\"];\n"
+ << __id << " -> " << _M_next << " [label=\"epsilon\"];\n";
+ break;
+ case _S_opcode_dummy:
+ break;
+ case _S_opcode_match:
+ __ostr << __id << " [label=\"" << __id << "\\nMATCH\"];\n"
+ << __id << " -> " << _M_next << " [label=\"<match>\"];\n";
+ break;
+ case _S_opcode_accept:
+ __ostr << __id << " [label=\"" << __id << "\\nACC\"];\n" ;
+ break;
+ default:
+ _GLIBCXX_DEBUG_ASSERT(false);
+ break;
}
+ return __ostr;
+ }
template<typename _CharT, typename _TraitsT>
- std::ostream& _NFA<_CharT, _TraitsT>::
- _M_dot(std::ostream& __ostr) const
+ std::ostream&
+ _NFA<_CharT, _TraitsT>::_M_dot(std::ostream& __ostr) const
{
__ostr << "digraph _Nfa {\n"
- << " rankdir=LR;\n";
+ " rankdir=LR;\n";
for (size_t __i = 0; __i < this->size(); ++__i)
- { this->at(__i)._M_dot(__ostr, __i); }
+ (*this)[__i]._M_dot(__ostr, __i);
__ostr << "}\n";
return __ostr;
}
#endif
template<typename _CharT, typename _TraitsT>
- _StateIdT _NFA<_CharT, _TraitsT>::
- _M_insert_backref(size_t __index)
+ _StateIdT
+ _NFA<_CharT, _TraitsT>::_M_insert_backref(size_t __index)
{
// To figure out whether a backref is valid, a stack is used to store
// unfinished sub-expressions. For example, when parsing
@@ -157,18 +155,18 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
// time, "\\2" is valid, but "\\1" and "\\3" are not.
if (__index >= _M_subexpr_count)
__throw_regex_error(regex_constants::error_backref);
- for (auto __it : _M_paren_stack)
+ for (auto __it : this->_M_paren_stack)
if (__index == __it)
__throw_regex_error(regex_constants::error_backref);
- _M_has_backref = true;
+ this->_M_has_backref = true;
_StateT __tmp(_S_opcode_backref);
__tmp._M_backref_index = __index;
- return _M_insert_state(__tmp);
+ return _M_insert_state(std::move(__tmp));
}
template<typename _CharT, typename _TraitsT>
- void _NFA<_CharT, _TraitsT>::
- _M_eliminate_dummy()
+ void
+ _NFA<_CharT, _TraitsT>::_M_eliminate_dummy()
{
for (auto& __it : *this)
{
@@ -185,8 +183,8 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
// Just apply DFS on the sequence and re-link their links.
template<typename _CharT, typename _TraitsT>
- _StateSeq<_CharT, _TraitsT> _StateSeq<_CharT, _TraitsT>::
- _M_clone()
+ _StateSeq<_CharT, _TraitsT>
+ _StateSeq<_CharT, _TraitsT>::_M_clone()
{
std::map<_StateIdT, _StateIdT> __m;
std::stack<_StateIdT> __stack;
@@ -55,8 +55,8 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
const _TraitsT& __traits, _FlagT __flags);
std::shared_ptr<_RegexT>
- _M_get_nfa() const
- { return make_shared<_RegexT>(_M_nfa); }
+ _M_get_nfa()
+ { return make_shared<_RegexT>(std::move(_M_nfa)); }
private:
typedef _Scanner<_FwdIter> _ScannerT;
@@ -147,11 +147,9 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
_M_assertion()
{
if (_M_match_token(_ScannerT::_S_token_line_begin))
- _M_stack.push(_StateSeqT(_M_nfa, _M_nfa.
- _M_insert_line_begin()));
+ _M_stack.push(_StateSeqT(_M_nfa, _M_nfa._M_insert_line_begin()));
else if (_M_match_token(_ScannerT::_S_token_line_end))
- _M_stack.push(_StateSeqT(_M_nfa, _M_nfa.
- _M_insert_line_end()));
+ _M_stack.push(_StateSeqT(_M_nfa, _M_nfa._M_insert_line_end()));
else if (_M_match_token(_ScannerT::_S_token_word_bound))
// _M_value[0] == 'n' means it's negtive, say "not word boundary".
_M_stack.push(_StateSeqT(_M_nfa, _M_nfa.
@@ -305,7 +303,7 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
_M_traits, _M_flags);
__matcher._M_add_character_class(_M_value);
_M_stack.push(_StateSeqT(_M_nfa,
- _M_nfa._M_insert_matcher(__matcher)));
+ _M_nfa._M_insert_matcher(std::move(__matcher))));
}
else if (_M_match_token(_ScannerT::_S_token_subexpr_no_group_begin))
{
@@ -343,7 +341,8 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
_BMatcherT __matcher(__neg, _M_traits, _M_flags);
while (!_M_match_token(_ScannerT::_S_token_bracket_end))
_M_expression_term(__matcher);
- _M_stack.push(_StateSeqT(_M_nfa, _M_nfa._M_insert_matcher(__matcher)));
+ _M_stack.push(_StateSeqT(_M_nfa,
+ _M_nfa._M_insert_matcher(std::move(__matcher))));
return true;
}
@@ -432,8 +431,8 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
}
template<typename _CharT, typename _TraitsT>
- bool _BracketMatcher<_CharT, _TraitsT>::
- operator()(_CharT __ch) const
+ bool
+ _BracketMatcher<_CharT, _TraitsT>::operator()(_CharT __ch) const
{
bool __ret = false;
if (_M_traits.isctype(__ch, _M_class_set)
@@ -53,7 +53,7 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
*/
template<typename _BiIter, typename _Alloc, typename _TraitsT,
- bool __dfs_mode>
+ bool __dfs_mode>
class _Executor
{
public:
@@ -117,9 +117,9 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
bool
_M_is_word(_CharT __ch) const
{
- static const _CharT __s = 'w';
+ static const _CharT __s[2] = { 'w' };
return _M_re._M_traits.isctype
- (__ch, _M_re._M_traits.lookup_classname(&__s, &__s+1));
+ (__ch, _M_re._M_traits.lookup_classname(__s, __s+1));
}
bool
This creates new base classes for the parts of _State and _NFA which are not dependent on template parameters, and replaces some copies with moves. 2013-11-08 Jonathan Wakely <jwakely.gcc@gmail.com> * include/bits/regex_automaton.h (__detail::_State): Split non-dependent parts into new _State_base. (__detail::_NFA): Likewise for _NFA_base. Use std::move() to avoid copies when inserting _MatcherT and _StateT objects. * include/bits/regex_automaton.tcc: Move member definitions to base class. Qualify dependent names. * include/bits/regex_compiler.h (__detail::_Compiler::_M_get_nfa): Make non-const and use std::move to avoid copying. * include/bits/regex_compiler.tcc: Likewise. * include/bits/regex_executor.h (__detail::_Executor::_M_is_word): Use array, so past-the-end iterator is valid. Tested x86_64-linux, committed to trunk. commit 3404a066d12f458e2b441798dbd70dfa21ef537d Author: Jonathan Wakely <jwakely.gcc@gmail.com> Date: Wed Nov 6 09:09:35 2013 +0000 * include/bits/regex_automaton.h (__detail::_State): Split non-dependent parts into new _State_base. (__detail::_NFA): Likewise for _NFA_base. Use std::move() to avoid copies when inserting _MatcherT and _StateT objects. * include/bits/regex_automaton.tcc: Move member definitions to base class. Qualify dependent names. * include/bits/regex_compiler.h (__detail::_Compiler::_M_get_nfa): Make non-const and use std::move to avoid copying. * include/bits/regex_compiler.tcc: Likewise. * include/bits/regex_executor.h (__detail::_Executor::_M_is_word): Use array, so past-the-end iterator is valid.